From yosefmel at post.tau.ac.il  Tue Jul  1 01:09:08 2008
From: yosefmel at post.tau.ac.il (Yosef Meller)
Date: Tue, 1 Jul 2008 08:09:08 +0300
Subject: [SciPy-user] single elements and arrays
In-Reply-To: <3d375d730806301157t13d8d3eejc62b1f9ced005ed7@mail.gmail.com>
References: <C24B3A1D-74FD-4531-BC93-34364B2F2670@columbia.edu>
	<02AA0899-7A57-4CD3-8DF3-388135BD5D0A@columbia.edu>
	<3d375d730806301157t13d8d3eejc62b1f9ced005ed7@mail.gmail.com>
Message-ID: <200807010809.08232.yosefmel@post.tau.ac.il>

On Monday 30 June 2008 21:57:18 Robert Kern wrote:
> On Mon, Jun 30, 2008 at 13:14, Gideon Simpson <grs2103 at columbia.edu> wrote:
> > foo finds a root where x is a parameter in the equation to be solved.
> > If x is an array, I iterate through the elements of the array.
>
> In that case, just special case it. Use numpy.isscalar() to do the test.

Or else, put this at the beginning:
x = numpy.atleast_1d(x)



From robert.kern at gmail.com  Tue Jul  1 01:13:58 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Tue, 1 Jul 2008 00:13:58 -0500
Subject: [SciPy-user] single elements and arrays
In-Reply-To: <200807010809.08232.yosefmel@post.tau.ac.il>
References: <C24B3A1D-74FD-4531-BC93-34364B2F2670@columbia.edu>
	<02AA0899-7A57-4CD3-8DF3-388135BD5D0A@columbia.edu>
	<3d375d730806301157t13d8d3eejc62b1f9ced005ed7@mail.gmail.com>
	<200807010809.08232.yosefmel@post.tau.ac.il>
Message-ID: <3d375d730806302213w7877cd8dl6f94540a66ac0da5@mail.gmail.com>

On Tue, Jul 1, 2008 at 00:09, Yosef Meller <yosefmel at post.tau.ac.il> wrote:
> On Monday 30 June 2008 21:57:18 Robert Kern wrote:
>> On Mon, Jun 30, 2008 at 13:14, Gideon Simpson <grs2103 at columbia.edu> wrote:
>> > foo finds a root where x is a parameter in the equation to be solved.
>> > If x is an array, I iterate through the elements of the array.
>>
>> In that case, just special case it. Use numpy.isscalar() to do the test.
>
> Or else, put this at the beginning:
> x = numpy.atleast_1d(x)

Presumably, he also wants to return a scalar if given a scalar. The
general outline would probably look like this:

def foo(x):
    xisscalar = numpy.isscalar(x)
    x = numpy.atleast_1d(x)
    y = ...
    if xisscalar:
        y = y[0]
    return y

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From peridot.faceted at gmail.com  Tue Jul  1 01:49:23 2008
From: peridot.faceted at gmail.com (Anne Archibald)
Date: Tue, 1 Jul 2008 01:49:23 -0400
Subject: [SciPy-user] single elements and arrays
In-Reply-To: <02AA0899-7A57-4CD3-8DF3-388135BD5D0A@columbia.edu>
References: <C24B3A1D-74FD-4531-BC93-34364B2F2670@columbia.edu>
	<3d375d730806301052k703db955iac318c6bcfa7f200@mail.gmail.com>
	<02AA0899-7A57-4CD3-8DF3-388135BD5D0A@columbia.edu>
Message-ID: <ce557a360806302249p7f5f07a7h6d086cac426ad3b7@mail.gmail.com>

2008/6/30 Gideon Simpson <grs2103 at columbia.edu>:
> foo finds a root where x is a parameter in the equation to be solved.
> If x is an array, I iterate through the elements of the array.

If you're actually *iterating*, in the sense of using a python loop to
go through every element, there will be substantial overhead -
something like a few tens or hundreds of floating-point operations for
every trip through the loop -  involved in looping. If the function
itself is reasonably fast and if you can write it in terms of vectors,
that will be much faster. It may also, with some care, operate
transparently on scalars.

However, if you can't - let's say it's a numerical root-finding using
brentq - then there's a handy tool to provide python-level looping:
the vectorize decorator:

@np.vectorize
def acos(y):
    return scipy.optimize.brentq(lambda x: y-np.cos(x),0,np.pi)

This makes acos behave a little like a ufunc from the user's point of
view: you can hand it a scalar or an array of arbitrary dimensionality
and it will iterate as appropriate. The iteration passes through
python, necessarily since the function being wrapped is in python, so
it won't be fast, but it is convenient.

Anne


From eiffleduarte at gmail.com  Tue Jul  1 09:30:07 2008
From: eiffleduarte at gmail.com (Marcus Vinicius Eiffle Duarte)
Date: Tue, 1 Jul 2008 10:30:07 -0300
Subject: [SciPy-user] Is there a way to use Sundials with scipy?
Message-ID: <280253810807010630k2e4fbdfey7018aa55641badfb@mail.gmail.com>

I am a new user of both python and scipy and was searching for a way to use
sundials with scipy.
I found some old messages in the history of the mailing list dating back to
2004, but nothing indicating wether such a wrapper was effectively
developed.
At the present, is it possible to use saundials with scipy?

Thanks,

Marcus Vinicius Eiffle Duarte
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080701/b360d48c/attachment.html>

From cimrman3 at ntc.zcu.cz  Tue Jul  1 10:13:46 2008
From: cimrman3 at ntc.zcu.cz (Robert Cimrman)
Date: Tue, 01 Jul 2008 16:13:46 +0200
Subject: [SciPy-user] ANN: SfePy 00.46.02
Message-ID: <486A3B9A.8060106@ntc.zcu.cz>

I am pleased announce the release of SfePy 00.46.02.

SfePy is a finite element analysis software in Python, based primarily
on Numpy and SciPy.

Mailing lists, issue tracking, mercurial repository: http://sfepy.org
Home page: http://sfepy.kme.zcu.cz

Major improvements:
- alternative short syntax for specifying essential boundary conditions, 
variables and regions
- manufactured solutions tests:
     - SymPy support
- site configuration now via script/config.py + site_cfg.py
- new solvers
- new terms

For more information on this release, see
http://sfepy.googlecode.com/svn/web/releases/004602_RELEASE_NOTES.txt

If you happen to come to Leipzig for EuroSciPy 2008, see you there!

Best regards,
Robert Cimrman & SfePy developers



From nwagner at iam.uni-stuttgart.de  Tue Jul  1 13:04:18 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Tue, 01 Jul 2008 19:04:18 +0200
Subject: [SciPy-user] Is there a way to use Sundials with scipy?
In-Reply-To: <280253810807010630k2e4fbdfey7018aa55641badfb@mail.gmail.com>
References: <280253810807010630k2e4fbdfey7018aa55641badfb@mail.gmail.com>
Message-ID: <web-104081430@uni-stuttgart.de>

On Tue, 1 Jul 2008 10:30:07 -0300
  "Marcus Vinicius Eiffle Duarte" <eiffleduarte at gmail.com> 
wrote:
> I am a new user of both python and scipy and was 
>searching for a way to use
> sundials with scipy.
> I found some old messages in the history of the mailing 
>list dating back to
> 2004, but nothing indicating wether such a wrapper was 
>effectively
> developed.
> At the present, is it possible to use saundials with 
>scipy?
> 
> Thanks,
> 
> Marcus Vinicius Eiffle Duarte

Hi Marcus,

you can use

svn co 
https://pysundials.svn.sourceforge.net/svnroot/pysundials/ 
pysundials
  
to get the latest version of pysundials.
Sundials is required of course.

* Download and untar the complete SUNDIALS suite.
* $ ./configure --enable-shared
* $ make && make install


* $ python setup.py install


Further information is given in pysundials/doc/README

Cheers,

          Nils
  


From travis at enthought.com  Wed Jul  2 10:32:43 2008
From: travis at enthought.com (Travis Vaught)
Date: Wed, 2 Jul 2008 09:32:43 -0500
Subject: [SciPy-user] Enthought Python Distribution
Message-ID: <78A3D219-9C3C-4B4C-BA77-8B63ED1191BF@enthought.com>

Greetings,

We're pleased to announce the beta release of the Enthought Python  
Distribution for *Mac OS X*.

http://www.enthought.com/products/epd.php

This release should safely install alongside other existing Python  
installations on your Mac.  With the Mac OS X platform support, EPD  
now provides a consistent scientific application tool set across three  
major platforms (Windows, RedHat Linux (32 and 64 bit) and OS X). This  
is a _beta_ release, so install at your own risk.  Please provide any  
feedback to info at enthought.com.  See the included EPD Readme.txt for  
instructions and known issues.

About EPD
---------
The Enthought Python Distribution (EPD) is a "kitchen-sink-included"  
distribution of the Python? Programming Language, including over 60  
additional tools and libraries. The EPD bundle includes the following  
major packages:

Python       Core Python
NumPy        Multidimensional arrays and fast numerics for Python
SciPy        Scientific Library for Python
Enthought Tool Suite (ETS)   A suite of tools including:
   Traits     Manifest typing, validation, visualization, delegation,  
etc.
   Mayavi     3D interactive data exploration environment.
   Chaco      Advanced 2D plotting toolkit for interactive 2D  
visualization.
   Kiva       2D drawing library in the spirit of DisplayPDF.
   Enable     Object-based canvas for interacting with 2D components  
and widgets.
Matplotlib   2D plotting library
wxPython     Cross-platform windowing and widget library.
Visualization Toolkit (VTK)  3D visualization framework

There are many more included packages as well.  There's a complete  
list here:

http://www.enthought.com/products/epdlibraries.php

License
-------
EPD is a bundle of software--every piece of which is available for  
free under various open-source licenses.  The bundle itself is offered  
as a free download to academic and individual hobbyist use.   
Commercial and non-degree granting institutions and agencies may  
purchase individual subscriptions for the bundle (http://www.enthought.com/products/order.php?ver=MacOSX 
) or contact Enthought to discuss an Enterprise license (http://www.enthought.com/products/enterprise.php 
).  Please see the FAQ for further explanation about how the software  
came together. (http://www.enthought.com/products/epdfaq.php)

Thanks,

Travis



From fperez.net at gmail.com  Wed Jul  2 13:10:10 2008
From: fperez.net at gmail.com (Fernando Perez)
Date: Wed, 2 Jul 2008 10:10:10 -0700
Subject: [SciPy-user] Enthought Python Distribution
In-Reply-To: <78A3D219-9C3C-4B4C-BA77-8B63ED1191BF@enthought.com>
References: <78A3D219-9C3C-4B4C-BA77-8B63ED1191BF@enthought.com>
Message-ID: <db6b5ecc0807021010q15606db8gcc00197023c4547b@mail.gmail.com>

Hey Travis,

On Wed, Jul 2, 2008 at 7:32 AM, Travis Vaught <travis at enthought.com> wrote:
> Greetings,
>
> We're pleased to announce the beta release of the Enthought Python
> Distribution for *Mac OS X*.

First of all, this is *fantastic*.  OSX has been a major pain for us
here in terms of distribution, so I am extremely happy and grateful to
you guys for putting this out.

Two minor requests: any chance in the next release, you could update
ipython and matplotlib?  You are shipping 0.8.1 and 0.91.2, both of
which are fairly old and the current versions of both do have
reasonably useful improvements and fixes (ipython has threading fixes
that impact plotting, and mpl has all the new mathtext code that gives
good math rendering without needing latex).

In any case, a big cheer from me :)

Regards,

f


From grs2103 at columbia.edu  Wed Jul  2 23:19:51 2008
From: grs2103 at columbia.edu (Gideon Simpson)
Date: Wed, 2 Jul 2008 23:19:51 -0400
Subject: [SciPy-user] 0.7 test results
Message-ID: <F43632A9-6ECD-4773-96A2-D23CD1472EC5@columbia.edu>

On an OS X 10.5.4 machine with fink python 2.5.2 and numpy 1.1.0, I  
get the following output when running scipy.test() with scipy  
0.7.0.dev4518:

/opt/lib/python2.5/site-packages/scipy/sparse/linalg/dsolve/ 
linsolve.py:20: DeprecationWarning: scipy.sparse.linalg.dsolve.umfpack  
will be removed, install scikits.umfpack instead
   ' install scikits.umfpack instead', DeprecationWarning )
/opt/lib/python2.5/site-packages/scipy/linsolve/__init__.py:4:  
DeprecationWarning: scipy.linsolve has moved to  
scipy.sparse.linalg.dsolve
   warn('scipy.linsolve has moved to scipy.sparse.linalg.dsolve',  
DeprecationWarning)
/opt/lib/python2.5/site-packages/scipy/splinalg/__init__.py:3:  
DeprecationWarning: scipy.splinalg has moved to scipy.sparse.linalg
   warn('scipy.splinalg has moved to scipy.sparse.linalg',  
DeprecationWarning)
....(1, 1) (0,)
(2, 2) (1,)
(3, 3) (3,)
(4, 4) (6,)
(5, 5) (10,)
(6, 6) (15,)
(7, 7) (21,)
(8, 8) (28,)
(9, 9) (36,)
................/opt/lib/python2.5/site-packages/scipy/cluster/vq.py: 
570: UserWarning: One of the clusters is empty. Re-run kmean with a  
different initialization.
   warnings.warn("One of the clusters is empty. "
exception raised as expected: One of the clusters is empty. Re-run  
kmean with a different initialization.
................................................................./opt/ 
lib/python2.5/site-packages/scipy/interpolate/fitpack2.py:479:  
UserWarning:
The coefficients of the spline returned have been computed as the
minimal norm least-squares solution of a (numerically) rank deficient
system (deficiency=7). If deficiency is large, the results may be
inaccurate. Deficiency may strongly depend on the value of eps.
   warnings.warn(message)
..../opt/lib/python2.5/site-packages/scipy/interpolate/fitpack2.py: 
420: UserWarning:
The required storage space exceeds the available storage space: nxest
or nyest too small, or s too small.
The weighted least-squares spline corresponds to the current set of
knots.
   warnings.warn(message)
.....................................................................................
Don't worry about a warning regarding the number of bytes read.
Warning: 1000000 bytes requested, 20 bytes read.
./opt/lib/python2.5/site-packages/numpy/lib/utils.py:114:  
DeprecationWarning: write_array is deprecated
   warnings.warn(str1, DeprecationWarning)
/opt/lib/python2.5/site-packages/numpy/lib/utils.py:114:  
DeprecationWarning: read_array is deprecated
   warnings.warn(str1, DeprecationWarning)
....................../opt/lib/python2.5/site-packages/numpy/lib/ 
utils.py:114: DeprecationWarning: npfile is deprecated
   warnings.warn(str1, DeprecationWarning)
............................caxpy:n=4
..caxpy:n=3
....ccopy:n=4
..ccopy:n=3
.............cscal:n=4
....cswap:n=4
..cswap:n=3
.....daxpy:n=4
..daxpy:n=3
....dcopy:n=4
..dcopy:n=3
.............dscal:n=4
....dswap:n=4
..dswap:n=3
.....saxpy:n=4
..saxpy:n=3
....scopy:n=4
..scopy:n=3
.............sscal:n=4
....sswap:n=4
..sswap:n=3
.....zaxpy:n=4
..zaxpy:n=3
....zcopy:n=4
..zcopy:n=3
.............zscal:n=4
....zswap:n=4
..zswap:n=3
..
****************************************************************
WARNING: clapack module is empty
-----------
See scipy/INSTALL.txt for troubleshooting.
Notes:
* If atlas library is not found by numpy/distutils/system_info.py,
   then scipy uses flapack instead of clapack.
****************************************************************

..........................................NO ATLAS INFO AVAILABLE
.........................................
****************************************************************
WARNING: cblas module is empty
-----------
See scipy/INSTALL.txt for troubleshooting.
Notes:
* If atlas library is not found by numpy/distutils/system_info.py,
   then scipy uses fblas instead of cblas.
****************************************************************

...............................................................................................caxpy:n 
=4
..caxpy:n=3
....ccopy:n=4
..ccopy:n=3
.............cscal:n=4
....cswap:n=4
..cswap:n=3
.....daxpy:n=4
..daxpy:n=3
....dcopy:n=4
..dcopy:n=3
.............dscal:n=4
....dswap:n=4
..dswap:n=3
.....saxpy:n=4
..saxpy:n=3
....scopy:n=4
..scopy:n=3
.............sscal:n=4
....sswap:n=4
..sswap:n=3
.....zaxpy:n=4
..zaxpy:n=3
....zcopy:n=4
..zcopy:n=3
.............zscal:n=4
....zswap:n=4
..zswap:n=3
....
****************************************************************
WARNING: clapack module is empty
-----------
See scipy/INSTALL.txt for troubleshooting.
Notes:
* If atlas library is not found by numpy/distutils/system_info.py,
   then scipy uses flapack instead of clapack.
****************************************************************

...Result may be inaccurate, approximate err = 1.23518201169e-08
...Result may be inaccurate, approximate err = 7.27595761418e-12
.....SSS 
........................................................................................................................................................................................................................................................................................................................................................................................................./opt 
/lib/python2.5/site-packages/scipy/ndimage/_registration.py:25:  
UserWarning: The registration code is under heavy development and  
therefore the public API will change in the future.  The NIPY group is  
actively working on this code, and has every intention of generalizing  
this for the Scipy community.  Use this module minimally, if at all,  
until it this warning is removed.
   warnings.warn(_msg, UserWarning)
...E..EEE./opt/lib/python2.5/site-packages/scipy/ndimage/_segmenter.py: 
30: UserWarning: The segmentation code is under heavy development and  
therefore the public API will change in the future.  The NIPY group is  
actively working on this code, and has every intention of generalizing  
this for the Scipy community.  Use this module minimally, if at all,  
until it this warning is removed.
   warnings.warn(_msg, UserWarning)
...F 
................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................F 
......................EE 
......................................................................................Ties 
  preclude use of exact statistic.
..Ties preclude use of exact statistic.
........................................................../opt/lib/ 
python2.5/site-packages/numpy/lib/function_base.py:166: FutureWarning:
         The semantics of histogram will be modified in
         release 1.2 to improve outlier handling. The new behavior can  
be
         obtained using new=True. Note that the new version accepts/ 
returns
         the bin edges instead of the left bin edges.
         Please read the docstring for more information.
   Please read the docstring for more information.""", FutureWarning)
/opt/lib/python2.5/site-packages/numpy/lib/function_base.py:181:  
FutureWarning:
                 Outliers handling will change in version 1.2.
                 Please read the docstring for details.
   Please read the docstring for details.""", FutureWarning)
................................................................................................warning 
: specified build_dir '_bad_path_' does not exist or is not writable.  
Trying default locations
...warning: specified build_dir '..' does not exist or is not  
writable. Trying default locations
..warning: specified build_dir '_bad_path_' does not exist or is not  
writable. Trying default locations
...warning: specified build_dir '..' does not exist or is not  
writable. Trying default locations
............................building extensions here: /Users/ 
gideon/.python25_compiled/m11
................................................................................................
======================================================================
ERROR: Execute a single test.  Returns a success boolean
----------------------------------------------------------------------
TypeError: exec_test() takes exactly 4 arguments (1 given)

======================================================================
ERROR: Run one test with arguments
----------------------------------------------------------------------
TypeError: run_test() takes exactly 3 arguments (1 given)

======================================================================
ERROR: Run many tests with a common setUp/tearDown.
----------------------------------------------------------------------
TypeError: run_tests() takes exactly 3 arguments (1 given)

======================================================================
ERROR: set_testMethodDoc (test_regression.ParametricTestCase)
----------------------------------------------------------------------
TypeError: set_testMethodDoc() takes exactly 2 arguments (1 given)

======================================================================
ERROR: test_huber (test_scale.TestScale)
----------------------------------------------------------------------
Traceback (most recent call last):
   File "/opt/lib/python2.5/site-packages/scipy/stats/models/tests/ 
test_scale.py", line 35, in test_huber
     m = scale.huber(X)
   File "/opt/lib/python2.5/site-packages/scipy/stats/models/robust/ 
scale.py", line 82, in __call__
     for donothing in self:
   File "/opt/lib/python2.5/site-packages/scipy/stats/models/robust/ 
scale.py", line 102, in next
     scale = N.sum(subset * (a - mu)**2, axis=self.axis) / (self.n *  
Huber.gamma - N.sum(1. - subset, axis=self.axis) * Huber.c**2)
   File "/opt/lib/python2.5/site-packages/numpy/core/fromnumeric.py",  
line 994, in sum
     return sum(axis, dtype, out)
TypeError: only length-1 arrays can be converted to Python scalars

======================================================================
ERROR: test_huberaxes (test_scale.TestScale)
----------------------------------------------------------------------
Traceback (most recent call last):
   File "/opt/lib/python2.5/site-packages/scipy/stats/models/tests/ 
test_scale.py", line 40, in test_huberaxes
     m = scale.huber(X, axis=0)
   File "/opt/lib/python2.5/site-packages/scipy/stats/models/robust/ 
scale.py", line 82, in __call__
     for donothing in self:
   File "/opt/lib/python2.5/site-packages/scipy/stats/models/robust/ 
scale.py", line 102, in next
     scale = N.sum(subset * (a - mu)**2, axis=self.axis) / (self.n *  
Huber.gamma - N.sum(1. - subset, axis=self.axis) * Huber.c**2)
   File "/opt/lib/python2.5/site-packages/numpy/core/fromnumeric.py",  
line 994, in sum
     return sum(axis, dtype, out)
TypeError: only length-1 arrays can be converted to Python scalars

======================================================================
FAIL: test_texture2 (test_segment.TestSegment)
----------------------------------------------------------------------
Traceback (most recent call last):
   File "/opt/lib/python2.5/site-packages/scipy/ndimage/tests/ 
test_segment.py", line 152, in test_texture2
     assert_array_almost_equal(tem0, truth_tem0, decimal=6)
   File "/opt/lib/python2.5/site-packages/numpy/testing/utils.py",  
line 255, in assert_array_almost_equal
     header='Arrays are not almost equal')
   File "/opt/lib/python2.5/site-packages/numpy/testing/utils.py",  
line 240, in assert_array_compare
     assert cond, msg
AssertionError:
Arrays are not almost equal

(mismatch 66.6666666667%)
  x: array([  0.00000000e+00,   0.00000000e+00,   0.00000000e+00,
          0.00000000e+00,   0.00000000e+00,   0.00000000e+00,
          1.91816598e-01,   1.02515288e-01,   9.30087343e-02,...
  y: array([ 0.        ,  0.        ,  0.        ,  0.        ,   
0.        ,
         0.        ,  0.13306101,  0.08511007,  0.05084148,  0.07550675,
         0.4334695 ,  0.03715914,  0.00289055,  0.02755581,  0.48142046,
         0.03137803,  0.00671277,  0.51568902,  0.01795249,   
0.49102375,  1.        ], dtype=float32)

======================================================================
FAIL: test_namespace (test_formula.TestFormula)
----------------------------------------------------------------------
Traceback (most recent call last):
   File "/opt/lib/python2.5/site-packages/scipy/stats/models/tests/ 
test_formula.py", line 119, in test_namespace
     self.assertEqual(xx.namespace, Y.namespace)
AssertionError: {} != {'Y': array([ 0,  2,  4,  6,  8, 10, 12, 14, 16,  
18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40,
        42, 44, 46, 48, 50, 52, 54, 56, 58, 60, 62, 64, 66, 68, 70,  
72, 74, 76, 78, 80, 82,
        84, 86, 88, 90, 92, 94, 96, 98]), 'X': array([ 0,  1,  2,  3,   
4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20,
        21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35,  
36, 37, 38, 39, 40, 41,
        42, 43, 44, 45, 46, 47, 48, 49])}

----------------------------------------------------------------------
Ran 2275 tests in 23.705s

FAILED (SKIP=3, errors=6, failures=2)




From nwagner at iam.uni-stuttgart.de  Thu Jul  3 05:24:01 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Thu, 03 Jul 2008 11:24:01 +0200
Subject: [SciPy-user] array manipulation
Message-ID: <web-104160546@uni-stuttgart.de>

Hi all,

How can I remove duplicate entries from an array ?

  
>>> Omega
array([ 157.08,  314.16,  471.24,  157.08,  157.08, 
 314.16,  314.16,
         471.24,  471.24])

I am looking for an array

Omega_new = array(([157.08,  314.16,  471.24]))

Nils


From robert.kern at gmail.com  Thu Jul  3 05:26:27 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Thu, 3 Jul 2008 04:26:27 -0500
Subject: [SciPy-user] array manipulation
In-Reply-To: <web-104160546@uni-stuttgart.de>
References: <web-104160546@uni-stuttgart.de>
Message-ID: <3d375d730807030226j6d11aea9t3eb16ab321629272@mail.gmail.com>

On Thu, Jul 3, 2008 at 04:24, Nils Wagner <nwagner at iam.uni-stuttgart.de> wrote:
> Hi all,
>
> How can I remove duplicate entries from an array ?
>
>
>>>> Omega
> array([ 157.08,  314.16,  471.24,  157.08,  157.08,
>  314.16,  314.16,
>         471.24,  471.24])
>
> I am looking for an array
>
> Omega_new = array(([157.08,  314.16,  471.24]))

numpy.unique1d()

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From lorenzo.isella at gmail.com  Thu Jul  3 10:39:41 2008
From: lorenzo.isella at gmail.com (Lorenzo Isella)
Date: Thu, 3 Jul 2008 16:39:41 +0200
Subject: [SciPy-user] Real Array Expressed as Complex Array
Message-ID: <a2b3004b0807030739r1032e8d4j25358927681475bb@mail.gmail.com>

Dear All,
I am bit puzzled: I was plotting a (rather complicated) analytical
potential for which an analytical form is available.
When asking to print out the value of the potential at the cut-off:

print "at the cut-off, the dimensionless potential takes the value, ",
pot_ext_dimensionless[-1]

I got the following:

at the cut-off, the dimensionless potential takes the value,
(-6.48829965957e-06+0j)

Now, since the potential I am coding via a function has to be a real
function, I checked that the real part was always zero (as it should).
Since the result was that the array pot_ext_dimensionless is real, how
comes that it is expressed as a complex array (though the imaginary
part is always zero)?
It is true, however, that the potential could become complex for
certain (physically unsound) choices of some parameters.
Any suggestions?
Cheers

Lorenzo


From kwmsmith at gmail.com  Thu Jul  3 11:14:16 2008
From: kwmsmith at gmail.com (Kurt Smith)
Date: Thu, 3 Jul 2008 10:14:16 -0500
Subject: [SciPy-user] Real Array Expressed as Complex Array
In-Reply-To: <a2b3004b0807030739r1032e8d4j25358927681475bb@mail.gmail.com>
References: <a2b3004b0807030739r1032e8d4j25358927681475bb@mail.gmail.com>
Message-ID: <c266a7930807030814v3268b9e5rca34b98bb7b84b57@mail.gmail.com>

On Thu, Jul 3, 2008 at 9:39 AM, Lorenzo Isella <lorenzo.isella at gmail.com> wrote:
> Dear All,
> I am bit puzzled: I was plotting a (rather complicated) analytical
> potential for which an analytical form is available.
> When asking to print out the value of the potential at the cut-off:
>
> print "at the cut-off, the dimensionless potential takes the value, ",
> pot_ext_dimensionless[-1]
>
> I got the following:
>
> at the cut-off, the dimensionless potential takes the value,
> (-6.48829965957e-06+0j)
>
> Now, since the potential I am coding via a function has to be a real
> function, I checked that the real part was always zero (as it should).

I think you mean "...the *imaginary* part was always zero..."

> Since the result was that the array pot_ext_dimensionless is real, how
> comes that it is expressed as a complex array (though the imaginary
> part is always zero)?

It all depends on how you're calculating the pot_ext_dimensionless
array; clearly somewhere in there an operation makes it complex.
You'll have to show us how it's calculated.

You can always access the (real,imaginary) part of a complex array
with (pot_ext_dimensionless.real, pot_ext_dimensionless.imag)

But be careful, these arrays are not contiguous (they're a view into
the complex array). That wrinkle has bitten me before, but I can't
quite recall the circumstances.  You can always make them contiguous
with numpy.ascontiguousarray().


> It is true, however, that the potential could become complex for
> certain (physically unsound) choices of some parameters.
> Any suggestions?
> Cheers
>
> Lorenzo
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>


From markbak at gmail.com  Thu Jul  3 13:15:35 2008
From: markbak at gmail.com (Mark Bakker)
Date: Thu, 3 Jul 2008 19:15:35 +0200
Subject: [SciPy-user] Enthought Python Distribution
Message-ID: <6946b9500807031015j392e8a20ic5b7636208e086d6@mail.gmail.com>

Hello Travis and other Enthought guru's -

I have benefited from the Enthought distribution for years, and really want
to switch to a Mac, so this is great news.

My main concern with the Enthought distribution is the frequency of new
releases. The guys at Python(x,y) are exceptionally quick at updating. That
may be hard to match (and cost a lot of effort). What are your plans at
releasing updates? Any chance we can look forward to a new version maybe 4
times a year or so?

Thanks again for putting these together,

Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080703/9373837b/attachment.html>

From lorenzo.isella at gmail.com  Thu Jul  3 16:27:22 2008
From: lorenzo.isella at gmail.com (Lorenzo Isella)
Date: Thu, 03 Jul 2008 22:27:22 +0200
Subject: [SciPy-user] Real Array Expressed as Complex Array
In-Reply-To: <mailman.15.1215104404.6261.scipy-user@scipy.org>
References: <mailman.15.1215104404.6261.scipy-user@scipy.org>
Message-ID: <486D362A.1020902@gmail.com>

Hello,
> 814v3268b9e5rca34b98bb7b84b57 at mail.gmail.com> Content-Type: 
> text/plain; charset=ISO-8859-1 On Thu, Jul 3, 2008 at 9:39 AM, Lorenzo 
> Isella <lorenzo.isella at gmail.com> wrote:
>> > Dear All,
>> > I am bit puzzled: I was plotting a (rather complicated) analytical
>> > potential for which an analytical form is available.
>> > When asking to print out the value of the potential at the cut-off:
>> >
>> > print "at the cut-off, the dimensionless potential takes the value, ",
>> > pot_ext_dimensionless[-1]
>> >
>> > I got the following:
>> >
>> > at the cut-off, the dimensionless potential takes the value,
>> > (-6.48829965957e-06+0j)
>> >
>> > Now, since the potential I am coding via a function has to be a real
>> > function, I checked that the real part was always zero (as it should).
>>     
>
> I think you mean "...the *imaginary* part was always zero..."
>   
Absolutely, a slip of the tongue.


>> > Since the result was that the array pot_ext_dimensionless is real, how
>> > comes that it is expressed as a complex array (though the imaginary
>> > part is always zero)?
>>     
>
> It all depends on how you're calculating the pot_ext_dimensionless
> array; clearly somewhere in there an operation makes it complex.
> You'll have to show us how it's calculated.
>   
I think I solved the problem: I introduced some real elements taken from 
an array that has also some complex entries into pot_ext_dimensionless; 
although all the elements of pot_ext_dimensionless are all real, somehow 
scipy retains memory of these, once-existing, complex entries.

> You can always access the (real,imaginary) part of a complex array
> with (pot_ext_dimensionless.real, pot_ext_dimensionless.imag)
>
> But be careful, these arrays are not contiguous (they're a view into
> the complex array). That wrinkle has bitten me before, but I can't
> quite recall the circumstances.  You can always make them contiguous
> with numpy.ascontiguousarray().
>   

This sounds important and not 100% clear to me. Do you mean that if I 
have a complex array z and call real.z, I do not get in general an array 
with the same length as z, since purely imaginary entries are "skipped" 
rather than appearing as entries with zero real part, as one would expect?
Many thanks

Lorenzo


From peridot.faceted at gmail.com  Thu Jul  3 16:57:00 2008
From: peridot.faceted at gmail.com (Anne Archibald)
Date: Thu, 3 Jul 2008 16:57:00 -0400
Subject: [SciPy-user] Real Array Expressed as Complex Array
In-Reply-To: <486D362A.1020902@gmail.com>
References: <mailman.15.1215104404.6261.scipy-user@scipy.org>
	<486D362A.1020902@gmail.com>
Message-ID: <ce557a360807031357y2c13be90m217bb08538d1b231@mail.gmail.com>

2008/7/3 Lorenzo Isella <lorenzo.isella at gmail.com>:
> Hello,
>> 814v3268b9e5rca34b98bb7b84b57 at mail.gmail.com> Content-Type:
>> text/plain; charset=ISO-8859-1 On Thu, Jul 3, 2008 at 9:39 AM, Lorenzo
>> Isella <lorenzo.isella at gmail.com> wrote:
>>> > Since the result was that the array pot_ext_dimensionless is real, how
>>> > comes that it is expressed as a complex array (though the imaginary
>>> > part is always zero)?
>>>
>>
>> It all depends on how you're calculating the pot_ext_dimensionless
>> array; clearly somewhere in there an operation makes it complex.
>> You'll have to show us how it's calculated.
>>
> I think I solved the problem: I introduced some real elements taken from
> an array that has also some complex entries into pot_ext_dimensionless;
> although all the elements of pot_ext_dimensionless are all real, somehow
> scipy retains memory of these, once-existing, complex entries.

The key idea is that arrays have a data type: that is, each numpy
array, upon creation, specifies the type of all its contents. So your
numpy arrays are marked as containing complex numbers. The fact that
the imaginary part of these complex numbers is approximately or
exactly zero isn't relevant; they are still stored as a pair of
floating-point numbers. If you prefer, you can think of their type as
"potentially complex numbers". In any case, such numbers are printed
as a+bj even if a or b is zero, and various arithmetic operations
treat them as complex numbers. If you know that the answer should be
real, and you want to represent them more conveniently and
efficiently, you can simply take the real part.

>> You can always access the (real,imaginary) part of a complex array
>> with (pot_ext_dimensionless.real, pot_ext_dimensionless.imag)
>>
>> But be careful, these arrays are not contiguous (they're a view into
>> the complex array). That wrinkle has bitten me before, but I can't
>> quite recall the circumstances.  You can always make them contiguous
>> with numpy.ascontiguousarray().
>>
>
> This sounds important and not 100% clear to me. Do you mean that if I
> have a complex array z and call real.z, I do not get in general an array
> with the same length as z, since purely imaginary entries are "skipped"
> rather than appearing as entries with zero real part, as one would expect?

No. Taking "X.real" is the same as the mathematical operation of
taking the real part:it throws away any imaginary part and interprets
what's left as real, whether it's zero or not. "Contiguous", in this
context, is a technical feature of numpy arrays that you should almost
never need to care about. (Numpy arrays can be homogeneous blocks of
memory, but they can also be homogeneous elements "strided" through
memory with other data in between. A few functions cannot deal with
this striding.)

Anne


From travis at enthought.com  Thu Jul  3 17:16:40 2008
From: travis at enthought.com (Travis Vaught)
Date: Thu, 3 Jul 2008 16:16:40 -0500
Subject: [SciPy-user] Enthought Python Distribution
In-Reply-To: <6946b9500807031015j392e8a20ic5b7636208e086d6@mail.gmail.com>
References: <6946b9500807031015j392e8a20ic5b7636208e086d6@mail.gmail.com>
Message-ID: <B066F1B9-0A34-491D-9FAA-49E26C35017F@enthought.com>

Hey Mark,

You're right.  Pierre has been extremely responsive--a great example  
to us all!

Our plan is to have a _minimum _ of two releases per year and your  
suggestion of 4 times is fairly likely.  It depends a lot on the  
frequency and urgency of updates to underlying libraries.  Any bugs in  
the installer itself should prompt urgent correction, so there may be  
updates due to this alone.

I'm glad you've found these things useful.  Let me know if you have  
any other questions.

Best,

Travis

On Jul 3, 2008, at 12:15 PM, Mark Bakker wrote:

> Hello Travis and other Enthought guru's -
>
> I have benefited from the Enthought distribution for years, and  
> really want to switch to a Mac, so this is great news.
>
> My main concern with the Enthought distribution is the frequency of  
> new releases. The guys at Python(x,y) are exceptionally quick at  
> updating. That may be hard to match (and cost a lot of effort). What  
> are your plans at releasing updates? Any chance we can look forward  
> to a new version maybe 4 times a year or so?
>
> Thanks again for putting these together,
>
> Mark
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080703/d125b1d5/attachment.html>

From travis at enthought.com  Thu Jul  3 17:26:13 2008
From: travis at enthought.com (Travis Vaught)
Date: Thu, 3 Jul 2008 16:26:13 -0500
Subject: [SciPy-user] Enthought Python Distribution
In-Reply-To: <db6b5ecc0807021010q15606db8gcc00197023c4547b@mail.gmail.com>
References: <78A3D219-9C3C-4B4C-BA77-8B63ED1191BF@enthought.com>
	<db6b5ecc0807021010q15606db8gcc00197023c4547b@mail.gmail.com>
Message-ID: <F2840526-8C56-471A-9354-D7251C8EC910@enthought.com>

Hey Fernando,

On Jul 2, 2008, at 12:10 PM, Fernando Perez wrote:

> Hey Travis,
>
> On Wed, Jul 2, 2008 at 7:32 AM, Travis Vaught <travis at enthought.com>  
> wrote:
>> Greetings,
>>
>> We're pleased to announce the beta release of the Enthought Python
>> Distribution for *Mac OS X*.
>
> First of all, this is *fantastic*.  OSX has been a major pain for us
> here in terms of distribution, so I am extremely happy and grateful to
> you guys for putting this out.

You're quite welcome.  I'm glad it's useful.  Let us know if you see  
any issues in the beta.

>
>
> Two minor requests: any chance in the next release, you could update
> ipython and matplotlib?  You are shipping 0.8.1 and 0.91.2, both of
> which are fairly old and the current versions of both do have
> reasonably useful improvements and fixes (ipython has threading fixes
> that impact plotting, and mpl has all the new mathtext code that gives
> good math rendering without needing latex).

I want to see these as well.  It should be no problem to get these in  
the next release.  In fact we'll upgrade to the latest versions of  
_everything_ as long as the developers recommend doing so and there  
are no compatibility issues.

>
>
> In any case, a big cheer from me :)

:-)


Thanks,

Travis

>
>
> Regards,
>
> f
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



From jh at physics.ucf.edu  Thu Jul  3 17:59:08 2008
From: jh at physics.ucf.edu (Joe Harrington)
Date: Thu, 03 Jul 2008 17:59:08 -0400
Subject: [SciPy-user] Summer Doc Marathon status report and request for more
	writers
Message-ID: <wl87ic2ocib.fsf@glup.physics.ucf.edu>

This is an interim status report on the Summer Documentation Marathon.
It is also an invitation and plea for all experienced users to
participate!  I am cross-posting in an effort to get broader
participation.  Please hold any discussion on the scipy-dev mailing
list.

As you know, our immediate goal is to produce first-draft docstrings
for the user-visible parts of Numpy in time for a release before Fall
classes (about 1 August).  The short version is: We are really moving
along!  But, we need *your* help to make it in time for August.

Here's the scoop:

1. We have all our infrastructure, standards, and procedure in place:
We have a wiki that makes editing the docs easy and even fun.  It
communicates directly with the numpy sources.  We have PDF and HTML
reference guides being generated essentially automatically:

http://sd-2116.dedibox.fr/pydocweb
http://mentat.za.net/numpy/refguide/NumPy.pdf
http://mentat.za.net/numpy/refguide/

The wiki front page contains or points to all you need to get started.

The wiki lets you pull down a docstring with a few mouse clicks, edit
on your machine, upload it, and see how it will look in its HTML
version right away.  You can also read everyone else's docstrings,
comment on them, see the status of the project, and so on.  The
formatted versions necessarily lag the docstrings on the wiki because
they are made whenever the docstrings are checked into the sources.

2. We have documented about 1/4 of numpy in a fairly professional way,
comparable to the reference pages of the major commercial packages.
The doc wiki is probably the next place to go if your question isn't
answered by the docstring in the current version's help(), since you
can look at the new docstrings we've generated, and they're *good*!

3. But, we're only 1/4 of the way there, we're halfway through the
summer, and some of the initial enthusiasm is waning.  The following
page tells the tale:

http://sd-2116.dedibox.fr/pydocweb/stats/

As you can see (you did click, right?  please click...), there are
2323 numpy objects with docstrings.  Of these, 1464 we deemed
"unimportant" (for now).  These are generally items not seen by
regular users.  This left 859 objects to document in this first pass.
We've done 24% of them at this writing.

Now, 24% is really exciting, and I'd like to take a moment to say a
public "Hooray!" for the team (no particular order):

St?fan van der Walt	Pauli Virtanen 	
Robert Hetland		Gael Varoquaux 	
Scott Sinclair		Alan Jackson 	
Tim Cera		Johann Cohen-Tanugi
David Huard		Keith Goodman

Together these ten have written around 7500 words of documentation on
the community's behalf, mainly as volunteers.

HOWEVER, we can all do the math.  We've spent one of our two months.
We are 1/4 of the way there.  Progress is slowing, and even if it
didn't, we wouldn't make it in time.  This is not a sprint, it's a
MARATHON.

			WE NEED YOUR HELP!

And we need it now.  Are you excited by the idea of having
documentation for numpy by the Fall release?  Of having docs that
answer your questions, that have *good* examples, that really save you
time?  If so, then please invest just a fraction of the time that
documentation will save you in the next year alone by signing up on
the wiki and writing some.  If each experienced user wrote just a few
pages, we'd be done!

If you don't think you know enough to write, then do some reviewing.
Are the docs readable?  Do you understand the examples?  Each
docstring on the wiki has an easy comment box waiting for your
thoughts.

You will have a reference guide in the next release of numpy!  I hope
you will help make it a complete one.  Sign up on the doc wiki today:

		http://sd-2116.dedibox.fr/pydocweb/

Thanks,

--jh-- and the SciPy Doc Team

Prof. Joseph Harrington
Department of Physics
MAP 414
4000 Central Florida Blvd.
University of Central Florida
Orlando, FL 32816-2385
(407) 823-3416 voice
(407) 823-5112 fax
(407) 823-2325 physics office
jh at physics.ucf.edu


From fperez.net at gmail.com  Thu Jul  3 18:40:58 2008
From: fperez.net at gmail.com (Fernando Perez)
Date: Thu, 3 Jul 2008 15:40:58 -0700
Subject: [SciPy-user] Enthought Python Distribution
In-Reply-To: <F2840526-8C56-471A-9354-D7251C8EC910@enthought.com>
References: <78A3D219-9C3C-4B4C-BA77-8B63ED1191BF@enthought.com>
	<db6b5ecc0807021010q15606db8gcc00197023c4547b@mail.gmail.com>
	<F2840526-8C56-471A-9354-D7251C8EC910@enthought.com>
Message-ID: <db6b5ecc0807031540h7df00a62sc33322dfd1e13fb4@mail.gmail.com>

On Thu, Jul 3, 2008 at 2:26 PM, Travis Vaught <travis at enthought.com> wrote:

> I want to see these as well.  It should be no problem to get these in
> the next release.  In fact we'll upgrade to the latest versions of
> _everything_ as long as the developers recommend doing so and there
> are no compatibility issues.

Any chance the next release might happen before scipy'08?  It would
make an enormous difference for those of us teaching the tutorials to
have this as an installation option.  Perhaps after waiting for the
outcome of the Mayavi sprint you guys are holding right now...

I know that if we could point all attendees to reasonably up to date
installers for the tools the tutorials will cover on Win32 and OSX,
things will be a LOT smoother than they were last year.  I think we
were really hurt last year by underestimating (I fully take the blame
for that) the importance of installation issues.  We burned valuable
time on that and probably left quite a few people behind simply
because debugging install problems on a one by one basis was
impossible under those circumstances.

In any case, thanks again for this!

Cheers,

f


From emanuele at relativita.com  Fri Jul  4 12:23:04 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Fri, 04 Jul 2008 18:23:04 +0200
Subject: [SciPy-user] [OpenOpt] lb issue
In-Reply-To: <48672993.1040001@scipy.org>
References: <48672993.1040001@scipy.org>
Message-ID: <486E4E68.4090004@relativita.com>

Yes Dmitrey you are right. After some thinking I successfully
implemented your advice in my code, and added contol and
inf. It seems to work quite well. Now "ralg" tries some attempts
in the forbidden area and then comes back. I hope that
generating the next 'x' (attempt) does not require lots of
computation to 'ralg', in the sense that wasting some attempts
does not cost too much time (especially for bigger problems,
like hundreds of variables).

I have another  question. Is there a way in OpenOpt to
tell the solver to optimize "just" on some variable/dimensions
and not all of them?
Example: assume my function takes 3 numbers and returns
1; I want to minimize it but with respect just the first 2 of the
3 input values; let's say that the initial guess of the 3rd value
is already OK or that I can't change it.
I'd like to pass something like a boolean vector [True, True, False]
to p.solve() o the NLP instance to say that the third parameter
should not be changed. In other words it can be thought as
another kind of constraint.

Obviously I can wrap my function in order to handle the
boolean vector and do the right thing, but I'm wondering
if OpenOpt is able to handle this kind of requests.

Thanks,

Emanuele


dmitrey wrote:
> I don't see any difficulties to map my advice to more difficult cases 
> (just don't forget to increase your ineq constraints by contol, for to 
> compare those ones with zero, not contol, and use inf, not the huge 
> value you have mentioned). Current ralg implementation doesn't need 
> objfunc value when point is outside of feasible region (i.e. if *any* 
> constraint is bigger than p.contol, not only lb-ub constraint). OO calls 
> objFunc outside feasible region just for to check some stop criteria, 
> yield iter output point text and possible graphics output.
>
> Regards, D.
>
> Emanuele Olivetti wrote:
>   
>> Unfortunately it is not so simple to map this advice to my
>> real situation, which is more complex than the
>> proof-of-concept example of the initial message. Returning
>> a big positive value when x is outside the bounds is an option
>> I considered some time ago but then discarded. But I'll think
>> more about it now.
>>
>> Ciao,
>>
>> Emanuele
>>
>> dmitrey wrote:
>>   
>>     
>>> Hi Emanuele,
>>>
>>> I could propose you temporary solution (see below), this one doesn't
>>> require updating oo from svn. However, usually ALGENCAN, ipopt and
>>> scipy_slsqp work much better for box-bound constrained problems (w/o
>>> other constraints) than current ralg implementation.
>>> D.
>>>
>>> import numpy as N
>>> from scikits.openopt import NLP
>>> from numpy import any, inf
>>> size = 100
>>> dimensions = 2
>>> data = N.random.rand(size,dimensions)-0.5
>>>
>>> contol = 1e-6
>>> lb=N.zeros(dimensions) + contol
>>>
>>> def f(x):
>>>    global data
>>>    if any(x<0):
>>>        #objective function is not defined here, let's use inf instead
>>>        #however, some iters will show objFunVa= inf in text output
>>>        # and graphic output is currently unavailable for the case
>>>        return inf
>>>    return N.dot(data**2,x.T)
>>>
>>> x0 = N.ones(dimensions)
>>> p = NLP(f,x0,lb=lb, contol = contol)
>>> p.solve('ralg')
>>> print p.ff,p.xf
>>>
>>>     
>>>       
>>
>>
>>   
>>     
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>   



From dmitrey.kroshko at scipy.org  Fri Jul  4 14:58:55 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Fri, 04 Jul 2008 21:58:55 +0300
Subject: [SciPy-user] [OpenOpt] lb issue
In-Reply-To: <486E4E68.4090004@relativita.com>
References: <48672993.1040001@scipy.org> <486E4E68.4090004@relativita.com>
Message-ID: <486E72EF.8000001@scipy.org>

hi Emanuele,

Emanuele Olivetti wrote:
> Yes Dmitrey you are right. After some thinking I successfully
> implemented your advice in my code, and added contol and
> inf. It seems to work quite well. Now "ralg" tries some attempts
> in the forbidden area and then comes back. I hope that
> generating the next 'x' (attempt) does not require lots of
> computation to 'ralg', in the sense that wasting some attempts
> does not cost too much time (especially for bigger problems,
> like hundreds of variables).
>   
each iteration with point infeasible due to box-bound constraints (if no 
other are present) eats no greater than  5*nVars^2 multiplication 
operations (+some evaluations of objective function while line search is 
going within feasible region, usually it's 2-4 steps, no matter how big 
is nVars), and for nVars < 1000 it's not a big number as I guess 
(however I don't know your situation). Also, there are some methods to 
reduce the number, but it's not implemented in current python ralg 
solver yet.
> I have another  question. Is there a way in OpenOpt to
> tell the solver to optimize "just" on some variable/dimensions
> and not all of them?
>   
I intend to implement "fixedVars" prob filed (similar to intVars and 
boolVars that are handled by some MILP solvers). However, it's not done yet.

Regards, D.


From hectorvd at gmail.com  Sat Jul  5 12:20:57 2008
From: hectorvd at gmail.com (Hector Villafuerte)
Date: Sat, 5 Jul 2008 10:20:57 -0600
Subject: [SciPy-user] EPD on Vista
Message-ID: <350f72300807050920r1d8625c3j355d2f20ff2fc1ed@mail.gmail.com>

Hi,
a colleague of mine is having troubles installing the Enthought Python
Distro on Windows Vista. I don't have access to Vista now (just
Windows XP, where EPD works like a charm). So basically my question:
has somebody successfully installed EPD on Vista? (also, it seems that
Vista is not currently supported).
As a side note: this other colleague had lots of trouble getting
Matlab to work when he "upgraded" to Vista... thought you would like
to know :)
Best,
-- 
 Hector


From robert.kern at gmail.com  Sat Jul  5 19:28:55 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Sat, 5 Jul 2008 18:28:55 -0500
Subject: [SciPy-user] EPD on Vista
In-Reply-To: <350f72300807050920r1d8625c3j355d2f20ff2fc1ed@mail.gmail.com>
References: <350f72300807050920r1d8625c3j355d2f20ff2fc1ed@mail.gmail.com>
Message-ID: <3d375d730807051628j55ab6b45l7b2fea32dbe80684@mail.gmail.com>

On Sat, Jul 5, 2008 at 11:20, Hector Villafuerte <hectorvd at gmail.com> wrote:
> Hi,
> a colleague of mine is having troubles installing the Enthought Python
> Distro on Windows Vista. I don't have access to Vista now (just
> Windows XP, where EPD works like a charm). So basically my question:
> has somebody successfully installed EPD on Vista? (also, it seems that
> Vista is not currently supported).

I think we tried once with an early unreleased pre-alpha, and it
worked. Vista is not officially supported, yet. However, we
(Enthought) would be interested in receiving bug reports on the
enthought-dev mailing list. There might be simple things we can do to
get it to work for your colleague.

  https://mail.enthought.com/mailman/listinfo/enthought-dev

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From yosefmel at post.tau.ac.il  Sun Jul  6 00:48:10 2008
From: yosefmel at post.tau.ac.il (Yosef Meller)
Date: Sun, 6 Jul 2008 07:48:10 +0300
Subject: [SciPy-user] array manipulation
In-Reply-To: <3d375d730807030226j6d11aea9t3eb16ab321629272@mail.gmail.com>
References: <web-104160546@uni-stuttgart.de>
	<3d375d730807030226j6d11aea9t3eb16ab321629272@mail.gmail.com>
Message-ID: <200807060748.10076.yosefmel@post.tau.ac.il>

On Thursday 03 July 2008 12:26:27 Robert Kern wrote:
> On Thu, Jul 3, 2008 at 04:24, Nils Wagner <nwagner at iam.uni-stuttgart.de> 
wrote:
> > How can I remove duplicate entries from an array ?
>
> numpy.unique1d()

Why is there both numpy.unique1d() and numpy.unique()? Their code seems very 
similar. And, if thery're different, it would be nice to see unique1d in 
http://www.scipy.org/Numpy_Example_List


From mwojc at p.lodz.pl  Sun Jul  6 10:17:06 2008
From: mwojc at p.lodz.pl (Marek Wojciechowski)
Date: Sun, 6 Jul 2008 16:17:06 +0200
Subject: [SciPy-user] scipy-0.6 on AIX 5.3
In-Reply-To: <mailman.45115.1213146608.27196.scipy-user@scipy.org>
References: <mailman.45115.1213146608.27196.scipy-user@scipy.org>
Message-ID: <200807061617.07505.mwojc@p.lodz.pl>

Hi!

I tried to install scipy in AIX 5.3 and i'm partially successful. For now i 
see no way to compile and use C++ code from scipy, so sparse and weave 
modules are not working. I used native IBM compilers: xlf_r, cc_r, xlc++_r. I 
compiled also properly blas and lapack for use with scipy.

I applied two dirty hacks in numpy.distutils.ccompiler (1.0.4) and some hacks 
in scipy itself. I'm attaching diffs because probably some changes could be 
applied to the development code...

Greetings
-- 
Marek Wojciechowski
-------------- next part --------------
A non-text attachment was scrubbed...
Name: scipy-0.6.0-aix.diff
Type: text/x-diff
Size: 60159 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080706/17b90905/attachment.diff>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ccompiler-aix.diff
Type: text/x-diff
Size: 1419 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080706/17b90905/attachment-0001.diff>

From gael.varoquaux at normalesup.org  Sun Jul  6 17:07:31 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Sun, 6 Jul 2008 23:07:31 +0200
Subject: [SciPy-user] Schedule for the SciPy08 conferencez
Message-ID: <20080706210731.GC25810@phare.normalesup.org>

We have received a large number of excellent contributions for papers for
the SciPy 2008 conference. The program committee has had to make a
difficult selection and we are happy to bring to you a preliminary
schedule:

Thursday
=========

**8:00** Registration/Breakfast

**8:55** Welcome (Travis Vaught)

**9:10** Keynote (Alex Martelli)

**10:00** State of SciPy (Travis Vaught, Jarrod Millman)

**10:40** -- Break --

**11:00** Sympy - Python library for symbolic mathematics: introduction
and applications (Ond?ej ?ertik)

**11:40** Interval arithmetic: Python implementation and applications
(Stefano Taschini)

**12:00** Experiences Using Scipy for Computer Vision Research (Damian
Eads)

**12:20** -- Lunch --

**1:40** The new NumPy documentation framework (St?fan Van der Walt)

**2:00** Matplotlib solves the riddle of the sphinx (Michael Droettboom)

**2:40** The SciPy documentation project (Joe Harrington)

**3:00** -- Break --

**3:40** Sage: creating a viable free Python-based open source
alternatice to Magma, Maple, Mathematica and Matlab (William Stein)

**4:20** Open space for lightning talks

Friday
========

**8:30** Breakfast

**9:00** Pysynphot: A Python Re-Implementation of a Legacy App in
Astronomy
(Perry Greenfield)

**9:40**  How the Large Synoptic Survey Telescope (LSST) is using
Python (Robert Lupton)

**10:00** Real-time Astronomical Time-series Classification and Broadcast
Pipeline (Dan Starr)

**10:20** Analysis and Visualization of Multi-Scale Astrophysical
Simulations using Python and NumPy (Matthew Turk)

**10:40** -- Break --

**11:00** Exploring network structure, dynamics, and function using
NetworkX (Aric Hagberg)

**11:40** Mayavi: Making 3D data visualization reusable (Prabhu
Ramachandran, Ga?l Varoquaux)

**12:00** Finite Element Modeling of Contact and Impact Problems Using
Python (Ryan Krauss)

**12:20** -- Lunch --

**2:00** PyCircuitScape: A Tool for Landscape Ecology (Viral Shah)

**2:20** Summarizing Complexity in High Dimensional Spaces (Karl Young)

**2:40** UFuncs: A generic function mechanism in Python (Travis Oliphant)

**3:20** -- Break --

**3:40** NumPy Optimization: Manual tuning and automated approaches (Evan
Patterson)

**4:00** Converting Python functions to dynamically-compiled C (Ilan
Schnell)

**4:20** unPython: Converting Python numerical programs into C
(Rahul Garg)

**4:40** Implementing the Grammar of Graphics for Python (Robert Kern)

**5:00** Ask the experts session.

A more detailled booklet including the abstract text will be available
soon.
We are looking forward to seeing you in Caltech,

Ga?l Varoquaux, on behalf of the program committee.

--
SciPy2008 conference.
Program committee

Anne Archibald, McGill University
Matthew Brett
Perry Greenfield, Space Telescope Science Institute
Charles Harris
Ryan Krauss, Southern Illinois University
Ga?l Varoquaux
St?fan van der Walt, University of Stellenbosch


From Joris.DeRidder at ster.kuleuven.be  Sun Jul  6 18:00:03 2008
From: Joris.DeRidder at ster.kuleuven.be (Joris De Ridder)
Date: Mon, 7 Jul 2008 00:00:03 +0200
Subject: [SciPy-user] array manipulation
In-Reply-To: <200807060748.10076.yosefmel@post.tau.ac.il>
References: <web-104160546@uni-stuttgart.de>
	<3d375d730807030226j6d11aea9t3eb16ab321629272@mail.gmail.com>
	<200807060748.10076.yosefmel@post.tau.ac.il>
Message-ID: <8019D79F-820A-4B50-83CE-A36FCD109DE8@ster.kuleuven.be>


On 6 Jul 2008, at 6:48 , Yosef Meller wrote:

> On Thursday 03 July 2008 12:26:27 Robert Kern wrote:
>> On Thu, Jul 3, 2008 at 04:24, Nils Wagner <nwagner at iam.uni-stuttgart.de 
>> >
> wrote:
>>> How can I remove duplicate entries from an array ?
>>
>> numpy.unique1d()
>
> Why is there both numpy.unique1d() and numpy.unique()? Their code  
> seems very
> similar. And, if thery're different, it would be nice to see  
> unique1d in
> http://www.scipy.org/Numpy_Example_List

Good question...
unique1d() supports returning the indices of the unique values while  
unique() does not, but that's the only difference I can see. Perhaps a  
somewhat more consistent implementation would only consist of unique()  
and argunique(). But changing this will perhaps break too much code.

Also, at the moment there seems to be no unique()-like function that  
doesn't first flatten its input array. That is, a function that is  
able to transform, for example, array([[1,2],[3,4],[1,2]]) into  
array([[1,2],[3,4]]).

Joris



Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm



From david at ar.media.kyoto-u.ac.jp  Sun Jul  6 22:53:41 2008
From: david at ar.media.kyoto-u.ac.jp (David Cournapeau)
Date: Mon, 07 Jul 2008 11:53:41 +0900
Subject: [SciPy-user] scipy-0.6 on AIX 5.3
In-Reply-To: <200807061617.07505.mwojc@p.lodz.pl>
References: <mailman.45115.1213146608.27196.scipy-user@scipy.org>
	<200807061617.07505.mwojc@p.lodz.pl>
Message-ID: <48718535.8000703@ar.media.kyoto-u.ac.jp>

Marek Wojciechowski wrote:
> Hi!
>
> I tried to install scipy in AIX 5.3 and i'm partially successful. For now i 
> see no way to compile and use C++ code from scipy, so sparse and weave 
> modules are not working. I used native IBM compilers: xlf_r, cc_r, xlc++_r. I 
> compiled also properly blas and lapack for use with scipy.
>
> I applied two dirty hacks in numpy.distutils.ccompiler (1.0.4) and some hacks 
> in scipy itself. I'm attaching diffs because probably some changes could be 
> applied to the development code...
>   

C++-style comments in C are bugs (I am guilty of most of them), but
should have been solved in the SVN trunk. Concerning the fortran code,
it seems that the AIX fortran compiler does not like special characters
in comments nor code lines starting with &...

I would consider using the Gnu compiler instead of AIX fortran compiler,

cheers,

David


From sgarcia at olfac.univ-lyon1.fr  Mon Jul  7 05:42:24 2008
From: sgarcia at olfac.univ-lyon1.fr (Samuel GARCIA)
Date: Mon, 07 Jul 2008 11:42:24 +0200
Subject: [SciPy-user] weave problems,
	weave_imp.o no such file or directory
In-Reply-To: <de1aa14a0806270649v3b041f0p59ef06cb54b4e30a@mail.gmail.com>
References: <de1aa14a0806270649v3b041f0p59ef06cb54b4e30a@mail.gmail.com>
Message-ID: <4871E500.6070704@olfac.univ-lyon1.fr>

Hi list,
I have exactly the same problem under Windows and the exelent 
Python(X,Y) distribution.

The problem is in the command line, there is no quote at the end just 
after -c and -o and so the path is splited.

The problem is new for me because before using this distrubution there 
was no space in my python path.
And now, yes. in 'c:\program files\...'

Does the weave maintener could fix this ? Or is it already fix but not 
include in the distribution ?

Thank

Samuel

S?ren Nielsen a ?crit :
> Hi,
>
> I've done a fresh install of MinGw 5.14, Python 2.4.4, Scipy 0.6.0, 
> Numpy 1.1.0 and numarray 1.5.2
>
> When I try to use weave i get this:
>
> C:\>test_weave.py
> <weave: compiling>
> running build_ext
> running build_src
> building extension "sc_552cccf5dbf4f6eadd273cdcbd5860523" sources
> customize Mingw32CCompiler
> customize Mingw32CCompiler using build_ext
> customize Mingw32CCompiler
> customize Mingw32CCompiler using build_ext
> building 'sc_552cccf5dbf4f6eadd273cdcbd5860523' extension
> compiling C++ sources
> C compiler: g++ -mno-cygwin -O2 -Wall
>
> compile options: '-IC:\Python24\lib\site-packages\scipy\weave 
> -IC:\Python24\lib\
> site-packages\scipy\weave\scxx 
> -IC:\Python24\lib\site-packages\numpy\core\includ
> e -IC:\Python24\include -IC:\Python24\PC -c'
> g++ -mno-cygwin -O2 -Wall -IC:\Python24\lib\site-packages\scipy\weave 
> -IC:\Pytho
> n24\lib\site-packages\scipy\weave\scxx 
> -IC:\Python24\lib\site-packages\numpy\cor
> e\include -IC:\Python24\include -IC:\Python24\PC -c 
> C:\Python24\lib\site-package
> s\scipy\weave\scxx\weave_imp.cpp -o 
> c:\docume~1\lisear~1\lokale~1\temp\Lise Arle
> th\python24_intermediate\compiler_08edc7e348e1c33f63a33ab500aef08e\Release\pytho
> n24\lib\site-packages\scipy\weave\scxx\weave_imp.o
> Found executable C:\MinGw\bin\g++.exe
> g++.exe: 
> Arleth\python24_intermediate\compiler_08edc7e348e1c33f63a33ab500aef08e\
> Release\python24\lib\site-packages\scipy\weave\scxx\weave_imp.o: No 
> such file or
>  directory
> Traceback (most recent call last):
>   File "C:\test_weave.py", line 341, in ?
>     main()
>   File "C:\test_weave.py", line 224, in main
>     weave.inline('printf("%d\\n",a);',['a'], verbose=2, 
> type_converters=converte
> rs.blitz)  #, compiler = 'msvc', verbpse=2, 
> type_converters=converters.blitz, au
> to_downcast=0) #'msvc' or 'gcc' or 'mingw32'
>   File "C:\Python24\Lib\site-packages\scipy\weave\inline_tools.py", 
> line 338, in
>  inline
>     auto_downcast = auto_downcast,
>   File "C:\Python24\Lib\site-packages\scipy\weave\inline_tools.py", 
> line 447, in
>  compile_function
>     verbose=verbose, **kw)
>   File "C:\Python24\Lib\site-packages\scipy\weave\ext_tools.py", line 
> 365, in co
> mpile
>     verbose = verbose, **kw)
>   File "C:\Python24\Lib\site-packages\scipy\weave\build_tools.py", 
> line 269, in
> build_extension
>     setup(name = module_name, ext_modules = [ext],verbose=verb)
>   File "C:\Python24\Lib\site-packages\numpy\distutils\core.py", line 
> 184, in set
> up
>     return old_setup(**new_attr)
>   File "C:\Python24\lib\distutils\core.py", line 166, in setup
>     raise SystemExit, "error: " + str(msg)
> distutils.errors.CompileError: error: Command "g++ -mno-cygwin -O2 
> -Wall -IC:\Py
> thon24\lib\site-packages\scipy\weave 
> -IC:\Python24\lib\site-packages\scipy\weave
> \scxx -IC:\Python24\lib\site-packages\numpy\core\include 
> -IC:\Python24\include -
> IC:\Python24\PC -c 
> C:\Python24\lib\site-packages\scipy\weave\scxx\weave_imp.cpp
> -o c:\docume~1\lisear~1\lokale~1\temp\Lise 
> Arleth\python24_intermediate\compiler
> _08edc7e348e1c33f63a33ab500aef08e\Release\python24\lib\site-packages\scipy\weave
> \scxx\weave_imp.o" failed with exit status 1
>
> the test_weave file was something i found on the scipy dev wiki... I 
> also tried some of my older files that also uses weave and they all 
> give the same error...
>
> Can anyone help me with this?
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user

-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Samuel Garcia
Laboratoire de Neurosciences Sensorielles, Comportement, Cognition.
CNRS - UMR5020 - Universite Claude Bernard LYON 1
Equipe logistique et technique
50, avenue Tony Garnier
69366 LYON Cedex 07
FRANCE
T?l : 04 37 28 74 64
Fax : 04 37 28 76 01
http://olfac.univ-lyon1.fr/unite/equipe-07/
http://neuralensemble.org/trac/OpenElectrophy
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080707/cde53e53/attachment.html>

From mwojc at p.lodz.pl  Mon Jul  7 09:06:46 2008
From: mwojc at p.lodz.pl (Marek Wojciechowski)
Date: Mon, 07 Jul 2008 15:06:46 +0200
Subject: [SciPy-user] scipy-0.6 on AIX 5.3
References: <mailman.45115.1213146608.27196.scipy-user@scipy.org>
	<200807061617.07505.mwojc@p.lodz.pl>
	<48718535.8000703@ar.media.kyoto-u.ac.jp>
Message-ID: <g4t44c$dqj$1@ger.gmane.org>

David Cournapeau wrote:

> Marek Wojciechowski wrote:
>> Hi!
>>
>> I tried to install scipy in AIX 5.3 and i'm partially successful. For now
>> i see no way to compile and use C++ code from scipy, so sparse and weave
>> modules are not working. I used native IBM compilers: xlf_r, cc_r,
>> xlc++_r. I compiled also properly blas and lapack for use with scipy.
>>
>> I applied two dirty hacks in numpy.distutils.ccompiler (1.0.4) and some
>> hacks in scipy itself. I'm attaching diffs because probably some changes
>> could be applied to the development code...
>>   
> 
> C++-style comments in C are bugs (I am guilty of most of them), but
> should have been solved in the SVN trunk. Concerning the fortran code,
> it seems that the AIX fortran compiler does not like special characters
> in comments nor code lines starting with &...
> 
> I would consider using the Gnu compiler instead of AIX fortran compiler,
> 

IBM Fortran compiler (xlf) worked very well for me. The only compilation
problem is with two functions in specfun.f, namely: CLQN and CLQMN.
However, this is because of the aggressive -O5 optimization flag (which is
set up by default on AIX in numpy.distutils). With -O3 whole code compiles
nicely.

BTW. i'm not sure if there is a way to get working g77 or gfortran on AIX.

Greetings,
-- 
Marek Wojciechowski



From soren.skou.nielsen at gmail.com  Mon Jul  7 15:31:03 2008
From: soren.skou.nielsen at gmail.com (=?ISO-8859-1?Q?S=F8ren_Nielsen?=)
Date: Mon, 7 Jul 2008 21:31:03 +0200
Subject: [SciPy-user] weave problems,
	weave_imp.o no such file or directory
In-Reply-To: <4871E500.6070704@olfac.univ-lyon1.fr>
References: <de1aa14a0806270649v3b041f0p59ef06cb54b4e30a@mail.gmail.com>
	<4871E500.6070704@olfac.univ-lyon1.fr>
Message-ID: <de1aa14a0807071231p50d3e655v3f65d3dcc6dfc3b8@mail.gmail.com>

I have a space in my path too... So thats why it works on some computers and
not on others...

If you find a solution I'd really like to hear it.. The activity level on
this mailing list is almost flatlining.

Soren

On Mon, Jul 7, 2008 at 11:42 AM, Samuel GARCIA <sgarcia at olfac.univ-lyon1.fr>
wrote:

>  Hi list,
> I have exactly the same problem under Windows and the exelent Python(X,Y)
> distribution.
>
> The problem is in the command line, there is no quote at the end just after
> -c and -o and so the path is splited.
>
> The problem is new for me because before using this distrubution there was
> no space in my python path.
> And now, yes. in 'c:\program files\...'
>
> Does the weave maintener could fix this ? Or is it already fix but not
> include in the distribution ?
>
> Thank
>
> Samuel
>
> S?ren Nielsen a ?crit :
>
> Hi,
>
> I've done a fresh install of MinGw 5.14, Python 2.4.4, Scipy 0.6.0, Numpy
> 1.1.0 and numarray 1.5.2
>
> When I try to use weave i get this:
>
> C:\>test_weave.py
> <weave: compiling>
> running build_ext
> running build_src
> building extension "sc_552cccf5dbf4f6eadd273cdcbd5860523" sources
> customize Mingw32CCompiler
> customize Mingw32CCompiler using build_ext
> customize Mingw32CCompiler
> customize Mingw32CCompiler using build_ext
> building 'sc_552cccf5dbf4f6eadd273cdcbd5860523' extension
> compiling C++ sources
> C compiler: g++ -mno-cygwin -O2 -Wall
>
> compile options: '-IC:\Python24\lib\site-packages\scipy\weave
> -IC:\Python24\lib\
> site-packages\scipy\weave\scxx
> -IC:\Python24\lib\site-packages\numpy\core\includ
> e -IC:\Python24\include -IC:\Python24\PC -c'
> g++ -mno-cygwin -O2 -Wall -IC:\Python24\lib\site-packages\scipy\weave
> -IC:\Pytho
> n24\lib\site-packages\scipy\weave\scxx
> -IC:\Python24\lib\site-packages\numpy\cor
> e\include -IC:\Python24\include -IC:\Python24\PC -c
> C:\Python24\lib\site-package
> s\scipy\weave\scxx\weave_imp.cpp -o c:\docume~1\lisear~1\lokale~1\temp\Lise
> Arle
>
> th\python24_intermediate\compiler_08edc7e348e1c33f63a33ab500aef08e\Release\pytho
> n24\lib\site-packages\scipy\weave\scxx\weave_imp.o
> Found executable C:\MinGw\bin\g++.exe
> g++.exe:
> Arleth\python24_intermediate\compiler_08edc7e348e1c33f63a33ab500aef08e\
> Release\python24\lib\site-packages\scipy\weave\scxx\weave_imp.o: No such
> file or
>  directory
> Traceback (most recent call last):
>   File "C:\test_weave.py", line 341, in ?
>     main()
>   File "C:\test_weave.py", line 224, in main
>     weave.inline('printf("%d\\n",a);',['a'], verbose=2,
> type_converters=converte
> rs.blitz)  #, compiler = 'msvc', verbpse=2,
> type_converters=converters.blitz, au
> to_downcast=0) #'msvc' or 'gcc' or 'mingw32'
>   File "C:\Python24\Lib\site-packages\scipy\weave\inline_tools.py", line
> 338, in
>  inline
>     auto_downcast = auto_downcast,
>   File "C:\Python24\Lib\site-packages\scipy\weave\inline_tools.py", line
> 447, in
>  compile_function
>     verbose=verbose, **kw)
>   File "C:\Python24\Lib\site-packages\scipy\weave\ext_tools.py", line 365,
> in co
> mpile
>     verbose = verbose, **kw)
>   File "C:\Python24\Lib\site-packages\scipy\weave\build_tools.py", line
> 269, in
> build_extension
>     setup(name = module_name, ext_modules = [ext],verbose=verb)
>   File "C:\Python24\Lib\site-packages\numpy\distutils\core.py", line 184,
> in set
> up
>     return old_setup(**new_attr)
>   File "C:\Python24\lib\distutils\core.py", line 166, in setup
>     raise SystemExit, "error: " + str(msg)
> distutils.errors.CompileError: error: Command "g++ -mno-cygwin -O2 -Wall
> -IC:\Py
> thon24\lib\site-packages\scipy\weave
> -IC:\Python24\lib\site-packages\scipy\weave
> \scxx -IC:\Python24\lib\site-packages\numpy\core\include
> -IC:\Python24\include -
> IC:\Python24\PC -c
> C:\Python24\lib\site-packages\scipy\weave\scxx\weave_imp.cpp
> -o c:\docume~1\lisear~1\lokale~1\temp\Lise
> Arleth\python24_intermediate\compiler
>
> _08edc7e348e1c33f63a33ab500aef08e\Release\python24\lib\site-packages\scipy\weave
> \scxx\weave_imp.o" failed with exit status 1
>
> the test_weave file was something i found on the scipy dev wiki... I also
> tried some of my older files that also uses weave and they all give the same
> error...
>
> Can anyone help me with this?
>
>
> ------------------------------
>
> _______________________________________________
> SciPy-user mailing listSciPy-user at scipy.orghttp://projects.scipy.org/mailman/listinfo/scipy-user
>
>
> --
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Samuel Garcia
> Laboratoire de Neurosciences Sensorielles, Comportement, Cognition.
> CNRS - UMR5020 - Universite Claude Bernard LYON 1
> Equipe logistique et technique
> 50, avenue Tony Garnier
> 69366 LYON Cedex 07
> FRANCE
> T?l : 04 37 28 74 64
> Fax : 04 37 28 76 01http://olfac.univ-lyon1.fr/unite/equipe-07/http://neuralensemble.org/trac/OpenElectrophy
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080707/e20d8235/attachment.html>

From jturner at gemini.edu  Mon Jul  7 20:36:48 2008
From: jturner at gemini.edu (James Turner)
Date: Mon, 07 Jul 2008 20:36:48 -0400
Subject: [SciPy-user] Portability problems (on Solaris)
Message-ID: <4872B6A0.8080706@gemini.edu>

Following my last post on numpy-discussion (Re: Ctypes required?
Fails to build.), I have found some additional portability
issues -- this time in SciPy (0.6.0) -- that cause my build to
fail on Solaris 9 with the Sun WorkShop 6 compiler.

Is this a good place to report these, or should I make a ticket
for them? They don't really need any troubleshooting -- just
changing (or not).

1. In the ndimage module, there is a C++ style comment
    at lines 144-145 of ni_interpolation.c.

2. The Sun ForTran compiler doesn't like the "END" syntax
    used in linalg/iterative -- for example
    "END SUBROUTINE sSTOPTEST2" fails but "END" on its
    own works (a shame, since the first syntax is nicer).
    This applies to all the ForTran files in that directory.

3. The main INSTALL.txt file says this: "Currently the SciPy
    build process does not use a C++ compiler, but the SciPy
    module Weave uses a C++ compiler at run time, so it is
    good to have a C++ compiler around as well". However,
    when I try to build SciPy with no C++ compiler available,
    the build stops at "scipy/sparese/sparsetools/
    sparsetools_wrap.cxx" with the error "c++: not found" and
    SciPy doesn't get installed. Do I need to tell it not to
    build certain modules?

Thanks,

James.



From robert.kern at gmail.com  Mon Jul  7 20:56:36 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Mon, 7 Jul 2008 19:56:36 -0500
Subject: [SciPy-user] Portability problems (on Solaris)
In-Reply-To: <4872B6A0.8080706@gemini.edu>
References: <4872B6A0.8080706@gemini.edu>
Message-ID: <3d375d730807071756q6057eea0pf0cc95730a176cfb@mail.gmail.com>

On Mon, Jul 7, 2008 at 19:36, James Turner <jturner at gemini.edu> wrote:
> Following my last post on numpy-discussion (Re: Ctypes required?
> Fails to build.), I have found some additional portability
> issues -- this time in SciPy (0.6.0) -- that cause my build to
> fail on Solaris 9 with the Sun WorkShop 6 compiler.
>
> Is this a good place to report these, or should I make a ticket
> for them? They don't really need any troubleshooting -- just
> changing (or not).
>
> 1. In the ndimage module, there is a C++ style comment
>    at lines 144-145 of ni_interpolation.c.

I can fix that. Unfortunately, there are a lot of new such comments in SVN.

> 2. The Sun ForTran compiler doesn't like the "END" syntax
>    used in linalg/iterative -- for example
>    "END SUBROUTINE sSTOPTEST2" fails but "END" on its
>    own works (a shame, since the first syntax is nicer).
>    This applies to all the ForTran files in that directory.

Isn't that standard Fortran-77? If it isn't, I'll change it. If it is,
I am going to have to insist on not supporting such broken compilers.

> 3. The main INSTALL.txt file says this: "Currently the SciPy
>    build process does not use a C++ compiler, but the SciPy
>    module Weave uses a C++ compiler at run time, so it is
>    good to have a C++ compiler around as well". However,
>    when I try to build SciPy with no C++ compiler available,
>    the build stops at "scipy/sparese/sparsetools/
>    sparsetools_wrap.cxx" with the error "c++: not found" and
>    SciPy doesn't get installed. Do I need to tell it not to
>    build certain modules?

The docs are old and wrong. scipy.sparse does require C++.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From dineshbvadhia at hotmail.com  Mon Jul  7 21:33:45 2008
From: dineshbvadhia at hotmail.com (Dinesh B Vadhia)
Date: Mon, 7 Jul 2008 18:33:45 -0700
Subject: [SciPy-user] A.mean(0) for sparse matrices
Message-ID: <BAY109-DS5C2C534A65A6EADFF12C1A3970@phx.gbl>

Binary sparse matrices are created as follows:

row = numpy.zeros(nnz, dtype='intc')
column = numpy.zeros(nnz, dtype='intc')
data = scipy.ones(nnz, dtype='intc')
A = sparse.csr_matrix((data, (row, column)), shape=(I,J))

Next, I want to get the mean of A with:

column_mean = A.mean(0)

I would expect column_mean to be a float vector but instead I get an integer vector.  Have I missed something?  

Dinesh

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080707/96827d05/attachment.html>

From ryanlists at gmail.com  Tue Jul  8 01:17:13 2008
From: ryanlists at gmail.com (Ryan Krauss)
Date: Tue, 8 Jul 2008 00:17:13 -0500
Subject: [SciPy-user] EPD on Vista
In-Reply-To: <3d375d730807051628j55ab6b45l7b2fea32dbe80684@mail.gmail.com>
References: <350f72300807050920r1d8625c3j355d2f20ff2fc1ed@mail.gmail.com>
	<3d375d730807051628j55ab6b45l7b2fea32dbe80684@mail.gmail.com>
Message-ID: <c5b438120807072217i2a7437e9w9214fe9795b19487@mail.gmail.com>

You might try right clicking on the installer and choosing run as
administrator.  This has solved some Vista install problems for my
students (though not with EPD).

On Sat, Jul 5, 2008 at 6:28 PM, Robert Kern <robert.kern at gmail.com> wrote:
> On Sat, Jul 5, 2008 at 11:20, Hector Villafuerte <hectorvd at gmail.com> wrote:
>> Hi,
>> a colleague of mine is having troubles installing the Enthought Python
>> Distro on Windows Vista. I don't have access to Vista now (just
>> Windows XP, where EPD works like a charm). So basically my question:
>> has somebody successfully installed EPD on Vista? (also, it seems that
>> Vista is not currently supported).
>
> I think we tried once with an early unreleased pre-alpha, and it
> worked. Vista is not officially supported, yet. However, we
> (Enthought) would be interested in receiving bug reports on the
> enthought-dev mailing list. There might be simple things we can do to
> get it to work for your colleague.
>
>  https://mail.enthought.com/mailman/listinfo/enthought-dev
>
> --
> Robert Kern
>
> "I have come to believe that the whole world is an enigma, a harmless
> enigma that is made terrible by our own mad attempt to interpret it as
> though it had an underlying truth."
>  -- Umberto Eco
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>


From wnbell at gmail.com  Tue Jul  8 06:08:10 2008
From: wnbell at gmail.com (Nathan Bell)
Date: Tue, 8 Jul 2008 05:08:10 -0500
Subject: [SciPy-user] A.mean(0) for sparse matrices
In-Reply-To: <BAY109-DS5C2C534A65A6EADFF12C1A3970@phx.gbl>
References: <BAY109-DS5C2C534A65A6EADFF12C1A3970@phx.gbl>
Message-ID: <d05265cb0807080308rc373787ica4b284939b93624@mail.gmail.com>

On Mon, Jul 7, 2008 at 8:33 PM, Dinesh B Vadhia
<dineshbvadhia at hotmail.com> wrote:
>
> Next, I want to get the mean of A with:
> column_mean = A.mean(0)
>
> I would expect column_mean to be a float vector but instead I get an integer
> vector.  Have I missed something?
>

Nope, it's a bug.

-- 
Nathan Bell wnbell at gmail.com
http://graphics.cs.uiuc.edu/~wnbell/


From twaite at berkeley.edu  Tue Jul  8 10:36:34 2008
From: twaite at berkeley.edu (Tom Waite)
Date: Tue, 8 Jul 2008 07:36:34 -0700
Subject: [SciPy-user] EPD on Vista
In-Reply-To: <3d375d730807051628j55ab6b45l7b2fea32dbe80684@mail.gmail.com>
References: <350f72300807050920r1d8625c3j355d2f20ff2fc1ed@mail.gmail.com>
	<3d375d730807051628j55ab6b45l7b2fea32dbe80684@mail.gmail.com>
Message-ID: <fe702a7c0807080736wda419abv12fcb32c2dbb89db@mail.gmail.com>

I will try EPD on my Vista 32 laptop. I have had problems with Vista. Cygwin
would not install fully (I would have to use task manager to kill some of
the installations), Matlab 6 would not install and there are problems with
older versions of MSVS (VC6 will install but project builds will have
problems).



On Sat, Jul 5, 2008 at 4:28 PM, Robert Kern <robert.kern at gmail.com> wrote:

> On Sat, Jul 5, 2008 at 11:20, Hector Villafuerte <hectorvd at gmail.com>
> wrote:
> > Hi,
> > a colleague of mine is having troubles installing the Enthought Python
> > Distro on Windows Vista. I don't have access to Vista now (just
> > Windows XP, where EPD works like a charm). So basically my question:
> > has somebody successfully installed EPD on Vista? (also, it seems that
> > Vista is not currently supported).
>
> I think we tried once with an early unreleased pre-alpha, and it
> worked. Vista is not officially supported, yet. However, we
> (Enthought) would be interested in receiving bug reports on the
> enthought-dev mailing list. There might be simple things we can do to
> get it to work for your colleague.
>
>  https://mail.enthought.com/mailman/listinfo/enthought-dev
>
> --
> Robert Kern
>
> "I have come to believe that the whole world is an enigma, a harmless
> enigma that is made terrible by our own mad attempt to interpret it as
> though it had an underlying truth."
>  -- Umberto Eco
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080708/e3f8f6f3/attachment.html>

From jturner at gemini.edu  Tue Jul  8 11:14:36 2008
From: jturner at gemini.edu (James Turner)
Date: Tue, 08 Jul 2008 11:14:36 -0400
Subject: [SciPy-user] Portability problems (on Solaris)
References: 3d375d730807071756q6057eea0pf0cc95730a176cfb@mail.gmail.com
Message-ID: <4873845C.4080904@gemini.edu>

> > 2. The Sun ForTran compiler doesn't like the "END" syntax
> >    used in linalg/iterative -- for example
> >    "END SUBROUTINE sSTOPTEST2" fails but "END" on its
> >    own works (a shame, since the first syntax is nicer).
> >    This applies to all the ForTran files in that directory.
>
> Isn't that standard Fortran-77? If it isn't, I'll change it. If
> it is, I am going to have to insist on not supporting such
> broken compilers.

I did a quick search for "Fortran 77 standard" and came up with
the following, which suggests that the syntax is just "END" without
extra characters but AFAICT doesn't explicitly forbid the latter.

  http://www.fortran.com/fortran/F77_std/rjcnf0001-sh-11.html#sh-11.14

> The docs are old and wrong. scipy.sparse does require C++.

OK, thanks. I'd be happy to update that paragraph (is it just
sparse that needs C++?) if there is some wiki page for editing
documentation like NumPy has? In the meantime I'm trying to get
a working C++ compiler installed (seemed for a moment like I
was finally almost finished building stuff...).

Cheers,

James.



From jturner at gemini.edu  Tue Jul  8 11:31:33 2008
From: jturner at gemini.edu (James Turner)
Date: Tue, 08 Jul 2008 11:31:33 -0400
Subject: [SciPy-user] Portability problems (on Solaris)
In-Reply-To: <4873845C.4080904@gemini.edu>
References: 3d375d730807071756q6057eea0pf0cc95730a176cfb@mail.gmail.com
	<4873845C.4080904@gemini.edu>
Message-ID: <48738855.8060207@gemini.edu>

>> The docs are old and wrong. scipy.sparse does require C++.
> 
> OK, thanks. I'd be happy to update that paragraph (is it just
> sparse that needs C++?) if there is some wiki page for editing
> documentation like NumPy has? In the meantime I'm trying to get
> a working C++ compiler installed (seemed for a moment like I
> was finally almost finished building stuff...).

I have updated the following wiki page, which is similar to the
INSTALL.txt file, but I'm not sure how to change the latter:

   http://scipy.org/Installing_SciPy/BuildingGeneral

Cheers,

James.



From isaulv at gmail.com  Tue Jul  8 16:31:51 2008
From: isaulv at gmail.com (Isaul Vargas)
Date: Tue, 8 Jul 2008 16:31:51 -0400
Subject: [SciPy-user] Scipy optimize fmin_l_bfgs_b gives me an obscure error
Message-ID: <12e2f0d90807081331q6a676115pc5c8f4854ab6425e@mail.gmail.com>

I am using Scipy 0.6 on windows vista with python 2.5.2 (same thing happens
in my linux setup), and when I use the function fmin_l_bfgs_b, I get this
obscure error:

ValueError: failed to initialize intent(inout) array -- input not fortran
contiguous

This is a part of the traceback within the function fmin_l_bfgs_b:
 197         _lbfgsb.setulb(m, x, low_bnd, upper_bnd, nbd, f, g, factr,
 198                        pgtol, wa, iwa, task, iprint, csave, lsave,
 199                        isave, dsave)


I have looked at the variables involved and they are all 1d arrays, so I
don't see how fortran ordering would apply.

Any ideas on how to fix this?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080708/c4b7df7b/attachment.html>

From nwagner at iam.uni-stuttgart.de  Tue Jul  8 17:37:12 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Tue, 08 Jul 2008 23:37:12 +0200
Subject: [SciPy-user] Scipy optimize fmin_l_bfgs_b gives me an obscure
 error
In-Reply-To: <12e2f0d90807081331q6a676115pc5c8f4854ab6425e@mail.gmail.com>
References: <12e2f0d90807081331q6a676115pc5c8f4854ab6425e@mail.gmail.com>
Message-ID: <web-104405419@uni-stuttgart.de>

On Tue, 8 Jul 2008 16:31:51 -0400
  "Isaul Vargas" <isaulv at gmail.com> wrote:
> I am using Scipy 0.6 on windows vista with python 2.5.2 
>(same thing happens
> in my linux setup), and when I use the function 
>fmin_l_bfgs_b, I get this
> obscure error:
> 
> ValueError: failed to initialize intent(inout) array -- 
>input not fortran
> contiguous
> 
> This is a part of the traceback within the function 
>fmin_l_bfgs_b:
> 197         _lbfgsb.setulb(m, x, low_bnd, upper_bnd, 
>nbd, f, g, factr,
> 198                        pgtol, wa, iwa, task, iprint, 
>csave, lsave,
> 199                        isave, dsave)
> 
> 
> I have looked at the variables involved and they are all 
>1d arrays, so I
> don't see how fortran ordering would apply.
> 
> Any ideas on how to fix this?
  
Please can you submit your example ?

Nils


From robert.kern at gmail.com  Tue Jul  8 18:11:39 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Tue, 8 Jul 2008 17:11:39 -0500
Subject: [SciPy-user] Portability problems (on Solaris)
In-Reply-To: <4873845C.4080904@gemini.edu>
References: <4873845C.4080904@gemini.edu>
Message-ID: <3d375d730807081511j32530f3dr19deb0032adb8b65@mail.gmail.com>

On Tue, Jul 8, 2008 at 10:14, James Turner <jturner at gemini.edu> wrote:
>> > 2. The Sun ForTran compiler doesn't like the "END" syntax
>> >    used in linalg/iterative -- for example
>> >    "END SUBROUTINE sSTOPTEST2" fails but "END" on its
>> >    own works (a shame, since the first syntax is nicer).
>> >    This applies to all the ForTran files in that directory.
>>
>> Isn't that standard Fortran-77? If it isn't, I'll change it. If
>> it is, I am going to have to insist on not supporting such
>> broken compilers.
>
> I did a quick search for "Fortran 77 standard" and came up with
> the following, which suggests that the syntax is just "END" without
> extra characters but AFAICT doesn't explicitly forbid the latter.
>
>  http://www.fortran.com/fortran/F77_std/rjcnf0001-sh-11.html#sh-11.14

Hmm. I can't find an explicit statement either, but I changed them anyways.

>> The docs are old and wrong. scipy.sparse does require C++.
>
> OK, thanks. I'd be happy to update that paragraph (is it just
> sparse that needs C++?)

scipy.stats.models seems to sort of optionally have some generated C++
code that it might possibly be able to ignore at runtime. It's a bit
vague. We should sort that situation out.

But otherwise, no. Just scipy.sparse.

> if there is some wiki page for editing
> documentation like NumPy has?

Not at the moment.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From anirudhvij at gmail.com  Wed Jul  9 09:03:24 2008
From: anirudhvij at gmail.com (anirudh vij)
Date: Wed, 9 Jul 2008 15:03:24 +0200
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
Message-ID: <1aa72c450807090603k1237c3dcm79d8a73558001b08@mail.gmail.com>

Hi,

I have a bunch of matlab scripts that I want to convert to
scipy/numpy/pylab. There is a nice resource at
mathesaurus.sourceforge.net

However, it is tedious and error prone to do these conversions manually. For eg:

1. there is no "end" in python. So all "end" statements from matlab
need to be deleted
2. the syntax "for i=1:10" has to be changed to "for i in range(1,11):"
3. special functions from matlab need to be converted to their scipy
equivalents.
4. For comments "%" must be replaced by "#"
5. for array scripting "(<index>)" should be replaced by "[<index>]"
6. Many others........

To my mind these examples are prime contenders for automation. A
reasonably small python/perl script can do the above(it will get more
complicated ofcourse, but the principle is the same).

I realize that such a script/module wont be a complete replacement. A
human will still need to look at the output python file and edit it.
But the script/module can remove the bullwork and leave the human
programmer to do more productive stuff.

Is there a pre-existing project that accom,plishes this kind of stuff.
If not, I would like to start one. Feedback/Suggestions will be very
helpful.


From david.huard at gmail.com  Wed Jul  9 09:44:06 2008
From: david.huard at gmail.com (David Huard)
Date: Wed, 9 Jul 2008 09:44:06 -0400
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <1aa72c450807090603k1237c3dcm79d8a73558001b08@mail.gmail.com>
References: <1aa72c450807090603k1237c3dcm79d8a73558001b08@mail.gmail.com>
Message-ID: <91cf711d0807090644n155c6881m47c026ba653229a2@mail.gmail.com>

Hi,

As a ex-matlab user, I remember being in the position where I would either
1. translate matlab codes to python, 2. rewrite the whole thing Python-like.
I always ended up rewriting the code from scratch because 1. it's a great
way to learn python, and 2. there things that you can do in Python that you
don't even dream are possible in matlab, so it's worth rethinking the code
structure.

Besides, a translator that does half the job will likely be seen rather as
buggy software than as a useful tool for translation. I suggest you look at
ways to interface matlab and python to launch matlab scripts from python.

Sorry to be so pessimistic,

David


2008/7/9 anirudh vij <anirudhvij at gmail.com>:

> Hi,
>
> I have a bunch of matlab scripts that I want to convert to
> scipy/numpy/pylab. There is a nice resource at
> mathesaurus.sourceforge.net
>
> However, it is tedious and error prone to do these conversions manually.
> For eg:
>
> 1. there is no "end" in python. So all "end" statements from matlab
> need to be deleted
> 2. the syntax "for i=1:10" has to be changed to "for i in range(1,11):"
> 3. special functions from matlab need to be converted to their scipy
> equivalents.
> 4. For comments "%" must be replaced by "#"
> 5. for array scripting "(<index>)" should be replaced by "[<index>]"
> 6. Many others........
>
> To my mind these examples are prime contenders for automation. A
> reasonably small python/perl script can do the above(it will get more
> complicated ofcourse, but the principle is the same).
>
> I realize that such a script/module wont be a complete replacement. A
> human will still need to look at the output python file and edit it.
> But the script/module can remove the bullwork and leave the human
> programmer to do more productive stuff.
>
> Is there a pre-existing project that accom,plishes this kind of stuff.
> If not, I would like to start one. Feedback/Suggestions will be very
> helpful.
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080709/95fb43ed/attachment.html>

From anirudhvij at gmail.com  Wed Jul  9 09:52:44 2008
From: anirudhvij at gmail.com (anirudh vij)
Date: Wed, 9 Jul 2008 15:52:44 +0200
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <91cf711d0807090644n155c6881m47c026ba653229a2@mail.gmail.com>
References: <1aa72c450807090603k1237c3dcm79d8a73558001b08@mail.gmail.com>
	<91cf711d0807090644n155c6881m47c026ba653229a2@mail.gmail.com>
Message-ID: <1aa72c450807090652t34a9b908le57cee0736cd7266@mail.gmail.com>

On Wed, Jul 9, 2008 at 3:44 PM, David Huard <david.huard at gmail.com> wrote:
> Hi,
>
> As a ex-matlab user, I remember being in the position where I would either
> 1. translate matlab codes to python, 2. rewrite the whole thing Python-like.
> I always ended up rewriting the code from scratch because 1. it's a great
> way to learn python, and 2. there things that you can do in Python that you
> don't even dream are possible in matlab, so it's worth rethinking the code
> structure.
>

I already know python(i think :) ). My purpose is to preprocess text
from matlab so that i dont have to do the bullwork of typing range(n)
everywhere where 1:n is used etc.

I agree python is millions of times superior to matlab. It has to be,
since it is a _real_ programming language, unlike matlab.

> Besides, a translator that does half the job will likely be seen rather as
> buggy software than as a useful tool for translation. I suggest you look at
> ways to interface matlab and python to launch matlab scripts from python.
>

I want the guys in my workplace to make the switch from matlab to
python. Thanks to large scale brainwashing(by me), they are receptive
to the idea :)

However, it is too much to expert matlab gurus to leave their
environment of choice and make the switch in a week. So, I hope to use
this tool/module to ease the translation. As a nice side effect, it
also help me right away in changing stuff from matlab to scipy.

> Sorry to be so pessimistic,
>

I can understand your pessimism. But this does not replace the human
programmer, but makes his life less dull (hopefully!), by doing the
boring work for him.
> David
>
>
> 2008/7/9 anirudh vij <anirudhvij at gmail.com>:
>>
>> Hi,
>>
>> I have a bunch of matlab scripts that I want to convert to
>> scipy/numpy/pylab. There is a nice resource at
>> mathesaurus.sourceforge.net
>>
>> However, it is tedious and error prone to do these conversions manually.
>> For eg:
>>
>> 1. there is no "end" in python. So all "end" statements from matlab
>> need to be deleted
>> 2. the syntax "for i=1:10" has to be changed to "for i in range(1,11):"
>> 3. special functions from matlab need to be converted to their scipy
>> equivalents.
>> 4. For comments "%" must be replaced by "#"
>> 5. for array scripting "(<index>)" should be replaced by "[<index>]"
>> 6. Many others........
>>
>> To my mind these examples are prime contenders for automation. A
>> reasonably small python/perl script can do the above(it will get more
>> complicated ofcourse, but the principle is the same).
>>
>> I realize that such a script/module wont be a complete replacement. A
>> human will still need to look at the output python file and edit it.
>> But the script/module can remove the bullwork and leave the human
>> programmer to do more productive stuff.
>>
>> Is there a pre-existing project that accom,plishes this kind of stuff.
>> If not, I would like to start one. Feedback/Suggestions will be very
>> helpful.
>> _______________________________________________
>> SciPy-user mailing list
>> SciPy-user at scipy.org
>> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>

cheers,
Anirudh.


From ivo.maljevic at gmail.com  Wed Jul  9 10:13:29 2008
From: ivo.maljevic at gmail.com (Ivo Maljevic)
Date: Wed, 9 Jul 2008 10:13:29 -0400
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <1aa72c450807090652t34a9b908le57cee0736cd7266@mail.gmail.com>
References: <1aa72c450807090603k1237c3dcm79d8a73558001b08@mail.gmail.com>
	<91cf711d0807090644n155c6881m47c026ba653229a2@mail.gmail.com>
	<1aa72c450807090652t34a9b908le57cee0736cd7266@mail.gmail.com>
Message-ID: <826c64da0807090713w424ae510hffcc8c6dbda360a3@mail.gmail.com>

If they are matlab experts than maybe it will be enough to give them the
matlab-python cross reference:

http://mathesaurus.sourceforge.net/matlab-python-xref.pdf

Ivo

2008/7/9 anirudh vij <anirudhvij at gmail.com>:

> On Wed, Jul 9, 2008 at 3:44 PM, David Huard <david.huard at gmail.com> wrote:
> > Hi,
> >
> > As a ex-matlab user, I remember being in the position where I would
> either
> > 1. translate matlab codes to python, 2. rewrite the whole thing
> Python-like.
> > I always ended up rewriting the code from scratch because 1. it's a great
> > way to learn python, and 2. there things that you can do in Python that
> you
> > don't even dream are possible in matlab, so it's worth rethinking the
> code
> > structure.
> >
>
> I already know python(i think :) ). My purpose is to preprocess text
> from matlab so that i dont have to do the bullwork of typing range(n)
> everywhere where 1:n is used etc.
>
> I agree python is millions of times superior to matlab. It has to be,
> since it is a _real_ programming language, unlike matlab.
>
> > Besides, a translator that does half the job will likely be seen rather
> as
> > buggy software than as a useful tool for translation. I suggest you look
> at
> > ways to interface matlab and python to launch matlab scripts from python.
> >
>
> I want the guys in my workplace to make the switch from matlab to
> python. Thanks to large scale brainwashing(by me), they are receptive
> to the idea :)
>
> However, it is too much to expert matlab gurus to leave their
> environment of choice and make the switch in a week. So, I hope to use
> this tool/module to ease the translation. As a nice side effect, it
> also help me right away in changing stuff from matlab to scipy.
>
> > Sorry to be so pessimistic,
> >
>
> I can understand your pessimism. But this does not replace the human
> programmer, but makes his life less dull (hopefully!), by doing the
> boring work for him.
> > David
> >
> >
> > 2008/7/9 anirudh vij <anirudhvij at gmail.com>:
> >>
> >> Hi,
> >>
> >> I have a bunch of matlab scripts that I want to convert to
> >> scipy/numpy/pylab. There is a nice resource at
> >> mathesaurus.sourceforge.net
> >>
> >> However, it is tedious and error prone to do these conversions manually.
> >> For eg:
> >>
> >> 1. there is no "end" in python. So all "end" statements from matlab
> >> need to be deleted
> >> 2. the syntax "for i=1:10" has to be changed to "for i in range(1,11):"
> >> 3. special functions from matlab need to be converted to their scipy
> >> equivalents.
> >> 4. For comments "%" must be replaced by "#"
> >> 5. for array scripting "(<index>)" should be replaced by "[<index>]"
> >> 6. Many others........
> >>
> >> To my mind these examples are prime contenders for automation. A
> >> reasonably small python/perl script can do the above(it will get more
> >> complicated ofcourse, but the principle is the same).
> >>
> >> I realize that such a script/module wont be a complete replacement. A
> >> human will still need to look at the output python file and edit it.
> >> But the script/module can remove the bullwork and leave the human
> >> programmer to do more productive stuff.
> >>
> >> Is there a pre-existing project that accom,plishes this kind of stuff.
> >> If not, I would like to start one. Feedback/Suggestions will be very
> >> helpful.
> >> _______________________________________________
> >> SciPy-user mailing list
> >> SciPy-user at scipy.org
> >> http://projects.scipy.org/mailman/listinfo/scipy-user
> >
> >
> > _______________________________________________
> > SciPy-user mailing list
> > SciPy-user at scipy.org
> > http://projects.scipy.org/mailman/listinfo/scipy-user
> >
> >
>
> cheers,
> Anirudh.
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080709/b16f696d/attachment.html>

From anirudhvij at gmail.com  Wed Jul  9 10:29:39 2008
From: anirudhvij at gmail.com (anirudh vij)
Date: Wed, 9 Jul 2008 16:29:39 +0200
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <826c64da0807090713w424ae510hffcc8c6dbda360a3@mail.gmail.com>
References: <1aa72c450807090603k1237c3dcm79d8a73558001b08@mail.gmail.com>
	<91cf711d0807090644n155c6881m47c026ba653229a2@mail.gmail.com>
	<1aa72c450807090652t34a9b908le57cee0736cd7266@mail.gmail.com>
	<826c64da0807090713w424ae510hffcc8c6dbda360a3@mail.gmail.com>
Message-ID: <1aa72c450807090729y259fc803yc27917e4f3d29ad7@mail.gmail.com>

On Wed, Jul 9, 2008 at 4:13 PM, Ivo Maljevic <ivo.maljevic at gmail.com> wrote:
> If they are matlab experts than maybe it will be enough to give them the
> matlab-python cross reference:
>
> http://mathesaurus.sourceforge.net/matlab-python-xref.pdf
>
Mathersaurus is great for guys creating new python code(where they can
use the cross-reference).

But conversion of legacy matlab code (which i have to do,
unfortunately) is still an issue. A python script could perhaps help
with that. I dont know whether such a script is better called a
preprocessor or a converter.

cheers,
Anirudh.


From rob.clewley at gmail.com  Wed Jul  9 11:28:15 2008
From: rob.clewley at gmail.com (Rob Clewley)
Date: Wed, 9 Jul 2008 11:28:15 -0400
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <1aa72c450807090729y259fc803yc27917e4f3d29ad7@mail.gmail.com>
References: <1aa72c450807090603k1237c3dcm79d8a73558001b08@mail.gmail.com>
	<91cf711d0807090644n155c6881m47c026ba653229a2@mail.gmail.com>
	<1aa72c450807090652t34a9b908le57cee0736cd7266@mail.gmail.com>
	<826c64da0807090713w424ae510hffcc8c6dbda360a3@mail.gmail.com>
	<1aa72c450807090729y259fc803yc27917e4f3d29ad7@mail.gmail.com>
Message-ID: <a749952d0807090828h45f2c43ct3609b28c555c9178@mail.gmail.com>

Unfortunately, basic "for" loop syntax and "end" statements are just
the tip of the iceberg in terms of conversion issues. Even if you had
a reasonable way to fix those, I think you'll find it much harder to
automatically convert array syntax. Least of which is indices starting
at 0, not 1. That can really change the presentation of an algorithm
in subtle ways.

I'm no expert in writing parsers, etc., but you might learn something
about what's necessary to convert from matlab by looking at the python
2to3 tool that is intended to help users automatically upgrade their
code from the python version 2.x series to python 3000. This is fairly
complex stuff, making use of a formal python grammar in an abstract
syntax tree. I suspect you won't achieve a reliable conversion of
matlab scripts without a similarly sophisticated approach, which would
be a lot of work for you.

Given your success with your colleagues so far, I'd suggest you keep
diligently working on the psychological angle until they make the
effort themselves to discover how easy it is to learn python. Maybe
dangle them a carrot, such as a proper object-oriented example
relevant to your work to show off what they are missing...

-Rob

On Wed, Jul 9, 2008 at 10:29 AM, anirudh vij <anirudhvij at gmail.com> wrote:
> On Wed, Jul 9, 2008 at 4:13 PM, Ivo Maljevic <ivo.maljevic at gmail.com> wrote:
>> If they are matlab experts than maybe it will be enough to give them the
>> matlab-python cross reference:
>>
>> http://mathesaurus.sourceforge.net/matlab-python-xref.pdf
>>
> Mathersaurus is great for guys creating new python code(where they can
> use the cross-reference).
>
> But conversion of legacy matlab code (which i have to do,
> unfortunately) is still an issue. A python script could perhaps help
> with that. I dont know whether such a script is better called a
> preprocessor or a converter.
>
> cheers,
> Anirudh.
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>


From pete.forman at westerngeco.com  Wed Jul  9 12:49:47 2008
From: pete.forman at westerngeco.com (Pete Forman)
Date: Wed, 09 Jul 2008 17:49:47 +0100
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
References: <1aa72c450807090603k1237c3dcm79d8a73558001b08@mail.gmail.com>
	<91cf711d0807090644n155c6881m47c026ba653229a2@mail.gmail.com>
	<1aa72c450807090652t34a9b908le57cee0736cd7266@mail.gmail.com>
	<826c64da0807090713w424ae510hffcc8c6dbda360a3@mail.gmail.com>
	<1aa72c450807090729y259fc803yc27917e4f3d29ad7@mail.gmail.com>
	<a749952d0807090828h45f2c43ct3609b28c555c9178@mail.gmail.com>
Message-ID: <7ibvdmtw.fsf@wgmail2.gatwick.eur.slb.com>

"Rob Clewley" <rob.clewley at gmail.com> writes:

 > I'm no expert in writing parsers, etc., but you might learn
 > something about what's necessary to convert from matlab ...

I've done a few and would recommend that general approach.  It will be
much easier doing things like changing to zero base indexes working
with an AST.  ANTLR would be my tool of choice but there are others.
Just avoid lex and friends, they are a lot of work compared with
modern parsers.
-- 
Pete Forman                -./\.-  Disclaimer: This post is originated
WesternGeco                  -./\.-   by myself and does not represent
pete.forman at westerngeco.com    -./\.-   the opinion of Schlumberger or
http://petef.22web.net           -./\.-   WesternGeco.



From millman at berkeley.edu  Wed Jul  9 13:04:46 2008
From: millman at berkeley.edu (Jarrod Millman)
Date: Wed, 9 Jul 2008 10:04:46 -0700
Subject: [SciPy-user] REMINDER: SciPy 2008 Early Registration ends in 2 days
Message-ID: <c7009a550807091004w48070735g70d082813e7250a0@mail.gmail.com>

Hello,

This is a reminder that early registration for SciPy 2008 ends in two
days  on Friday, July 11th.  To register, please see:
 http://conference.scipy.org/to_register

This year's conference has two days for tutorials, two days of
presentations, and ends with a two day coding sprint.  If you want to
learn more see my blog post:
http://jarrodmillman.blogspot.com/2008/07/scipy-2008-conference-program-posted.html

Cheers,

-- 
Jarrod Millman
Computational Infrastructure for Research Labs
10 Giannini Hall, UC Berkeley
phone: 510.643.4014
http://cirl.berkeley.edu/


From isaulv at gmail.com  Wed Jul  9 13:24:59 2008
From: isaulv at gmail.com (Isaul Vargas)
Date: Wed, 9 Jul 2008 13:24:59 -0400
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
Message-ID: <12e2f0d90807091024p42aa7977vced0ae210a0a9c50@mail.gmail.com>

There are some issues in converting Matlab code to Python.
Matlab has "dynamic arrays". Suppose you have a 1x 3 array, and you want to
add a new row via this syntax:
a[4,:] = b;
This won't work in python. You will have to preallocated the array size
somehow.
Second, it's the 0 based indexing. Generally you shift indices by -1, but
sometimes you may have to add a 1 to the end range .

Also concatenation can be a pain.

Hope this helps.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080709/c4d76567/attachment.html>

From bjracine at glosten.com  Wed Jul  9 13:42:25 2008
From: bjracine at glosten.com (Benjamin Racine )
Date: Wed, 09 Jul 2008 10:42:25 -0700
Subject: [SciPy-user] Fwd: pep8.py
References: <48749568.7EC6.00A7.0@glosten.com>
Message-ID: <487495F1.7EC6.00A7.0@glosten.com>

I am new,

Anybody care to tell me where I
should put the pep8.py style conformance
checking script in order
that ipython will see it when i invoke:

run pep8.py test.py

I have installed from python(x,y) on windows.

Thanks,
Ben R.


From anirudhvij at gmail.com  Wed Jul  9 13:59:15 2008
From: anirudhvij at gmail.com (anirudh vij)
Date: Wed, 9 Jul 2008 19:59:15 +0200
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <12e2f0d90807091024p42aa7977vced0ae210a0a9c50@mail.gmail.com>
References: <12e2f0d90807091024p42aa7977vced0ae210a0a9c50@mail.gmail.com>
Message-ID: <1aa72c450807091059j2d373f95ia290097cafd1b35c@mail.gmail.com>

On Wed, Jul 9, 2008 at 7:24 PM, Isaul Vargas <isaulv at gmail.com> wrote:
> There are some issues in converting Matlab code to Python.
Tons of them infact :)
The idea is to handle the easy stuff via the script, and the harder
ones manually.

> Matlab has "dynamic arrays". Suppose you have a 1x 3 array, and you want to
> add a new row via this syntax:
> a[4,:] = b;
> This won't work in python. You will have to preallocated the array size
> somehow.
> Second, it's the 0 based indexing. Generally you shift indices by -1, but
> sometimes you may have to add a 1 to the end range .
>
This can be handled i think.

> Also concatenation can be a pain.
>
Currently using hstack and vstack
> Hope this helps.

I am working on the script right now. Will post it somewhere once the
thing starts working. Currently doing very basic stuff.
 As Pete pointed out, for a full conversion, something like ANTLR will
be needed.

cheers,
Anirudh.


From robert.kern at gmail.com  Wed Jul  9 15:31:34 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Wed, 9 Jul 2008 14:31:34 -0500
Subject: [SciPy-user] Fwd: pep8.py
In-Reply-To: <487495F1.7EC6.00A7.0@glosten.com>
References: <48749568.7EC6.00A7.0@glosten.com>
	<487495F1.7EC6.00A7.0@glosten.com>
Message-ID: <3d375d730807091231m15aa7a08x2931d85f9233919a@mail.gmail.com>

On Wed, Jul 9, 2008 at 12:42, Benjamin Racine <bjracine at glosten.com> wrote:
> I am new,
>
> Anybody care to tell me where I
> should put the pep8.py style conformance
> checking script in order
> that ipython will see it when i invoke:
>
> run pep8.py test.py
>
> I have installed from python(x,y) on windows.

You will want to ask IPython questions on the IPython mailing list,
but basically, you need to give the full path to the pep8.py file.
%run does not search your PATH or anything.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From isaulv at gmail.com  Wed Jul  9 19:17:58 2008
From: isaulv at gmail.com (Isaul Vargas)
Date: Wed, 9 Jul 2008 19:17:58 -0400
Subject: [SciPy-user] Scipy optimize fmin_l_bfgs_b gives me an obscure
	error
Message-ID: <12e2f0d90807091617u711d7fbbp9baf1e7de698d786@mail.gmail.com>

Here is how my function looks like:

xt =  scipy.optimize.fmin_l_bfgs_b(obj_grad_func, xcur, args =
(b,h,Beta,R,wR,wh,muh, alpha_b, beta_b, BN, sp), maxfun = 1000)


the args are mostly arrays except for BN and SP which are dicts that contain
various flags.

In the debugger, I tried the following:
xcur is a 1,3 array with the values 1,1,0.
I check the flags of xcur and I see it is not fortran contiguous.
I do a copy: xcur2 = xcur.copy('f')
and now I get this error message:
*** error: failed in converting 10th argument `wa' of _lbfgsb.setulb to
C/Fortra
n array

There is no variable 'wa' in any of my functions, so I am not sure why this
failing.

Also, how do I setup the bounds, I am not clear on how to set it. I have a
list of upper bounds (an array of 3 values) and a list of lower bounds Not
sure how to set them.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080709/d6767ed4/attachment.html>

From nwagner at iam.uni-stuttgart.de  Thu Jul 10 02:06:43 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Thu, 10 Jul 2008 08:06:43 +0200
Subject: [SciPy-user] Scipy optimize fmin_l_bfgs_b gives me an
 obscure	error
In-Reply-To: <12e2f0d90807091617u711d7fbbp9baf1e7de698d786@mail.gmail.com>
References: <12e2f0d90807091617u711d7fbbp9baf1e7de698d786@mail.gmail.com>
Message-ID: <web-104473733@uni-stuttgart.de>

On Wed, 9 Jul 2008 19:17:58 -0400
  "Isaul Vargas" <isaulv at gmail.com> wrote:
> Here is how my function looks like:
> 
> xt =  scipy.optimize.fmin_l_bfgs_b(obj_grad_func, xcur, 
>args =
> (b,h,Beta,R,wR,wh,muh, alpha_b, beta_b, BN, sp), maxfun 
>= 1000)
> 
> 
> the args are mostly arrays except for BN and SP which 
>are dicts that contain
> various flags.
> 
> In the debugger, I tried the following:
> xcur is a 1,3 array with the values 1,1,0.
> I check the flags of xcur and I see it is not fortran 
>contiguous.
> I do a copy: xcur2 = xcur.copy('f')
> and now I get this error message:
> *** error: failed in converting 10th argument `wa' of 
>_lbfgsb.setulb to
> C/Fortra
> n array
> 
> There is no variable 'wa' in any of my functions, so I 
>am not sure why this
> failing.
> 
> Also, how do I setup the bounds, I am not clear on how 
>to set it. I have a
> list of upper bounds (an array of 3 values) and a list 
>of lower bounds Not
> sure how to set them.

For lower and upper bounda:

bounds = [(2000.,2100.),(1800.,1850.),(1600.,1630.)]
x_lbl,f,d = optimize.fmin_l_bfgs_b(func1, x0, fprime=None, 
args=(), approx_grad=1, bounds=bounds)

Nils


From nwagner at iam.uni-stuttgart.de  Thu Jul 10 02:10:59 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Thu, 10 Jul 2008 08:10:59 +0200
Subject: [SciPy-user] Scipy optimize fmin_l_bfgs_b gives me an
 obscure	error
In-Reply-To: <12e2f0d90807091617u711d7fbbp9baf1e7de698d786@mail.gmail.com>
References: <12e2f0d90807091617u711d7fbbp9baf1e7de698d786@mail.gmail.com>
Message-ID: <web-104473892@uni-stuttgart.de>

On Wed, 9 Jul 2008 19:17:58 -0400
  "Isaul Vargas" <isaulv at gmail.com> wrote:
> Here is how my function looks like:
> 
> xt =  scipy.optimize.fmin_l_bfgs_b(obj_grad_func, xcur, 
>args =
> (b,h,Beta,R,wR,wh,muh, alpha_b, beta_b, BN, sp), maxfun 
>= 1000)
> 
> 
> the args are mostly arrays except for BN and SP which 
>are dicts that contain
> various flags.
> 
> In the debugger, I tried the following:
> xcur is a 1,3 array with the values 1,1,0.
> I check the flags of xcur and I see it is not fortran 
>contiguous.
> I do a copy: xcur2 = xcur.copy('f')
> and now I get this error message:
> *** error: failed in converting 10th argument `wa' of 
>_lbfgsb.setulb to
> C/Fortra
> n array
>
You can find the corresponding file (lbfgsb.pyf) in

scipy/scipy/optimize/lbfgsb


!    -*- f90 -*-
python module _lbfgsb ! in
     interface  ! in :_lbfgsb
         subroutine 
setulb(n,m,x,l,u,nbd,f,g,factr,pgtol,wa,iwa,task,iprint,csave,lsave,isave,dsave) 
! in :lbfsgb:routines.f
             integer 
intent(in),optional,check(len(x)>=n),depend(x) :: n=len(x)
             integer intent(in) :: m
             double precision dimension(n),intent(inout) 
:: x
             double precision 
dimension(n),depend(n),intent(in) :: l
             double precision 
dimension(n),depend(n),intent(in) :: u
             integer dimension(n),depend(n),intent(in) :: 
nbd
             double precision intent(inout) :: f
             double precision 
dimension(n),depend(n),intent(inout) :: g
             double precision intent(in) :: factr
             double precision intent(in) :: pgtol
             double precision 
dimension(2*m*n+4*n+12*m*m+12*m),depend(n,m),intent(inout) 
:: wa
             integer dimension(3 * 
n),depend(n),intent(inout) :: iwa
             character*60 intent(inout) :: task
             integer intent(in) :: iprint
             character*60 intent(inout) :: csave
             logical dimension(4),intent(inout) :: lsave
             integer dimension(44),intent(inout) :: isave
             double precision dimension(29),intent(inout) 
:: dsave
         end subroutine setulb
     end interface
end python module _lbfgsb

Nils

> There is no variable 'wa' in any of my functions, so I 
>am not sure why this
> failing.
> 
> Also, how do I setup the bounds, I am not clear on how 
>to set it. I have a
> list of upper bounds (an array of 3 values) and a list 
>of lower bounds Not
> sure how to set them.
  


From silva at lma.cnrs-mrs.fr  Thu Jul 10 09:13:55 2008
From: silva at lma.cnrs-mrs.fr (Fabrice Silva)
Date: Thu, 10 Jul 2008 15:13:55 +0200
Subject: [SciPy-user] odepack or f2py error
Message-ID: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>

Hi,
Using odeint, I've got an "Internal error" when the maximum number of
steps is reached :
         lsoda--  at current t (=r1), mxstep (=i1) steps   
               taken on this call before reaching tout     
              In above message,  I1 =      2000
        At line 1991 of file scipy/integrate/odepack/ddasrt.f (unit = 6,
        file = 'stdout')
        Internal Error: printf is broken
              In above message,  R1 =f
        fab at laptop$

as "printf is broken", I can not know the instant when it breaks...
I'm not sure the internal error is due to odeint or to f2py, as it seems
that I've already seen this problem out of odeint.

Any suggestion ?
-- 
Fabrice Silva



From pav at iki.fi  Thu Jul 10 15:30:33 2008
From: pav at iki.fi (Pauli Virtanen)
Date: Thu, 10 Jul 2008 19:30:33 +0000 (UTC)
Subject: [SciPy-user] odepack or f2py error
References: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>
Message-ID: <g55o0o$b0i$1@ger.gmane.org>

Thu, 10 Jul 2008 15:13:55 +0200, Fabrice Silva wrote:

> Hi,
> Using odeint, I've got an "Internal error" when the maximum number of
> steps is reached :
>          lsoda--  at current t (=r1), mxstep (=i1) steps
>                taken on this call before reaching tout
>               In above message,  I1 =      2000
>         At line 1991 of file scipy/integrate/odepack/ddasrt.f (unit = 6,
>         file = 'stdout')
>         Internal Error: printf is broken
>               In above message,  R1 =f
>         fab at laptop$
> 
> as "printf is broken", I can not know the instant when it breaks... I'm
> not sure the internal error is due to odeint or to f2py, as it seems
> that I've already seen this problem out of odeint.
> 
> Any suggestion ?

What Fortran compiler are you using? Some old version of gfortran perhaps?

I think this "Internal Error:" message is from your Fortran compiler. 
Googling for it brings up several messages referring to gfortran, so it's 
likely this is a gfortran bug (possibly fixed in newer versions).

-- 
Pauli Virtanen



From astrofitz at gmail.com  Thu Jul 10 19:19:49 2008
From: astrofitz at gmail.com (Michael Fitzgerald)
Date: Thu, 10 Jul 2008 16:19:49 -0700
Subject: [SciPy-user] patch for ndimage.generic_filter
Message-ID: <477CA520-C850-47AD-9888-92DDC968BBF3@gmail.com>


Hello,

I found a small bug in ndimage.generic_filter() that arises when  
using the 'size' keyword with multidimensional arrays.  Patch attached.

Mike

-------------- next part --------------
A non-text attachment was scrubbed...
Name: filter.diff
Type: application/octet-stream
Size: 580 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080710/e6c51365/attachment.obj>
-------------- next part --------------



From callen at gemini.edu  Fri Jul 11 21:26:48 2008
From: callen at gemini.edu (C. Allen)
Date: Fri, 11 Jul 2008 15:26:48 -1000
Subject: [SciPy-user] HDULists sharing HDU entries
Message-ID: <1215826008.5560.24.camel@phact.hi.gemini.edu>

I have been making a system for handling astronomical data, the actual
fits manipulation using pyfits. Please correct me if there is another
list this question would be more appropriate on, I did not see a pyfits
list at the project page.

When I started this system I was new to python (but experienced in many
other languages) and in the process I have done some things I am not
sure about. I of course test my ideas to make sure they work, but of
course, such tests can miss some condition in which the idiom ultimately
is bound to fail horribly.

So... I'm curious if something I'm doing with HDULists is ok.

Basically, we have an abstraction for our data, lets call it AstroData
because that's what we call it.  It contains a member which is the
HDUList as returned by pyfits when opening a MEF.

I also have some syntax that allows one to get an AstroData instance of
just sub-selections of extensions in the MEF, e.g.:

if ad = AstroData(fitsfilename) then (e.g.) ad["SCI"] returns a separate
AstroData instance from 'ad'. In this case I build the HDUList out of
the HDU entries in the original instance.  I make a python list with the
PHU and whatever extensions are being selected (in the sample case,
those where EXTNAME="SCI"), and initialize a new HDUList instance with
this list.  Then this unique HDUList is stored in the new AstroData
instance.  So each AstroData instance has it's own HDUList, but the HDU
entries in it are shared.

Everything works fine, to write the constructed HDUList is to save just
those extensions listed, as desired, and I can pass the HDUList around
as normal to anything that wants HDUList instances. Of course, if I
change any of the shared extensions, it will be seen in all HDULists
that refer to it, which is confusing to some but is exactly what we want
in this case.  

All this runs perfectly fine in practice. Does anyone know any reason
this might get me into trouble? It seems I have nothing to worry about,
but I have a nagging idea that perhaps the HDU is more tightly linked to
the list containing it in some way that I do not yet know.

thanks,
craig
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080711/3379f1fb/attachment.html>

From chanley at stsci.edu  Fri Jul 11 21:36:20 2008
From: chanley at stsci.edu (Christopher Hanley)
Date: Fri, 11 Jul 2008 21:36:20 -0400
Subject: [SciPy-user] HDULists sharing HDU entries
In-Reply-To: <1215826008.5560.24.camel@phact.hi.gemini.edu>
References: <1215826008.5560.24.camel@phact.hi.gemini.edu>
Message-ID: <48780A94.1030303@stsci.edu>

C. Allen wrote:
> I have been making a system for handling astronomical data, the actual 
> fits manipulation using pyfits. Please correct me if there is another 
> list this question would be more appropriate on, I did not see a pyfits 
> list at the project page.
> 
> When I started this system I was new to python (but experienced in many 
> other languages) and in the process I have done some things I am not 
> sure about. I of course test my ideas to make sure they work, but of 
> course, such tests can miss some condition in which the idiom ultimately 
> is bound to fail horribly.
> 
> So... I'm curious if something I'm doing with HDULists is ok.
> 
> Basically, we have an abstraction for our data, lets call it AstroData 
> because that's what we call it.  It contains a member which is the 
> HDUList as returned by pyfits when opening a MEF.
> 
> I also have some syntax that allows one to get an AstroData instance of 
> just sub-selections of extensions in the MEF, e.g.:
> 
> if ad = AstroData(fitsfilename) then (e.g.) ad["SCI"] returns a separate 
> AstroData instance from 'ad'. In this case I build the HDUList out of 
> the HDU entries in the original instance.  I make a python list with the 
> PHU and whatever extensions are being selected (in the sample case, 
> those where EXTNAME="SCI"), and initialize a new HDUList instance with 
> this list.  Then this unique HDUList is stored in the new AstroData 
> instance.  So each AstroData instance has it's own HDUList, but the HDU 
> entries in it are shared.
> 
> Everything works fine, to write the constructed HDUList is to save just 
> those extensions listed, as desired, and I can pass the HDUList around 
> as normal to anything that wants HDUList instances. Of course, if I 
> change any of the shared extensions, it will be seen in all HDULists 
> that refer to it, which is confusing to some but is exactly what we want 
> in this case. 
> 
> All this runs perfectly fine in practice. Does anyone know any reason 
> this might get me into trouble? It seems I have nothing to worry about, 
> but I have a nagging idea that perhaps the HDU is more tightly linked to 
> the list containing it in some way that I do not yet know.
> 
> thanks,
> craig
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user

Hi Craig,

The more appropriate list is probably astropy at scipy.org.

As to your question.  Assuming that I understand what you are 
attempting, I don't think you should run into any problems.

Chris


-- 
Christopher Hanley
Systems Software Engineer
Space Telescope Science Institute
3700 San Martin Drive
Baltimore MD, 21218
(410) 338-4338


From callen at gemini.edu  Fri Jul 11 21:53:05 2008
From: callen at gemini.edu (C. Allen)
Date: Fri, 11 Jul 2008 15:53:05 -1000
Subject: [SciPy-user] HDULists sharing HDU entries
In-Reply-To: <48780A94.1030303@stsci.edu>
References: <1215826008.5560.24.camel@phact.hi.gemini.edu>
	<48780A94.1030303@stsci.edu>
Message-ID: <1215827585.5560.27.camel@phact.hi.gemini.edu>

thanks chris

On Fri, 2008-07-11 at 21:36 -0400, Christopher Hanley wrote:

> C. Allen wrote:
> > I have been making a system for handling astronomical data, the actual 
> > fits manipulation using pyfits. Please correct me if there is another 
> > list this question would be more appropriate on, I did not see a pyfits 
> > list at the project page.
> > 
> > When I started this system I was new to python (but experienced in many 
> > other languages) and in the process I have done some things I am not 
> > sure about. I of course test my ideas to make sure they work, but of 
> > course, such tests can miss some condition in which the idiom ultimately 
> > is bound to fail horribly.
> > 
> > So... I'm curious if something I'm doing with HDULists is ok.
> > 
> > Basically, we have an abstraction for our data, lets call it AstroData 
> > because that's what we call it.  It contains a member which is the 
> > HDUList as returned by pyfits when opening a MEF.
> > 
> > I also have some syntax that allows one to get an AstroData instance of 
> > just sub-selections of extensions in the MEF, e.g.:
> > 
> > if ad = AstroData(fitsfilename) then (e.g.) ad["SCI"] returns a separate 
> > AstroData instance from 'ad'. In this case I build the HDUList out of 
> > the HDU entries in the original instance.  I make a python list with the 
> > PHU and whatever extensions are being selected (in the sample case, 
> > those where EXTNAME="SCI"), and initialize a new HDUList instance with 
> > this list.  Then this unique HDUList is stored in the new AstroData 
> > instance.  So each AstroData instance has it's own HDUList, but the HDU 
> > entries in it are shared.
> > 
> > Everything works fine, to write the constructed HDUList is to save just 
> > those extensions listed, as desired, and I can pass the HDUList around 
> > as normal to anything that wants HDUList instances. Of course, if I 
> > change any of the shared extensions, it will be seen in all HDULists 
> > that refer to it, which is confusing to some but is exactly what we want 
> > in this case. 
> > 
> > All this runs perfectly fine in practice. Does anyone know any reason 
> > this might get me into trouble? It seems I have nothing to worry about, 
> > but I have a nagging idea that perhaps the HDU is more tightly linked to 
> > the list containing it in some way that I do not yet know.
> > 
> > thanks,
> > craig
> > 
> > 
> > ------------------------------------------------------------------------
> > 
> > _______________________________________________
> > SciPy-user mailing list
> > SciPy-user at scipy.org
> > http://projects.scipy.org/mailman/listinfo/scipy-user
> 
> Hi Craig,
> 
> The more appropriate list is probably astropy at scipy.org.
> 
> As to your question.  Assuming that I understand what you are 
> attempting, I don't think you should run into any problems.
> 
> Chris
> 
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080711/2b60854e/attachment.html>

From silva at lma.cnrs-mrs.fr  Sat Jul 12 07:32:29 2008
From: silva at lma.cnrs-mrs.fr (Fabrice Silva)
Date: Sat, 12 Jul 2008 13:32:29 +0200
Subject: [SciPy-user] odepack or f2py error
In-Reply-To: <g55o0o$b0i$1@ger.gmane.org>
References: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>
	<g55o0o$b0i$1@ger.gmane.org>
Message-ID: <1215862349.3033.3.camel@Portable-s2m.cnrs-mrs.fr>

Le jeudi 10 juillet 2008 ? 19:30 +0000, Pauli Virtanen a ?crit :
> What Fortran compiler are you using? Some old version of gfortran
> perhaps?

I do have gfortran installed

        fab at Portable-s2m:~$ gfortran --version
        GNU Fortran (Debian 4.3.1-6) 4.3.1

How can I check gfortran is used by f2py ?

> I think this "Internal Error:" message is from your Fortran compiler. 
> Googling for it brings up several messages referring to gfortran, so
> it's likely this is a gfortran bug (possibly fixed in newer versions).

I am using debian unstable, by weekly upgrade. I remember using printf
calls for debugging last year without any error.
I have looked at the gfortran bugzilla, but I did not see any bugs which
seems similar to this one...
> 
-- 
Fabrice Silva <silva at lma.cnrs-mrs.fr>
LMA UPR CNRS 7051 - ?quipe S2M



From pav at iki.fi  Sat Jul 12 08:22:36 2008
From: pav at iki.fi (Pauli Virtanen)
Date: Sat, 12 Jul 2008 12:22:36 +0000 (UTC)
Subject: [SciPy-user] odepack or f2py error
References: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>
	<g55o0o$b0i$1@ger.gmane.org>
	<1215862349.3033.3.camel@Portable-s2m.cnrs-mrs.fr>
Message-ID: <g5a7mc$dk8$1@ger.gmane.org>

Sat, 12 Jul 2008 13:32:29 +0200, Fabrice Silva wrote:

> Le jeudi 10 juillet 2008 ? 19:30 +0000, Pauli Virtanen a ?crit :
>> What Fortran compiler are you using? Some old version of gfortran
>> perhaps?
> 
> I do have gfortran installed
> 
>         fab at Portable-s2m:~$ gfortran --version GNU Fortran (Debian
>         4.3.1-6) 4.3.1
> 
> How can I check gfortran is used by f2py ?

Look at the output when scipy is compiled, or try to ldd scipy/integrate/
vode.so and check if it's linked with libgfortran.

>> I think this "Internal Error:" message is from your Fortran compiler.
>> Googling for it brings up several messages referring to gfortran, so
>> it's likely this is a gfortran bug (possibly fixed in newer versions).
> 
> I am using debian unstable, by weekly upgrade. I remember using printf
> calls for debugging last year without any error. I have looked at the
> gfortran bugzilla, but I did not see any bugs which seems similar to
> this one...

It would help if you can write a minimal testcase for this, ie. what 
arguments to give to odeint so that it produces this error.

It's probably best to continue the discussion here:

	http://scipy.org/scipy/scipy/ticket/696

so that the information is archived on a single place.

-- 
Pauli Virtanen



From travis at enthought.com  Sat Jul 12 11:34:57 2008
From: travis at enthought.com (Travis Vaught)
Date: Sat, 12 Jul 2008 10:34:57 -0500
Subject: [SciPy-user] SciPy 2008 Registration Deadline Extended
Message-ID: <3A00FC0B-B463-4308-B0CF-7D6F3195E609@enthought.com>

Greetings,

The merchant account processor that we use for the SciPy Conference  
online registration has been experiencing some inexplicable problems  
authorizing some registrations.  Apologies to those who have struggled  
to register and have not been successful.  Because of the problems,  
we're extending the early-bird rates through Monday at midnight  
Central Time.  If you experience any problems registering, please give  
us a call during business hours Monday (9:00am - 5:00pm Central -  
512.536.1057).

http://conference.scipy.org/

For those of you who have set up an account on the conference site,  
but have not yet registered, I encourage you to do so in time to take  
advantage of the lower rates.  I also encourage everyone to make sure  
you've specified which tutorial track, T-shirt size, whether you'll  
attend the sprint, and meal preferences in your profile (http://conference.scipy.org/profile 
).

Please send me an email if you have any questions.

Best,

Travis




From williamhpurcell at gmail.com  Sun Jul 13 15:59:47 2008
From: williamhpurcell at gmail.com (William Purcell)
Date: Sun, 13 Jul 2008 14:59:47 -0500
Subject: [SciPy-user] signal.lti(A,B,C,D) with D=0
Message-ID: <d9186c3b0807131259v3e5732e3j8c0536095424b9b6@mail.gmail.com>

I am trying to set up state space representation of a system using
signal.lti. The system has no feedforward so D = 0. I've tried the three
options listed in the code below to represent D.

The only way I can get it to work is option 2 if C has 1 row. If C has more
than 1 row it won't work.

Any thoughts?
-Bill

Code
---------------------------------------------------------------
from numpy import ones,matrix
from scipy import signal
r = ones(3)
A = matrix([r,r,r])
B = matrix([r]).T
C = matrix([r,r])

#three options of D to make it 0
#1) D=0
#2) D = matrix([0,0]).T
#3) D = None

D = 0
#D = None
#D = matrix([0,0]).T

Gss = signal.lti(A,B,C,D)
-----------------------------------------------------------

Tracebacks
-----------------------------------------------------------
Option 1
/usr/lib/python2.5/site-packages/scipy/signal/ltisys.py in abcd_normalize(A,
B, C, D)
    101         raise ValueError, "A and C must have the same number of
columns."
    102     if MD != MC:
--> 103         raise ValueError, "C and D must have the same number of
rows."
    104     if ND != NB:
    105         raise ValueError, "B and D must have the same number of
columns."

<type 'exceptions.ValueError'>: C and D must have the same number of rows.
WARNING: Failure executing file: <testlti.py>

Option 2 (with C as two rows...if C is a single row I do not get this
traceback)
/usr/lib/python2.5/site-packages/scipy/signal/ltisys.py in ss2tf(A, B, C, D,
input)
    147
    148     num_states = A.shape[0]
--> 149     type_test = A[:,0] + B[:,0] + C[0,:] + D
    150     num = numpy.zeros((nout, num_states+1), type_test.dtype)
    151     for k in range(nout):

<type 'exceptions.ValueError'>: shape mismatch: objects cannot be broadcast
to a single shape
WARNING: Failure executing file: <testlti.py>

Option 3
same as 1
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080713/a656db19/attachment.html>

From contact at pythonxy.com  Sun Jul 13 13:19:08 2008
From: contact at pythonxy.com (Pierre Raybaut)
Date: Sun, 13 Jul 2008 19:19:08 +0200
Subject: [SciPy-user] [python(x,y)] New Release: 2.0.0
Message-ID: <629b08a40807131019o36a7c35cw19ee640bbbcc173@mail.gmail.com>

Hi all,

Python(x,y) 2.0.0 is now available on http://www.pythonxy.com.

Changes history
07 -12 -2008 - Version 2.0.0 :

    * Added:
        o ITK 3.6 (WrapITK) - Open-source software system for image
processing (leading-edge segmentation and registration algorithms)
        o Windows installer: new component manager system (updating is
easier and more flexible, install and uninstall process are much
cleaner, ...)
        o Windows installer: Python(x,y) may now be installed silently
(/S option), and the non-silent installation now allows the user to
keep working on the machine during the process
    * Updated:
        o Matplotlib 0.98.1 (see release notes)
        o SymPy 0.6.0
        o GDAL 1.5.2
        o PySerial 2.4

Regards,
Pierre Raybaut


From ryanlists at gmail.com  Mon Jul 14 09:41:43 2008
From: ryanlists at gmail.com (Ryan Krauss)
Date: Mon, 14 Jul 2008 08:41:43 -0500
Subject: [SciPy-user] signal.lti(A,B,C,D) with D=0
In-Reply-To: <d9186c3b0807131259v3e5732e3j8c0536095424b9b6@mail.gmail.com>
References: <d9186c3b0807131259v3e5732e3j8c0536095424b9b6@mail.gmail.com>
Message-ID: <c5b438120807140641o50ffe49cx6a8f9ec75768884d@mail.gmail.com>

This seems like a bug in ltisys.  I think this line
149     type_test = A[:,0] + B[:,0] + C[0,:] + D

should be
149     type_test = A[:,0] + B[:,0] + C[0,:] + D[0,:]

for a multipe-input/multiple-output system with i inputs, n states,
and m outputs, A should be n by n, B should be n by i, C should be m
by n, and D should be m by i.

(actually, because of this code a few lines above:
    # make MOSI from possibly MOMI system.
    if B.shape[-1] != 0:
        B = B[:,input]
    B.shape = (B.shape[0],1)
    if D.shape[-1] != 0:
        D = D[:,input]

I guess line 149 should be
149     type_test = A[:,0] + B[:,0] + C[0,:] + D[0]
)

But making this change exposes another problem:

/usr/lib/python2.5/site-packages/scipy/signal/ltisys.py in
__init__(self, *args, **kwords)
    224                                 self.__dict__['D'] =
abcd_normalize(*args)
    225             self.__dict__['zeros'], self.__dict__['poles'], \
--> 226                                     self.__dict__['gain'] =
ss2zpk(*args)
    227             self.__dict__['num'], self.__dict__['den'] = ss2tf(*args)
    228             self.inputs = self.B.shape[-1]

/usr/lib/python2.5/site-packages/scipy/signal/ltisys.py in ss2zpk(A,
B, C, D, input)
    183     """
    184     Pdb().set_trace()
--> 185     return tf2zpk(*ss2tf(A,B,C,D,input=input))
    186
    187 class lti(object):

/usr/lib/python2.5/site-packages/scipy/signal/filter_design.py in tf2zpk(b, a)
    128     k = b[0]
    129     b /= b[0]
--> 130     z = roots(b)
    131     p = roots(a)
    132     return z, p, k

/usr/lib/python2.5/site-packages/numpy/lib/polynomial.py in roots(p)
     98     p = atleast_1d(p)
     99     if len(p.shape) != 1:
--> 100         raise ValueError,"Input must be a rank-1 array."
    101
    102     # find non-zero array entries

<type 'exceptions.ValueError'>: Input must be a rank-1 array.
WARNING: Failure executing file: <proj2.py>

In [2]: %debug
> /usr/lib/python2.5/site-packages/numpy/lib/polynomial.py(100)roots()
     99     if len(p.shape) != 1:
--> 100         raise ValueError,"Input must be a rank-1 array."
    101

ipdb> print p
[[  1.00000000e+00   1.00000000e+00   1.00000000e+00   1.00000000e+00
    1.00000000e+00   1.00000000e+00]
 [  1.81898940e-11  -3.06222400e+00  -2.79196150e+02  -5.87518033e+03
   -1.59843721e+04   1.91386789e-07]
 [  1.75910240e+01   1.60479206e+03   3.37698682e+04   9.12618857e+04
   -1.23219975e+04  -3.29828641e+04]]


ss2tf seems to correctly handle MIMO (or at least SIMO systems)
correctly and returns one denominator polynomial with several (m)
numerator polynomials.  But tf2zpk in filter_design.py does not seem
able to handle more than siso systems (which makes sense, it is
expecting a transfer function which is just a SISO system).

How should this be fixed?

I understand why ss2tf converts a MIMO system to SIMO - trying to
represent a mulitple input, multiple output system with a transfer
function has some limitations.  I think the real offender is in the
__init__ method of signal.lti:

elif N == 4:
...
self.__dict__['zeros'], self.__dict__['poles'], \
                                    self.__dict__['gain'] = ss2zpk(*args)

For a MIMO system, what should the zpk representation be?  For a SIMO
system, I would expect a vector of poles and a matrix of zeros.  This
seems to be in line with what ss2tf does.

It seems like the init method could find the pole by passing C[0,:]
and D[0,:] to ss2zpk.  The zeros and gains could then be found by
calling ss2zpk in some vectorized way or simply in a for loop for each
row of C and D.  But there may well be a more elegant solution.

Any thoughts?

Ryan

On Sun, Jul 13, 2008 at 2:59 PM, William Purcell
<williamhpurcell at gmail.com> wrote:
> I am trying to set up state space representation of a system using
> signal.lti. The system has no feedforward so D = 0. I've tried the three
> options listed in the code below to represent D.
>
> The only way I can get it to work is option 2 if C has 1 row. If C has more
> than 1 row it won't work.
>
> Any thoughts?
> -Bill
>
> Code
> ---------------------------------------------------------------
> from numpy import ones,matrix
> from scipy import signal
> r = ones(3)
> A = matrix([r,r,r])
> B = matrix([r]).T
> C = matrix([r,r])
>
> #three options of D to make it 0
> #1) D=0
> #2) D = matrix([0,0]).T
> #3) D = None
>
> D = 0
> #D = None
> #D = matrix([0,0]).T
>
> Gss = signal.lti(A,B,C,D)
> -----------------------------------------------------------
>
> Tracebacks
> -----------------------------------------------------------
> Option 1
> /usr/lib/python2.5/site-packages/scipy/signal/ltisys.py in abcd_normalize(A,
> B, C, D)
>     101         raise ValueError, "A and C must have the same number of
> columns."
>     102     if MD != MC:
> --> 103         raise ValueError, "C and D must have the same number of
> rows."
>     104     if ND != NB:
>     105         raise ValueError, "B and D must have the same number of
> columns."
>
> <type 'exceptions.ValueError'>: C and D must have the same number of rows.
> WARNING: Failure executing file: <testlti.py>
>
> Option 2 (with C as two rows...if C is a single row I do not get this
> traceback)
> /usr/lib/python2.5/site-packages/scipy/signal/ltisys.py in ss2tf(A, B, C, D,
> input)
>     147
>     148     num_states = A.shape[0]
> --> 149     type_test = A[:,0] + B[:,0] + C[0,:] + D
>     150     num = numpy.zeros((nout, num_states+1), type_test.dtype)
>     151     for k in range(nout):
>
> <type 'exceptions.ValueError'>: shape mismatch: objects cannot be broadcast
> to a single shape
> WARNING: Failure executing file: <testlti.py>
>
> Option 3
> same as 1
>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>


From emanuele at relativita.com  Mon Jul 14 12:42:05 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Mon, 14 Jul 2008 18:42:05 +0200
Subject: [SciPy-user] [OpenOpt] Segmentation fault
Message-ID: <487B81DD.5000608@relativita.com>

Dear all,

Whenever I run example 'nlp_1.py' of OpenOpt (both tarball 0.18
and svn) on both my linux boxes I get "Segmentation fault"!
Or better, I get it 9 times out of ten (due to random
initialization without fixed seed, I guess).

My configuration:
box1: Ubuntu Gutsy, amd64 (Intel C2D)
box2: Ubuntu Hardy, amd64 (Intel QuadCore)

Python, Numpy, Scipy, Matplotlib etc. are standard versions
from Ubuntu repositories.

Can anyone confirm/reproduce this behavior?

I've a report that in i386 debian it seems to work flawlessly.

Thanks,

Emanuele



From emanuele at relativita.com  Mon Jul 14 13:06:47 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Mon, 14 Jul 2008 19:06:47 +0200
Subject: [SciPy-user] [OpenOpt] Segmentation fault
In-Reply-To: <487B81DD.5000608@relativita.com>
References: <487B81DD.5000608@relativita.com>
Message-ID: <487B87A7.8030400@relativita.com>

Emanuele Olivetti wrote:
> Dear all,
>
> Whenever I run example 'nlp_1.py' of OpenOpt (both tarball 0.18
> and svn) on both my linux boxes I get "Segmentation fault"!
> Or better, I get it 9 times out of ten (due to random
> initialization without fixed seed, I guess).
>
> My configuration:
> box1: Ubuntu Gutsy, amd64 (Intel C2D)
> box2: Ubuntu Hardy, amd64 (Intel QuadCore)
>
> Python, Numpy, Scipy, Matplotlib etc. are standard versions
> from Ubuntu repositories.
>
> Can anyone confirm/reproduce this behavior?
>
> I've a report that in i386 debian it seems to work flawlessly.
>
>   

Here is the a full log on box1:
----
/usr/lib/python2.5/site-packages/scikits/openopt/examples# python nlp_1.py
OpenOpt checks user-supplied gradient df (shape: (150,) )
according to prob.diffInt = [  1.00000000e-07]
lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
will be shown
max(abs(df_user - df_numerical)) = 1.72111294887e-05
(is registered in df number 67)
========================
OpenOpt checks user-supplied gradient dc (shape: (2, 150) )
according to prob.diffInt = [  1.00000000e-07]
lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
will be shown
max(abs(dc_user - dc_numerical)) = 8.42460444801e-05
(is registered in dc number 0)
========================
OpenOpt checks user-supplied gradient dh (shape: (2, 150) )
according to prob.diffInt = [  1.00000000e-07]
lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
will be shown
max(abs(dh_user - dh_numerical)) = 0.00043945014113
(is registered in dh number 149)
========================
-----------------------------------------------------
solver: ralg   problem: unnamed   goal: minimum
 iter    objFunVal    log10(maxResidual)  
    0  8.596e+03               3.91
Segmentation fault (core dumped)
----

The core dump is available here:
http://sra.fbk.eu/people/olivetti/tmp/core.bz2


HTH,

Emanuele



From nwagner at iam.uni-stuttgart.de  Mon Jul 14 13:21:45 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Mon, 14 Jul 2008 19:21:45 +0200
Subject: [SciPy-user] [OpenOpt] Segmentation fault
In-Reply-To: <487B87A7.8030400@relativita.com>
References: <487B81DD.5000608@relativita.com> <487B87A7.8030400@relativita.com>
Message-ID: <web-104660948@uni-stuttgart.de>

On Mon, 14 Jul 2008 19:06:47 +0200
  Emanuele Olivetti <emanuele at relativita.com> wrote:
> Emanuele Olivetti wrote:
>> Dear all,
>>
>> Whenever I run example 'nlp_1.py' of OpenOpt (both 
>>tarball 0.18
>> and svn) on both my linux boxes I get "Segmentation 
>>fault"!
>> Or better, I get it 9 times out of ten (due to random
>> initialization without fixed seed, I guess).
>>
>> My configuration:
>> box1: Ubuntu Gutsy, amd64 (Intel C2D)
>> box2: Ubuntu Hardy, amd64 (Intel QuadCore)
>>
>> Python, Numpy, Scipy, Matplotlib etc. are standard 
>>versions
>> from Ubuntu repositories.
>>
>> Can anyone confirm/reproduce this behavior?
>>
>> I've a report that in i386 debian it seems to work 
>>flawlessly.
>>
>>   
> 
> Here is the a full log on box1:
> ----
> /usr/lib/python2.5/site-packages/scikits/openopt/examples# 
>python nlp_1.py
> OpenOpt checks user-supplied gradient df (shape: (150,) 
>)
> according to prob.diffInt = [  1.00000000e-07]
> lines with 1 - info_user/info_numerical greater than 
>maxViolation = 0.01
> will be shown
> max(abs(df_user - df_numerical)) = 1.72111294887e-05
> (is registered in df number 67)
> ========================
> OpenOpt checks user-supplied gradient dc (shape: (2, 
>150) )
> according to prob.diffInt = [  1.00000000e-07]
> lines with 1 - info_user/info_numerical greater than 
>maxViolation = 0.01
> will be shown
> max(abs(dc_user - dc_numerical)) = 8.42460444801e-05
> (is registered in dc number 0)
> ========================
> OpenOpt checks user-supplied gradient dh (shape: (2, 
>150) )
> according to prob.diffInt = [  1.00000000e-07]
> lines with 1 - info_user/info_numerical greater than 
>maxViolation = 0.01
> will be shown
> max(abs(dh_user - dh_numerical)) = 0.00043945014113
> (is registered in dh number 149)
> ========================
> -----------------------------------------------------
> solver: ralg   problem: unnamed   goal: minimum
> iter    objFunVal    log10(maxResidual)  
>    0  8.596e+03               3.91
> Segmentation fault (core dumped)
> ----
> 
> The core dump is available here:
> http://sra.fbk.eu/people/olivetti/tmp/core.bz2
> 
> 
> HTH,
> 
> Emanuele
> 
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user

Hi Emanuele,

I am using  r1129 of openopt and r5410 (numpy) on a 64 bit 
linux bos (open SuSe 10.2).
nlp_1.py works for me but nlp_3.py segfaults within the 
algencan part. IIRC this is a known issue.
solver: algencan   problem: nlp3   goal: maximum
  iter    objFunVal    log10(maxResidual)
     0  -1.640e+02               0.81

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 47292235252112 (LWP 5534)]
0x00002b0314b4af03 in free () from /lib64/libc.so.6
(gdb) bt
#0  0x00002b0314b4af03 in free () from /lib64/libc.so.6
#1  0x00002b032a717bd9 in memFree () from 
/home/nwagner/TANGO/ALGENCAN/pywrapper.so
#2  0x00002b032a719b1b in pywrapper_solver () from 
/home/nwagner/TANGO/ALGENCAN/pywrapper.so

Nils



From silva at lma.cnrs-mrs.fr  Mon Jul 14 14:12:21 2008
From: silva at lma.cnrs-mrs.fr (Fabrice Silva)
Date: Mon, 14 Jul 2008 20:12:21 +0200
Subject: [SciPy-user] odepack or f2py error
In-Reply-To: <g5a7mc$dk8$1@ger.gmane.org>
References: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>
	<g55o0o$b0i$1@ger.gmane.org>
	<1215862349.3033.3.camel@Portable-s2m.cnrs-mrs.fr>
	<g5a7mc$dk8$1@ger.gmane.org>
Message-ID: <1216059141.3007.5.camel@Portable-s2m.cnrs-mrs.fr>

Le samedi 12 juillet 2008 ? 12:22 +0000, Pauli Virtanen a ?crit :
> Sat, 12 Jul 2008 13:32:29 +0200, Fabrice Silva wrote:
> > How can I check gfortran is used by f2py ?
> Look at the output when scipy is compiled, or try to ldd scipy/integrate/
> vode.so and check if it's linked with libgfortran.

I am using scipy from debian packages, so I did not compile it. The ldd
command on either vode.so or my share object from f2py gives the
following ouput :
        ldd /usr/lib/python2.5/site-packages/scipy/integrate/vode.so
        	linux-gate.so.1 =>  (0xb7f25000)
        	libblas.so.3gf => /usr/lib/atlas/libblas.so.3gf (0xb7b83000)
        	libgfortran.so.3 => /usr/lib/libgfortran.so.3 (0xb7ad1000)
        	libm.so.6 => /lib/i686/cmov/libm.so.6 (0xb7aab000)
        	libgcc_s.so.1 => /lib/libgcc_s.so.1 (0xb7a9e000)
        	libc.so.6 => /lib/i686/cmov/libc.so.6 (0xb7943000)
        	/lib/ld-linux.so.2 (0xb7f26000)

> 	http://scipy.org/scipy/scipy/ticket/696

You may see that on this ticket, I've provided an example of small
computation where the solver reach the mxstep number of iteration, but
the error message is nicely displayed. The small example is also written
in fortran and compiled with f2py in a identical manner as I have done
with my study case...

-- 
Fabrice Silva <silva at lma.cnrs-mrs.fr>
LMA UPR CNRS 7051 - ?quipe S2M



From dmitrey.kroshko at scipy.org  Mon Jul 14 16:37:34 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Mon, 14 Jul 2008 23:37:34 +0300
Subject: [SciPy-user] [OpenOpt] Segmentation fault
In-Reply-To: <487B87A7.8030400@relativita.com>
References: <487B81DD.5000608@relativita.com> <487B87A7.8030400@relativita.com>
Message-ID: <487BB90E.9090200@scipy.org>

hi Emanuele,
I have no idea and I have no experience of working with core dump files.
I can only guess that it's related to numpy (since OO have no other than 
python code). IIRC there ware some troubles with numpy version from 
channels (IIRC release 1.0.5) (however, other troubles with other 
output), AFAIK both Nils and I have numpy from latest svn.

Regards, D.

Emanuele Olivetti wrote:
> Emanuele Olivetti wrote:
>   
>> Dear all,
>>
>> Whenever I run example 'nlp_1.py' of OpenOpt (both tarball 0.18
>> and svn) on both my linux boxes I get "Segmentation fault"!
>> Or better, I get it 9 times out of ten (due to random
>> initialization without fixed seed, I guess).
>>
>> My configuration:
>> box1: Ubuntu Gutsy, amd64 (Intel C2D)
>> box2: Ubuntu Hardy, amd64 (Intel QuadCore)
>>
>> Python, Numpy, Scipy, Matplotlib etc. are standard versions
>> from Ubuntu repositories.
>>
>> Can anyone confirm/reproduce this behavior?
>>
>> I've a report that in i386 debian it seems to work flawlessly.
>>
>>   
>>     
>
> Here is the a full log on box1:
> ----
> /usr/lib/python2.5/site-packages/scikits/openopt/examples# python nlp_1.py
> OpenOpt checks user-supplied gradient df (shape: (150,) )
> according to prob.diffInt = [  1.00000000e-07]
> lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
> will be shown
> max(abs(df_user - df_numerical)) = 1.72111294887e-05
> (is registered in df number 67)
> ========================
> OpenOpt checks user-supplied gradient dc (shape: (2, 150) )
> according to prob.diffInt = [  1.00000000e-07]
> lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
> will be shown
> max(abs(dc_user - dc_numerical)) = 8.42460444801e-05
> (is registered in dc number 0)
> ========================
> OpenOpt checks user-supplied gradient dh (shape: (2, 150) )
> according to prob.diffInt = [  1.00000000e-07]
> lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
> will be shown
> max(abs(dh_user - dh_numerical)) = 0.00043945014113
> (is registered in dh number 149)
> ========================
> -----------------------------------------------------
> solver: ralg   problem: unnamed   goal: minimum
>  iter    objFunVal    log10(maxResidual)  
>     0  8.596e+03               3.91
> Segmentation fault (core dumped)
> ----
>
> The core dump is available here:
> http://sra.fbk.eu/people/olivetti/tmp/core.bz2
>
>
> HTH,
>
> Emanuele
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
>
>   



From dmitrey.kroshko at scipy.org  Mon Jul 14 16:39:12 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Mon, 14 Jul 2008 23:39:12 +0300
Subject: [SciPy-user] [OpenOpt] Segmentation fault
In-Reply-To: <web-104660948@uni-stuttgart.de>
References: <487B81DD.5000608@relativita.com> <487B87A7.8030400@relativita.com>
	<web-104660948@uni-stuttgart.de>
Message-ID: <487BB970.9070209@scipy.org>

Hi Nils,
what algencan version do you have?
I have published a post in my blog today that is related to algencan 
2.0.3 and it works well for me.

Regards, D.

Nils Wagner wrote:
> Hi Emanuele,
>
> I am using  r1129 of openopt and r5410 (numpy) on a 64 bit 
> linux bos (open SuSe 10.2).
> nlp_1.py works for me but nlp_3.py segfaults within the 
> algencan part. IIRC this is a known issue.
> solver: algencan   problem: nlp3   goal: maximum
>   iter    objFunVal    log10(maxResidual)
>      0  -1.640e+02               0.81
>
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 47292235252112 (LWP 5534)]
> 0x00002b0314b4af03 in free () from /lib64/libc.so.6
> (gdb) bt
> #0  0x00002b0314b4af03 in free () from /lib64/libc.so.6
> #1  0x00002b032a717bd9 in memFree () from 
> /home/nwagner/TANGO/ALGENCAN/pywrapper.so
> #2  0x00002b032a719b1b in pywrapper_solver () from 
> /home/nwagner/TANGO/ALGENCAN/pywrapper.so
>
> Nils
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
>
>   



From forrest at physics.Auburn.EDU  Mon Jul 14 17:16:57 2008
From: forrest at physics.Auburn.EDU (Phil Forrest)
Date: Mon, 14 Jul 2008 16:16:57 -0500
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
Message-ID: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>

Hello,

I'm a non-Python using system administrator attempting to install scipy on
Sun Solaris 10 AMD64 architecture (IBM Blade Server).

Sun Studio 12 is installed in default, vanilla mode (/opt/SUNWspro)

I was able to install numpy with no apparent (no yet tested) problems.

First things: In the file "INSTALL.txt" there is a line stating the
following:

"See http://new.scipy.org/Wiki/Installing_SciPy for updates of this
document."

When I go to the above URL, I get a misconfigured web page (500: Internal
Server Error).

The crux of my problem is that the scipy build is complaining that it cannot
find the F77 Compatibility library:

/opt/SUNWspro/bin/f90 -Bdynamic -G -Bdynamic -G
build/temp.solaris-2.10-i86pc-2.4/build/src.solaris-2.10-i86pc-2.4/scipy/fft
pack/_fftpackmodule.o
build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zfft.o
build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/drfft.o
build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zrfft.o
build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zfftnd.o
build/temp.solaris-2.10-i86pc-2.4/build/src.solaris-2.10-i86pc-2.4/fortranob
ject.o -Lbuild/temp.solaris-2.10-i86pc-2.4 -ldfftpack -lfsu -lsunmath -lmvec
-lf77compat -o build/lib.solaris-2.10-i86pc-2.4/scipy/fftpack/_fftpack.so
ld: fatal: library -lf77compat: not found
ld: fatal: File processing errors. No output written to
build/lib.solaris-2.10-i86pc-2.4/scipy/fftpack/_fftpack.so
ld: fatal: library -lf77compat: not found
ld: fatal: File processing errors. No output written to
build/lib.solaris-2.10-i86pc-2.4/scipy/fftpack/_fftpack.so
error: Command "/opt/SUNWspro/bin/f90 -Bdynamic -G -Bdynamic -G
build/temp.solaris-2.10-i86pc-2.4/build/src.solaris-2.10-i86pc-2.4/scipy/fft
pack/_fftpackmodule.o
build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zfft.o
build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/drfft.o
build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zrfft.o
build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zfftnd.o
build/temp.solaris-2.10-i86pc-2.4/build/src.solaris-2.10-i86pc-2.4/fortranob
ject.o -Lbuild/temp.solaris-2.10-i86pc-2.4 -ldfftpack -lfsu -lsunmath -lmvec
-lf77compat -o build/lib.solaris-2.10-i86pc-2.4/scipy/fftpack/_fftpack.so"
failed with exit status 1
#

Here are some environment variables I have defined:
# env | grep BLAS
BLAS=/opt/SUNWspro/lib/libsunperf.so
# env | grep LAPACK
LAPACK=/opt/SUNWspro/lib/libsunmath.so

Here is my crle:

# crle  

Configuration file [version 4]: /var/ld/ld.config  
  Default Library Path (ELF):
/lib:/usr/lib:/opt/SUNWspro/lib:/usr/local/lib:/opt/sfw/lib
  Trusted Directories (ELF):    /lib/secure:/usr/lib/secure  (system
default)

Command line:
  crle -c /var/ld/ld.config -l
/lib:/usr/lib:/opt/SUNWspro/lib:/usr/local/lib:/opt/sfw/lib

# crle -64

Configuration file [version 4]: /var/ld/64/ld.config  
  Default Library Path (ELF):   /lib/64:/usr/lib/64:/opt/SUNWspro/lib/amd64
  Trusted Directories (ELF):    /lib/secure/64:/usr/lib/secure/64  (system
default)

Command line:
  crle -64 -c /var/ld/64/ld.config -l
/lib/64:/usr/lib/64:/opt/SUNWspro/lib/amd64


Can anyone shed some light on how to resolve the "ld: fatal: library
-lf77compat: not found" error? Is the scipy build attempting calls that
force F77 mode/calls?

I would appreciate any documentation URL that might help clear this up.

Thanks,
Phil




From michael.abshoff at googlemail.com  Mon Jul 14 17:37:36 2008
From: michael.abshoff at googlemail.com (Michael Abshoff)
Date: Mon, 14 Jul 2008 14:37:36 -0700
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
In-Reply-To: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>
Message-ID: <487BC720.8060904@gmail.com>

Phil Forrest wrote:
> Hello,

Hi Phil,

> I'm a non-Python using system administrator attempting to install scipy on
> Sun Solaris 10 AMD64 architecture (IBM Blade Server).
> 
> Sun Studio 12 is installed in default, vanilla mode (/opt/SUNWspro)
> 
> I was able to install numpy with no apparent (no yet tested) problems.
> 
> First things: In the file "INSTALL.txt" there is a line stating the
> following:
> 
> "See http://new.scipy.org/Wiki/Installing_SciPy for updates of this
> document."
> 
> When I go to the above URL, I get a misconfigured web page (500: Internal
> Server Error).
> 
> The crux of my problem is that the scipy build is complaining that it cannot
> find the F77 Compatibility library:

This is comping from numpy:

-bash-3.00$ /usr/sfw/bin/ggrep -r f77compat *
site-packages/numpy/distutils/fcompiler/sun.py: 
opt.extend(['fsu','sunmath','mvec','f77compat'])

You might want to throw it out there. Depending on which symbols are 
missing at link or module import time you might need to add some libs 
there. I guess by the time I have build Python, numpy and scipy on that 
box using the Sun compiler you will have fixed the problem :) If you 
have problems I can poke around in case it is not fixed in numpy 1.1.x.

<SNIP>


> Here are some environment variables I have defined:
> # env | grep BLAS
> BLAS=/opt/SUNWspro/lib/libsunperf.so
> # env | grep LAPACK
> LAPACK=/opt/SUNWspro/lib/libsunmath.so
> 
> Here is my crle:
> 
> # crle  
> 
> Configuration file [version 4]: /var/ld/ld.config  
>   Default Library Path (ELF):
> /lib:/usr/lib:/opt/SUNWspro/lib:/usr/local/lib:/opt/sfw/lib
>   Trusted Directories (ELF):    /lib/secure:/usr/lib/secure  (system
> default)
> 
> Command line:
>   crle -c /var/ld/ld.config -l
> /lib:/usr/lib:/opt/SUNWspro/lib:/usr/local/lib:/opt/sfw/lib
> 
> # crle -64
> 
> Configuration file [version 4]: /var/ld/64/ld.config  
>   Default Library Path (ELF):   /lib/64:/usr/lib/64:/opt/SUNWspro/lib/amd64
>   Trusted Directories (ELF):    /lib/secure/64:/usr/lib/secure/64  (system
> default)
> 
> Command line:
>   crle -64 -c /var/ld/64/ld.config -l
> /lib/64:/usr/lib/64:/opt/SUNWspro/lib/amd64
> 
> 
> Can anyone shed some light on how to resolve the "ld: fatal: library
> -lf77compat: not found" error? Is the scipy build attempting calls that
> force F77 mode/calls?

Sun Studio 12 does not ship a liblf77compat.[so|a]. I would guess this 
is a bug in numpy and it should probably be fixed there.

> I would appreciate any documentation URL that might help clear this up.
> 
> Thanks,
> Phil
> 

Cheers,

Michael
> 
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
> 



From robert.kern at gmail.com  Mon Jul 14 17:41:47 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Mon, 14 Jul 2008 16:41:47 -0500
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
In-Reply-To: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>
Message-ID: <3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>

On Mon, Jul 14, 2008 at 16:16, Phil Forrest <forrest at physics.auburn.edu> wrote:
> Hello,
>
> I'm a non-Python using system administrator attempting to install scipy on
> Sun Solaris 10 AMD64 architecture (IBM Blade Server).
>
> Sun Studio 12 is installed in default, vanilla mode (/opt/SUNWspro)
>
> I was able to install numpy with no apparent (no yet tested) problems.
>
> First things: In the file "INSTALL.txt" there is a line stating the
> following:
>
> "See http://new.scipy.org/Wiki/Installing_SciPy for updates of this
> document."
>
> When I go to the above URL, I get a misconfigured web page (500: Internal
> Server Error).

Sorry about that. The new URL is http://www.scipy.org/Installing_SciPy
. The document has been fixed.

> The crux of my problem is that the scipy build is complaining that it cannot
> find the F77 Compatibility library:
>
> /opt/SUNWspro/bin/f90 -Bdynamic -G -Bdynamic -G
> build/temp.solaris-2.10-i86pc-2.4/build/src.solaris-2.10-i86pc-2.4/scipy/fft
> pack/_fftpackmodule.o
> build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zfft.o
> build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/drfft.o
> build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zrfft.o
> build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zfftnd.o
> build/temp.solaris-2.10-i86pc-2.4/build/src.solaris-2.10-i86pc-2.4/fortranob
> ject.o -Lbuild/temp.solaris-2.10-i86pc-2.4 -ldfftpack -lfsu -lsunmath -lmvec
> -lf77compat -o build/lib.solaris-2.10-i86pc-2.4/scipy/fftpack/_fftpack.so
> ld: fatal: library -lf77compat: not found
> ld: fatal: File processing errors. No output written to
> build/lib.solaris-2.10-i86pc-2.4/scipy/fftpack/_fftpack.so
> ld: fatal: library -lf77compat: not found
> ld: fatal: File processing errors. No output written to
> build/lib.solaris-2.10-i86pc-2.4/scipy/fftpack/_fftpack.so
> error: Command "/opt/SUNWspro/bin/f90 -Bdynamic -G -Bdynamic -G
> build/temp.solaris-2.10-i86pc-2.4/build/src.solaris-2.10-i86pc-2.4/scipy/fft
> pack/_fftpackmodule.o
> build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zfft.o
> build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/drfft.o
> build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zrfft.o
> build/temp.solaris-2.10-i86pc-2.4/scipy/fftpack/src/zfftnd.o
> build/temp.solaris-2.10-i86pc-2.4/build/src.solaris-2.10-i86pc-2.4/fortranob
> ject.o -Lbuild/temp.solaris-2.10-i86pc-2.4 -ldfftpack -lfsu -lsunmath -lmvec
> -lf77compat -o build/lib.solaris-2.10-i86pc-2.4/scipy/fftpack/_fftpack.so"
> failed with exit status 1
> #
>
> Here are some environment variables I have defined:
> # env | grep BLAS
> BLAS=/opt/SUNWspro/lib/libsunperf.so
> # env | grep LAPACK
> LAPACK=/opt/SUNWspro/lib/libsunmath.so
>
> Here is my crle:
>
> # crle
>
> Configuration file [version 4]: /var/ld/ld.config
>  Default Library Path (ELF):
> /lib:/usr/lib:/opt/SUNWspro/lib:/usr/local/lib:/opt/sfw/lib
>  Trusted Directories (ELF):    /lib/secure:/usr/lib/secure  (system
> default)
>
> Command line:
>  crle -c /var/ld/ld.config -l
> /lib:/usr/lib:/opt/SUNWspro/lib:/usr/local/lib:/opt/sfw/lib
>
> # crle -64
>
> Configuration file [version 4]: /var/ld/64/ld.config
>  Default Library Path (ELF):   /lib/64:/usr/lib/64:/opt/SUNWspro/lib/amd64
>  Trusted Directories (ELF):    /lib/secure/64:/usr/lib/secure/64  (system
> default)
>
> Command line:
>  crle -64 -c /var/ld/64/ld.config -l
> /lib/64:/usr/lib/64:/opt/SUNWspro/lib/amd64
>
>
> Can anyone shed some light on how to resolve the "ld: fatal: library
> -lf77compat: not found" error? Is the scipy build attempting calls that
> force F77 mode/calls?

All of our FORTRAN code is FORTRAN-77.

> I would appreciate any documentation URL that might help clear this up.

I'm afraid that we don't have much in the way of Sun-specific
information, nor am I much familiar with the lay of the land, here. Is
the FORTRAN-77 support an optional install? Or is your version of the
compiler supposed to work without any extra libraries like f77compat
any more? If this is the case and you can figure out which version no
longer requires f77compat, we can put the appropriate check into
numpy. Please provide the output of "f90 -V".

Thanks.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From michael.abshoff at googlemail.com  Mon Jul 14 18:08:56 2008
From: michael.abshoff at googlemail.com (Michael Abshoff)
Date: Mon, 14 Jul 2008 15:08:56 -0700
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
In-Reply-To: <3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>
	<3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>
Message-ID: <487BCE78.3040403@gmail.com>

Robert Kern wrote:
> On Mon, Jul 14, 2008 at 16:16, Phil Forrest <forrest at physics.auburn.edu> wrote:

<SNIP>

>> Can anyone shed some light on how to resolve the "ld: fatal: library
>> -lf77compat: not found" error? Is the scipy build attempting calls that
>> force F77 mode/calls?
> 
> All of our FORTRAN code is FORTRAN-77.
> 
>> I would appreciate any documentation URL that might help clear this up.
> 
> I'm afraid that we don't have much in the way of Sun-specific
> information, nor am I much familiar with the lay of the land, here. Is
> the FORTRAN-77 support an optional install? Or is your version of the
> compiler supposed to work without any extra libraries like f77compat
> any more? If this is the case and you can figure out which version no
> longer requires f77compat, we can put the appropriate check into
> numpy. 

I am not sure when it changed. When I look at

http://projects.scipy.org/scipy/numpy/ticket/671

it seemt that up to and including Release 6 of the Sun compiler it was 
still valid.

> Please provide the output of "f90 -V".

This is SunStudio12-200709:

-bash-3.00$ ./f77  -V
NOTICE: Invoking ./f90 -f77 -ftrap=%none -V
f90: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
Usage: f90 [ options ] files.  Use 'f90 -flags' for details
-bash-3.00$ ./f90  -V
f90: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
Usage: f90 [ options ] files.  Use 'f90 -flags' for details
-bash-3.00$ ./f95  -V
f95: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
Usage: f95 [ options ] files.  Use 'f95 -flags' for details

I currently do not have access to older Sun Studio releases personally, 
but I could ask some people if they could check. This is unlikely to 
work, but I will try.

> Thanks.
> 


Cheers,

Michael


From robert.kern at gmail.com  Mon Jul 14 18:42:49 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Mon, 14 Jul 2008 17:42:49 -0500
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
In-Reply-To: <487BCE78.3040403@gmail.com>
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>
	<3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>
	<487BCE78.3040403@gmail.com>
Message-ID: <3d375d730807141542i1086c63fx9a43e0bad275fb7c@mail.gmail.com>

On Mon, Jul 14, 2008 at 17:08, Michael Abshoff
<michael.abshoff at googlemail.com> wrote:
> Robert Kern wrote:
>> On Mon, Jul 14, 2008 at 16:16, Phil Forrest <forrest at physics.auburn.edu> wrote:
>
> <SNIP>
>
>>> Can anyone shed some light on how to resolve the "ld: fatal: library
>>> -lf77compat: not found" error? Is the scipy build attempting calls that
>>> force F77 mode/calls?
>>
>> All of our FORTRAN code is FORTRAN-77.
>>
>>> I would appreciate any documentation URL that might help clear this up.
>>
>> I'm afraid that we don't have much in the way of Sun-specific
>> information, nor am I much familiar with the lay of the land, here. Is
>> the FORTRAN-77 support an optional install? Or is your version of the
>> compiler supposed to work without any extra libraries like f77compat
>> any more? If this is the case and you can figure out which version no
>> longer requires f77compat, we can put the appropriate check into
>> numpy.
>
> I am not sure when it changed. When I look at
>
> http://projects.scipy.org/scipy/numpy/ticket/671
>
> it seemt that up to and including Release 6 of the Sun compiler it was
> still valid.
>
>> Please provide the output of "f90 -V".
>
> This is SunStudio12-200709:
>
> -bash-3.00$ ./f77  -V
> NOTICE: Invoking ./f90 -f77 -ftrap=%none -V
> f90: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
> -bash-3.00$ ./f90  -V
> f90: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
> -bash-3.00$ ./f95  -V
> f95: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
> Usage: f95 [ options ] files.  Use 'f95 -flags' for details
>
> I currently do not have access to older Sun Studio releases personally,
> but I could ask some people if they could check. This is unlikely to
> work, but I will try.

Manuals sometimes give a little history. If it got dropped in version
8.0, it's probably documented in there, at least as a bit of
advertising. Are the manuals publicly accessible online anywhere? I'm
happy to do some searching myself.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From michael.abshoff at googlemail.com  Mon Jul 14 19:09:33 2008
From: michael.abshoff at googlemail.com (Michael Abshoff)
Date: Mon, 14 Jul 2008 16:09:33 -0700
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
In-Reply-To: <3d375d730807141542i1086c63fx9a43e0bad275fb7c@mail.gmail.com>
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>	<3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>	<487BCE78.3040403@gmail.com>
	<3d375d730807141542i1086c63fx9a43e0bad275fb7c@mail.gmail.com>
Message-ID: <487BDCAD.4090907@gmail.com>

Robert Kern wrote:

<SNIP>

>>> Please provide the output of "f90 -V".
>> This is SunStudio12-200709:
>>
>> -bash-3.00$ ./f77  -V
>> NOTICE: Invoking ./f90 -f77 -ftrap=%none -V
>> f90: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
>> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
>> -bash-3.00$ ./f90  -V
>> f90: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
>> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
>> -bash-3.00$ ./f95  -V
>> f95: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
>> Usage: f95 [ options ] files.  Use 'f95 -flags' for details
>>
>> I currently do not have access to older Sun Studio releases personally,
>> but I could ask some people if they could check. This is unlikely to
>> work, but I will try.
> 
> Manuals sometimes give a little history. If it got dropped in version
> 8.0, it's probably documented in there, at least as a bit of
> advertising. Are the manuals publicly accessible online anywhere? I'm
> happy to do some searching myself.
> 

Hi Robert,

caught a lucky break and found a box with Sun Studio 11: It ships a 
libf77compat.so.

To identify:

mabshoff at cerberus: /usr/local/SUNWspro/bin>./f77 -V
NOTICE: Invoking ./f90 -f77 -ftrap=%none -V
f90: Sun Fortran 95 8.2 Patch 121019-04 2006/10/26
Usage: f90 [ options ] files.  Use 'f90 -flags' for details
mabshoff at cerberus: /usr/local/SUNWspro/bin>./f90 -V
f90: Sun Fortran 95 8.2 Patch 121019-04 2006/10/26
Usage: f90 [ options ] files.  Use 'f90 -flags' for details
mabshoff at cerberus: /usr/local/SUNWspro/bin>./f95 -V
f95: Sun Fortran 95 8.2 Patch 121019-04 2006/10/26
Usage: f95 [ options ] files.  Use 'f95 -flags' for details

Hope this helps.

It seems that Phil is not poking around. I have some odd numpy 1.0.4cvs 
failure with gcc in 32 bit mode on Solaris 10 Intel when computing roots 
of polynomials. Since Sage will finally upgrade to numpy 1.1 soon I will 
try to see if I can reproduce the problem there and for fun I can also 
build numpy 1.1 + scipy in 64 bit mode with Sun Forte. So if there are 
any patches let me know and I can try them out. Otherwise I will likely 
do some crude patching to get the flags right and let you guys work out 
the fines details :)

Cheers,

Michael


From robert.kern at gmail.com  Mon Jul 14 19:35:24 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Mon, 14 Jul 2008 18:35:24 -0500
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
In-Reply-To: <487BDCAD.4090907@gmail.com>
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>
	<3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>
	<487BCE78.3040403@gmail.com>
	<3d375d730807141542i1086c63fx9a43e0bad275fb7c@mail.gmail.com>
	<487BDCAD.4090907@gmail.com>
Message-ID: <3d375d730807141635l667ecc94j4cea55c0c4251909@mail.gmail.com>

On Mon, Jul 14, 2008 at 18:09, Michael Abshoff
<michael.abshoff at googlemail.com> wrote:
> Robert Kern wrote:
>
> <SNIP>
>
>>>> Please provide the output of "f90 -V".
>>> This is SunStudio12-200709:
>>>
>>> -bash-3.00$ ./f77  -V
>>> NOTICE: Invoking ./f90 -f77 -ftrap=%none -V
>>> f90: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
>>> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
>>> -bash-3.00$ ./f90  -V
>>> f90: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
>>> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
>>> -bash-3.00$ ./f95  -V
>>> f95: Sun Fortran 95 8.3 SunOS_i386 Patch 127002-01 2007/07/18
>>> Usage: f95 [ options ] files.  Use 'f95 -flags' for details
>>>
>>> I currently do not have access to older Sun Studio releases personally,
>>> but I could ask some people if they could check. This is unlikely to
>>> work, but I will try.
>>
>> Manuals sometimes give a little history. If it got dropped in version
>> 8.0, it's probably documented in there, at least as a bit of
>> advertising. Are the manuals publicly accessible online anywhere? I'm
>> happy to do some searching myself.
>>
>
> Hi Robert,
>
> caught a lucky break and found a box with Sun Studio 11: It ships a
> libf77compat.so.
>
> To identify:
>
> mabshoff at cerberus: /usr/local/SUNWspro/bin>./f77 -V
> NOTICE: Invoking ./f90 -f77 -ftrap=%none -V
> f90: Sun Fortran 95 8.2 Patch 121019-04 2006/10/26
> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
> mabshoff at cerberus: /usr/local/SUNWspro/bin>./f90 -V
> f90: Sun Fortran 95 8.2 Patch 121019-04 2006/10/26
> Usage: f90 [ options ] files.  Use 'f90 -flags' for details
> mabshoff at cerberus: /usr/local/SUNWspro/bin>./f95 -V
> f95: Sun Fortran 95 8.2 Patch 121019-04 2006/10/26
> Usage: f95 [ options ] files.  Use 'f95 -flags' for details
>
> Hope this helps.

Yup. We can do a version check. For >= 8.3, we don't put in f77compat.
Can you and Phil check that if you modify that line of code in sun.py
to omit f77compat on your Sun Fortran 95 8.3 boxes, that it works?
(Ugh, that sentence got away from me, but too late now!)

> It seems that Phil is not poking around. I have some odd numpy 1.0.4cvs
> failure with gcc in 32 bit mode on Solaris 10 Intel when computing roots
> of polynomials. Since Sage will finally upgrade to numpy 1.1 soon I will
> try to see if I can reproduce the problem there and for fun I can also
> build numpy 1.1 + scipy in 64 bit mode with Sun Forte. So if there are
> any patches let me know and I can try them out.

I'm not sure if any are relevant, but we have a branch for the
upcoming 1.1.1 bugfix release:

  http://svn.scipy.org/svn/numpy/branches/1.1.x/

> Otherwise I will likely
> do some crude patching to get the flags right and let you guys work out
> the fines details :)

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From michael.abshoff at googlemail.com  Mon Jul 14 22:35:57 2008
From: michael.abshoff at googlemail.com (Michael Abshoff)
Date: Mon, 14 Jul 2008 19:35:57 -0700
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
In-Reply-To: <3d375d730807141635l667ecc94j4cea55c0c4251909@mail.gmail.com>
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>	<3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>	<487BCE78.3040403@gmail.com>	<3d375d730807141542i1086c63fx9a43e0bad275fb7c@mail.gmail.com>	<487BDCAD.4090907@gmail.com>
	<3d375d730807141635l667ecc94j4cea55c0c4251909@mail.gmail.com>
Message-ID: <487C0D0D.5040904@gmail.com>

Robert Kern wrote:

<SNIP>

Hi Robert,

> Yup. We can do a version check. For >= 8.3, we don't put in f77compat.
> Can you and Phil check that if you modify that line of code in sun.py
> to omit f77compat on your Sun Fortran 95 8.3 boxes, that it works?
> (Ugh, that sentence got away from me, but too late now!)

:) - I am working on it, but it will be tomorrow until I am done.

>> It seems that Phil is not poking around. I have some odd numpy 1.0.4cvs
>> failure with gcc in 32 bit mode on Solaris 10 Intel when computing roots
>> of polynomials. Since Sage will finally upgrade to numpy 1.1 soon I will
>> try to see if I can reproduce the problem there and for fun I can also
>> build numpy 1.1 + scipy in 64 bit mode with Sun Forte. So if there are
>> any patches let me know and I can try them out.
> 
> I'm not sure if any are relevant, but we have a branch for the
> upcoming 1.1.1 bugfix release:
> 
>   http://svn.scipy.org/svn/numpy/branches/1.1.x/

Yeah, I am aware of that, but I just build Python 2.5.2 and numpy 1.1.0 
and expect for the borken ctypes all tests pass on x86-64 Solaris with 
Sun Studio 12 in 64 bit mode. Unfortunately I do not have the SunPerf 
lib installed for x86-64, so I will have to do the whole thing on Sparc 
again before attempting to come up with a set of working flags. I have 
an ATLAS+Lapack build on that box, but I used gfortran, so it won't be 
any good in debugging this problem.

>> Otherwise I will likely
>> do some crude patching to get the flags right and let you guys work out
>> the fines details :)
> 

One more thing: Sin Phil wants to build a 64 bit version he ought to set

export 
BLAS=/usr/local/SunStudio12-200709/x86_64-SunOS/SUNWspro/lib/amd64/libsunmath.so

Notice .../lib/amd64/... since lib contains the 32 bit version of the 
library and lib/amd64 the 64 bit version.

And for the SunPerf lib itself: There are a bunch of them on my Sparc box:

-bash-3.00$ find /usr/local/SunStudio12-200709/ -name "lib*perf*"
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/bin/libsunperf_check
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/bin/libsunperf_check
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/lib/v9/libsuniperf.a
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/lib/libsuniperf.a
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/lib/sparcfmaf/libsuniperf.a
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/lib/v7/libsuniperf.a
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/lib/v8/libsuniperf.a
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/lib/v8a/libsuniperf.a
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/lib/v8plus/libsuniperf.a
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/lib/v8plusa/libsuniperf.a
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/lib/v8plusb/libsuniperf.a
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/lib/v9a/libsuniperf.a
/usr/local/SunStudio12-200709/sparc-SunOS/SUNWspro/prod/lib/v9b/libsuniperf.a

And the same applies here: the v9 is 64 bit while v8 and v7 are 32 bit. 
Hope this helps. If anybody cares I can send my notes on the build 
(including Python) to the list so that someone interested can stuff them 
in the wiki. There seems to be no mention of Solaris on

   http://www.scipy.org/Installing_SciPy
   http://www.scipy.org/Installing_SciPy/BuildingGeneral
   http://www.scipy.org/FAQ

I also send some build info on 64 bit OSX out a while ago and somebody 
mentioned that those might also be well placed in the wiki.

Cheers,

Michael


From robert.kern at gmail.com  Mon Jul 14 22:55:47 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Mon, 14 Jul 2008 21:55:47 -0500
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
In-Reply-To: <487C0D0D.5040904@gmail.com>
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>
	<3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>
	<487BCE78.3040403@gmail.com>
	<3d375d730807141542i1086c63fx9a43e0bad275fb7c@mail.gmail.com>
	<487BDCAD.4090907@gmail.com>
	<3d375d730807141635l667ecc94j4cea55c0c4251909@mail.gmail.com>
	<487C0D0D.5040904@gmail.com>
Message-ID: <3d375d730807141955l641a1866s24e2103c1de134d@mail.gmail.com>

On Mon, Jul 14, 2008 at 21:35, Michael Abshoff
<michael.abshoff at googlemail.com> wrote:
> Hope this helps. If anybody cares I can send my notes on the build
> (including Python) to the list so that someone interested can stuff them
> in the wiki.

Yes, please! However, I don't think anyone who hasn't built numpyor
scipy on Solaris will have the motivation (or background, probably) to
pretty them up for the wiki, so they'll probably get lost if you don't
do it yourself. But toss 'em over anyways. Actually, if you have notes
on building anything in the Sage stack on any platform, I'm personally
interested in seeing them.

> There seems to be no mention of Solaris on
>
>   http://www.scipy.org/Installing_SciPy
>   http://www.scipy.org/Installing_SciPy/BuildingGeneral
>   http://www.scipy.org/FAQ
>
> I also send some build info on 64 bit OSX out a while ago and somebody
> mentioned that those might also be well placed in the wiki.

To this list? I'll have to go back through the archives.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From michael.abshoff at googlemail.com  Mon Jul 14 23:21:53 2008
From: michael.abshoff at googlemail.com (Michael Abshoff)
Date: Mon, 14 Jul 2008 20:21:53 -0700
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
In-Reply-To: <3d375d730807141955l641a1866s24e2103c1de134d@mail.gmail.com>
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>	<3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>	<487BCE78.3040403@gmail.com>	<3d375d730807141542i1086c63fx9a43e0bad275fb7c@mail.gmail.com>	<487BDCAD.4090907@gmail.com>	<3d375d730807141635l667ecc94j4cea55c0c4251909@mail.gmail.com>	<487C0D0D.5040904@gmail.com>
	<3d375d730807141955l641a1866s24e2103c1de134d@mail.gmail.com>
Message-ID: <487C17D1.2080704@gmail.com>

Robert Kern wrote:
> On Mon, Jul 14, 2008 at 21:35, Michael Abshoff
> <michael.abshoff at googlemail.com> wrote:

Hi Robert,

>> Hope this helps. If anybody cares I can send my notes on the build
>> (including Python) to the list so that someone interested can stuff them
>> in the wiki.
> 
> Yes, please! However, I don't think anyone who hasn't built numpyor
> scipy on Solaris will have the motivation (or background, probably) to
> pretty them up for the wiki, so they'll probably get lost if you don't
> do it yourself. But toss 'em over anyways. Actually, if you have notes
> on building anything in the Sage stack on any platform, I'm personally
> interested in seeing them.

Ok, besides the 64 bit OSX there is little that is working and of 
interest to you guys. The main issue is still getting the ctypes 
extension to build at all. For various reasons what is in-tree in python 
is crap and lacking 64 bit Darwin support completely and I have not had 
time to come up with a clean patch for ffcall (or is it fficall?).

For 64 bit OSX Scipy there was one failure with one extension that was 
compiled in 32 bit mode even on 64 bit. I would assume someone is 
cleverly clearing the CFLAGS or CXXFLAGS and then shooting me in the 
foot, but maybe I myself was dumb and overlooked something.

Solaris in 32 and 64 bit mode is mostly done in regards to python, numpy 
and scipy. I promised Stefan to run a build bot on Solaris and I have 
started to put together eight builds, i.e.

  a) Sparc, Sparc 64, x86 and x86-64

each with

  b) gcc with gfortran or cc with SunPerflib

So those are eight build bots in total. But progress there has also been 
hampered by me working on other Sage issues.

What will hopefully happen next months is a 64 bit ATLAS+Netlib.org 
Lapack+numpy+scipy on Vista 64bit, but we will see how that goes. If I 
have trouble with ATLAS I will probably use the Intel MKL for now to get 
up and running on the 64 bit native Sage on Windows port.

>> There seems to be no mention of Solaris on
>>
>>   http://www.scipy.org/Installing_SciPy
>>   http://www.scipy.org/Installing_SciPy/BuildingGeneral
>>   http://www.scipy.org/FAQ
>>
>> I also send some build info on 64 bit OSX out a while ago and somebody
>> mentioned that those might also be well placed in the wiki.
> 
> To this list? I'll have to go back through the archives.
> 

It was sent to the numpy list IIRC. If they don't turn up in the 
archives I can send you a copy off list.

Cheers,

Michael


From jh at physics.ucf.edu  Tue Jul 15 00:16:18 2008
From: jh at physics.ucf.edu (Joe Harrington)
Date: Tue, 15 Jul 2008 00:16:18 -0400
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <mailman.13.1215709207.15246.scipy-user@scipy.org>
	(scipy-user-request@scipy.org)
References: <mailman.13.1215709207.15246.scipy-user@scipy.org>
Message-ID: <wl8tzerhje5.fsf@glup.physics.ucf.edu>

Automatic conversion of at least the boring stuff is crucial.  No
matter how educational rewriting your code may be, if you have 100,000
lines of tools and utilities to convert, you generally can't.  Our
IDL-based group has made good use of i2py, which may at least provide
some educational examples for converting MATLAB, if not a usable
framework.  The general thing is to covert what is easy, and at least
flag what is difficult.  It doesn't do a mapping of function names to
function names or anything like that.  Here's the URL, hope it helps.

http://software.pseudogreen.org/i2py/

If you do a general parser approach, it would be useful to do it in
such a way that the front end can be ripped out and modified or
replaced with one for another language.

I've added a section on the scipy.org wiki under Topical Software for
conversion programs.  When yours is ready, please put in a link there.

--jh--


From nwagner at iam.uni-stuttgart.de  Tue Jul 15 02:17:21 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Tue, 15 Jul 2008 08:17:21 +0200
Subject: [SciPy-user] [OpenOpt] Segmentation fault
In-Reply-To: <487BB970.9070209@scipy.org>
References: <487B81DD.5000608@relativita.com> <487B87A7.8030400@relativita.com>
	<web-104660948@uni-stuttgart.de> <487BB970.9070209@scipy.org>
Message-ID: <web-104680240@uni-stuttgart.de>

On Mon, 14 Jul 2008 23:39:12 +0300
  dmitrey <dmitrey.kroshko at scipy.org> wrote:
> Hi Nils,
> what algencan version do you have?
> I have published a post in my blog today that is related 
>to algencan 
> 2.0.3 and it works well for me.
> 
> Regards, D.
> 
Hi Dmitrey,

I am using the old algencan < 2.0.

I will upgrade asap.

Cheers,
           Nils
  


From dmitrey.kroshko at scipy.org  Tue Jul 15 02:44:42 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Tue, 15 Jul 2008 09:44:42 +0300
Subject: [SciPy-user] [OpenOpt] Segmentation fault
In-Reply-To: <web-104680240@uni-stuttgart.de>
References: <487B81DD.5000608@relativita.com>
	<487B87A7.8030400@relativita.com>	<web-104660948@uni-stuttgart.de>
	<487BB970.9070209@scipy.org> <web-104680240@uni-stuttgart.de>
Message-ID: <487C475A.2050701@scipy.org>

NB you should call lower-case algencan for v >= 2.0 and upper-case 
ALGENCAN for v 1.0. Ensure you have single pywrapper.so file in 
PYTHONPATH (i.e. remove old pywrapper.so).

It's easy to do via
 >>> import pywrapper
 >>> pywrapper.__file__
'/usr/lib/python2.5/site-packages/pywrapper.so'
$ [sudo ]rm -rf ...

On the other hand, for my tests ALGENCAN 1.0 works better. Of course, it 
maintenance sooner or later will be ceased.
Regards, D.


Nils Wagner wrote:
> On Mon, 14 Jul 2008 23:39:12 +0300
>   dmitrey <dmitrey.kroshko at scipy.org> wrote:
>   
>> Hi Nils,
>> what algencan version do you have?
>> I have published a post in my blog today that is related 
>> to algencan 
>> 2.0.3 and it works well for me.
>>
>> Regards, D.
>>
>>     
> Hi Dmitrey,
>
> I am using the old algencan < 2.0.
>
> I will upgrade asap.
>
> Cheers,
>            Nils
>   
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
>
>   



From hoytak at gmail.com  Tue Jul 15 03:24:06 2008
From: hoytak at gmail.com (Hoyt Koepke)
Date: Tue, 15 Jul 2008 10:24:06 +0300
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <wl8tzerhje5.fsf@glup.physics.ucf.edu>
References: <mailman.13.1215709207.15246.scipy-user@scipy.org>
	<wl8tzerhje5.fsf@glup.physics.ucf.edu>
Message-ID: <4db580fd0807150024v79726315j7570b0785783381d@mail.gmail.com>

I know I'm a little late to the main discussion, but FYI there is a
project, pym, which claims to do some conversion:
http://code.google.com/p/pym/  It's only available from svn, and
there's no claim about how well it works, but it might be a good
starting point.  I believe it's MIT licensed, so no issue there.  Plus
the authors might be good contacts.

--Hoyt




On Tue, Jul 15, 2008 at 7:16 AM, Joe Harrington <jh at physics.ucf.edu> wrote:
> Automatic conversion of at least the boring stuff is crucial.  No
> matter how educational rewriting your code may be, if you have 100,000
> lines of tools and utilities to convert, you generally can't.  Our
> IDL-based group has made good use of i2py, which may at least provide
> some educational examples for converting MATLAB, if not a usable
> framework.  The general thing is to covert what is easy, and at least
> flag what is difficult.  It doesn't do a mapping of function names to
> function names or anything like that.  Here's the URL, hope it helps.
>
> http://software.pseudogreen.org/i2py/
>
> If you do a general parser approach, it would be useful to do it in
> such a way that the front end can be ripped out and modified or
> replaced with one for another language.
>
> I've added a section on the scipy.org wiki under Topical Software for
> conversion programs.  When yours is ready, please put in a link there.
>
> --jh--
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
+++++++++++++++++++++++++++++++++++
Hoyt Koepke
UBC Department of Computer Science
http://www.cs.ubc.ca/~hoytak/
hoytak at gmail.com
+++++++++++++++++++++++++++++++++++


From emanuele at relativita.com  Tue Jul 15 05:50:35 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Tue, 15 Jul 2008 11:50:35 +0200
Subject: [SciPy-user] [OpenOpt] Segmentation fault
In-Reply-To: <487BB90E.9090200@scipy.org>
References: <487B81DD.5000608@relativita.com>
	<487B87A7.8030400@relativita.com> <487BB90E.9090200@scipy.org>
Message-ID: <487C72EB.9040709@relativita.com>

Updated to latest svn (numpy, scipy, etc.). Now nlp_1.py works.
I came across this bug because of troubles in my code. Now
let's go back and fix mine :D

Thanks,

Emanuele

dmitrey wrote:
> hi Emanuele,
> I have no idea and I have no experience of working with core dump files.
> I can only guess that it's related to numpy (since OO have no other than 
> python code). IIRC there ware some troubles with numpy version from 
> channels (IIRC release 1.0.5) (however, other troubles with other 
> output), AFAIK both Nils and I have numpy from latest svn.
>
> Regards, D.
>
> Emanuele Olivetti wrote:
>   
>> Emanuele Olivetti wrote:
>>   
>>     
>>> Dear all,
>>>
>>> Whenever I run example 'nlp_1.py' of OpenOpt (both tarball 0.18
>>> and svn) on both my linux boxes I get "Segmentation fault"!
>>> Or better, I get it 9 times out of ten (due to random
>>> initialization without fixed seed, I guess).
>>>
>>> My configuration:
>>> box1: Ubuntu Gutsy, amd64 (Intel C2D)
>>> box2: Ubuntu Hardy, amd64 (Intel QuadCore)
>>>
>>> Python, Numpy, Scipy, Matplotlib etc. are standard versions
>>> from Ubuntu repositories.
>>>
>>> Can anyone confirm/reproduce this behavior?
>>>
>>> I've a report that in i386 debian it seems to work flawlessly.
>>>
>>>   
>>>     
>>>       
>> Here is the a full log on box1:
>> ----
>> /usr/lib/python2.5/site-packages/scikits/openopt/examples# python nlp_1.py
>> OpenOpt checks user-supplied gradient df (shape: (150,) )
>> according to prob.diffInt = [  1.00000000e-07]
>> lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
>> will be shown
>> max(abs(df_user - df_numerical)) = 1.72111294887e-05
>> (is registered in df number 67)
>> ========================
>> OpenOpt checks user-supplied gradient dc (shape: (2, 150) )
>> according to prob.diffInt = [  1.00000000e-07]
>> lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
>> will be shown
>> max(abs(dc_user - dc_numerical)) = 8.42460444801e-05
>> (is registered in dc number 0)
>> ========================
>> OpenOpt checks user-supplied gradient dh (shape: (2, 150) )
>> according to prob.diffInt = [  1.00000000e-07]
>> lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
>> will be shown
>> max(abs(dh_user - dh_numerical)) = 0.00043945014113
>> (is registered in dh number 149)
>> ========================
>> -----------------------------------------------------
>> solver: ralg   problem: unnamed   goal: minimum
>>  iter    objFunVal    log10(maxResidual)  
>>     0  8.596e+03               3.91
>> Segmentation fault (core dumped)
>> ----
>>
>> The core dump is available here:
>> http://sra.fbk.eu/people/olivetti/tmp/core.bz2
>>
>>
>> HTH,
>>
>> Emanuele
>>
>> _______________________________________________
>> SciPy-user mailing list
>> SciPy-user at scipy.org
>> http://projects.scipy.org/mailman/listinfo/scipy-user
>>
>>
>>
>>   
>>     
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>   



From contact at pythonxy.com  Tue Jul 15 13:53:05 2008
From: contact at pythonxy.com (Pierre Raybaut)
Date: Tue, 15 Jul 2008 19:53:05 +0200
Subject: [SciPy-user] [python(x,y)] Call for donations
In-Reply-To: <629b08a40807150338y42dab8bbxab15a7d990afdf26@mail.gmail.com>
References: <629b08a40807150338y42dab8bbxab15a7d990afdf26@mail.gmail.com>
Message-ID: <629b08a40807151053j139b51b8g5992d9859705789c@mail.gmail.com>

Hi all,

As you may already know, Python(x,y) is a free Python / Qt / Eclipse
distribution for scientific computing, numerical simulations, data
analysis and visualization, ...

The basic idea of Python(x,y) was to freely share with other people a
ready-to-use Python distribution being a viable alternative to MATLAB
or IDL, i.e. providing documentation, development environment and
interactive consoles in an integrated way with each other. Almost
everything was already available to make it, one had only to gather
all these tools, documents, libraries, and put them together in one
self-sufficient software (and scripting the installation process to
configure automatically all these softwares).

At first, I did this for me (in my free time) because it was a good
way to install Python from scratch.
Then I quickly decided to distribute it freely - and this takes a lot
of time, but I was very glad that other people could benefit from my
work.
However, distributing freely Python(x,y) paradoxically takes money too
(which is mine by the way because this is a private initiative). Not
much though. But still, hosting 500Mb of data with a traffic limit of
24Gb per month is not free.

Now Python(x,y) is a victim of its own success (traffic is increasing
each month : 37Gb last month, which is over the limit but my web
hosting provider has a flexible quota policy) and the new release
(2.0.0) allow to distribute easily Python(x,y) individual packages
(even on computers with only Python 2.5 installed and for users who
don't want to install the distribution but only some packages) but
500Mb are not enough...

I could pay for it because it's really not much (it's clearly not a
question of money but more likely a question of principle), but I've
already paid for the current website and spend a lot of time on this,
so I thought that I could share my debts as well as my work :)

So, feel free to donate on the Python(x,y) website.
Thanks !
Pierre


From anirudhvij at gmail.com  Tue Jul 15 14:31:14 2008
From: anirudhvij at gmail.com (anirudh vij)
Date: Tue, 15 Jul 2008 20:31:14 +0200
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <4db580fd0807150024v79726315j7570b0785783381d@mail.gmail.com>
References: <mailman.13.1215709207.15246.scipy-user@scipy.org>
	<wl8tzerhje5.fsf@glup.physics.ucf.edu>
	<4db580fd0807150024v79726315j7570b0785783381d@mail.gmail.com>
Message-ID: <1aa72c450807151131s51636acayf46d736aa802c193@mail.gmail.com>

On Tue, Jul 15, 2008 at 9:24 AM, Hoyt Koepke <hoytak at gmail.com> wrote:
> I know I'm a little late to the main discussion, but FYI there is a
> project, pym, which claims to do some conversion:
> http://code.google.com/p/pym/  It's only available from svn, and
> there's no claim about how well it works, but it might be a good
> starting point.  I believe it's MIT licensed, so no issue there.  Plus
> the authors might be good contacts.
>
> --Hoyt
>

thanks for this. Your google-fu is stronger than mine :)


Work on my own script has been sluggish, since there is a conference
going on here. I am looking at the pym code right now to see how
usable it is.
>
>
>
> On Tue, Jul 15, 2008 at 7:16 AM, Joe Harrington <jh at physics.ucf.edu> wrote:
>> Automatic conversion of at least the boring stuff is crucial.  No
>> matter how educational rewriting your code may be, if you have 100,000
>> lines of tools and utilities to convert, you generally can't.  Our
>> IDL-based group has made good use of i2py, which may at least provide
>> some educational examples for converting MATLAB, if not a usable
>> framework.  The general thing is to covert what is easy, and at least
>> flag what is difficult.  It doesn't do a mapping of function names to
>> function names or anything like that.  Here's the URL, hope it helps.
>>
>> http://software.pseudogreen.org/i2py/
>>
>> If you do a general parser approach, it would be useful to do it in
>> such a way that the front end can be ripped out and modified or
>> replaced with one for another language.
>>
>> I've added a section on the scipy.org wiki under Topical Software for
>> conversion programs.  When yours is ready, please put in a link there.
>>
Will do. Meanwhile, should a link to pym be put up on the wiki?


From forrest at physics.Auburn.EDU  Tue Jul 15 16:13:15 2008
From: forrest at physics.Auburn.EDU (Phil Forrest)
Date: Tue, 15 Jul 2008 15:13:15 -0500
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
In-Reply-To: <487C0D0D.5040904@gmail.com>
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>	<3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>	<487BCE78.3040403@gmail.com>	<3d375d730807141542i1086c63fx9a43e0bad275fb7c@mail.gmail.com>	<487BDCAD.4090907@gmail.com>	<3d375d730807141635l667ecc94j4cea55c0c4251909@mail.gmail.com>
	<487C0D0D.5040904@gmail.com>
Message-ID: <015401c8e6b7$36f552f0$a4dff8d0$@auburn.edu>

Robert Kern wrote:

>> Yup. We can do a version check. For >= 8.3, we don't put in f77compat.
>> Can you and Phil check that if you modify that line of code in sun.py
>> to omit f77compat on your Sun Fortran 95 8.3 boxes, that it works?
>> (Ugh, that sentence got away from me, but too late now!)

Hey Guys,

I did not have any success on a Sun Studio 12 (8.3) scipy build with
f77compat removed from the sun.py in the numpy tree. The numpy build went
fine (I cleaned the tree, downloaded the source, built via 'python setup.py
install')

I didn't want to flood this list with compile error messages, so please let
me know if someone would like me to send out the build failure output for
scipy.

>One more thing: Sin Phil wants to build a 64 bit version he ought to set

I do not need (at the moment) a 64-bit build. Which is why my environment
variables for BLAS and LAPACK point to 32 bit libs:

BLAS=/opt/SUNWspro/lib/libsunperf.so
LAPACK=/opt/SUNWspro/lib/libsunmath.so

>export
BLAS=/usr/local/SunStudio12-200709/x86_64-SunOS/SUNWspro/lib/amd64/libsunmat
h.so
>Notice .../lib/amd64/... since lib contains the 32 bit version of the
library and lib/amd64 the 64 bit version.

Again, I know about as much of python as my 2 year old nephew. I'm
attempting this install at the request of a researcher who needs it.

Thanks,
Phil




From robert.kern at gmail.com  Tue Jul 15 16:19:28 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Tue, 15 Jul 2008 15:19:28 -0500
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
In-Reply-To: <015401c8e6b7$36f552f0$a4dff8d0$@auburn.edu>
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>
	<3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>
	<487BCE78.3040403@gmail.com>
	<3d375d730807141542i1086c63fx9a43e0bad275fb7c@mail.gmail.com>
	<487BDCAD.4090907@gmail.com>
	<3d375d730807141635l667ecc94j4cea55c0c4251909@mail.gmail.com>
	<487C0D0D.5040904@gmail.com>
	<015401c8e6b7$36f552f0$a4dff8d0$@auburn.edu>
Message-ID: <3d375d730807151319x405f6966i947357bb49cbdd19@mail.gmail.com>

On Tue, Jul 15, 2008 at 15:13, Phil Forrest <forrest at physics.auburn.edu> wrote:
> Robert Kern wrote:
>
>>> Yup. We can do a version check. For >= 8.3, we don't put in f77compat.
>>> Can you and Phil check that if you modify that line of code in sun.py
>>> to omit f77compat on your Sun Fortran 95 8.3 boxes, that it works?
>>> (Ugh, that sentence got away from me, but too late now!)
>
> Hey Guys,
>
> I did not have any success on a Sun Studio 12 (8.3) scipy build with
> f77compat removed from the sun.py in the numpy tree. The numpy build went
> fine (I cleaned the tree, downloaded the source, built via 'python setup.py
> install')
>
> I didn't want to flood this list with compile error messages, so please let
> me know if someone would like me to send out the build failure output for
> scipy.

You can send it to me in private email.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From s.mientki at ru.nl  Tue Jul 15 16:24:16 2008
From: s.mientki at ru.nl (Stef Mientki)
Date: Tue, 15 Jul 2008 22:24:16 +0200
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <1aa72c450807151131s51636acayf46d736aa802c193@mail.gmail.com>
References: <mailman.13.1215709207.15246.scipy-user@scipy.org>
	<wl8tzerhje5.fsf@glup.physics.ucf.edu>
	<4db580fd0807150024v79726315j7570b0785783381d@mail.gmail.com>
	<1aa72c450807151131s51636acayf46d736aa802c193@mail.gmail.com>
Message-ID: <487D0770.8060400@ru.nl>



anirudh vij wrote:
> On Tue, Jul 15, 2008 at 9:24 AM, Hoyt Koepke <hoytak at gmail.com> wrote:
>   
>> I know I'm a little late to the main discussion, but FYI there is a
>> project, pym, which claims to do some conversion:
>> http://code.google.com/p/pym/  It's only available from svn, and
>> there's no claim about how well it works, but it might be a good
>> starting point.  I believe it's MIT licensed, so no issue there.  Plus
>> the authors might be good contacts.
>>
>> --Hoyt
>>
>>     
>
> thanks for this. Your google-fu is stronger than mine :)
>
>   
I couldn't find any files
(don't understand anything of SVN)
could you please attach the downloaded files to this list ?

thanks,
Stef
> Work on my own script has been sluggish, since there is a conference
> going on here. I am looking at the pym code right now to see how
> usable it is.
>   
>>
>> On Tue, Jul 15, 2008 at 7:16 AM, Joe Harrington <jh at physics.ucf.edu> wrote:
>>     
>>> Automatic conversion of at least the boring stuff is crucial.  No
>>> matter how educational rewriting your code may be, if you have 100,000
>>> lines of tools and utilities to convert, you generally can't.  Our
>>> IDL-based group has made good use of i2py, which may at least provide
>>> some educational examples for converting MATLAB, if not a usable
>>> framework.  The general thing is to covert what is easy, and at least
>>> flag what is difficult.  It doesn't do a mapping of function names to
>>> function names or anything like that.  Here's the URL, hope it helps.
>>>
>>> http://software.pseudogreen.org/i2py/
>>>
>>> If you do a general parser approach, it would be useful to do it in
>>> such a way that the front end can be ripped out and modified or
>>> replaced with one for another language.
>>>
>>> I've added a section on the scipy.org wiki under Topical Software for
>>> conversion programs.  When yours is ready, please put in a link there.
>>>
>>>       
> Will do. Meanwhile, should a link to pym be put up on the wiki?
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
>   


From cournape at gmail.com  Tue Jul 15 16:28:16 2008
From: cournape at gmail.com (david)
Date: Tue, 15 Jul 2008 20:28:16 +0000 (UTC)
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>
	<3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>
	<487BCE78.3040403@gmail.com>
	<3d375d730807141542i1086c63fx9a43e0bad275fb7c@mail.gmail.com>
	<487BDCAD.4090907@gmail.com>
	<3d375d730807141635l667ecc94j4cea55c0c4251909@mail.gmail.com>
	<487C0D0D.5040904@gmail.com>
	<015401c8e6b7$36f552f0$a4dff8d0$@auburn.edu>
	<3d375d730807151319x405f6966i947357bb49cbdd19@mail.gmail.com>
Message-ID: <loom.20080715T202032-72@post.gmane.org>

> > me know if someone would like me to send out the build failure output for
> > scipy.

One thing which I had problems with on solaris (open solaris with sun studio
express, that is a bit different than the version we are talking about here)
was C++ code in sparsetools. Some weird missing symbols related to RTTI (g++
builds and link the thing ok).

Except that, I could build and test numpy and scipy correctly with Sun studio
express and sun perf  libraries, but using numscons. I did not feel like
supporting sunperf in numpy.distutils,

cheers,

David










From robert.kern at gmail.com  Tue Jul 15 16:44:27 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Tue, 15 Jul 2008 15:44:27 -0500
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <487D0770.8060400@ru.nl>
References: <mailman.13.1215709207.15246.scipy-user@scipy.org>
	<wl8tzerhje5.fsf@glup.physics.ucf.edu>
	<4db580fd0807150024v79726315j7570b0785783381d@mail.gmail.com>
	<1aa72c450807151131s51636acayf46d736aa802c193@mail.gmail.com>
	<487D0770.8060400@ru.nl>
Message-ID: <3d375d730807151344m6d85a2d8j8d6bdefd3af4349b@mail.gmail.com>

On Tue, Jul 15, 2008 at 15:24, Stef Mientki <s.mientki at ru.nl> wrote:
>
> anirudh vij wrote:
>> On Tue, Jul 15, 2008 at 9:24 AM, Hoyt Koepke <hoytak at gmail.com> wrote:
>>
>>> I know I'm a little late to the main discussion, but FYI there is a
>>> project, pym, which claims to do some conversion:
>>> http://code.google.com/p/pym/  It's only available from svn, and
>>> there's no claim about how well it works, but it might be a good
>>> starting point.  I believe it's MIT licensed, so no issue there.  Plus
>>> the authors might be good contacts.
>>>
>>> --Hoyt
>>
>> thanks for this. Your google-fu is stronger than mine :)
>>
> I couldn't find any files
> (don't understand anything of SVN)
> could you please attach the downloaded files to this list ?

Um, not to the list, please. You can browse the raw files here, if you wish:

  http://pym.googlecode.com/svn/trunk/

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From s.mientki at ru.nl  Tue Jul 15 17:13:41 2008
From: s.mientki at ru.nl (Stef Mientki)
Date: Tue, 15 Jul 2008 23:13:41 +0200
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <3d375d730807151344m6d85a2d8j8d6bdefd3af4349b@mail.gmail.com>
References: <mailman.13.1215709207.15246.scipy-user@scipy.org>
	<wl8tzerhje5.fsf@glup.physics.ucf.edu>
	<4db580fd0807150024v79726315j7570b0785783381d@mail.gmail.com>
	<1aa72c450807151131s51636acayf46d736aa802c193@mail.gmail.com>
	<487D0770.8060400@ru.nl>
	<3d375d730807151344m6d85a2d8j8d6bdefd3af4349b@mail.gmail.com>
Message-ID: <487D1305.5020506@ru.nl>



Robert Kern wrote:
> On Tue, Jul 15, 2008 at 15:24, Stef Mientki <s.mientki at ru.nl> wrote:
>   
>> anirudh vij wrote:
>>     
>>> On Tue, Jul 15, 2008 at 9:24 AM, Hoyt Koepke <hoytak at gmail.com> wrote:
>>>
>>>       
>>>> I know I'm a little late to the main discussion, but FYI there is a
>>>> project, pym, which claims to do some conversion:
>>>> http://code.google.com/p/pym/  It's only available from svn, and
>>>> there's no claim about how well it works, but it might be a good
>>>> starting point.  I believe it's MIT licensed, so no issue there.  Plus
>>>> the authors might be good contacts.
>>>>
>>>> --Hoyt
>>>>         
>>> thanks for this. Your google-fu is stronger than mine :)
>>>
>>>       
>> I couldn't find any files
>> (don't understand anything of SVN)
>> could you please attach the downloaded files to this list ?
>>     
>
> Um, not to the list, please. You can browse the raw files here, if you wish:
>
>   http://pym.googlecode.com/svn/trunk/
>   
thanks,
now I can see some files.
cheers,
Stef


From cournape at gmail.com  Tue Jul 15 17:27:31 2008
From: cournape at gmail.com (david)
Date: Tue, 15 Jul 2008 21:27:31 +0000 (UTC)
Subject: [SciPy-user] -lf77compat failure - SciPy on AMD64 Sun Studio
References: <00d101c8e5f6$f2992f20$d7cb8d60$@auburn.edu>
	<3d375d730807141441x4c7a1044wca49fafe653578ca@mail.gmail.com>
	<487BCE78.3040403@gmail.com>
	<3d375d730807141542i1086c63fx9a43e0bad275fb7c@mail.gmail.com>
	<487BDCAD.4090907@gmail.com>
	<3d375d730807141635l667ecc94j4cea55c0c4251909@mail.gmail.com>
	<487C0D0D.5040904@gmail.com>
	<015401c8e6b7$36f552f0$a4dff8d0$@auburn.edu>
	<3d375d730807151319x405f6966i947357bb49cbdd19@mail.gmail.com>
	<loom.20080715T202032-72@post.gmane.org>
Message-ID: <loom.20080715T211531-249@post.gmane.org>

david <cournape <at> gmail.com> writes:

> 
> > > me know if someone would like me to send out the build failure output for
> > > scipy.
> 
> One thing which I had problems with on solaris (open solaris with sun studio
> express, that is a bit different than the version we are talking about here)
> was C++ code in sparsetools. Some weird missing symbols related to RTTI (g++
> builds and link the thing ok).

Ok, I found out the problem: when using some features of C++, the sun C++ 
compiler needs additional libraries... The one I needed so far was Crun. 
Appending -lCrun to the C++ link command seems to work.

Now, I got some segfaults in sparse unit tests, but that may be caused by recent 
changes to it.

cheers,

David






From cournape at gmail.com  Tue Jul 15 17:43:04 2008
From: cournape at gmail.com (david)
Date: Tue, 15 Jul 2008 21:43:04 +0000 (UTC)
Subject: [SciPy-user] scipy-0.6 on AIX 5.3
References: <mailman.45115.1213146608.27196.scipy-user@scipy.org>
	<200807061617.07505.mwojc@p.lodz.pl>
	<48718535.8000703@ar.media.kyoto-u.ac.jp>
	<g4t44c$dqj$1@ger.gmane.org>
Message-ID: <loom.20080715T214020-754@post.gmane.org>

Marek Wojciechowski <mwojc <at> p.lodz.pl> writes:

> 
>
> 
> IBM Fortran compiler (xlf) worked very well for me.

Well, it does not work here.

> The only compilation
> problem is with two functions in specfun.f, namely: CLQN and CLQMN.
> However, this is because of the aggressive -O5 optimization flag (which is
> set up by default on AIX in numpy.distutils). With -O3 whole code compiles
> nicely.
> 

That's strange, since the errors are syntax related. In particular, from your
 output, it seemed like some comments were causing problems.

> BTW. i'm not sure if there is a way to get working g77 or gfortran on AIX.

Yes. When I googled for your problem on AIX, I saw similar problems 
compiling octave, and using gnu fortran compilers did solve the problem.

But anyway, since changing the optimization level worked, I guess there is no 
need for that anymore.

cheers,

David








From silva at lma.cnrs-mrs.fr  Tue Jul 15 17:51:17 2008
From: silva at lma.cnrs-mrs.fr (Fabrice Silva)
Date: Tue, 15 Jul 2008 23:51:17 +0200
Subject: [SciPy-user] odepack or f2py error
In-Reply-To: <1216059141.3007.5.camel@Portable-s2m.cnrs-mrs.fr>
References: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>
	<g55o0o$b0i$1@ger.gmane.org>
	<1215862349.3033.3.camel@Portable-s2m.cnrs-mrs.fr>
	<g5a7mc$dk8$1@ger.gmane.org>
	<1216059141.3007.5.camel@Portable-s2m.cnrs-mrs.fr>
Message-ID: <1216158677.2943.2.camel@Portable-s2m.cnrs-mrs.fr>

On ticket 696 (http://scipy.org/scipy/scipy/ticket/696), you can find a
minimal example of code reproducing the problem. It does not use f2py
anymore, but the trouble seems to be linked to the import of pylab...

-- 
Fabrice Silva



From hoytak at gmail.com  Wed Jul 16 02:42:54 2008
From: hoytak at gmail.com (Hoyt Koepke)
Date: Wed, 16 Jul 2008 09:42:54 +0300
Subject: [SciPy-user] Automatic MATLAB to scipy/numpy/pylab conversion
In-Reply-To: <1aa72c450807151131s51636acayf46d736aa802c193@mail.gmail.com>
References: <mailman.13.1215709207.15246.scipy-user@scipy.org>
	<wl8tzerhje5.fsf@glup.physics.ucf.edu>
	<4db580fd0807150024v79726315j7570b0785783381d@mail.gmail.com>
	<1aa72c450807151131s51636acayf46d736aa802c193@mail.gmail.com>
Message-ID: <4db580fd0807152342o1e9cae2atb4b3937a109788b@mail.gmail.com>

I only knew about it from a few months ago, when I was looking for a
similar tool, and it seems it was higher ranked then.  I couldn't find
it except for googling the exact name...

--Hoyt

On Tue, Jul 15, 2008 at 9:31 PM, anirudh vij <anirudhvij at gmail.com> wrote:
> On Tue, Jul 15, 2008 at 9:24 AM, Hoyt Koepke <hoytak at gmail.com> wrote:
>> I know I'm a little late to the main discussion, but FYI there is a
>> project, pym, which claims to do some conversion:
>> http://code.google.com/p/pym/  It's only available from svn, and
>> there's no claim about how well it works, but it might be a good
>> starting point.  I believe it's MIT licensed, so no issue there.  Plus
>> the authors might be good contacts.
>>
>> --Hoyt
>>
>
> thanks for this. Your google-fu is stronger than mine :)
>
>
> Work on my own script has been sluggish, since there is a conference
> going on here. I am looking at the pym code right now to see how
> usable it is.
>>
>>
>>
>> On Tue, Jul 15, 2008 at 7:16 AM, Joe Harrington <jh at physics.ucf.edu> wrote:
>>> Automatic conversion of at least the boring stuff is crucial.  No
>>> matter how educational rewriting your code may be, if you have 100,000
>>> lines of tools and utilities to convert, you generally can't.  Our
>>> IDL-based group has made good use of i2py, which may at least provide
>>> some educational examples for converting MATLAB, if not a usable
>>> framework.  The general thing is to covert what is easy, and at least
>>> flag what is difficult.  It doesn't do a mapping of function names to
>>> function names or anything like that.  Here's the URL, hope it helps.
>>>
>>> http://software.pseudogreen.org/i2py/
>>>
>>> If you do a general parser approach, it would be useful to do it in
>>> such a way that the front end can be ripped out and modified or
>>> replaced with one for another language.
>>>
>>> I've added a section on the scipy.org wiki under Topical Software for
>>> conversion programs.  When yours is ready, please put in a link there.
>>>
> Will do. Meanwhile, should a link to pym be put up on the wiki?
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
+++++++++++++++++++++++++++++++++++
Hoyt Koepke
UBC Department of Computer Science
http://www.cs.ubc.ca/~hoytak/
hoytak at gmail.com
+++++++++++++++++++++++++++++++++++


From haase at msg.ucsf.edu  Wed Jul 16 07:47:56 2008
From: haase at msg.ucsf.edu (Sebastian Haase)
Date: Wed, 16 Jul 2008 13:47:56 +0200
Subject: [SciPy-user] [python(x,y)] Call for donations
In-Reply-To: <629b08a40807151053j139b51b8g5992d9859705789c@mail.gmail.com>
References: <629b08a40807150338y42dab8bbxab15a7d990afdf26@mail.gmail.com>
	<629b08a40807151053j139b51b8g5992d9859705789c@mail.gmail.com>
Message-ID: <bc657ead0807160447x7285bd06p54403f1038d14f06@mail.gmail.com>

On Tue, Jul 15, 2008 at 7:53 PM, Pierre Raybaut <contact at pythonxy.com> wrote:
> Hi all,
>
> As you may already know, Python(x,y) is a free Python / Qt / Eclipse
> distribution for scientific computing, numerical simulations, data
> analysis and visualization, ...
>
> The basic idea of Python(x,y) was to freely share with other people a
> ready-to-use Python distribution being a viable alternative to MATLAB
> or IDL, i.e. providing documentation, development environment and
> interactive consoles in an integrated way with each other. Almost
> everything was already available to make it, one had only to gather
> all these tools, documents, libraries, and put them together in one
> self-sufficient software (and scripting the installation process to
> configure automatically all these softwares).
>
> At first, I did this for me (in my free time) because it was a good
> way to install Python from scratch.
> Then I quickly decided to distribute it freely - and this takes a lot
> of time, but I was very glad that other people could benefit from my
> work.
> However, distributing freely Python(x,y) paradoxically takes money too
> (which is mine by the way because this is a private initiative). Not
> much though. But still, hosting 500Mb of data with a traffic limit of
> 24Gb per month is not free.
>
> Now Python(x,y) is a victim of its own success (traffic is increasing
> each month : 37Gb last month, which is over the limit but my web
> hosting provider has a flexible quota policy) and the new release
> (2.0.0) allow to distribute easily Python(x,y) individual packages
> (even on computers with only Python 2.5 installed and for users who
> don't want to install the distribution but only some packages) but
> 500Mb are not enough...
>
> I could pay for it because it's really not much (it's clearly not a
> question of money but more likely a question of principle), but I've
> already paid for the current website and spend a lot of time on this,
> so I thought that I could share my debts as well as my work :)
>
> So, feel free to donate on the Python(x,y) website.
> Thanks !
> Pierre

Did you look into free hosting solutions - like google code !?

- Sebastian Haase


From dave.hirschfeld at gmail.com  Wed Jul 16 09:33:14 2008
From: dave.hirschfeld at gmail.com (Dave)
Date: Wed, 16 Jul 2008 13:33:14 +0000 (UTC)
Subject: [SciPy-user] OpenOpt svn broken on Windows
Message-ID: <loom.20080716T132953-79@post.gmane.org>

Under BrasilOpt there are two files ALGENCAN_oo.py and algencan_oo.py. Windows
filenames are not case sensitive so this causes the svn checkout to fail.

-Dave



From dmitrey.kroshko at scipy.org  Wed Jul 16 11:02:29 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Wed, 16 Jul 2008 18:02:29 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <loom.20080716T132953-79@post.gmane.org>
References: <loom.20080716T132953-79@post.gmane.org>
Message-ID: <487E0D85.5040709@scipy.org>

I don't know what is easiest solution here, currently you could use 
zip-file from OO installation page (~ 10 Mb, contains all scikits from 
latest svn automatically generated by Trac).

Regards, D.

Dave wrote:
> Under BrasilOpt there are two files ALGENCAN_oo.py and algencan_oo.py. Windows
> filenames are not case sensitive so this causes the svn checkout to fail.
>
> -Dave
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
>
>   



From william.ratcliff at gmail.com  Wed Jul 16 11:25:19 2008
From: william.ratcliff at gmail.com (william ratcliff)
Date: Wed, 16 Jul 2008 11:25:19 -0400
Subject: [SciPy-user] [python(x,y)] Call for donations
In-Reply-To: <bc657ead0807160447x7285bd06p54403f1038d14f06@mail.gmail.com>
References: <629b08a40807150338y42dab8bbxab15a7d990afdf26@mail.gmail.com>
	<629b08a40807151053j139b51b8g5992d9859705789c@mail.gmail.com>
	<bc657ead0807160447x7285bd06p54403f1038d14f06@mail.gmail.com>
Message-ID: <827183970807160825l232363f0p8893d33edce09b0f@mail.gmail.com>

Google code is actually very convenient to use :>

On Wed, Jul 16, 2008 at 7:47 AM, Sebastian Haase <haase at msg.ucsf.edu> wrote:

> On Tue, Jul 15, 2008 at 7:53 PM, Pierre Raybaut <contact at pythonxy.com>
> wrote:
> > Hi all,
> >
> > As you may already know, Python(x,y) is a free Python / Qt / Eclipse
> > distribution for scientific computing, numerical simulations, data
> > analysis and visualization, ...
> >
> > The basic idea of Python(x,y) was to freely share with other people a
> > ready-to-use Python distribution being a viable alternative to MATLAB
> > or IDL, i.e. providing documentation, development environment and
> > interactive consoles in an integrated way with each other. Almost
> > everything was already available to make it, one had only to gather
> > all these tools, documents, libraries, and put them together in one
> > self-sufficient software (and scripting the installation process to
> > configure automatically all these softwares).
> >
> > At first, I did this for me (in my free time) because it was a good
> > way to install Python from scratch.
> > Then I quickly decided to distribute it freely - and this takes a lot
> > of time, but I was very glad that other people could benefit from my
> > work.
> > However, distributing freely Python(x,y) paradoxically takes money too
> > (which is mine by the way because this is a private initiative). Not
> > much though. But still, hosting 500Mb of data with a traffic limit of
> > 24Gb per month is not free.
> >
> > Now Python(x,y) is a victim of its own success (traffic is increasing
> > each month : 37Gb last month, which is over the limit but my web
> > hosting provider has a flexible quota policy) and the new release
> > (2.0.0) allow to distribute easily Python(x,y) individual packages
> > (even on computers with only Python 2.5 installed and for users who
> > don't want to install the distribution but only some packages) but
> > 500Mb are not enough...
> >
> > I could pay for it because it's really not much (it's clearly not a
> > question of money but more likely a question of principle), but I've
> > already paid for the current website and spend a lot of time on this,
> > so I thought that I could share my debts as well as my work :)
> >
> > So, feel free to donate on the Python(x,y) website.
> > Thanks !
> > Pierre
>
> Did you look into free hosting solutions - like google code !?
>
> - Sebastian Haase
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080716/d26cb2bd/attachment.html>

From aisaac at american.edu  Wed Jul 16 11:52:44 2008
From: aisaac at american.edu (Alan G Isaac)
Date: Wed, 16 Jul 2008 11:52:44 -0400
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487E0D85.5040709@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E0D85.5040709@scipy.org>
Message-ID: <Mahogany-0.67.0-1064-20080716-115244.00@american.edu>

On Wed, 16 Jul 2008, dmitrey apparently wrote:
> I don't know what is easiest solution here, currently you 
> could use zip-file from OO installation page (~ 10 Mb, 
> contains all scikits from latest svn automatically 
> generated by Trac). 

Hi Dmitrey,

Since this is a real problem for cross platform use,
it needs to be addressed.

As a rule, two files with the same name expect for
capitalization is a bad idea.  Clearly these names
do not distinguish the file content.

Additionally, the two files contain serious DRY/SPOT
violations <URL:http://en.wikipedia.org/wiki/Don%27t_repeat_yourself>
so it looks like this bug report offers a chance to address
those as well.

Cheers,
Alan Isaac





From dmitrey.kroshko at scipy.org  Wed Jul 16 12:18:15 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Wed, 16 Jul 2008 19:18:15 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <Mahogany-0.67.0-1064-20080716-115244.00@american.edu>
References: <loom.20080716T132953-79@post.gmane.org>	<487E0D85.5040709@scipy.org>
	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>
Message-ID: <487E1F47.9020303@scipy.org>

Since OO has solvers "ALGENCAN" (for ALGENCAN v 1.0) and "algencan" (for 
ALGENCAN v 2.0) openopt must have the files ALGENCAN_oo.py and 
algencan_oo.py (because calling solver (that is done in OO by name) 
depend on file names).

I could just remove ALGENCAN (in anyway in future ALGENCAN v 1.0 support 
will be ceased), but currently some tests made by me show v 1.0 works 
better than 2.0.

So I intend to remove ALGENCAN_oo.py for next OO release, and doing it 
right now isn't  a good idea.

D.

Alan G Isaac wrote:
> On Wed, 16 Jul 2008, dmitrey apparently wrote:
>   
>> I don't know what is easiest solution here, currently you 
>> could use zip-file from OO installation page (~ 10 Mb, 
>> contains all scikits from latest svn automatically 
>> generated by Trac). 
>>     
>
> Hi Dmitrey,
>
> Since this is a real problem for cross platform use,
> it needs to be addressed.
>
> As a rule, two files with the same name expect for
> capitalization is a bad idea.  Clearly these names
> do not distinguish the file content.
>
> Additionally, the two files contain serious DRY/SPOT
> violations <URL:http://en.wikipedia.org/wiki/Don%27t_repeat_yourself>
> so it looks like this bug report offers a chance to address
> those as well.
>
> Cheers,
> Alan Isaac
>
>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
>
>   



From matthieu.brucher at gmail.com  Wed Jul 16 12:28:16 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Wed, 16 Jul 2008 18:28:16 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487E1F47.9020303@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E0D85.5040709@scipy.org>
	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>
	<487E1F47.9020303@scipy.org>
Message-ID: <e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>

Hi,

As Alan states it, you could use a diffrent name. Why not use
algencan2_oo.py ? With the current registry pattern in OO, the name of
the solver is not dependent on the actual implementation file.
Besides, you could resfactor the code by using some patterns, as Alan
suggested it ;)

Matthieu

2008/7/16 dmitrey <dmitrey.kroshko at scipy.org>:
> Since OO has solvers "ALGENCAN" (for ALGENCAN v 1.0) and "algencan" (for
> ALGENCAN v 2.0) openopt must have the files ALGENCAN_oo.py and
> algencan_oo.py (because calling solver (that is done in OO by name)
> depend on file names).
>
> I could just remove ALGENCAN (in anyway in future ALGENCAN v 1.0 support
> will be ceased), but currently some tests made by me show v 1.0 works
> better than 2.0.
>
> So I intend to remove ALGENCAN_oo.py for next OO release, and doing it
> right now isn't  a good idea.
>
> D.
>
> Alan G Isaac wrote:
>> On Wed, 16 Jul 2008, dmitrey apparently wrote:
>>
>>> I don't know what is easiest solution here, currently you
>>> could use zip-file from OO installation page (~ 10 Mb,
>>> contains all scikits from latest svn automatically
>>> generated by Trac).
>>>
>>
>> Hi Dmitrey,
>>
>> Since this is a real problem for cross platform use,
>> it needs to be addressed.
>>
>> As a rule, two files with the same name expect for
>> capitalization is a bad idea.  Clearly these names
>> do not distinguish the file content.
>>
>> Additionally, the two files contain serious DRY/SPOT
>> violations <URL:http://en.wikipedia.org/wiki/Don%27t_repeat_yourself>
>> so it looks like this bug report offers a chance to address
>> those as well.
>>
>> Cheers,
>> Alan Isaac
>>
>>
>>
>> _______________________________________________
>> SciPy-user mailing list
>> SciPy-user at scipy.org
>> http://projects.scipy.org/mailman/listinfo/scipy-user
>>
>>
>>
>>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From dmitrey.kroshko at scipy.org  Wed Jul 16 12:37:06 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Wed, 16 Jul 2008 19:37:06 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>
References: <loom.20080716T132953-79@post.gmane.org>	<487E0D85.5040709@scipy.org>	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>	<487E1F47.9020303@scipy.org>
	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>
Message-ID: <487E23B2.9060507@scipy.org>

Matthieu Brucher wrote:
> Hi,
>
> As Alan states it, you could use a diffrent name. Why not use
> algencan2_oo.py ?
I don't think using r = p.solve('algencan2') is a good idea, even 
"algencan" is too long to type. Also, there is a number of those 
algencan 2.0beta, 2.0.1beta, ..., 2.0.4 beta and the number will continue.

Also it means after new A. version I should remove old files each time, 
making some users code not working.
>  With the current registry pattern in OO, the name of
> the solver is not dependent on the actual implementation file.
>   
I don't understand what do you mean here, see runProbSolver.py, solver 
is extracting from py-file with same name.

> Besides, you could resfactor the code by using some patterns, as Alan
> suggested it ;)
>   
I intend just to remove ALGENCAN_oo.py, because I don't see needs to 
continue it maintenance in future.
D.


From matthieu.brucher at gmail.com  Wed Jul 16 12:47:58 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Wed, 16 Jul 2008 18:47:58 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487E23B2.9060507@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E0D85.5040709@scipy.org>
	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>
	<487E1F47.9020303@scipy.org>
	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>
	<487E23B2.9060507@scipy.org>
Message-ID: <e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>

2008/7/16 dmitrey <dmitrey.kroshko at scipy.org>:
> Matthieu Brucher wrote:
>> Hi,
>>
>> As Alan states it, you could use a diffrent name. Why not use
>> algencan2_oo.py ?
> I don't think using r = p.solve('algencan2') is a good idea, even
> "algencan" is too long to type. Also, there is a number of those
> algencan 2.0beta, 2.0.1beta, ..., 2.0.4 beta and the number will continue.
>
> Also it means after new A. version I should remove old files each time,
> making some users code not working.

Algecan2 is supposed to have a clear interface, isn't it ? So you only
have one wrapper per major version. This is done in other packages as
well. And it is clearer that this is for algecan 2.x and not for
algecan 1.x.

>>  With the current registry pattern in OO, the name of
>> the solver is not dependent on the actual implementation file.
>>
> I don't understand what do you mean here, see runProbSolver.py, solver
> is extracting from py-file with same name.

It wasn't the case some time ago. But you still can do what I mean.
I see that you can't register a new solver in OO on the fly and use
your wrappers. Why didn't you make __solverPaths__ directly a
dictionary that you populate then ? This way, you always have one
__solverPaths__ that is not different accross modules, as it could be
expected to be.

>> Besides, you could resfactor the code by using some patterns, as Alan
>> suggested it ;)
>>
> I intend just to remove ALGENCAN_oo.py, because I don't see needs to
> continue it maintenance in future.

Of course ;) but you should try to reuse your code at the beginning as well ;)

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From dmitrey.kroshko at scipy.org  Wed Jul 16 12:58:56 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Wed, 16 Jul 2008 19:58:56 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>
References: <loom.20080716T132953-79@post.gmane.org>	<487E0D85.5040709@scipy.org>	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>	<487E1F47.9020303@scipy.org>	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>	<487E23B2.9060507@scipy.org>
	<e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>
Message-ID: <487E28D0.7070505@scipy.org>

Matthieu Brucher wrote:
> 2008/7/16 dmitrey <dmitrey.kroshko at scipy.org>:
>   
>> Matthieu Brucher wrote:
>>     
>>> Hi,
>>>
>>> As Alan states it, you could use a diffrent name. Why not use
>>> algencan2_oo.py ?
>>>       
>> I don't think using r = p.solve('algencan2') is a good idea, even
>> "algencan" is too long to type. Also, there is a number of those
>> algencan 2.0beta, 2.0.1beta, ..., 2.0.4 beta and the number will continue.
>>
>> Also it means after new A. version I should remove old files each time,
>> making some users code not working.
>>     
>
> Algecan2 is supposed to have a clear interface, isn't it ? So you only
> have one wrapper per major version. This is done in other packages as
> well. And it is clearer that this is for algecan 2.x and not for
> algecan 1.x.
>   
I meant not 2.0.1 vs 2.0.3 but 2.x vs 3.x vs ... as well
>   
>>>  With the current registry pattern in OO, the name of
>>> the solver is not dependent on the actual implementation file.
>>>
>>>       
>> I don't understand what do you mean here, see runProbSolver.py, solver
>> is extracting from py-file with same name.
>>     
>
> It wasn't the case some time ago. But you still can do what I mean.
> I see that you can't register a new solver in OO on the fly and use
> your wrappers. Why didn't you make __solverPaths__ directly a
> dictionary that you populate then ? This way, you always have one
> __solverPaths__ that is not different accross modules, as it could be
> expected to be.
>   
Because the solution you propose turn out to be a headache, especially 
for me - and that's the main reason why I have removed your dictionary 
__solverPaths__.
Each time I created solvers like "ralgELS", "ralg2" etc (with some 
modifications to compare results with origin ralg) I have got a error 
during execution and have to go to runProbSolver.py file to fix the 
issue (as well as for connecting a new solver). Also, I was not fond of 
those dozens of additional lines of code in rPS.py
>   
>>> Besides, you could resfactor the code by using some patterns, as Alan
>>> suggested it ;)
>>>
>>>       
>> I intend just to remove ALGENCAN_oo.py, because I don't see needs to
>> continue it maintenance in future.
>>     
>
> Of course ;) but you should try to reuse your code at the beginning as well ;)
>   
I don't know what code do you mean here, I had already moved some code 
from ALGENCAN_oo.py to algencan_oo.py. Also, only 1 version of algencan 
can be used with OO because both versions produce pywrapper.so file, and 
only 1 can be used from PYTHONPATH. Of course, I can create a workaround 
but I have lots of other things to do, especially to complite my GSoC 
assigned tasks.
D.



From aisaac at american.edu  Wed Jul 16 13:15:36 2008
From: aisaac at american.edu (Alan G Isaac)
Date: Wed, 16 Jul 2008 13:15:36 -0400
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487E28D0.7070505@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E0D85.5040709@scipy.org>
	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>
	<487E1F47.9020303@scipy.org>
	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>
	<487E23B2.9060507@scipy.org><e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>
	<487E28D0.7070505@scipy.org>
Message-ID: <Mahogany-0.67.0-948-20080716-131536.00@american.edu>

> Matthieu Brucher wrote:
>> but you should try to reuse your code at the beginning as 
>> well ;)


On Wed, 16 Jul 2008, dmitrey apparently wrote:
> I don't know what code do you mean here, I had already moved some code 
> from ALGENCAN_oo.py to algencan_oo.py.


I think Matthieu is just referring to the DRY/SPOT 
violations I pointed out.  Generally speaking, you should 
not have two files containing duplicate code.

Just for example, one might refactor as follows:

        algencan_common.py
        (for code found in both current files)

        algencan.py
        (support for current version (unnumbered))

        algencan1.py (support for version 1)

The last two would import from the first as needed.
This is not a "proposal" per se, just an illustration.

Cheers,
Alan Isaac





From matthieu.brucher at gmail.com  Wed Jul 16 13:23:14 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Wed, 16 Jul 2008 19:23:14 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487E28D0.7070505@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E0D85.5040709@scipy.org>
	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>
	<487E1F47.9020303@scipy.org>
	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>
	<487E23B2.9060507@scipy.org>
	<e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>
	<487E28D0.7070505@scipy.org>
Message-ID: <e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>

>> Algecan2 is supposed to have a clear interface, isn't it ? So you only
>> have one wrapper per major version. This is done in other packages as
>> well. And it is clearer that this is for algecan 2.x and not for
>> algecan 1.x.
>>
> I meant not 2.0.1 vs 2.0.3 but 2.x vs 3.x vs ... as well

algecan 3 will break your current model. I think that the fft wrappers
for scipy are also named fftw2 and fftw3, so why not using it as well
?

>> It wasn't the case some time ago. But you still can do what I mean.
>> I see that you can't register a new solver in OO on the fly and use
>> your wrappers. Why didn't you make __solverPaths__ directly a
>> dictionary that you populate then ? This way, you always have one
>> __solverPaths__ that is not different accross modules, as it could be
>> expected to be.
>>
> Because the solution you propose turn out to be a headache, especially
> for me - and that's the main reason why I have removed your dictionary
> __solverPaths__.
> Each time I created solvers like "ralgELS", "ralg2" etc (with some
> modifications to compare results with origin ralg) I have got a error
> during execution and have to go to runProbSolver.py file to fix the
> issue (as well as for connecting a new solver). Also, I was not fond of
> those dozens of additional lines of code in rPS.py

I didn't say that you shouldn't have done this, you obviously can, you
have done it and it works. What I'm saying is that there were several
purposes for this registry :
- not polluting sys.path
- allow a clean way of adding a solver

For the moment, I think you have two places where you can do the second part :
- ooMisc.py
- runProbSolver.py
Each time, it uses __solverPaths__. The funny thing is that
__solverPaths__ in ooMisc is in fact never used or modified. If it
were a dictionary, it would be of some use.
If you modify ooMic.__solverPaths__ after loading OO, you can't add a
new solver, as it equals None. And you cannot add it before, as you
have to load OO to get access to it. So I think it is
runProbSolver.__solverPaths__ that should be used, but then,
ooMic.__solverPaths__ really is just junk code that is of no use. Or
have I missed something ?

>> Of course ;) but you should try to reuse your code at the beginning as well ;)
>>
> I don't know what code do you mean here, I had already moved some code
> from ALGENCAN_oo.py to algencan_oo.py.

It is one of the basis of agile development. Never copy code. What is
cool with Python is that you can import the pieces you want. For
instance, you can factor these code lines in a new package that you
will import in each solver. This is especially important when you find
a bug in one wrapper and you forget to apply it in the second wrapper.
And it will work great with algecan3 if you can also reuse the same
parts, as you will have a much shorter file to modify (and thus less
headhache ;))

Also, only 1 version of algencan
> can be used with OO because both versions produce pywrapper.so file, and
> only 1 can be used from PYTHONPATH. Of course, I can create a workaround
> but I have lots of other things to do, especially to complite my GSoC
> assigned tasks.

I didn't know that there was only one filename. Is it something you
build or is it something algecan provides ?

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From matthieu.brucher at gmail.com  Wed Jul 16 13:24:52 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Wed, 16 Jul 2008 19:24:52 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487E28D0.7070505@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E0D85.5040709@scipy.org>
	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>
	<487E1F47.9020303@scipy.org>
	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>
	<487E23B2.9060507@scipy.org>
	<e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>
	<487E28D0.7070505@scipy.org>
Message-ID: <e76aa17f0807161024s260d092jfb558ce9f29f7dd2@mail.gmail.com>

> I don't know what code do you mean here, I had already moved some code
> from ALGENCAN_oo.py to algencan_oo.py. Also, only 1 version of algencan
> can be used with OO because both versions produce pywrapper.so file, and
> only 1 can be used from PYTHONPATH. Of course, I can create a workaround
> but I have lots of other things to do, especially to complite my GSoC
> assigned tasks.

In fact, refactoring is also what you did when you rewrote the registry part ;)

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From aisaac at american.edu  Wed Jul 16 13:51:33 2008
From: aisaac at american.edu (Alan G Isaac)
Date: Wed, 16 Jul 2008 13:51:33 -0400
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487E23B2.9060507@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E0D85.5040709@scipy.org>
	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>
	<487E1F47.9020303@scipy.org><e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>
	<487E23B2.9060507@scipy.org>
Message-ID: <Mahogany-0.67.0-948-20080716-135133.00@american.edu>

On Wed, 16 Jul 2008, dmitrey apparently wrote:
> I intend just to remove ALGENCAN_oo.py, because I don't 
> see needs to continue it maintenance in future. 

I understand, but

- there is a problem right now: SVN is not working for
  Windows users.  This is not acceptable.
- the same problem will reappear in the future (new ALGENCAN 
  versions), so you might as well pick a solution now.

See my previous message for what might be a simple solution.

Cheers,
Alan Isaac





From dmitrey.kroshko at scipy.org  Wed Jul 16 14:19:23 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Wed, 16 Jul 2008 21:19:23 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <Mahogany-0.67.0-948-20080716-135133.00@american.edu>
References: <loom.20080716T132953-79@post.gmane.org>	<487E0D85.5040709@scipy.org>	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>	<487E1F47.9020303@scipy.org><e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>	<487E23B2.9060507@scipy.org>
	<Mahogany-0.67.0-948-20080716-135133.00@american.edu>
Message-ID: <487E3BAB.3070402@scipy.org>

Alan G Isaac wrote:
> On Wed, 16 Jul 2008, dmitrey apparently wrote:
>   
>> I intend just to remove ALGENCAN_oo.py, because I don't 
>> see needs to continue it maintenance in future. 
>>     
>
> I understand, but
>
> - there is a problem right now: SVN is not working for
>   Windows users.  This is not acceptable.
> - the same problem will reappear in the future (new ALGENCAN 
>   versions), so you might as well pick a solution now.
>   
There will be no problem because I intend to provide just single version 
of algencan (latest) in algencan_oo.py and nothing more.
As for Windows users, I guess they can wait till next OO release. These 
2 files are in repository for several months but Dave was first (and I 
guess he will be last till next OO release) who encountered the problem, 
because very seldom Windows users download OO via svn. In anyway, they 
have a number of other download possibilities - zip-file, tarballs etc.
D.


From dmitrey.kroshko at scipy.org  Wed Jul 16 14:19:26 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Wed, 16 Jul 2008 21:19:26 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>
References: <loom.20080716T132953-79@post.gmane.org>	<487E0D85.5040709@scipy.org>	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>	<487E1F47.9020303@scipy.org>	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>	<487E23B2.9060507@scipy.org>	<e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>	<487E28D0.7070505@scipy.org>
	<e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>
Message-ID: <487E3BAE.1000004@scipy.org>

Matthieu Brucher wrote:
>>> Algecan2 is supposed to have a clear interface, isn't it ? So you only
>>> have one wrapper per major version. This is done in other packages as
>>> well. And it is clearer that this is for algecan 2.x and not for
>>> algecan 1.x.
>>>
>>>       
>> I meant not 2.0.1 vs 2.0.3 but 2.x vs 3.x vs ... as well
>>     
>
> algecan 3 will break your current model. I think that the fft wrappers
> for scipy are also named fftw2 and fftw3, so why not using it as well
> ?
>   
I don't think ALGENCAN 3 will have other than ALGENCAN 2 API. If it will 
be different, I'll change algencan_oo.py once again. In any way I don't 
intend to maintain several ALGENCAN versions.
>   
>>> It wasn't the case some time ago. But you still can do what I mean.
>>> I see that you can't register a new solver in OO on the fly and use
>>> your wrappers. Why didn't you make __solverPaths__ directly a
>>> dictionary that you populate then ? This way, you always have one
>>> __solverPaths__ that is not different accross modules, as it could be
>>> expected to be.
>>>
>>>       
>> Because the solution you propose turn out to be a headache, especially
>> for me - and that's the main reason why I have removed your dictionary
>> __solverPaths__.
>> Each time I created solvers like "ralgELS", "ralg2" etc (with some
>> modifications to compare results with origin ralg) I have got a error
>> during execution and have to go to runProbSolver.py file to fix the
>> issue (as well as for connecting a new solver). Also, I was not fond of
>> those dozens of additional lines of code in rPS.py
>>     
>
> I didn't say that you shouldn't have done this, you obviously can, you
> have done it and it works. What I'm saying is that there were several
> purposes for this registry :
> - not polluting sys.path
> - allow a clean way of adding a solver
>   
Adding a solver have been never more clear than current style: just add 
_oo file to any directory from /solvers, and it will work. Also, it was 
the way it works before your changes.
> For the moment, I think you have two places where you can do the second part :
> - ooMisc.py
> - runProbSolver.py
> Each time, it uses __solverPaths__. The funny thing is that
> __solverPaths__ in ooMisc is in fact never used or modified. 

> If it
> were a dictionary, it would be of some use.
> If you modify ooMic.__solverPaths__ after loading OO, you can't add a
> new solver, as it equals None. And you cannot add it before, as you
> have to load OO to get access to it. So I think it is
> runProbSolver.__solverPaths__ that should be used, but then,
> ooMic.__solverPaths__ really is just junk code that is of no use. Or
> have I missed something ?
>   
__solverPaths__ is exported by runProbSolver.py, and than converted from None to
dict. I intended to init it in runProbSolver.py but some issues had
appeared (Python reported of bugs) so I had moved it to ooMisc.py

>>> Of course ;) but you should try to reuse your code at the beginning as well ;)
>>>
>>>       
>> I don't know what code do you mean here, I had already moved some code
>> from ALGENCAN_oo.py to algencan_oo.py.
>>     
>
> It is one of the basis of agile development. Never copy code. What is
> cool with Python is that you can import the pieces you want. For
> instance, you can factor these code lines in a new package that you
> will import in each solver. This is especially important when you find
> a bug in one wrapper and you forget to apply it in the second wrapper.
> And it will work great with algecan3 if you can also reuse the same
> parts, as you will have a much shorter file to modify (and thus less
> headhache ;))
>   
I understand the DRY issue, in fact, there is some code related to 
CVXOPT that is used by several solvers. But I when I created 
algencan_oo.py I intended to remove ALGENCAN_oo.py as soon as new file 
will be ready. However, I decided to remain it in repository for some 
more time, because old ALGENCAN passes some tests better than new ver. 
In anyway, soon it will be removed and the problem will gone away.
> Also, only 1 version of algencan
>   
>> can be used with OO because both versions produce pywrapper.so file, and
>> only 1 can be used from PYTHONPATH. Of course, I can create a workaround
>> but I have lots of other things to do, especially to complite my GSoC
>> assigned tasks.
>>     
>
> I didn't know that there was only one filename. Is it something you
> build or is it something algecan provides ?
>   
latter. pywrapper.so it generated by ALGENCAN. As I had mentioned, I 
could create a workaround but I have lots of other things to do.
D.


From matthieu.brucher at gmail.com  Wed Jul 16 14:43:03 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Wed, 16 Jul 2008 20:43:03 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487E3BAE.1000004@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E0D85.5040709@scipy.org>
	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>
	<487E1F47.9020303@scipy.org>
	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>
	<487E23B2.9060507@scipy.org>
	<e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>
	<487E28D0.7070505@scipy.org>
	<e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>
	<487E3BAE.1000004@scipy.org>
Message-ID: <e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>

>> I didn't say that you shouldn't have done this, you obviously can, you
>> have done it and it works. What I'm saying is that there were several
>> purposes for this registry :
>> - not polluting sys.path
>> - allow a clean way of adding a solver
>>
> Adding a solver have been never more clear than current style: just add
> _oo file to any directory from /solvers, and it will work. Also, it was
> the way it works before your changes.

You are still missing the point after so many months....

> __solverPaths__ is exported by runProbSolver.py, and than converted from None to
> dict. I intended to init it in runProbSolver.py but some issues had
> appeared (Python reported of bugs) so I had moved it to ooMisc.py

It has not been moved to ooMisc. The actual registry is only available
in runProbSolver. Because you are initializing it in runProbSolver and
not in ooMisc.

>> I didn't know that there was only one filename. Is it something you
>> build or is it something algecan provides ?
>>
> latter. pywrapper.so it generated by ALGENCAN. As I had mentioned, I
> could create a workaround but I have lots of other things to do.

And it is not your job to try to get a workaround (IMHO) ;)

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From cdcasey at gmail.com  Wed Jul 16 15:04:22 2008
From: cdcasey at gmail.com (chris)
Date: Wed, 16 Jul 2008 14:04:22 -0500
Subject: [SciPy-user] [python(x,y)] Call for donations
In-Reply-To: <bc657ead0807160447x7285bd06p54403f1038d14f06@mail.gmail.com>
References: <629b08a40807150338y42dab8bbxab15a7d990afdf26@mail.gmail.com>
	<629b08a40807151053j139b51b8g5992d9859705789c@mail.gmail.com>
	<bc657ead0807160447x7285bd06p54403f1038d14f06@mail.gmail.com>
Message-ID: <ead057790807161204p34776e70obcae6a0816e6de63@mail.gmail.com>

On Wed, Jul 16, 2008 at 6:47 AM, Sebastian Haase <haase at msg.ucsf.edu> wrote:
>
> Did you look into free hosting solutions - like google code !?
>

Or, as far as I know, sourceforge.


From aisaac at american.edu  Wed Jul 16 15:09:27 2008
From: aisaac at american.edu (Alan G Isaac)
Date: Wed, 16 Jul 2008 15:09:27 -0400
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487E3BAB.3070402@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E0D85.5040709@scipy.org>
	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>
	<487E1F47.9020303@scipy.org><e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>
	<487E23B2.9060507@scipy.org><Mahogany-0.67.0-948-20080716-135133.00@american.edu>
	<487E3BAB.3070402@scipy.org>
Message-ID: <Mahogany-0.67.0-948-20080716-150927.00@american.edu>

On Wed, 16 Jul 2008, dmitrey apparently wrote:
> very seldom Windows users download OO via svn 

You may have just learned why: it does not work.

What is the problem with the approach I suggested?
It looks like 15 minutes (max) to fix something broken,
which is 15 minutes well spent.  Or am I overlooking
another problem?

As I said, fixing this should pay benefits in the future as 
well, since you can always keep old versions around in 
numbered form.  You are very likely to have this need again.

Cheers,
Alan Isaac





From matthieu.brucher at gmail.com  Wed Jul 16 15:07:35 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Wed, 16 Jul 2008 21:07:35 +0200
Subject: [SciPy-user] [python(x,y)] Call for donations
In-Reply-To: <ead057790807161204p34776e70obcae6a0816e6de63@mail.gmail.com>
References: <629b08a40807150338y42dab8bbxab15a7d990afdf26@mail.gmail.com>
	<629b08a40807151053j139b51b8g5992d9859705789c@mail.gmail.com>
	<bc657ead0807160447x7285bd06p54403f1038d14f06@mail.gmail.com>
	<ead057790807161204p34776e70obcae6a0816e6de63@mail.gmail.com>
Message-ID: <e76aa17f0807161207w53159ca8x301f3c61edfb9b95@mail.gmail.com>

2008/7/16 chris <cdcasey at gmail.com>:
> On Wed, Jul 16, 2008 at 6:47 AM, Sebastian Haase <haase at msg.ucsf.edu> wrote:
>>
>> Did you look into free hosting solutions - like google code !?
>>
>
> Or, as far as I know, sourceforge.

or Launchpad, or gna, ...

-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From dmitrey.kroshko at scipy.org  Wed Jul 16 15:32:56 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Wed, 16 Jul 2008 22:32:56 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
References: <loom.20080716T132953-79@post.gmane.org>	<487E0D85.5040709@scipy.org>	<Mahogany-0.67.0-1064-20080716-115244.00@american.edu>	<487E1F47.9020303@scipy.org>	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>	<487E23B2.9060507@scipy.org>	<e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>	<487E28D0.7070505@scipy.org>	<e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>	<487E3BAE.1000004@scipy.org>
	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
Message-ID: <487E4CE8.5000609@scipy.org>

Matthieu Brucher wrote:
>> __solverPaths__ is exported by runProbSolver.py, and than converted from None to
>> dict. I intended to init it in runProbSolver.py but some issues had
>> appeared (Python reported of bugs) so I had moved it to ooMisc.py
>>     
>
> It has not been moved to ooMisc. The actual registry is only available
> in runProbSolver. Because you are initializing it in runProbSolver and
> not in ooMisc.
>   
What is not been moved to ooMisc?! I have said I encountered bugs while 
creating __solverPaths__ in runProbSolver, so now it's created in ooMisc 
(as None). Of course, then I redefine it as Python dict (in runProbSolver).
>   
>>> I didn't know that there was only one filename. Is it something you
>>> build or is it something algecan provides ?
>>>
>>>       
>> latter. pywrapper.so it generated by ALGENCAN. As I had mentioned, I
>> could create a workaround but I have lots of other things to do.
>>     
>
> And it is not your job to try to get a workaround (IMHO) ;)
>   
Neither connecting IPOPT nor ALGENCAN 2.x version  (as well as some 
other work done during my GSoC 2008) are not mentioned in my 
GSoC-assigned schedule (still I had spent some more time because some OO 
users were very interested in these issues). So it's not my job to get 
ALGENCAN workaround.

If my GSoC mentors will insist on getting OO svn work for windows w/o 
wait for next OO release, for to economy my time I'll just remove 
ALGENCAN_oo.py, it will take much less efforts.
D.


From pav at iki.fi  Wed Jul 16 15:52:54 2008
From: pav at iki.fi (Pauli Virtanen)
Date: Wed, 16 Jul 2008 19:52:54 +0000 (UTC)
Subject: [SciPy-user] odepack or f2py error
References: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>
	<g55o0o$b0i$1@ger.gmane.org>
	<1215862349.3033.3.camel@Portable-s2m.cnrs-mrs.fr>
	<g5a7mc$dk8$1@ger.gmane.org>
	<1216059141.3007.5.camel@Portable-s2m.cnrs-mrs.fr>
	<1216158677.2943.2.camel@Portable-s2m.cnrs-mrs.fr>
Message-ID: <g5ljim$kqm$1@ger.gmane.org>

Tue, 15 Jul 2008 23:51:17 +0200, Fabrice Silva wrote:

> On ticket 696 (http://scipy.org/scipy/scipy/ticket/696), you can find a
> minimal example of code reproducing the problem. It does not use f2py
> anymore, but the trouble seems to be linked to the import of pylab...

gcc bug #36857. Can be worked around by

	export LC_ALL=C

in shell before running your codes.

(The scipy ticket can be closed.)

-- 
Pauli Virtanen



From matthieu.brucher at gmail.com  Wed Jul 16 16:01:30 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Wed, 16 Jul 2008 22:01:30 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487E4CE8.5000609@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E1F47.9020303@scipy.org>
	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>
	<487E23B2.9060507@scipy.org>
	<e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>
	<487E28D0.7070505@scipy.org>
	<e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>
	<487E3BAE.1000004@scipy.org>
	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
	<487E4CE8.5000609@scipy.org>
Message-ID: <e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>

2008/7/16 dmitrey <dmitrey.kroshko at scipy.org>:
> Matthieu Brucher wrote:
>>> __solverPaths__ is exported by runProbSolver.py, and than converted from None to
>>> dict. I intended to init it in runProbSolver.py but some issues had
>>> appeared (Python reported of bugs) so I had moved it to ooMisc.py
>>>
>>
>> It has not been moved to ooMisc. The actual registry is only available
>> in runProbSolver. Because you are initializing it in runProbSolver and
>> not in ooMisc.
>>
> What is not been moved to ooMisc?! I have said I encountered bugs while
> creating __solverPaths__ in runProbSolver, so now it's created in ooMisc
> (as None). Of course, then I redefine it as Python dict (in runProbSolver).

Try and get the type of ooMisc.__solverPaths__. It will always be
None. Why ? Because you import the variable None as __solverPaths__ in
rPS. Then you modify the reference, but not the underlying variable.

Try this:

module1.py

s = "Something"

module2.py

import module1.s

module1.s = "Something else"

module3.py

from module1 import s
s = "Something else"

And now

>>> import module1
>>> module1.s
Something
>>> import module2
>>> module1.s
Something else

Try the same with module3
>>> import module1
>>> module1.s
Something
>>> import module3
>>> module1.s
Something

This is exactly your case.

This question showed up on the python-list ML. At first, I didn't
understand either, but if you're thinking about references to
variables, you can understand what is really going on.

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From aisaac at american.edu  Wed Jul 16 16:42:52 2008
From: aisaac at american.edu (Alan G Isaac)
Date: Wed, 16 Jul 2008 16:42:52 -0400
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
References: <loom.20080716T132953-79@post.gmane.org><487E1F47.9020303@scipy.org><e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com><487E23B2.9060507@scipy.org><e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com><487E28D0.7070505@scipy.org><e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com><487E3BAE.1000004@scipy.org><e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com><487E4CE8.5000609@scipy.org>
	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
Message-ID: <Mahogany-0.67.0-948-20080716-164252.00@american.edu>

On Wed, 16 Jul 2008, Matthieu Brucher apparently wrote:
> you modify the reference, but not the underlying variable. 

This may be helpful:
http://effbot.org/zone/import-confusion.htm#what-does-python-do

Importing a module creates a module object
and then executes the module's code in this object's namespace.
Like other objects, this object allows dynamic attribute 
assignment. (E.g., you could say ``module1.t='test'``
even though module1 does not define ``t``.)

But for each module you can only import once.
(Ignoring ``reload``.)  Subsequent imports statements
will therefore not recover the assignments made in the module.

One thing that follows is that
``from module1 import s``
is alot like
import module1
``s = module1.s``
(except the latter has ``module1`` in the namespace).

fwiw,
Alan





From astrofitz at gmail.com  Wed Jul 16 17:44:23 2008
From: astrofitz at gmail.com (Michael Fitzgerald)
Date: Wed, 16 Jul 2008 14:44:23 -0700
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <loom.20080716T132953-79@post.gmane.org>
References: <loom.20080716T132953-79@post.gmane.org>
Message-ID: <2df6e3390807161444x527080b2gf75cbc0cd1f66f55@mail.gmail.com>

On Jul 16, 2008, at 6:33 AM, Dave wrote:

Under BrasilOpt there are two files ALGENCAN_oo.py and algencan_oo.py.
Windows
filenames are not case sensitive so this causes the svn checkout to fail.


This is also the case under OSX with most HFS+ file systems (case
sensitivity is disabled by default).

Mike
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080716/5cf6d959/attachment.html>

From silva at lma.cnrs-mrs.fr  Wed Jul 16 21:06:19 2008
From: silva at lma.cnrs-mrs.fr (Fabrice Silva)
Date: Thu, 17 Jul 2008 03:06:19 +0200
Subject: [SciPy-user] odepack or f2py error
In-Reply-To: <g5ljim$kqm$1@ger.gmane.org>
References: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>
	<g55o0o$b0i$1@ger.gmane.org>
	<1215862349.3033.3.camel@Portable-s2m.cnrs-mrs.fr>
	<g5a7mc$dk8$1@ger.gmane.org>
	<1216059141.3007.5.camel@Portable-s2m.cnrs-mrs.fr>
	<1216158677.2943.2.camel@Portable-s2m.cnrs-mrs.fr>
	<g5ljim$kqm$1@ger.gmane.org>
Message-ID: <1216256779.3088.6.camel@Portable-s2m.cnrs-mrs.fr>

Le mercredi 16 juillet 2008 ? 19:52 +0000, Pauli Virtanen a ?crit :
> gcc bug #36857. Can be worked around by
> 
> 	export LC_ALL=C
> in shell before running your codes.
> (The scipy ticket can be closed.)

If this bug is linked to locale setting and decimal separator symbol,
how can it be explained that without pylab import, the displayed results
is 

         lsoda--  at current t (=r1), mxstep (=i1) steps   
               taken on this call before reaching tout     
              In above message,  I1 =         2
              In above message,  R1 =  0.4064399263119E-05
        Excess work done on this call (perhaps wrong Dfun type).
        Run with full_output = 1 to get quantitative information.

and with pylab import, it crashes as follows:

         lsoda--  at current t (=r1), mxstep (=i1) steps   
               taken on this call before reaching tout     
              In above message,  I1 =         2
              In above message,  R1 =
        At line 1991 of file scipy/integrate/odepack/ddasrt.f (unit = 6, file = 'stdout')
        Internal Error: printf is broken

with the example file Solv.py given in Ticket #696 ?

In the first example, we can see that the separator is '.' . Does pylab
modify this desktop setting ?
-- 
Fabrice Silva



From silva at lma.cnrs-mrs.fr  Wed Jul 16 21:35:38 2008
From: silva at lma.cnrs-mrs.fr (Fabrice Silva)
Date: Thu, 17 Jul 2008 03:35:38 +0200
Subject: [SciPy-user] odepack or f2py error
In-Reply-To: <1216256779.3088.6.camel@Portable-s2m.cnrs-mrs.fr>
References: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>
	<g55o0o$b0i$1@ger.gmane.org>
	<1215862349.3033.3.camel@Portable-s2m.cnrs-mrs.fr>
	<g5a7mc$dk8$1@ger.gmane.org>
	<1216059141.3007.5.camel@Portable-s2m.cnrs-mrs.fr>
	<1216158677.2943.2.camel@Portable-s2m.cnrs-mrs.fr>
	<g5ljim$kqm$1@ger.gmane.org>
	<1216256779.3088.6.camel@Portable-s2m.cnrs-mrs.fr>
Message-ID: <1216258538.3088.10.camel@Portable-s2m.cnrs-mrs.fr>

Le jeudi 17 juillet 2008 ? 03:06 +0200, Fabrice Silva a ?crit :
> Le mercredi 16 juillet 2008 ? 19:52 +0000, Pauli Virtanen a ?crit :
> > gcc bug #36857. Can be worked around by
> > 
> > 	export LC_ALL=C
> > in shell before running your codes.
> > (The scipy ticket can be closed.)
> 
> If this bug is linked to locale setting and decimal separator symbol,
> how can it be explained that without pylab import[...]

Another comment on this bug :
http://telin.ugent.be/~slippens/drupal/pylabllocaleproblem
which is really due to pylab.
-- 
Fabrice Silva



From robert.kern at gmail.com  Wed Jul 16 20:09:59 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Wed, 16 Jul 2008 19:09:59 -0500
Subject: [SciPy-user] odepack or f2py error
In-Reply-To: <1216256779.3088.6.camel@Portable-s2m.cnrs-mrs.fr>
References: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>
	<g55o0o$b0i$1@ger.gmane.org>
	<1215862349.3033.3.camel@Portable-s2m.cnrs-mrs.fr>
	<g5a7mc$dk8$1@ger.gmane.org>
	<1216059141.3007.5.camel@Portable-s2m.cnrs-mrs.fr>
	<1216158677.2943.2.camel@Portable-s2m.cnrs-mrs.fr>
	<g5ljim$kqm$1@ger.gmane.org>
	<1216256779.3088.6.camel@Portable-s2m.cnrs-mrs.fr>
Message-ID: <3d375d730807161709q22b2685el10d32695c0e268c0@mail.gmail.com>

On Wed, Jul 16, 2008 at 20:06, Fabrice Silva <silva at lma.cnrs-mrs.fr> wrote:
> Le mercredi 16 juillet 2008 ? 19:52 +0000, Pauli Virtanen a ?crit :
>> gcc bug #36857. Can be worked around by
>>
>>       export LC_ALL=C
>> in shell before running your codes.
>> (The scipy ticket can be closed.)
>
> If this bug is linked to locale setting and decimal separator symbol,
> how can it be explained that without pylab import, the displayed results
> is
>
>         lsoda--  at current t (=r1), mxstep (=i1) steps
>               taken on this call before reaching tout
>              In above message,  I1 =         2
>              In above message,  R1 =  0.4064399263119E-05
>        Excess work done on this call (perhaps wrong Dfun type).
>        Run with full_output = 1 to get quantitative information.
>
> and with pylab import, it crashes as follows:
>
>         lsoda--  at current t (=r1), mxstep (=i1) steps
>               taken on this call before reaching tout
>              In above message,  I1 =         2
>              In above message,  R1 =
>        At line 1991 of file scipy/integrate/odepack/ddasrt.f (unit = 6, file = 'stdout')
>        Internal Error: printf is broken
>
> with the example file Solv.py given in Ticket #696 ?
>
> In the first example, we can see that the separator is '.' . Does pylab
> modify this desktop setting ?

Can you try doing this:

  import locale
  encoding = locale.getpreferredencoding()

  # ... whatever crashes

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco

From matthieu.brucher at gmail.com  Thu Jul 17 01:52:37 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Thu, 17 Jul 2008 07:52:37 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487E4CE8.5000609@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E1F47.9020303@scipy.org>
	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>
	<487E23B2.9060507@scipy.org>
	<e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>
	<487E28D0.7070505@scipy.org>
	<e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>
	<487E3BAE.1000004@scipy.org>
	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
	<487E4CE8.5000609@scipy.org>
Message-ID: <e76aa17f0807162252x3939234cu67e062df4b55aa8a@mail.gmail.com>

> Neither connecting IPOPT nor ALGENCAN 2.x version  (as well as some
> other work done during my GSoC 2008) are not mentioned in my
> GSoC-assigned schedule (still I had spent some more time because some OO
> users were very interested in these issues). So it's not my job to get
> ALGENCAN workaround.

I've just noticed this.
If you want people to help you, provide a simple way of doing so.
Write a little tutorial about registering a basic solver and then you
may have help.
I don't think that maintaining a package is only a GSoC work. You will
always have work that you did not plan to do. But modifying an
external package (algecan) is not the same as enhancing your own
package. The first is not your primary job. The latter is.

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From dmitrey.kroshko at scipy.org  Thu Jul 17 02:24:08 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Thu, 17 Jul 2008 09:24:08 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
References: <loom.20080716T132953-79@post.gmane.org>	<487E1F47.9020303@scipy.org>	<e76aa17f0807160928t4210651fga7a26957f38ef026@mail.gmail.com>	<487E23B2.9060507@scipy.org>	<e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>	<487E28D0.7070505@scipy.org>	<e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>	<487E3BAE.1000004@scipy.org>	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>	<487E4CE8.5000609@scipy.org>
	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
Message-ID: <487EE588.1000204@scipy.org>

Matthieu Brucher wrote:
> 2008/7/16 dmitrey <dmitrey.kroshko at scipy.org>:
>   
>> Matthieu Brucher wrote:
>>     
>>>> __solverPaths__ is exported by runProbSolver.py, and than converted from None to
>>>> dict. I intended to init it in runProbSolver.py but some issues had
>>>> appeared (Python reported of bugs) so I had moved it to ooMisc.py
>>>>
>>>>         
>>> It has not been moved to ooMisc. The actual registry is only available
>>> in runProbSolver. Because you are initializing it in runProbSolver and
>>> not in ooMisc.
>>>
>>>       
>> What is not been moved to ooMisc?! I have said I encountered bugs while
>> creating __solverPaths__ in runProbSolver, so now it's created in ooMisc
>> (as None). Of course, then I redefine it as Python dict (in runProbSolver).
>>     
>
> Try and get the type of ooMisc.__solverPaths__. It will always be
> None. Why ? Because you import the variable None as __solverPaths__ in
> rPS. Then you modify the reference, but not the underlying variable.
>   
Do you want to inform me something is wrong with my code?
I have no needs to import ooMisc.__solverPaths__ each time. It is 
imported only once per Python session, no matter how many solvers were 
called, as well as it is redefined in rPS.py to dictionary just a single 
time per session.

>But modifying an external package (algecan) is not the same as enhancing your own package. The first is not your primary job. The latter is.

I just said " it's not my job to get ALGENCAN workaround." I have to do 
my GSoC milestones from OOTimeLine and no more, so it's is up to me to 
decide to do any additional job or not to do, aa well as  choosing which 
one.

D.


From matthieu.brucher at gmail.com  Thu Jul 17 02:31:44 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Thu, 17 Jul 2008 08:31:44 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487EE588.1000204@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E23B2.9060507@scipy.org>
	<e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>
	<487E28D0.7070505@scipy.org>
	<e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>
	<487E3BAE.1000004@scipy.org>
	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
	<487E4CE8.5000609@scipy.org>
	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
	<487EE588.1000204@scipy.org>
Message-ID: <e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>

>> Try and get the type of ooMisc.__solverPaths__. It will always be
>> None. Why ? Because you import the variable None as __solverPaths__ in
>> rPS. Then you modify the reference, but not the underlying variable.
>>
> Do you want to inform me something is wrong with my code?

Yes...

> I have no needs to import ooMisc.__solverPaths__ each time. It is
> imported only once per Python session, no matter how many solvers were
> called, as well as it is redefined in rPS.py to dictionary just a single
> time per session.

What is the purpose of ooMisc.__solverPaths__ ? If it should be a
dictionary once scikits.openopt is imported, you have an error in your
program. If it is not, you have some junk code that can be deleted.

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From dmitrey.kroshko at scipy.org  Thu Jul 17 02:43:47 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Thu, 17 Jul 2008 09:43:47 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>
References: <loom.20080716T132953-79@post.gmane.org>	<487E23B2.9060507@scipy.org>	<e76aa17f0807160947v37668096l6acc8185ad999898@mail.gmail.com>	<487E28D0.7070505@scipy.org>	<e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>	<487E3BAE.1000004@scipy.org>	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>	<487E4CE8.5000609@scipy.org>	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>	<487EE588.1000204@scipy.org>
	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>
Message-ID: <487EEA23.1020105@scipy.org>

Matthieu Brucher wrote:
>> I have no needs to import ooMisc.__solverPaths__ each time. It is
>> imported only once per Python session, no matter how many solvers were
>> called, as well as it is redefined in rPS.py to dictionary just a single
>> time per session.
>>     
>
> What is the purpose of ooMisc.__solverPaths__ ? If it should be a
> dictionary once scikits.openopt is imported, you have an error in your
> program. If it is not, you have some junk code that can be deleted.
>   
Why it should be a dictionary once scikits.openopt is imported? What's 
wrong now? scikits.openopt doesn't contain __solverPaths__, it's 
auxiliary variable used deeply in Kernel only, and anyone don't have to 
care about this variable value.

Currently I removed import  __solverPaths__, add __solverPaths__= None 
in rPS.py file and it works ok with my Python 2.5 but I got a error in 
2.4, that's why I had moved __solverPaths__ to ooMisc. I intend to 
remove it (from svn) later, when Python 2.4 will be too obsolete.
D.


From matthieu.brucher at gmail.com  Thu Jul 17 03:02:53 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Thu, 17 Jul 2008 09:02:53 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487EEA23.1020105@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E28D0.7070505@scipy.org>
	<e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>
	<487E3BAE.1000004@scipy.org>
	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
	<487E4CE8.5000609@scipy.org>
	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
	<487EE588.1000204@scipy.org>
	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>
	<487EEA23.1020105@scipy.org>
Message-ID: <e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>

>> What is the purpose of ooMisc.__solverPaths__ ? If it should be a
>> dictionary once scikits.openopt is imported, you have an error in your
>> program. If it is not, you have some junk code that can be deleted.
>>
> Why it should be a dictionary once scikits.openopt is imported? What's
> wrong now? scikits.openopt doesn't contain __solverPaths__, it's
> auxiliary variable used deeply in Kernel only, and anyone don't have to
> care about this variable value.

There are two parts in what I'm saying:
- the first is that there is a langage (Python) misunderstanding, see
the link Alan provided sooner in the discussion, probably leading to
mistakes in the future,
- the second is that you have to expose it if you want people to help you.

Let me explain this.
You are complaining that you have to connect several optimizers
yourself to OO. This is sad, because, as you've stated, you don't have
much time, you have to work on the GSoC. So your best interest and the
best interest of the OO users is that additional people show up and
help you.
So what is the best course of action? Not everyone wants to have
access to the SVN. Enabling adding solvers on the fly solves this (and
you already stated before that you don't like people committing
changes to the SVN without your approval, so this IS the only way of
properly doing things). People install OO from the latest tarball.
Then they can create their own solver, add some tests to check if it
is working and then come back to you. This way, you will have feedback
from the community as well, and you won't feel alone developing the
package.

So if you want people to help you, please advertise __solverPaths__
and write a small tutorial.
If you don't want people helping you, please stop saying that it is
not your job to connect solver X or solver Y. It is, as you are the
only developer of this specific part of the package.

> Currently I removed import  __solverPaths__, add __solverPaths__= None
> in rPS.py file and it works ok with my Python 2.5 but I got a error in
> 2.4, that's why I had moved __solverPaths__ to ooMisc. I intend to
> remove it (from svn) later, when Python 2.4 will be too obsolete.

This is strange. I'll try to find out what is happening and I'll let you know ;)

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From dmitrey.kroshko at scipy.org  Thu Jul 17 03:19:12 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Thu, 17 Jul 2008 10:19:12 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>
References: <loom.20080716T132953-79@post.gmane.org>	<487E28D0.7070505@scipy.org>	<e76aa17f0807161023w64c2bbbft666d4332aa1d72bf@mail.gmail.com>	<487E3BAE.1000004@scipy.org>	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>	<487E4CE8.5000609@scipy.org>	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>	<487EE588.1000204@scipy.org>	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>	<487EEA23.1020105@scipy.org>
	<e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>
Message-ID: <487EF270.3050901@scipy.org>

Matthieu Brucher wrote:
> There are two parts in what I'm saying:
> - the first is that there is a langage (Python) misunderstanding, see
> the link Alan provided sooner in the discussion, probably leading to
> mistakes in the future,
> - the second is that you have to expose it if you want people to help you.
>
> Let me explain this.
> You are complaining that you have to connect several optimizers
> yourself to OO. 
Matthieu, IIRC I had been never complaining that I have to connect 
several optimizers to OO. As well as I don't need any help, it would be 
enough for me to feel myself OK would anyone don't commit his changes 
w/o even notification to me (and than I had to find bugs). IIRC there 
was a time I had proposed you to sent tarball with your proposed changes 
to me before committing to svn (to check for bugs) but you ignored it. 
The things like these make me crazy much more than doing OO jobs by 
myself (latter is preferred by me).
> This is sad, because, as you've stated, you don't have
> much time, you have to work on the GSoC. So your best interest and the
> best interest of the OO users is that additional people show up and
> help you.
> So what is the best course of action? Not everyone wants to have
> access to the SVN. Enabling adding solvers on the fly solves this (and
> you already stated before that you don't like people committing
> changes to the SVN without your approval, so this IS the only way of
> properly doing things). People install OO from the latest tarball.
> Then they can create their own solver, add some tests to check if it
> is working and then come back to you. This way, you will have feedback
> from the community as well, and you won't feel alone developing the
> package.
>
> So if you want people to help you, please advertise __solverPaths__
> and write a small tutorial.
> If you don't want people helping you, please stop saying that it is
> not your job to connect solver X or solver Y. 
Please give me a link to my message where I say "it is not my job to 
connect solver X or solver Y".

> It is, as you are the
> only developer of this specific part of the package.
>
>   
>> Currently I removed import  __solverPaths__, add __solverPaths__= None
>> in rPS.py file and it works ok with my Python 2.5 but I got a error in
>> 2.4, that's why I had moved __solverPaths__ to ooMisc. I intend to
>> remove it (from svn) later, when Python 2.4 will be too obsolete.
>>     
>
> This is strange. I'll try to find out what is happening and I'll let you know ;)
>
> Matthieu
>   

Let me also remember you that OO had been started even before GSOC 2007, 
and in GSoC 2008 rules (Python dept) they clarified that all code 
produced by student is his own one (still it should be available with 
BSD license). If you want to fork openopt to other package and develop 
that one by yourself - please inform me ASAP.
D.


From matthieu.brucher at gmail.com  Thu Jul 17 03:37:30 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Thu, 17 Jul 2008 09:37:30 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487EF270.3050901@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E3BAE.1000004@scipy.org>
	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
	<487E4CE8.5000609@scipy.org>
	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
	<487EE588.1000204@scipy.org>
	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>
	<487EEA23.1020105@scipy.org>
	<e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>
	<487EF270.3050901@scipy.org>
Message-ID: <e76aa17f0807170037j293b6c42o2af9f2d11eeae06e@mail.gmail.com>

>> Let me explain this.
>> You are complaining that you have to connect several optimizers
>> yourself to OO.

> Matthieu, IIRC I had been never complaining that I have to connect
> several optimizers to OO.

This is what you were saying some hours ago:

"> And it is not your job to try to get a workaround (IMHO) ;)
>
Neither connecting IPOPT nor ALGENCAN 2.x version  (as well as some
other work done during my GSoC 2008) are not mentioned in my
GSoC-assigned schedule (still I had spent some more time because some OO
users were very interested in these issues). So it's not my job to get
ALGENCAN workaround."

Sorry, but for me, this is complaining. And I'm not talking about the
mails you sent some months ago (in which you were saying that you
wanted to get paid to connect some solvers).

> Please give me a link to my message where I say "it is not my job to
> connect solver X or solver Y".

See my above quote.

> Let me also remember you that OO had been started even before GSOC 2007,
> and in GSoC 2008 rules (Python dept) they clarified that all code
> produced by student is his own one (still it should be available with
> BSD license). If you want to fork openopt to other package and develop
> that one by yourself - please inform me ASAP.

Let me also inform you that OO is a scikit to which I contributed a
lot as well. And to which I will keep contributing (generic framework
+ some other lines).

But I'll there, because, as always, it is not possible to discuss
something with you. I only wanted to provide you a simple and
efficient way to have additional contribution.

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From pav at iki.fi  Thu Jul 17 03:49:45 2008
From: pav at iki.fi (Pauli Virtanen)
Date: Thu, 17 Jul 2008 07:49:45 +0000 (UTC)
Subject: [SciPy-user] odepack or f2py error
References: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>
	<g55o0o$b0i$1@ger.gmane.org>
	<1215862349.3033.3.camel@Portable-s2m.cnrs-mrs.fr>
	<g5a7mc$dk8$1@ger.gmane.org>
	<1216059141.3007.5.camel@Portable-s2m.cnrs-mrs.fr>
	<1216158677.2943.2.camel@Portable-s2m.cnrs-mrs.fr>
	<g5ljim$kqm$1@ger.gmane.org>
	<1216256779.3088.6.camel@Portable-s2m.cnrs-mrs.fr>
Message-ID: <g5mtip$eq7$1@ger.gmane.org>

Thu, 17 Jul 2008 03:06:19 +0200, Fabrice Silva wrote:

> Le mercredi 16 juillet 2008 ? 19:52 +0000, Pauli Virtanen a ?crit :
>> gcc bug #36857. Can be worked around by
>> 
>> 	export LC_ALL=C
>> in shell before running your codes.
>> (The scipy ticket can be closed.)
> 
> If this bug is linked to locale setting and decimal separator symbol,
> how can it be explained that without pylab import, the displayed results
> is
[clip]
> and with pylab import, it crashes as follows:
[clip]

As said in the ticket comments, pylab imports gtk, which calls setlocale 
and sets the active locale according to the environment variables. This 
does not occur if you don't import pylab or call setlocale through some 
other route.

-- 
Pauli Virtanen



From robert.kern at gmail.com  Thu Jul 17 03:50:25 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Thu, 17 Jul 2008 02:50:25 -0500
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <e76aa17f0807170037j293b6c42o2af9f2d11eeae06e@mail.gmail.com>
References: <loom.20080716T132953-79@post.gmane.org>
	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
	<487E4CE8.5000609@scipy.org>
	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
	<487EE588.1000204@scipy.org>
	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>
	<487EEA23.1020105@scipy.org>
	<e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>
	<487EF270.3050901@scipy.org>
	<e76aa17f0807170037j293b6c42o2af9f2d11eeae06e@mail.gmail.com>
Message-ID: <3d375d730807170050m15aebd4es93c37a0258b8609d@mail.gmail.com>

I have asked both of you before to take your personal arguments off
the list. Keep it civil, or keep it in private email.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From dmitrey.kroshko at scipy.org  Thu Jul 17 03:55:15 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Thu, 17 Jul 2008 10:55:15 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <e76aa17f0807170037j293b6c42o2af9f2d11eeae06e@mail.gmail.com>
References: <loom.20080716T132953-79@post.gmane.org>	<487E3BAE.1000004@scipy.org>	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>	<487E4CE8.5000609@scipy.org>	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>	<487EE588.1000204@scipy.org>	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>	<487EEA23.1020105@scipy.org>	<e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>	<487EF270.3050901@scipy.org>
	<e76aa17f0807170037j293b6c42o2af9f2d11eeae06e@mail.gmail.com>
Message-ID: <487EFAE3.9020503@scipy.org>

Matthieu Brucher wrote:
>>> Let me explain this.
>>> You are complaining that you have to connect several optimizers
>>> yourself to OO.
>>>       
>
>   
>> Matthieu, IIRC I had been never complaining that I have to connect
>> several optimizers to OO.
>>     
>
> This is what you were saying some hours ago:
>
> "> And it is not your job to try to get a workaround (IMHO) ;)
>   
> Neither connecting IPOPT nor ALGENCAN 2.x version  (as well as some
> other work done during my GSoC 2008) are not mentioned in my
> GSoC-assigned schedule (still I had spent some more time because some OO
> users were very interested in these issues). So it's not my job to get
> ALGENCAN workaround."
>
> Sorry, but for me, this is complaining.
Maybe for you, but not for me (and I guess for others as well).
>  And I'm not talking about the
> mails you sent some months ago (in which you were saying that you
> wanted to get paid to connect some solvers).
>   
Why don't talk? Give a link to the messages (because I can't remember 
those ones), let's discuss it.
>   
>> Please give me a link to my message where I say "it is not my job to
>> connect solver X or solver Y".
>>     
>
> See my above quote.
>   
I see your quote above and don't find the words.

>   
>> Let me also remember you that OO had been started even before GSOC 2007,
>> and in GSoC 2008 rules (Python dept) they clarified that all code
>> produced by student is his own one (still it should be available with
>> BSD license). If you want to fork openopt to other package and develop
>> that one by yourself - please inform me ASAP.
>>     
>
> Let me also inform you that OO is a scikit to which I contributed a
> lot as well. And to which I will keep contributing (generic framework
> + some other lines).
>   
Let me remember you when I had been insisted to connect GenericOpt to 
OpenOpt I said "GenericOpt should stay available as separate package", 
and I have obtained agree (and AFAIK it remains available as separate 
package). So please enough speculating on GenericOpt. You have 
GenericOpt page in scikits webserver, any time you can remove GenericOpt 
from OpenOpt svn and continue to develop and use it as a standalone 
package (I would gladly hear from you of this decision).

The situation is like I commit something to Linux Kernel and then I 
commit whatever I want to any Linux files saying "I'm Linux developer 
too, I had committed something so I will continue more and more, no 
matter what you think".

D.


From gael.varoquaux at normalesup.org  Thu Jul 17 04:02:28 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Thu, 17 Jul 2008 10:02:28 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487EFAE3.9020503@scipy.org>
References: <e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
	<487E4CE8.5000609@scipy.org>
	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
	<487EE588.1000204@scipy.org>
	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>
	<487EEA23.1020105@scipy.org>
	<e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>
	<487EF270.3050901@scipy.org>
	<e76aa17f0807170037j293b6c42o2af9f2d11eeae06e@mail.gmail.com>
	<487EFAE3.9020503@scipy.org>
Message-ID: <20080717080228.GD616@phare.normalesup.org>

Guys. Quit it.

Ga?l


From silva at lma.cnrs-mrs.fr  Thu Jul 17 06:52:15 2008
From: silva at lma.cnrs-mrs.fr (Fabrice Silva)
Date: Thu, 17 Jul 2008 12:52:15 +0200
Subject: [SciPy-user] odepack or f2py error
In-Reply-To: <3d375d730807161709q22b2685el10d32695c0e268c0@mail.gmail.com>
References: <1215695635.3278.7.camel@Portable-s2m.cnrs-mrs.fr>
	<g55o0o$b0i$1@ger.gmane.org>
	<1215862349.3033.3.camel@Portable-s2m.cnrs-mrs.fr>
	<g5a7mc$dk8$1@ger.gmane.org>
	<1216059141.3007.5.camel@Portable-s2m.cnrs-mrs.fr>
	<1216158677.2943.2.camel@Portable-s2m.cnrs-mrs.fr>
	<g5ljim$kqm$1@ger.gmane.org>
	<1216256779.3088.6.camel@Portable-s2m.cnrs-mrs.fr>
	<3d375d730807161709q22b2685el10d32695c0e268c0@mail.gmail.com>
Message-ID: <1216291935.2996.9.camel@Portable-s2m.cnrs-mrs.fr>

Le mercredi 16 juillet 2008 ? 19:09 -0500, Robert Kern a ?crit :
> Can you try doing this:
> 
        import locale
        encoding = locale.getpreferredencoding()

'Preferred encoding' seems not to be affected by pylab import.

        import locale
        print locale.getpreferredencoding()
        import pylab
        print locale.getpreferredencoding()

gives
        UTF-8
        UTF-8
and the script crashes or not as well depending pylab is import or not.

Is that what you were expecting ?
> 
-- 
Fabrice Silva



From stefan at sun.ac.za  Thu Jul 17 09:08:02 2008
From: stefan at sun.ac.za (=?ISO-8859-1?Q?St=E9fan_van_der_Walt?=)
Date: Thu, 17 Jul 2008 15:08:02 +0200
Subject: [SciPy-user] patch for ndimage.generic_filter
In-Reply-To: <477CA520-C850-47AD-9888-92DDC968BBF3@gmail.com>
References: <477CA520-C850-47AD-9888-92DDC968BBF3@gmail.com>
Message-ID: <9457e7c80807170608m1ae4daeap8aeb18b52b18b4bb@mail.gmail.com>

Hey Mike,

2008/7/11 Michael Fitzgerald <astrofitz at gmail.com>:
> I found a small bug in ndimage.generic_filter() that arises when using the
> 'size' keyword with multidimensional arrays.  Patch attached.

The bug you found indicates that we have inadequate unit tests.  Would
you mind writing one that excercised generic_filter, so that this
doesn't happen again?

Thanks
St?fan


From aisaac at american.edu  Thu Jul 17 11:54:47 2008
From: aisaac at american.edu (Alan G Isaac)
Date: Thu, 17 Jul 2008 11:54:47 -0400
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487F6203.9030403@ukr.net>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E3BAE.1000004@scipy.org>
	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
	<487E4CE8.5000609@scipy.org>
	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
	<487EE588.1000204@scipy.org>
	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>
	<487EEA23.1020105@scipy.org>
	<e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>
	<487EF270.3050901@scipy.org><e76aa17f0807170037j293b6c42o2af9f2d11eeae06e@mail.gmail.com>
	<487EFAE3.9020503@scipy.org>
	<Mahogany-0.67.0-1792-20080717-091450.00@american.edu>
	<487F6203.9030403@ukr.net>
Message-ID: <Mahogany-0.67.0-1792-20080717-115447.00@american.edu>

OpenOpt SVN is once again usable by Windows and Mac users.
(Thanks Dmitrey!) Please report any problems.

Cheers,
Alan Isaac





From dmitrey.kroshko at scipy.org  Thu Jul 17 12:46:57 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Thu, 17 Jul 2008 19:46:57 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <Mahogany-0.67.0-1792-20080717-115447.00@american.edu>
References: <loom.20080716T132953-79@post.gmane.org>	<487E3BAE.1000004@scipy.org>	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>	<487E4CE8.5000609@scipy.org>	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>	<487EE588.1000204@scipy.org>	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>	<487EEA23.1020105@scipy.org>	<e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>	<487EF270.3050901@scipy.org><e76aa17f0807170037j293b6c42o2af9f2d11eeae06e@mail.gmail.com>	<487EFAE3.9020503@scipy.org>	<Mahogany-0.67.0-1792-20080717-091450.00@american.edu>	<487F6203.9030403@ukr.net>
	<Mahogany-0.67.0-1792-20080717-115447.00@american.edu>
Message-ID: <487F7781.3040904@scipy.org>

Alan G Isaac wrote:
> OpenOpt SVN is once again usable by Windows and Mac users.
> (Thanks Dmitrey!) Please report any problems.
>   
however, there are no reasons to thank because I just removed 
ALGENCAN_oo.py from repository, as it has been mentioned in OO blog. So 
now only ALGENCAN v 2.0.3 or later can be used.

Regards, D.


From forrest at physics.Auburn.EDU  Thu Jul 17 16:53:35 2008
From: forrest at physics.Auburn.EDU (Phil Forrest)
Date: Thu, 17 Jul 2008 15:53:35 -0500
Subject: [SciPy-user] sparsetools involvement? (was -lf77compat build
	failure)
Message-ID: <001301c8e84f$2dfdd750$89f985f0$@auburn.edu>

Hey Guys,

After reviewing the compilation failures, Robert Kern suggested I post the
following error messages to this list. Again, the original problem has to do
with no "f77 compatibility" library included with Sun Studio 12 (8.3) - at
least on Sun Solaris x86. The lack of this library caused the scipy build to
fail.

Below is a portion of the error messages from building scipy with
"f77compat" REMOVED from sun.py in the numpy tree/build.

Thanks,
Phil

> building 'scipy.sparse._sparsetools' extension compiling C++ sources C 
> compiler: /usr/lib/python2.4/pyCC -DNDEBUG
>
> compile options: '-Iscipy/sparse/sparsetools 
> -I/usr/lib/python2.4/site-packages/numpy/core/include
> -I/usr/include/python2.4 -c'
> pyCC: scipy/sparse/sparsetools/sparsetools_wrap.cxx
>
"/usr/lib/python2.4/site-packages/numpy/core/include/numpy/__multiarray_api.
> h", line 960: Warning: String literal converted to char* in formal 
> argument name in call to PyImport_ImportModule(char*).
>
"/usr/lib/python2.4/site-packages/numpy/core/include/numpy/__multiarray_api.
> h", line 963: Warning: String literal converted to char* in formal 
> argument
> 2 in call to PyObject_GetAttrString(_object*, char*).
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2602: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2603: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2604: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2605: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2606: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2607: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2608: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2609: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2610: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2611: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2612: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2614: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools.h", line 409: Error: multiplies 
> is not a member of std.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> While instantiating "csr_elmul_csr<int, int>(const int, const int, 
> const int*, const int*, const int*, const int*, const int*, const 
> int*, std::vector<int>*, std::vector<int>*, std::vector<int>*)".
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> Instantiated from non-template code.
> "scipy/sparse/sparsetools/sparsetools.h", line 409: Error: Unexpected 
> type name "T" encountered.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> While instantiating "csr_elmul_csr<int, int>(const int, const int, 
> const int*, const int*, const int*, const int*, const int*, const 
> int*, std::vector<int>*, std::vector<int>*, std::vector<int>*)".
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> Instantiated from non-template code.
> "scipy/sparse/sparsetools/sparsetools.h", line 409: Error: Operand 
> expected instead of ")".
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> While instantiating "csr_elmul_csr<int, int>(const int, const int, 
> const int*, const int*, const int*, const int*, const int*, const 
> int*, std::vector<int>*, std::vector<int>*, std::vector<int>*)".
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> Instantiated from non-template code.
> 3 Error(s) and 14 Warning(s) detected.
>
"/usr/lib/python2.4/site-packages/numpy/core/include/numpy/__multiarray_api.
> h", line 960: Warning: String literal converted to char* in formal 
> argument name in call to PyImport_ImportModule(char*).
>
"/usr/lib/python2.4/site-packages/numpy/core/include/numpy/__multiarray_api.
> h", line 963: Warning: String literal converted to char* in formal 
> argument
> 2 in call to PyObject_GetAttrString(_object*, char*).
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2602: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2603: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2604: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2605: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2606: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2607: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2608: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2609: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2610: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2611: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2612: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 2614: Warning: 
> String literal converted to char* in initialization.
> "scipy/sparse/sparsetools/sparsetools.h", line 409: Error: multiplies 
> is not a member of std.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> While instantiating "csr_elmul_csr<int, int>(const int, const int, 
> const int*, const int*, const int*, const int*, const int*, const 
> int*, std::vector<int>*, std::vector<int>*, std::vector<int>*)".
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> Instantiated from non-template code.
> "scipy/sparse/sparsetools/sparsetools.h", line 409: Error: Unexpected 
> type name "T" encountered.
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> While instantiating "csr_elmul_csr<int, int>(const int, const int, 
> const int*, const int*, const int*, const int*, const int*, const 
> int*, std::vector<int>*, std::vector<int>*, std::vector<int>*)".
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> Instantiated from non-template code.
> "scipy/sparse/sparsetools/sparsetools.h", line 409: Error: Operand 
> expected instead of ")".
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> While instantiating "csr_elmul_csr<int, int>(const int, const int, 
> const int*, const int*, const int*, const int*, const int*, const 
> int*, std::vector<int>*, std::vector<int>*, std::vector<int>*)".
> "scipy/sparse/sparsetools/sparsetools_wrap.cxx", line 15087:     Where:
> Instantiated from non-template code.
> 3 Error(s) and 14 Warning(s) detected.
> error: Command "/usr/lib/python2.4/pyCC -DNDEBUG 
> -Iscipy/sparse/sparsetools 
> -I/usr/lib/python2.4/site-packages/numpy/core/include
> -I/usr/include/python2.4 -c 
> scipy/sparse/sparsetools/sparsetools_wrap.cxx -o
build/temp.solaris-2.10-i86pc-2.4/scipy/sparse/sparsetools/sparsetools_wrap.
> o" failed with exit status 3
> #




From washakie at gmail.com  Fri Jul 18 01:37:31 2008
From: washakie at gmail.com (John [H2O])
Date: Thu, 17 Jul 2008 22:37:31 -0700 (PDT)
Subject: [SciPy-user]  help assigning arrays - global model output
Message-ID: <18522938.post@talk.nabble.com>


Hello, I am trying to fill an array reading in unformatted binary data :-/

I actually have most the code working now, and I have a F2Py module which
reads the binary efficiently. My problem is assigning the arrays into the
python multidimensional array.

This works:
G={}
for date in dates:
    data = F2Py.module.getdata()
    # data is shape (180,360,3)
    G[date] = data

What I WANT is this:

G=zeros((180,360,3,len(dates)))
for i in range(len(dates)):
    data = F2Py.module.getdata()
    # data is shape (180,360,3)
    G[:,:,:,i] = data

But I get a shape mismatch error, something about not broadcasting an
array.... 

Could someone please provide comment on how I need to do this assignment?

Thank you.
-- 
View this message in context: http://www.nabble.com/help-assigning-arrays---global-model-output-tp18522938p18522938.html
Sent from the Scipy-User mailing list archive at Nabble.com.



From robert.kern at gmail.com  Fri Jul 18 01:40:24 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Fri, 18 Jul 2008 00:40:24 -0500
Subject: [SciPy-user] help assigning arrays - global model output
In-Reply-To: <18522938.post@talk.nabble.com>
References: <18522938.post@talk.nabble.com>
Message-ID: <3d375d730807172240i33727681g508a8a6003da222a@mail.gmail.com>

On Fri, Jul 18, 2008 at 00:37, John [H2O] <washakie at gmail.com> wrote:
>
> Hello, I am trying to fill an array reading in unformatted binary data :-/
>
> I actually have most the code working now, and I have a F2Py module which
> reads the binary efficiently. My problem is assigning the arrays into the
> python multidimensional array.
>
> This works:
> G={}
> for date in dates:
>    data = F2Py.module.getdata()
>    # data is shape (180,360,3)
>    G[date] = data
>
> What I WANT is this:
>
> G=zeros((180,360,3,len(dates)))
> for i in range(len(dates)):
>    data = F2Py.module.getdata()
>    # data is shape (180,360,3)
>    G[:,:,:,i] = data
>
> But I get a shape mismatch error, something about not broadcasting an
> array....
>
> Could someone please provide comment on how I need to do this assignment?

Please double-check your shapes by printing out the actual shapes:


G=zeros((180,360,3,len(dates)))
for i in range(len(dates)):
   data = F2Py.module.getdata()
   print 'data: %s' % (data.shape,)
   print 'G[:,:,:,i]: %s' % (G[:,:,:,i].shape,)
   # data is shape (180,360,3)
   G[:,:,:,i] = data

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From washakie at gmail.com  Fri Jul 18 02:07:14 2008
From: washakie at gmail.com (John [H2O])
Date: Thu, 17 Jul 2008 23:07:14 -0700 (PDT)
Subject: [SciPy-user] help assigning arrays - global model output
In-Reply-To: <3d375d730807172240i33727681g508a8a6003da222a@mail.gmail.com>
References: <18522938.post@talk.nabble.com>
	<3d375d730807172240i33727681g508a8a6003da222a@mail.gmail.com>
Message-ID: <18523212.post@talk.nabble.com>




Robert Kern-2 wrote:
> 
> 
> Please double-check your shapes by printing out the actual shapes:
> 

Seems I must have been doing something wrong earlier! You're correct, I just
rewrote it in the 'old' or desired method without using the dict, and it
works! I must have been doing something wrong earlier... now to just figure
out which dimensions to sum across. 

Mind providing some insight on how to generate the average? What I have is
an array, with 20 days worth of global model output at three layers
(180,360,3) so my array is actually shape (180,360,3,20). I want the average
over the 20 days of the sum of the three vertical layers. I would also like
the complete sum.

Thanks in advance if you're so inclined!
-- 
View this message in context: http://www.nabble.com/help-assigning-arrays---global-model-output-tp18522938p18523212.html
Sent from the Scipy-User mailing list archive at Nabble.com.



From robert.kern at gmail.com  Fri Jul 18 02:30:10 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Fri, 18 Jul 2008 01:30:10 -0500
Subject: [SciPy-user] help assigning arrays - global model output
In-Reply-To: <18523212.post@talk.nabble.com>
References: <18522938.post@talk.nabble.com>
	<3d375d730807172240i33727681g508a8a6003da222a@mail.gmail.com>
	<18523212.post@talk.nabble.com>
Message-ID: <3d375d730807172330h440cd0f6ya3dc9f6292fc813e@mail.gmail.com>

On Fri, Jul 18, 2008 at 01:07, John [H2O] <washakie at gmail.com> wrote:
> Robert Kern-2 wrote:
>> Please double-check your shapes by printing out the actual shapes:
>
> Seems I must have been doing something wrong earlier! You're correct, I just
> rewrote it in the 'old' or desired method without using the dict, and it
> works! I must have been doing something wrong earlier... now to just figure
> out which dimensions to sum across.
>
> Mind providing some insight on how to generate the average? What I have is
> an array, with 20 days worth of global model output at three layers
> (180,360,3) so my array is actually shape (180,360,3,20). I want the average
> over the 20 days of the sum of the three vertical layers.

G.sum(axis=2).mean(axis=2)

> I would also like
> the complete sum.

Do you mean the sum over all of the layers and all of the days?

  G.sum(axis=2).sum(axis=2)

Or did you mean something else?

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From washakie at gmail.com  Fri Jul 18 02:39:49 2008
From: washakie at gmail.com (John [H2O])
Date: Thu, 17 Jul 2008 23:39:49 -0700 (PDT)
Subject: [SciPy-user]  sum / average of multidimensional arrays
Message-ID: <18523480.post@talk.nabble.com>


Hello, I have an array, with 20 days worth of global model output at three
layers (180,360,3) so my array is actually shape (180,360,3,20). 

I'm trying to produce a few new arrays:

1) The sum of the three layers for all 20 days, so shape (180,360,20)

and

2) The sum of all 20 days AND all three layers, so shape (180,360)

and

3) The average of all 20 days at each layer, so shape (180,360,3)

Anyone mind providing some insight? Thanks!
-- 
View this message in context: http://www.nabble.com/sum---average-of-multidimensional-arrays-tp18523480p18523480.html
Sent from the Scipy-User mailing list archive at Nabble.com.



From lbolla at gmail.com  Fri Jul 18 04:07:35 2008
From: lbolla at gmail.com (lorenzo bolla)
Date: Fri, 18 Jul 2008 09:07:35 +0100
Subject: [SciPy-user] sum / average of multidimensional arrays
In-Reply-To: <18523480.post@talk.nabble.com>
References: <18523480.post@talk.nabble.com>
Message-ID: <80c99e790807180107i489b0d2fr874f784e16413172@mail.gmail.com>

You can use the "axis" keyword in numpy.sum() and numpy.mean().
L.

On Fri, Jul 18, 2008 at 7:39 AM, John [H2O] <washakie at gmail.com> wrote:

>
> Hello, I have an array, with 20 days worth of global model output at three
> layers (180,360,3) so my array is actually shape (180,360,3,20).
>
> I'm trying to produce a few new arrays:
>
> 1) The sum of the tohree layers for all 20 days, so shape (180,360,20)
>
> and
>
> 2) The sum of all 20 days AND all three layers, so shape (180,360)
>
> and
>
> 3) The average of all 20 days at each layer, so shape (180,360,3)
>
> Anyone mind providing some insight? Thanks!
> --
> View this message in context:
> http://www.nabble.com/sum---average-of-multidimensional-arrays-tp18523480p18523480.html
> Sent from the Scipy-User mailing list archive at Nabble.com.
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
"Whereof one cannot speak, thereof one must be silent." -- Ludwig
Wittgenstein
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080718/b5513a0c/attachment.html>

From nwagner at iam.uni-stuttgart.de  Fri Jul 18 06:14:37 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Fri, 18 Jul 2008 12:14:37 +0200
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <487F7781.3040904@scipy.org>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E3BAE.1000004@scipy.org>
	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
	<487E4CE8.5000609@scipy.org>
	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
	<487EE588.1000204@scipy.org>
	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>
	<487EEA23.1020105@scipy.org>
	<e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>
	<487EF270.3050901@scipy.org>
	<e76aa17f0807170037j293b6c42o2af9f2d11eeae06e@mail.gmail.com>
	<487EFAE3.9020503@scipy.org>
	<Mahogany-0.67.0-1792-20080717-091450.00@american.edu>
	<487F6203.9030403@ukr.net>
	<Mahogany-0.67.0-1792-20080717-115447.00@american.edu>
	<487F7781.3040904@scipy.org>
Message-ID: <web-104874383@uni-stuttgart.de>

On Thu, 17 Jul 2008 19:46:57 +0300
  dmitrey <dmitrey.kroshko at scipy.org> wrote:
> Alan G Isaac wrote:
>> OpenOpt SVN is once again usable by Windows and Mac 
>>users
>> (Thanks Dmitrey!) Please report any problems.
>>   
> however, there are no reasons to thank because I just 
>removed 
> ALGENCAN_oo.py from repository, as it has been mentioned 
>in OO blog. So 
> now only ALGENCAN v 2.0.3 or later can be used.
> 
> Regards, D.
  
Hi Dmitrey,

I have some trouble to install tne new ALGENCAN version
on my linux box.

Please can you add some notes to your blog.

Thanks in advance

                      Nils


From dmitrey.kroshko at scipy.org  Fri Jul 18 07:24:22 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Fri, 18 Jul 2008 14:24:22 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <web-104874383@uni-stuttgart.de>
References: <loom.20080716T132953-79@post.gmane.org>	<487E3BAE.1000004@scipy.org>	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>	<487E4CE8.5000609@scipy.org>	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>	<487EE588.1000204@scipy.org>	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>	<487EEA23.1020105@scipy.org>	<e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>	<487EF270.3050901@scipy.org>	<e76aa17f0807170037j293b6c42o2af9f2d11eeae06e@mail.gmail.com>	<487EFAE3.9020503@scipy.org>	<Mahogany-0.67.0-1792-20080717-091450.00@american.edu>	<487F6203.9030403@ukr.net>	<Mahogany-0.67.0-1792-20080717-115447.00@american.edu>	<487F7781.3040904@scipy.org>
	<web-104874383@uni-stuttgart.de>
Message-ID: <48807D66.2060002@scipy.org>

Hi Nils,
w/o details on the troubles you have encountered I can hardly solve it.

Have you look at detailed instructions here:
http://openopt.blogspot.com/2008/06/connecting-algencan-2x-beta.html
?

chief idea - ensure you have pywrapper.so on your PYTHONPATH and old 
file pywrapper.so from algencan v 1.0 has been removed.

I will mention the moment in NLP page.

Regards, D.

Nils Wagner wrote:
> Hi Dmitrey,
>
> I have some trouble to install tne new ALGENCAN version
> on my linux box.
>
> Please can you add some notes to your blog.
>
> Thanks in advance
>
>                       Nils
>
>   



From wnbell at gmail.com  Fri Jul 18 09:37:46 2008
From: wnbell at gmail.com (Nathan Bell)
Date: Fri, 18 Jul 2008 08:37:46 -0500
Subject: [SciPy-user] sparsetools involvement? (was -lf77compat build
	failure)
In-Reply-To: <001301c8e84f$2dfdd750$89f985f0$@auburn.edu>
References: <001301c8e84f$2dfdd750$89f985f0$@auburn.edu>
Message-ID: <d05265cb0807180637s6307814em99b2103890de5920@mail.gmail.com>

On Thu, Jul 17, 2008 at 3:53 PM, Phil Forrest
<forrest at physics.auburn.edu> wrote:
>
> After reviewing the compilation failures, Robert Kern suggested I post the
> following error messages to this list. Again, the original problem has to do
> with no "f77 compatibility" library included with Sun Studio 12 (8.3) - at
> least on Sun Solaris x86. The lack of this library caused the scipy build to
> fail.
>
> Below is a portion of the error messages from building scipy with
> "f77compat" REMOVED from sun.py in the numpy tree/build.
>

Phil, could you try compiling SciPy from SVN?  Much of sparsetools has
changed since the last release.

I believe there was a missing #include <algorithm> in sparsetools.h in
the release you're using.

-- 
Nathan Bell wnbell at gmail.com
http://graphics.cs.uiuc.edu/~wnbell/


From washakie at gmail.com  Fri Jul 18 12:50:44 2008
From: washakie at gmail.com (John [H2O])
Date: Fri, 18 Jul 2008 09:50:44 -0700 (PDT)
Subject: [SciPy-user] sum / average of multidimensional arrays
In-Reply-To: <80c99e790807180107i489b0d2fr874f784e16413172@mail.gmail.com>
References: <18523480.post@talk.nabble.com>
	<80c99e790807180107i489b0d2fr874f784e16413172@mail.gmail.com>
Message-ID: <18533448.post@talk.nabble.com>



lorenzo bolla wrote:
> 
> You can use the "axis" keyword in numpy.sum() and numpy.mean().
> L.
> 

Thank you. 


-- 
View this message in context: http://www.nabble.com/sum---average-of-multidimensional-arrays-tp18523480p18533448.html
Sent from the Scipy-User mailing list archive at Nabble.com.



From Dharhas.Pothina at twdb.state.tx.us  Fri Jul 18 15:05:34 2008
From: Dharhas.Pothina at twdb.state.tx.us (Dharhas Pothina)
Date: Fri, 18 Jul 2008 14:05:34 -0500
Subject: [SciPy-user] Help interpolating equally spaced 2D gridded data to
	a	different grid.
Message-ID: <4880A32E0200009B00014590@GWWEB.twdb.state.tx.us>

Hi,

I have a regular grid of data :

lat_old.shape = (277, 349)
lon_old.shape = (277, 349)
precip_old.shape = (8, 277, 349) # the first axis is time 

I need to interpolate precip_old to a new regular grid (lat,lon) where 

lat.shape = (204, 160)
lon.shape = (204, 160)

Basically I need to calculate  precip_new where precip_new.shape = (8, 204, 160)

I've had a look at the cookbook page for 'N-D interpolation for equally-spaced data' and also some of the documentation for interp2D etc but am having trouble understanding what I need to do.

Any help is appreciated.

thanks

- dharhas


From dmitrey.kroshko at scipy.org  Fri Jul 18 15:08:37 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Fri, 18 Jul 2008 22:08:37 +0300
Subject: [SciPy-user] OpenOpt svn broken on Windows
In-Reply-To: <web-104893376@uni-stuttgart.de>
References: <loom.20080716T132953-79@post.gmane.org>
	<487E3BAE.1000004@scipy.org>
	<e76aa17f0807161143h5819a33by591b867728ca6004@mail.gmail.com>
	<487E4CE8.5000609@scipy.org>
	<e76aa17f0807161301w5ed11ac5ga6b4f331aaa65643@mail.gmail.com>
	<487EE588.1000204@scipy.org>
	<e76aa17f0807162331o700cd70fh9e782614d158968d@mail.gmail.com>
	<487EEA23.1020105@scipy.org>
	<e76aa17f0807170002s320ff03emb6619c030af8a580@mail.gmail.com>
	<487EF270.3050901@scipy.org>
	<e76aa17f0807170037j293b6c42o2af9f2d11eeae06e@mail.gmail.com>
	<487EFAE3.9020503@scipy.org>
	<Mahogany-0.67.0-1792-20080717-091450.00@american.edu>
	<487F6203.9030403@ukr.net>
	<Mahogany-0.67.0-1792-20080717-115447.00@american.edu>
	<487F7781.3040904@scipy.org> <web-104874383@uni-stuttgart.de>
	<48807D66.2060002@scipy.org> <web-104893376@uni-stuttgart.de>
Message-ID: <4880EA35.5020604@scipy.org>

hi Nils,
I don't know for sure yet but it can be actual:
http://openopt.blogspot.com/2008/07/algencan-diffint-issue.html

Regards, D.

Nils Wagner wrote:
>
> Hi Dmitrey,
>
> It works fine for me now with algencan-2.0.4-beta.
> I have used a modified Makefile together with
>
> make algencan-py
>
> Then I have modified my PYTHONPATH.
>  
> However the new algencan shows a poor convergence
> behavior wrt nlp_3.py. Any idea ?
>



From emanuele at relativita.com  Sat Jul 19 11:29:30 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Sat, 19 Jul 2008 17:29:30 +0200
Subject: [SciPy-user] bug in cho_factor?
Message-ID: <4882085A.2040107@relativita.com>

In scipy v0.7.0.dev4543 (svn) the documentation of
scipy.linalg.cho_factor says that cho_factor returns a tuple
made of:
----
    c : array, shape (M, M)
        Upper- or lower-triangular Cholesky factor of A
    lower : array, shape (M, M)
        Flag indicating whether the factor is lower or upper triangular
----
But the following simple example shows that 'c' is not triangular
and the result of cho_factor() is pretty different from that of
cholesky()!
----
import numpy as N
import scipy.linalg as SL
A = N.array([[2,1],[1,2]])
c_wrong,lower = SL.cho_factor(A)
print c_wrong
c_correct = SL.cholesky(A)
print c_correct
print c_wrong==c_correct
----
The output is:
----
[[ 1.41421356  0.70710678]
 [ 1.          1.22474487]]
[[ 1.41421356  0.70710678]
 [ 0.          1.22474487]]
[[ True  True]
 [False  True]]
----
Why c_wrong[1,0] is not zero? And why is it 1.0 ?

I went mad tracking this issue in my code 8-|

Regards,

Emanuele

P.S.: there is a type in cho_factor documentation: 'lower' is
not an 'array' but an 'int'.



From lbolla at gmail.com  Sat Jul 19 12:06:41 2008
From: lbolla at gmail.com (Lorenzo Bolla)
Date: Sat, 19 Jul 2008 17:06:41 +0100
Subject: [SciPy-user] Help interpolating equally spaced 2D gridded
	data	to a different grid.
In-Reply-To: <4880A32E0200009B00014590@GWWEB.twdb.state.tx.us>
References: <4880A32E0200009B00014590@GWWEB.twdb.state.tx.us>
Message-ID: <20080719160631.GA6332@lollo-laptop>

something like this should work (adjust the dimensions as you wish):

----------------------------------------
import numpy
from scipy.interpolate import interp2d

x = numpy.linspace(0, 1, 27)
y = numpy.linspace(0, 1, 34)

lat_old, lon_old = numpy.meshgrid(y, x)
precip_old = numpy.sin(lat_old + lon_old)

print lat_old.shape
print lon_old.shape
print precip_old.shape

x_new = numpy.linspace(0, 1, 20)
y_new = numpy.linspace(0, 1, 16)

lat, lon = numpy.meshgrid(y_new, x_new)
precip_interp = interp2d(lat_old, lon_old, precip_old)
precip = precip_interp(y_new, x_new)

print lat.shape
print lon.shape
print precip.shape
----------------------------------------

hth,
L.



On Fri, Jul 18, 2008 at 02:05:34PM -0500, Dharhas Pothina wrote:
> Hi,
> 
> I have a regular grid of data :
> 
> lat_old.shape = (277, 349)
> lon_old.shape = (277, 349)
> precip_old.shape = (8, 277, 349) # the first axis is time 
> 
> I need to interpolate precip_old to a new regular grid (lat,lon) where 
> 
> lat.shape = (204, 160)
> lon.shape = (204, 160)
> 
> Basically I need to calculate  precip_new where precip_new.shape = (8, 204, 160)
> 
> I've had a look at the cookbook page for 'N-D interpolation for equally-spaced data' and also some of the documentation for interp2D etc but am having trouble understanding what I need to do.
> 
> Any help is appreciated.
> 
> thanks
> 
> - dharhas
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user


From warren.weckesser at gmail.com  Sat Jul 19 12:28:55 2008
From: warren.weckesser at gmail.com (Warren Weckesser)
Date: Sat, 19 Jul 2008 12:28:55 -0400
Subject: [SciPy-user] bug in cho_factor?
In-Reply-To: <4882085A.2040107@relativita.com>
References: <4882085A.2040107@relativita.com>
Message-ID: <114880320807190928h29f43122ye85f060181c2d39b@mail.gmail.com>

It appears that cho_factor is a wrapper for the lapack routine potrf.
This function doesn't zero out the irrelevant part of the matrix; it
appears to leave it untouched.  In your case, the upper triangular
part of your answer is correct; the subdiagonal part is irrelevant
(and is apparently left over from the input matrix).

Apparently this is not a problem, since the output of cho_factor is
intended to be used in cho_solve (as the docstring states).  Presumably
cho_solve ignores the irrelevant parts of the input matrix.

Warren

On Sat, Jul 19, 2008 at 11:29 AM, Emanuele Olivetti <emanuele at relativita.com>
wrote:

> In scipy v0.7.0.dev4543 (svn) the documentation of
> scipy.linalg.cho_factor says that cho_factor returns a tuple
> made of:
> ----
>    c : array, shape (M, M)
>        Upper- or lower-triangular Cholesky factor of A
>    lower : array, shape (M, M)
>        Flag indicating whether the factor is lower or upper triangular
> ----
> But the following simple example shows that 'c' is not triangular
> and the result of cho_factor() is pretty different from that of
> cholesky()!
> ----
> import numpy as N
> import scipy.linalg as SL
> A = N.array([[2,1],[1,2]])
> c_wrong,lower = SL.cho_factor(A)
> print c_wrong
> c_correct = SL.cholesky(A)
> print c_correct
> print c_wrong==c_correct
> ----
> The output is:
> ----
> [[ 1.41421356  0.70710678]
>  [ 1.          1.22474487]]
> [[ 1.41421356  0.70710678]
>  [ 0.          1.22474487]]
> [[ True  True]
>  [False  True]]
> ----
> Why c_wrong[1,0] is not zero? And why is it 1.0 ?
>
> I went mad tracking this issue in my code 8-|
>
> Regards,
>
> Emanuele
>
> P.S.: there is a type in cho_factor documentation: 'lower' is
> not an 'array' but an 'int'.
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080719/0cb6493f/attachment.html>

From warren.weckesser at gmail.com  Sat Jul 19 13:17:48 2008
From: warren.weckesser at gmail.com (Warren Weckesser)
Date: Sat, 19 Jul 2008 13:17:48 -0400
Subject: [SciPy-user] bug in cho_factor?
In-Reply-To: <4882085A.2040107@relativita.com>
References: <4882085A.2040107@relativita.com>
Message-ID: <114880320807191017j4684602ak9cbe80439d82d5f@mail.gmail.com>

Emanuele,

After a closer look at the code, it appears that the only difference
between cholesky() and cho_factor() is that cho_factor() does not
set the irrelevant terms to zero.  So, if you are only going to
use the result to call cho_solve(), you can use cho_factor(), but
if you need the correct square triangular matrix with zeros in the
subdiagonal, use cholesky().

Warren


On Sat, Jul 19, 2008 at 11:29 AM, Emanuele Olivetti <emanuele at relativita.com>
wrote:

> In scipy v0.7.0.dev4543 (svn) the documentation of
> scipy.linalg.cho_factor says that cho_factor returns a tuple
> made of:
> ----
>    c : array, shape (M, M)
>        Upper- or lower-triangular Cholesky factor of A
>    lower : array, shape (M, M)
>        Flag indicating whether the factor is lower or upper triangular
> ----
> But the following simple example shows that 'c' is not triangular
> and the result of cho_factor() is pretty different from that of
> cholesky()!
> ----
> import numpy as N
> import scipy.linalg as SL
> A = N.array([[2,1],[1,2]])
> c_wrong,lower = SL.cho_factor(A)
> print c_wrong
> c_correct = SL.cholesky(A)
> print c_correct
> print c_wrong==c_correct
> ----
> The output is:
> ----
> [[ 1.41421356  0.70710678]
>  [ 1.          1.22474487]]
> [[ 1.41421356  0.70710678]
>  [ 0.          1.22474487]]
> [[ True  True]
>  [False  True]]
> ----
> Why c_wrong[1,0] is not zero? And why is it 1.0 ?
>
> I went mad tracking this issue in my code 8-|
>
> Regards,
>
> Emanuele
>
> P.S.: there is a type in cho_factor documentation: 'lower' is
> not an 'array' but an 'int'.
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080719/44e380d0/attachment.html>

From emanuele at relativita.com  Sat Jul 19 13:50:47 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Sat, 19 Jul 2008 19:50:47 +0200
Subject: [SciPy-user] bug in cho_factor?
In-Reply-To: <114880320807191017j4684602ak9cbe80439d82d5f@mail.gmail.com>
References: <4882085A.2040107@relativita.com>
	<114880320807191017j4684602ak9cbe80439d82d5f@mail.gmail.com>
Message-ID: <48822977.80601@relativita.com>

Thanks a lot for the explanation. I believe cho_factor's docstring
should be updated in order to mention these facts. It is definitely
unexpected that the result of the two decompositions are different
and can cause problems like I had (a couple of hours spent). A
clear "Warning" should fit. Consider that U,lower=cho_factor(A)
outputs an U that does not satisfy A==N.dot(U.T,U) !!

Do you know why cho_factor does not zeros out the matrix?
Is it for performance reasons? Otherwise there are no reasons
for using cho_factor() instead of cholesky().

Best Regards,

Emanuele


Warren Weckesser wrote:
> Emanuele,
>
> After a closer look at the code, it appears that the only difference
> between cholesky() and cho_factor() is that cho_factor() does not
> set the irrelevant terms to zero.  So, if you are only going to
> use the result to call cho_solve(), you can use cho_factor(), but
> if you need the correct square triangular matrix with zeros in the
> subdiagonal, use cholesky().
>
> Warren
>
>
> On Sat, Jul 19, 2008 at 11:29 AM, Emanuele Olivetti
> <emanuele at relativita.com <mailto:emanuele at relativita.com>> wrote:
>
>     In scipy v0.7.0.dev4543 (svn) the documentation of
>     scipy.linalg.cho_factor says that cho_factor returns a tuple
>     made of:
>     ----
>        c : array, shape (M, M)
>            Upper- or lower-triangular Cholesky factor of A
>        lower : array, shape (M, M)
>            Flag indicating whether the factor is lower or upper triangular
>     ----
>     But the following simple example shows that 'c' is not triangular
>     and the result of cho_factor() is pretty different from that of
>     cholesky()!
>     ----
>     import numpy as N
>     import scipy.linalg as SL
>     A = N.array([[2,1],[1,2]])
>     c_wrong,lower = SL.cho_factor(A)
>     print c_wrong
>     c_correct = SL.cholesky(A)
>     print c_correct
>     print c_wrong==c_correct
>     ----
>     The output is:
>     ----
>     [[ 1.41421356  0.70710678]
>      [ 1.          1.22474487]]
>     [[ 1.41421356  0.70710678]
>      [ 0.          1.22474487]]
>     [[ True  True]
>      [False  True]]
>     ----
>     Why c_wrong[1,0] is not zero? And why is it 1.0 ?
>
>     I went mad tracking this issue in my code 8-|
>
>     Regards,
>
>     Emanuele
>
>     P.S.: there is a type in cho_factor documentation: 'lower' is
>     not an 'array' but an 'int'.
>
>     _______________________________________________
>     SciPy-user mailing list
>     SciPy-user at scipy.org <mailto:SciPy-user at scipy.org>
>     http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>   



From warren.weckesser at gmail.com  Sat Jul 19 14:24:29 2008
From: warren.weckesser at gmail.com (Warren Weckesser)
Date: Sat, 19 Jul 2008 14:24:29 -0400
Subject: [SciPy-user] bug in cho_factor?
In-Reply-To: <48822977.80601@relativita.com>
References: <4882085A.2040107@relativita.com>
	<114880320807191017j4684602ak9cbe80439d82d5f@mail.gmail.com>
	<48822977.80601@relativita.com>
Message-ID: <114880320807191124q453b288y2ab29f11d3def6c7@mail.gmail.com>

Hi Emanuele,

On Sat, Jul 19, 2008 at 1:50 PM, Emanuele Olivetti <emanuele at relativita.com>
wrote:

> Thanks a lot for the explanation. I believe cho_factor's docstring
> should be updated in order to mention these facts. It is definitely
> unexpected that the result of the two decompositions are different
> and can cause problems like I had (a couple of hours spent). A
> clear "Warning" should fit. Consider that U,lower=cho_factor(A)
> outputs an U that does not satisfy A==N.dot(U.T,U) !!


I agree--the docstring description of the return matrix c is wrong.
c is a matrix whose upper or lower (depending on the parameter
lower) triangular part gives the Cholesky factor, but c itself is not
triangular.


> Do you know why cho_factor does not zeros out the matrix?
> Is it for performance reasons?


That would be my guess, and it makes sense (why zero out elements
that will be ignored by cho_solve()?), but you'd have to ask the author of
the
code to be sure.

Warren
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080719/9e0dd849/attachment.html>

From emanuele at relativita.com  Sat Jul 19 17:25:27 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Sat, 19 Jul 2008 23:25:27 +0200
Subject: [SciPy-user] bug in cho_factor?
In-Reply-To: <114880320807191124q453b288y2ab29f11d3def6c7@mail.gmail.com>
References: <4882085A.2040107@relativita.com>	<114880320807191017j4684602ak9cbe80439d82d5f@mail.gmail.com>	<48822977.80601@relativita.com>
	<114880320807191124q453b288y2ab29f11d3def6c7@mail.gmail.com>
Message-ID: <48825BC7.5090904@relativita.com>

In the meanwhile I've submitted ticket #704 about this issue and
attached a tentative patch for the docstring.

I hope someone will review, improve it and commit.

Regards,

Emanuele

Warren Weckesser wrote:
> Hi Emanuele,
>
> On Sat, Jul 19, 2008 at 1:50 PM, Emanuele Olivetti
> <emanuele at relativita.com <mailto:emanuele at relativita.com>> wrote:
>
>     Thanks a lot for the explanation. I believe cho_factor's docstring
>     should be updated in order to mention these facts. It is definitely
>     unexpected that the result of the two decompositions are different
>     and can cause problems like I had (a couple of hours spent). A
>     clear "Warning" should fit. Consider that U,lower=cho_factor(A)
>     outputs an U that does not satisfy A==N.dot(U.T,U) !!
>
>
> I agree--the docstring description of the return matrix c is wrong.
> c is a matrix whose upper or lower (depending on the parameter
> lower) triangular part gives the Cholesky factor, but c itself is not
> triangular.
>  
>
>     Do you know why cho_factor does not zeros out the matrix?
>     Is it for performance reasons?
>
>
> That would be my guess, and it makes sense (why zero out elements
> that will be ignored by cho_solve()?), but you'd have to ask the
> author of the
> code to be sure.
>
> Warren
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>   



From rpmuller at gmail.com  Sun Jul 20 14:24:43 2008
From: rpmuller at gmail.com (Rick Muller)
Date: Sun, 20 Jul 2008 12:24:43 -0600
Subject: [SciPy-user] Problems while trying to build scipy with umfpack on
	Intel MacBook Pro
Message-ID: <e8a72dc40807201124p669c2544sf40baa9bcbabb218@mail.gmail.com>

I'm trying to get a scipy build that contains umfpack on an Intel MacBook
Pro. I believe that I've compiled and installed umfpack correctly.


I appended the output I get from the build to this message. In short, I get
some warnings about having multiply defined "-arch" flags, but I'm not clear
whether this is the fatal error or not. At the end of the install process, I
have what looks like a umfpack directory:

$ ls  /Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/
site-packages/scipy/linsolve/umfpack/
__init__.py     _umfpack.py     info.pyc        tests/
__init__.pyc    _umfpack.pyc    setup.py        umfpack.py
__umfpack.so    info.py         setup.pyc       umfpack.pyc

But for some reason I can't import it:

s886301{rmuller}[1]: import scipy.linsolve.umfpack as um
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call
last)

/Users/rmuller/<ipython console> in <module>()
AttributeError: 'module' object has no attribute 'umfpack'


Any suggestions?


mkl_info:
  libraries mkl,vml,guide not found in
/Library/Frameworks/Python.framework/Versions/2.5/lib
  libraries mkl,vml,guide not found in /usr/local/lib
  libraries mkl,vml,guide not found in /usr/lib
  NOT AVAILABLE

fftw3_info:
  libraries fftw3 not found in
/Library/Frameworks/Python.framework/Versions/2.5/lib
  libraries fftw3 not found in /usr/local/lib
  libraries fftw3 not found in /usr/lib
  fftw3 not found
  NOT AVAILABLE

fftw2_info:
  libraries rfftw,fftw not found in
/Library/Frameworks/Python.framework/Versions/2.5/lib
  libraries rfftw,fftw not found in /usr/local/lib
  libraries rfftw,fftw not found in /usr/lib
  fftw2 not found
  NOT AVAILABLE

dfftw_info:
  libraries drfftw,dfftw not found in
/Library/Frameworks/Python.framework/Versions/2.5/lib
  libraries drfftw,dfftw not found in /usr/local/lib
  libraries drfftw,dfftw not found in /usr/lib
  dfftw not found
  NOT AVAILABLE

djbfft_info:
  NOT AVAILABLE

blas_opt_info:
  FOUND:
    extra_link_args = ['-Wl,-framework', '-Wl,Accelerate']
    define_macros = [('NO_ATLAS_INFO', 3)]
    extra_compile_args = ['-faltivec',
'-I/System/Library/Frameworks/vecLib.framework/Headers']

lapack_opt_info:
  FOUND:
    extra_link_args = ['-Wl,-framework', '-Wl,Accelerate']
    define_macros = [('NO_ATLAS_INFO', 3)]
    extra_compile_args = ['-faltivec']

non-existing path in 'scipy/linsolve': 'tests'
umfpack_info:
  libraries umfpack not found in
/Library/Frameworks/Python.framework/Versions/2.5/lib
amd_info:
  libraries amd not found in
/Library/Frameworks/Python.framework/Versions/2.5/lib
  libraries amd not found in /usr/local/lib
  libraries amd not found in /usr/lib
  NOT AVAILABLE

  FOUND:
    libraries = ['umfpack']
    library_dirs = ['/usr/local/lib']
    swig_opts = ['-I/usr/local/include']
    define_macros = [('SCIPY_UMFPACK_H', None)]
    include_dirs = ['/usr/local/include']

running build
running scons
customize UnixCCompiler
Found executable /usr/bin/gcc
customize NAGFCompiler
Could not locate executable f95
customize AbsoftFCompiler
Could not locate executable f90
Could not locate executable f77
customize IBMFCompiler
Could not locate executable xlf90
Could not locate executable xlf
customize IntelFCompiler
Could not locate executable ifort
Could not locate executable ifc
customize GnuFCompiler
Could not locate executable g77
customize Gnu95FCompiler
Found executable /usr/local/bin/gfortran
customize Gnu95FCompiler
customize UnixCCompiler
customize UnixCCompiler using scons
running config_cc
unifing config_cc, config, build_clib, build_ext, build commands --compiler
options
running config_fc
unifing config_fc, config, build_clib, build_ext, build commands --fcompiler
options
running build_src
building py_modules sources
building library "dfftpack" sources
building library "linpack_lite" sources
building library "mach" sources
building library "quadpack" sources
building library "odepack" sources
building library "fitpack" sources
building library "superlu_src" sources
building library "odrpack" sources
building library "minpack" sources
building library "rootfind" sources
building library "c_misc" sources
building library "cephes" sources
building library "mach" sources
building library "toms" sources
building library "amos" sources
building library "cdf" sources
building library "specfun" sources
building library "statlib" sources
building extension "scipy.cluster._vq" sources
building extension "scipy.fftpack._fftpack" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.fftpack.convolve" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.integrate._quadpack" sources
building extension "scipy.integrate._odepack" sources
building extension "scipy.integrate.vode" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.interpolate._fitpack" sources
building extension "scipy.interpolate.dfitpack" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
  adding
'build/src.macosx-10.3-fat-2.5/scipy/interpolate/dfitpack-f2pywrappers.f' to
sources.
building extension "scipy.io.numpyio" sources
building extension "scipy.lib.blas.fblas" sources
f2py options: ['skip:', ':']
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
  adding
'build/src.macosx-10.3-fat-2.5/build/src.macosx-10.3-fat-2.5/scipy/lib/blas/fblas-f2pywrappers.f'
to sources.
building extension "scipy.lib.blas.cblas" sources
  adding 'build/src.macosx-10.3-fat-2.5/scipy/lib/blas/cblas.pyf' to
sources.
f2py options: ['skip:', ':']
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.lib.lapack.flapack" sources
f2py options: ['skip:', ':']
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.lib.lapack.clapack" sources
  adding 'build/src.macosx-10.3-fat-2.5/scipy/lib/lapack/clapack.pyf' to
sources.
f2py options: ['skip:', ':']
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.lib.lapack.calc_lwork" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.lib.lapack.atlas_version" sources
building extension "scipy.linalg.fblas" sources
  adding 'build/src.macosx-10.3-fat-2.5/scipy/linalg/fblas.pyf' to sources.
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
  adding
'build/src.macosx-10.3-fat-2.5/build/src.macosx-10.3-fat-2.5/scipy/linalg/fblas-f2pywrappers.f'
to sources.
building extension "scipy.linalg.cblas" sources
  adding 'build/src.macosx-10.3-fat-2.5/scipy/linalg/cblas.pyf' to sources.
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.linalg.flapack" sources
  adding 'build/src.macosx-10.3-fat-2.5/scipy/linalg/flapack.pyf' to
sources.
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
  adding
'build/src.macosx-10.3-fat-2.5/build/src.macosx-10.3-fat-2.5/scipy/linalg/flapack-f2pywrappers.f'
to sources.
building extension "scipy.linalg.clapack" sources
  adding 'build/src.macosx-10.3-fat-2.5/scipy/linalg/clapack.pyf' to
sources.
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.linalg._flinalg" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.linalg.calc_lwork" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.linalg.atlas_version" sources
building extension "scipy.linalg._iterative" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.linsolve._zsuperlu" sources
building extension "scipy.linsolve._dsuperlu" sources
building extension "scipy.linsolve._csuperlu" sources
building extension "scipy.linsolve._ssuperlu" sources
building extension "scipy.linsolve.umfpack.__umfpack" sources
  adding 'scipy/linsolve/umfpack/umfpack.i' to sources.
building extension "scipy.odr.__odrpack" sources
building extension "scipy.optimize._minpack" sources
building extension "scipy.optimize._zeros" sources
building extension "scipy.optimize._lbfgsb" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.optimize.moduleTNC" sources
building extension "scipy.optimize._cobyla" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.optimize.minpack2" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.signal.sigtools" sources
building extension "scipy.signal.spline" sources
building extension "scipy.sparse._sparsetools" sources
building extension "scipy.special._cephes" sources
building extension "scipy.special.specfun" sources
f2py options: ['--no-wrap-functions']
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.stats.statlib" sources
f2py options: ['--no-wrap-functions']
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.stats.futil" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
building extension "scipy.stats.mvn" sources
f2py options: []
  adding 'build/src.macosx-10.3-fat-2.5/fortranobject.c' to sources.
  adding 'build/src.macosx-10.3-fat-2.5' to include_dirs.
  adding 'build/src.macosx-10.3-fat-2.5/scipy/stats/mvn-f2pywrappers.f' to
sources.
building extension "scipy.ndimage._nd_image" sources
building data_files sources
running build_py
copying build/src.macosx-10.3-fat-2.5/scipy/__config__.py ->
build/lib.macosx-10.3-fat-2.5/scipy
creating build/lib.macosx-10.3-fat-2.5/scipy/linsolve/umfpack
copying scipy/linsolve/umfpack/__init__.py ->
build/lib.macosx-10.3-fat-2.5/scipy/linsolve/umfpack
copying scipy/linsolve/umfpack/info.py ->
build/lib.macosx-10.3-fat-2.5/scipy/linsolve/umfpack
copying scipy/linsolve/umfpack/setup.py ->
build/lib.macosx-10.3-fat-2.5/scipy/linsolve/umfpack
copying scipy/linsolve/umfpack/umfpack.py ->
build/lib.macosx-10.3-fat-2.5/scipy/linsolve/umfpack
copying build/src.macosx-10.3-fat-2.5/scipy/linsolve/umfpack/_umfpack.py ->
build/lib.macosx-10.3-fat-2.5/scipy/linsolve/umfpack
running build_clib
customize UnixCCompiler
customize UnixCCompiler using build_clib
customize NAGFCompiler
customize AbsoftFCompiler
customize IBMFCompiler
customize IntelFCompiler
customize GnuFCompiler
customize Gnu95FCompiler
customize Gnu95FCompiler
customize Gnu95FCompiler using build_clib
running build_ext
customize UnixCCompiler
customize UnixCCompiler using build_ext
library 'mach' defined more than once, overwriting build_info
{'sources': ['scipy/integrate/mach/d1mach.f',
'scipy/integrate/mach/i1mach.f', 'scipy/integrate/mach/r1mach.f',
'scipy/integrate/mach/xerror.f'], 'config_fc': {'noopt':
('scipy/integrate/setup.pyc', 1)}, 'source_languages': ['f77']}...
with
{'sources': ['scipy/special/mach/d1mach.f', 'scipy/special/mach/i1mach.f',
'scipy/special/mach/r1mach.f', 'scipy/special/mach/xerror.f'], 'config_fc':
{'noopt': ('scipy/special/setup.pyc', 1)}, 'source_languages': ['f77']}...
extending extension 'scipy.linsolve._zsuperlu' defined_macros with
[('USE_VENDOR_BLAS', 1)]
extending extension 'scipy.linsolve._dsuperlu' defined_macros with
[('USE_VENDOR_BLAS', 1)]
extending extension 'scipy.linsolve._csuperlu' defined_macros with
[('USE_VENDOR_BLAS', 1)]
extending extension 'scipy.linsolve._ssuperlu' defined_macros with
[('USE_VENDOR_BLAS', 1)]
customize UnixCCompiler
customize UnixCCompiler using build_ext
customize NAGFCompiler
customize AbsoftFCompiler
customize IBMFCompiler
customize IntelFCompiler
customize GnuFCompiler
customize Gnu95FCompiler
customize Gnu95FCompiler
customize Gnu95FCompiler using build_ext
building 'scipy.linsolve.umfpack.__umfpack' extension
compiling C sources
C compiler: gcc -arch ppc -arch i386 -isysroot
/Developer/SDKs/MacOSX10.4u.sdk -fno-strict-aliasing -Wno-long-double
-no-cpp-precomp -mno-fused-madd -fno-common -dynamic -DNDEBUG -g -O3

compile options: '-DSCIPY_UMFPACK_H -DNO_ATLAS_INFO=3 -I/usr/local/include
-I/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/numpy/core/include
-I/Library/Frameworks/Python.framework/Versions/2.5/include/python2.5 -c'
extra options: '-faltivec
-I/System/Library/Frameworks/vecLib.framework/Headers'
gcc -arch i386 -arch ppc -isysroot /Developer/SDKs/MacOSX10.4u.sdk -g
-bundle -undefined dynamic_lookup
build/temp.macosx-10.3-fat-2.5/build/src.macosx-10.3-fat-2.5/scipy/linsolve/umfpack/_umfpack_wrap.o
-L/usr/local/lib -Lbuild/temp.macosx-10.3-fat-2.5 -lumfpack -o
build/lib.macosx-10.3-fat-2.5/scipy/linsolve/umfpack/__umfpack.so
-Wl,-framework -Wl,Accelerate
/usr/bin/ld: for architecture ppc
/usr/bin/ld: warning /usr/local/lib/libumfpack.a archive's cputype (7,
architecture i386) does not match cputype (18) for specified -arch flag: ppc
(can't load from it)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080720/62474842/attachment.html>

From rpmuller at gmail.com  Sun Jul 20 17:26:22 2008
From: rpmuller at gmail.com (Rick Muller)
Date: Sun, 20 Jul 2008 15:26:22 -0600
Subject: [SciPy-user] Problems while trying to build scipy with umfpack
	on Intel MacBook Pro
In-Reply-To: <e8a72dc40807201124p669c2544sf40baa9bcbabb218@mail.gmail.com>
References: <e8a72dc40807201124p669c2544sf40baa9bcbabb218@mail.gmail.com>
Message-ID: <e8a72dc40807201426w2899bb19j59497f76c40d6ae1@mail.gmail.com>

On Sun, Jul 20, 2008 at 12:24 PM, Rick Muller <rpmuller at gmail.com> wrote:

> I'm trying to get a scipy build that contains umfpack on an Intel MacBook
> Pro. I believe that I've compiled and installed umfpack correctly.
>
>
> I appended the output I get from the build to this message. In short, I get
> some warnings about having multiply defined "-arch" flags, but I'm not clear
> whether this is the fatal error or not. At the end of the install process, I
> have what looks like a umfpack directory:
>

I'm replying to my own message to give a little more information, just in
case it helps at all. Oddly enough, if I chdir to the relevant place, I can
import umfpack:

$ cd
/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/site-packages/scipy/linsolve/umfpack/
s886301{umfpack}508$ python
Python 2.5 (r25:51918, Sep 19 2006, 08:49:13)
[GCC 4.0.1 (Apple Computer, Inc. build 5341)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import umfpack as um

Does this make sense to anyone?

Thanks in advance for any help.

Rick
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080720/9e026eb4/attachment.html>

From didier.rano at gmail.com  Sun Jul 20 22:56:14 2008
From: didier.rano at gmail.com (didier rano)
Date: Sun, 20 Jul 2008 22:56:14 -0400
Subject: [SciPy-user] Statistics advise with scipy
Message-ID: <da09baf80807201956r6f92db05vd5262a4236bc3ae2@mail.gmail.com>

Hi all,
I am using the TimeSeries module with scipy for several months. I am very
impressive by this module. It is very easy to use.

I have a problem to filter only relevant data. Sometimes, my data retrieved
are not in good quality (very too high value or very too low value). I've
tried to use median filter, but data could be irrelevant during a long
period. I know that the pattern "strong change in a short time" is not
possible normally. Then, how to filter it correctly ?

You may see the data in
http://spreadsheets.google.com/pub?key=pF2qPjwUpy_-1m7FbqafDXQ

Thanks

Didier Rano
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080720/8eb8f7bd/attachment.html>

From qing at qpeng.org  Mon Jul 21 04:43:30 2008
From: qing at qpeng.org (Qing Peng)
Date: Mon, 21 Jul 2008 01:43:30 -0700
Subject: [SciPy-user] Mode "reflect" may yield incorrect results on
	boundaries
Message-ID: <be020a860807210143v1e3036e8r7a56c9ca0f668fdb@mail.gmail.com>

Hi there,

I carefully followed the instructions to install numpy-1.1.0 and
scipy-0.6.0, after lapack,atlas,amd,umfpack,fftw3 installation. The cite.cfg
is the following.
[DEFAULT]
library_dirs =
/home/users3/cmt/qing/usr/local/lib:/home/users3/cmt/qing/usr/local/gcc-4.3.1/lib64
include_dirs =
/home/users3/cmt/qing/usr/local/include:/home/users3/cmt/qing/usr/local/gcc-4.3.1/include


[atlas]
library_dirs = /home/users3/cmt/qing/usr/local/lib
atlas_libs = lapack, f77blas, cblas, atlas

[amd]
library_dirs =
/home/users3/cmt/qing/usr/local/lib:/home/users3/cmt/qing/usr/local/gcc-4.3.1/lib64
include_dirs =
/home/users3/cmt/qing/usr/local/include:/home/users3/cmt/qing/usr/local/gcc-4.3.1/include
amd_libs = amd

[umfpack]
library_dirs =
/home/users3/cmt/qing/usr/local/lib:/home/users3/cmt/qing/usr/local/gcc-4.3.1/lib64
include_dirs =
/home/users3/cmt/qing/usr/local/include:/home/users3/cmt/qing/usr/local/gcc-4.3.1/include
umfpack_libs = umfpack, gfortran

[fftw3]
library_dirs = /home/users3/cmt/qing/usr/local/fftw-3.1.2/lib
include_dirs = /home/users3/cmt/qing/usr/local/fftw-3.1.2/include
fftw3_libs = fftw3

"python setup.py install"  successfully install numpy and scipy.
But when I run the test: it gives:
-bash-2.05b$   python -c 'import numpy; import scipy; scipy.test(1)'
Failed importing scipy.linsolve.umfpack: 'module' object has no attribute
'umfpack'
  Found 9/9 tests for scipy.cluster.tests.test_vq
  Found 18/18 tests for scipy.fftpack.tests.test_basic
  Found 4/4 tests for scipy.fftpack.tests.test_helper
  Found 20/20 tests for scipy.fftpack.tests.test_pseudo_diffs
  Found 1/1 tests for scipy.integrate.tests.test_integrate
  Found 3/3 tests for scipy.integrate.tests.test_quadrature
  Found 10/10 tests for scipy.integrate.tests.test_quadpack
  Found 6/6 tests for scipy.tests.test_fitpack
  Found 6/6 tests for scipy.tests.test_interpolate
  Found 28/28 tests for scipy.io.tests.test_mio
  Found 5/5 tests for scipy.io.tests.test_npfile
  Found 4/4 tests for scipy.io.tests.test_array_import
  Found 4/4 tests for scipy.io.tests.test_recaster
  Found 13/13 tests for scipy.io.tests.test_mmio
  Found 128/128 tests for scipy.lib.blas.tests.test_fblas
  Found 16/16 tests for scipy.lib.blas.tests.test_blas
  Found 42/42 tests for scipy.lib.lapack.tests.test_lapack
  Found 128/128 tests for scipy.linalg.tests.test_fblas
  Found 6/6 tests for scipy.linalg.tests.test_iterative
  Found 72/72 tests for scipy.linalg.tests.test_decomp
  Found 4/4 tests for scipy.linalg.tests.test_lapack
  Found 41/41 tests for scipy.linalg.tests.test_basic
  Found 7/7 tests for scipy.linalg.tests.test_matfuncs
<module 'scipy.linalg.fblas' from
'/home/users3/cmt/qing/usr/local/python/lib/python2.5/site-packages/scipy/linalg/fblas.so'>
  Found 16/16 tests for scipy.linalg.tests.test_blas
Failed importing
/home/users3/cmt/qing/usr/local/python/lib/python2.5/site-packages/scipy/linsolve/umfpack/tests/test_umfpack.py:
'module' object has no attribute 'umfpack'
  Found 2/2 tests for scipy.maxentropy.tests.test_maxentropy
Failed importing
/home/users3/cmt/qing/usr/local/python/lib/python2.5/site-packages/scipy/misc/tests/test_pilutil.py:
No module named PIL.Image
  Found 399/399 tests for scipy.ndimage.tests.test_ndimage
  Found 5/5 tests for scipy.odr.tests.test_odr
  Found 8/8 tests for scipy.optimize.tests.test_optimize
  Found 4/4 tests for scipy.optimize.tests.test_zeros
  Found 1/1 tests for scipy.optimize.tests.test_cobyla
  Found 10/10 tests for scipy.optimize.tests.test_nonlin
  Found 5/5 tests for scipy.signal.tests.test_signaltools
  Found 4/4 tests for scipy.signal.tests.test_wavelets
  Found 152/152 tests for scipy.sparse.tests.test_sparse
  Found 3/3 tests for scipy.special.tests.test_spfun_stats
  Found 342/342 tests for scipy.special.tests.test_basic
  Found 107/107 tests for scipy.stats.tests.test_stats
  Found 10/10 tests for scipy.stats.tests.test_morestats
  Found 73/73 tests for scipy.stats.tests.test_distributions
  Found 9/9 tests for scipy.weave.tests.test_build_tools
  Found 0/0 tests for scipy.weave.tests.test_scxx_dict
Failed importing
/home/users3/cmt/qing/usr/local/python/lib/python2.5/site-packages/scipy/weave/tests/test_wx_spec.py:
Could not locate wxPython base directory.
  Found 3/3 tests for scipy.weave.tests.test_standard_array_spec
  Found 0/0 tests for scipy.weave.tests.test_inline_tools
  Found 0/0 tests for scipy.weave.tests.test_c_spec
  Found 1/1 tests for scipy.weave.tests.test_ast_tools
  Found 0/0 tests for scipy.weave.tests.test_scxx_sequence
  Found 16/16 tests for scipy.weave.tests.test_slice_handler
  Found 0/0 tests for scipy.weave.tests.test_scxx_object
  Found 2/2 tests for scipy.weave.tests.test_blitz_tools
building extensions here: /home/users3/cmt/qing/.python25_compiled/m5
  Found 1/1 tests for scipy.weave.tests.test_ext_tools
  Found 26/26 tests for scipy.weave.tests.test_catalog
  Found 74/74 tests for scipy.weave.tests.test_size_check
.../home/users3/cmt/qing/usr/local/python/lib/python2.5/site-packages/scipy/cluster/vq.py:477:
UserWarning: One of the clusters is empty. Re-run kmean with a different
initialization.
  warnings.warn("One of the clusters is empty. "
exception raised as expected: One of the clusters is empty. Re-run kmean
with a different initialization.
................................................Residual: 1.05006987327e-07
............../home/users3/cmt/qing/usr/local/python/lib/python2.5/site-packages/scipy/interpolate/fitpack2.py:458:
UserWarning:
The coefficients of the spline returned have been computed as the
minimal norm least-squares solution of a (numerically) rank deficient
system (deficiency=7). If deficiency is large, the results may be
inaccurate. Deficiency may strongly depend on the value of eps.
  warnings.warn(message)
.............................................
Don't worry about a warning regarding the number of bytes read.
Warning: 1000000 bytes requested, 20 bytes read.
........................caxpy:n=4
..caxpy:n=3
....ccopy:n=4
..ccopy:n=3
.............cscal:n=4
....cswap:n=4
..cswap:n=3
.....daxpy:n=4
..daxpy:n=3
....dcopy:n=4
..dcopy:n=3
.............dscal:n=4
....dswap:n=4
..dswap:n=3
.....saxpy:n=4
..saxpy:n=3
....scopy:n=4
..scopy:n=3
.............sscal:n=4
....sswap:n=4
..sswap:n=3
.....zaxpy:n=4
..zaxpy:n=3
....zcopy:n=4
..zcopy:n=3
.............zscal:n=4
....zswap:n=4
..zswap:n=3
............................................FF.................caxpy:n=4
..caxpy:n=3
....ccopy:n=4
..ccopy:n=3
.............cscal:n=4
....cswap:n=4
..cswap:n=3
.....daxpy:n=4
..daxpy:n=3
....dcopy:n=4
..dcopy:n=3
.............dscal:n=4
....dswap:n=4
..dswap:n=3
.....saxpy:n=4
..saxpy:n=3
....scopy:n=4
..scopy:n=3
.............sscal:n=4
....sswap:n=4
..sswap:n=3
.....zaxpy:n=4
..zaxpy:n=3
....zcopy:n=4
..zcopy:n=3
.............zscal:n=4
....zswap:n=4
..zswap:n=3
..............................................................................................................................Result
may be inaccurate, approximate err = 2.53191024037e-09
...Result may be inaccurate, approximate err = 3.85396349383e-12
............................................................................................................................/home/users3/cmt/qing/usr/local/python/lib/python2.5/site-packages/scipy/ndimage/interpolation.py:41:
UserWarning: Mode "reflect" may yield incorrect results on boundaries.
Please use "mirror" instead.
  warnings.warn('Mode "reflect" may yield incorrect results on '
.............................................................................................Illegal
instruction

After google with the key words: 'Mode "reflect" may yield incorrect results
on boundaries', it looks like a bug, which said been fixed in year 2007. I
just wondering where I go wrong.

Thanks

-- Qing
=============
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080721/84eb62a9/attachment.html>

From mwojc at p.lodz.pl  Mon Jul 21 05:13:53 2008
From: mwojc at p.lodz.pl (Marek Wojciechowski)
Date: Mon, 21 Jul 2008 11:13:53 +0200
Subject: [SciPy-user] optimize.fmin_tnc - how to catch its output?
Message-ID: <g61jl9$hq0$1@ger.gmane.org>

Hallo!
I tried to catch fmin_tnc output to a file in the following way:

    file = open('messages', 'w')
    stdout = sys.stdout
    stderr = sys.stderr
    sys.stdout = file
    sys.stderr = file
    res = optimize.fmin_tnc(..., messages=1)
    file.close()
    sys.stdout = stdout
    sys.stderr = stderr

But this does not work. I realized that fmin_tnc calls C routine which
probably uses its own stdout, stderr. Is there a way to catch its output? 

Greetings,
-- 
Marek Wojciechowski



From emanuele at relativita.com  Mon Jul 21 05:24:56 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Mon, 21 Jul 2008 11:24:56 +0200
Subject: [SciPy-user] fastest 'solve' for triangular matrix?
Message-ID: <488455E8.60900@relativita.com>

Dear All,

Is there a better/faster way (provided by scipy) to solve a linear system
Lx=b (when L is lower-triangular) then the following?

piv = N.arange(L.shape[0])
x = scipy.linalg.lu_solve((L,piv),b)

The linear system is trivial to solve but using 'solve' is much much slower
and I was no t able to find something ad-hoc for triangular matrices.

Best,

Emanuele




From emanuele at relativita.com  Mon Jul 21 06:12:47 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Mon, 21 Jul 2008 12:12:47 +0200
Subject: [SciPy-user] fastest 'solve' for triangular matrix?
In-Reply-To: <488455E8.60900@relativita.com>
References: <488455E8.60900@relativita.com>
Message-ID: <4884611F.5020106@relativita.com>

Oops, the previous code was incorrect (it referred to the upper-triangular
case, not the lower-triangular). Here is the correct version:

# L = lower-triangular matrix
piv = N.arange(L.shape[0])
x = scipy.linalg.lu_solve((L.T,piv),b,trans=1)


And so in case of upper-triangular matrix, Ux=b, it is:

# U = upper-triangular matrix
piv = N.arange(L.shape[0])
x = scipy.linalg.lu_solve((U,piv),b,trans=0)

Emanuele

P.S.: for some reasons the 'upper' case is consistently faster
than the 'lower' one...

Emanuele Olivetti wrote:
> Dear All,
>
> Is there a better/faster way (provided by scipy) to solve a linear system
> Lx=b (when L is lower-triangular) then the following?
>
> piv = N.arange(L.shape[0])
> x = scipy.linalg.lu_solve((L,piv),b)
>
> The linear system is trivial to solve but using 'solve' is much much slower
> and I was no t able to find something ad-hoc for triangular matrices.
>
> Best,
>
> Emanuele
>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



From emanuele at relativita.com  Mon Jul 21 08:22:48 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Mon, 21 Jul 2008 14:22:48 +0200
Subject: [SciPy-user] [OpenOpt] evaluation of f(x) and df(x)
Message-ID: <48847F98.7030600@relativita.com>

Dear All and Dmitrey,

in my code the evaluation of f(x) and df(x) shares many
intermediate steps. I'd like to re-use what is computed
inside f(x) to evaluate df(x) more efficiently, during f(x)
optimization. Then is it _always_ true that, when OpenOpt
evaluates df(x) at a certain x=x^*, f(x) too was previously
evaluated at x=x^*? And in case f(x) was evaluated multiple
times before evaluating df(x), is it true that the last x at
which f(x) was evaluated (before computing df(x=x^*))
was x=x^*?

If these assumptions holds (as it seems from preliminary
tests on NLP using ralg), the extra code to take advantage
of this fact is extremely simple.

Best,

Emanuele

P.S.: if the previous assumptions are false in general, I'd
like to know it they are true at least for the NLP case.



From dmitrey.kroshko at scipy.org  Mon Jul 21 08:35:48 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Mon, 21 Jul 2008 15:35:48 +0300
Subject: [SciPy-user] [OpenOpt] evaluation of f(x) and df(x)
In-Reply-To: <48847F98.7030600@relativita.com>
References: <48847F98.7030600@relativita.com>
Message-ID: <488482A4.5060601@scipy.org>

Hi Emanuele,

if df(x1) is obtained via finite-difference calculations then f(x1) is 
stored and compared during next call to f / df, and vice versa: if f(x1) 
is called then the value obtained is stored and compared during next 
call to f and/or finite-difference df.

At least it is intended so, I can take more precise look if you have 
noticed it doesn't work properly.

Regards, D.

Emanuele Olivetti wrote:
> Dear All and Dmitrey,
>
> in my code the evaluation of f(x) and df(x) shares many
> intermediate steps. I'd like to re-use what is computed
> inside f(x) to evaluate df(x) more efficiently, during f(x)
> optimization. Then is it _always_ true that, when OpenOpt
> evaluates df(x) at a certain x=x^*, f(x) too was previously
> evaluated at x=x^*? And in case f(x) was evaluated multiple
> times before evaluating df(x), is it true that the last x at
> which f(x) was evaluated (before computing df(x=x^*))
> was x=x^*?
>
> If these assumptions holds (as it seems from preliminary
> tests on NLP using ralg), the extra code to take advantage
> of this fact is extremely simple.
>
> Best,
>
> Emanuele
>
> P.S.: if the previous assumptions are false in general, I'd
> like to know it they are true at least for the NLP case.
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
>
>   



From stefan at sun.ac.za  Mon Jul 21 08:53:28 2008
From: stefan at sun.ac.za (=?ISO-8859-1?Q?St=E9fan_van_der_Walt?=)
Date: Mon, 21 Jul 2008 14:53:28 +0200
Subject: [SciPy-user] Mode "reflect" may yield incorrect results on
	boundaries
In-Reply-To: <be020a860807210143v1e3036e8r7a56c9ca0f668fdb@mail.gmail.com>
References: <be020a860807210143v1e3036e8r7a56c9ca0f668fdb@mail.gmail.com>
Message-ID: <9457e7c80807210553y1c729d48n4eadf31f62553210@mail.gmail.com>

2008/7/21 Qing Peng <qing at qpeng.org>:
> After google with the key words: 'Mode "reflect" may yield incorrect results
> on boundaries', it looks like a bug, which said been fixed in year 2007. I
> just wondering where I go wrong.

That's true, but unfortunately we haven't released in the meantime.
Hopefully, 0.7 should be out by September.  In the meantime, building
the latest sources from SVN should solve your problem.

Regards
St?fan


From Dharhas.Pothina at twdb.state.tx.us  Mon Jul 21 08:53:55 2008
From: Dharhas.Pothina at twdb.state.tx.us (Dharhas Pothina)
Date: Mon, 21 Jul 2008 07:53:55 -0500
Subject: [SciPy-user] Help interpolating equally spaced
	2D	griddeddata	to a different grid.
In-Reply-To: <20080719160631.GA6332@lollo-laptop>
References: <4880A32E0200009B00014590@GWWEB.twdb.state.tx.us>
	<20080719160631.GA6332@lollo-laptop>
Message-ID: <48844093.63BA.009B.0@twdb.state.tx.us>


Thanks I'll try that. I guess I didn't realize interp2d returns something that can be used as a function and was wondering where to specify ynew,xnew in the interp2d function call.

- dharhas

>>> Lorenzo Bolla <lbolla at gmail.com> 7/19/2008 11:06 AM >>>
something like this should work (adjust the dimensions as you wish):

----------------------------------------
import numpy
from scipy.interpolate import interp2d

x = numpy.linspace(0, 1, 27)
y = numpy.linspace(0, 1, 34)

lat_old, lon_old = numpy.meshgrid(y, x)
precip_old = numpy.sin(lat_old + lon_old)

print lat_old.shape
print lon_old.shape
print precip_old.shape

x_new = numpy.linspace(0, 1, 20)
y_new = numpy.linspace(0, 1, 16)

lat, lon = numpy.meshgrid(y_new, x_new)
precip_interp = interp2d(lat_old, lon_old, precip_old)
precip = precip_interp(y_new, x_new)

print lat.shape
print lon.shape
print precip.shape
----------------------------------------

hth,
L.



On Fri, Jul 18, 2008 at 02:05:34PM -0500, Dharhas Pothina wrote:
> Hi,
> 
> I have a regular grid of data :
> 
> lat_old.shape = (277, 349)
> lon_old.shape = (277, 349)
> precip_old.shape = (8, 277, 349) # the first axis is time 
> 
> I need to interpolate precip_old to a new regular grid (lat,lon) where 
> 
> lat.shape = (204, 160)
> lon.shape = (204, 160)
> 
> Basically I need to calculate  precip_new where precip_new.shape = (8, 204, 160)
> 
> I've had a look at the cookbook page for 'N-D interpolation for equally-spaced data' and also some of the documentation for interp2D etc but am having trouble understanding what I need to do.
> 
> Any help is appreciated.
> 
> thanks
> 
> - dharhas
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org 
> http://projects.scipy.org/mailman/listinfo/scipy-user 
_______________________________________________
SciPy-user mailing list
SciPy-user at scipy.org 
http://projects.scipy.org/mailman/listinfo/scipy-user



From emanuele at relativita.com  Mon Jul 21 10:04:35 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Mon, 21 Jul 2008 16:04:35 +0200
Subject: [SciPy-user] [OpenOpt] evaluation of f(x) and df(x)
In-Reply-To: <488482A4.5060601@scipy.org>
References: <48847F98.7030600@relativita.com> <488482A4.5060601@scipy.org>
Message-ID: <48849773.3060907@relativita.com>

Hi Dmitrey,

I do not understand if your message answers my question so let me
reformulate. I have the exact gradient df(x) implemented in my code
so I don't use finite differences.

In my problem, In order to compute the gradient df(x=x1), I'd like to
take advantage of intermediate results of f(x=x1)'s compuation.
The re-use of these results is trivial to implement if
the sequence of function calls made by OpenOpt is, e.g., like this:
f(x0), df(x0), f(x1), f(x2), df(x2), f(x3), df(x3).... . Instead the
implementation could become quite difficult if the sequence would
be like this: f(x0), f(x1), df(x0), f(x2), f(x3), f(x4), df(x3),...
(i.e., the sequence of f / df is not evaluated on the same values).

Is OpenOpt working as in the first case?

Thanks,

Emanuele

dmitrey wrote:
> Hi Emanuele,
>
> if df(x1) is obtained via finite-difference calculations then f(x1) is 
> stored and compared during next call to f / df, and vice versa: if f(x1) 
> is called then the value obtained is stored and compared during next 
> call to f and/or finite-difference df.
>
> At least it is intended so, I can take more precise look if you have 
> noticed it doesn't work properly.
>
> Regards, D.
>
> Emanuele Olivetti wrote:
>   
>> Dear All and Dmitrey,
>>
>> in my code the evaluation of f(x) and df(x) shares many
>> intermediate steps. I'd like to re-use what is computed
>> inside f(x) to evaluate df(x) more efficiently, during f(x)
>> optimization. Then is it _always_ true that, when OpenOpt
>> evaluates df(x) at a certain x=x^*, f(x) too was previously
>> evaluated at x=x^*? And in case f(x) was evaluated multiple
>> times before evaluating df(x), is it true that the last x at
>> which f(x) was evaluated (before computing df(x=x^*))
>> was x=x^*?
>>
>> If these assumptions holds (as it seems from preliminary
>> tests on NLP using ralg), the extra code to take advantage
>> of this fact is extremely simple.
>>
>> Best,
>>
>> Emanuele
>>
>> P.S.: if the previous assumptions are false in general, I'd
>> like to know it they are true at least for the NLP case.
>>
>> _______________________________________________
>> SciPy-user mailing list
>> SciPy-user at scipy.org
>> http://projects.scipy.org/mailman/listinfo/scipy-user
>>
>>
>>
>>   
>>     
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>   



From dmitrey.kroshko at scipy.org  Mon Jul 21 10:58:32 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Mon, 21 Jul 2008 17:58:32 +0300
Subject: [SciPy-user] [OpenOpt] evaluation of f(x) and df(x)
In-Reply-To: <48849773.3060907@relativita.com>
References: <48847F98.7030600@relativita.com> <488482A4.5060601@scipy.org>
	<48849773.3060907@relativita.com>
Message-ID: <4884A418.5000006@scipy.org>

Hi Emanuele,
no, openopt will never hold in memory more then 1 point value, because 
it's too expensive to check all those previous values, as well as to 
store in memory so much data (sometimes nVars are very large). The 
proble you have mentioned is too specific, so it assumes user will take 
care of the situation. As for solvers (especially those which use 
derivatives), they will hardly use previous points (if so it may be 
likely considered as a bug).

To prevent the bugs I use openopt Point concept (in my ralg solver). 
Instead of handling in current workspace so lots of variables (f,df, 
f_prev, df_prev, c_prev, h_prev, dh_prev, etc, + any kinds of linear, + 
possible 2nd derivatives) I use something like this:
iterPoint = p.point(x)
if I need f(x), df(x), dc(x), maxResidual(x) etc I use
iterPoint.f(), iterPoint.df(), iterPoint.dc(), iterPoint.mr() etc
and I'm sure these values will not be recalculated.

Still I'm not sure in my current Point having f will benefit of having 
df and wise versa (if some calculations with other points had been done 
between them).

As for you as a user, if you want to take usage of OO Point, it could be 
done via
p = NLP()
p.args = p
and then using p.point in your objFunc/constraints (still I'm not sure 
it will not bring endless recursive cycle).
Or, you could write something like this by yourself in your code. Point 
class is situated in /Kernel/Point.py.

Regards, D.


Emanuele Olivetti wrote:
> Hi Dmitrey,
>
> I do not understand if your message answers my question so let me
> reformulate. I have the exact gradient df(x) implemented in my code
> so I don't use finite differences.
>
> In my problem, In order to compute the gradient df(x=x1), I'd like to
> take advantage of intermediate results of f(x=x1)'s compuation.
> The re-use of these results is trivial to implement if
> the sequence of function calls made by OpenOpt is, e.g., like this:
> f(x0), df(x0), f(x1), f(x2), df(x2), f(x3), df(x3).... . Instead the
> implementation could become quite difficult if the sequence would
> be like this: f(x0), f(x1), df(x0), f(x2), f(x3), f(x4), df(x3),...
> (i.e., the sequence of f / df is not evaluated on the same values).
>
> Is OpenOpt working as in the first case?
>
> Thanks,
>
> Emanuele
>
> dmitrey wrote:
>   
>> Hi Emanuele,
>>
>> if df(x1) is obtained via finite-difference calculations then f(x1) is 
>> stored and compared during next call to f / df, and vice versa: if f(x1) 
>> is called then the value obtained is stored and compared during next 
>> call to f and/or finite-difference df.
>>
>> At least it is intended so, I can take more precise look if you have 
>> noticed it doesn't work properly.
>>
>> Regards, D.
>>
>> Emanuele Olivetti wrote:
>>   
>>     
>>> Dear All and Dmitrey,
>>>
>>> in my code the evaluation of f(x) and df(x) shares many
>>> intermediate steps. I'd like to re-use what is computed
>>> inside f(x) to evaluate df(x) more efficiently, during f(x)
>>> optimization. Then is it _always_ true that, when OpenOpt
>>> evaluates df(x) at a certain x=x^*, f(x) too was previously
>>> evaluated at x=x^*? And in case f(x) was evaluated multiple
>>> times before evaluating df(x), is it true that the last x at
>>> which f(x) was evaluated (before computing df(x=x^*))
>>> was x=x^*?
>>>
>>> If these assumptions holds (as it seems from preliminary
>>> tests on NLP using ralg), the extra code to take advantage
>>> of this fact is extremely simple.
>>>
>>> Best,
>>>
>>> Emanuele
>>>
>>> P.S.: if the previous assumptions are false in general, I'd
>>> like to know it they are true at least for the NLP case.
>>>
>>> _______________________________________________
>>> SciPy-user mailing list
>>> SciPy-user at scipy.org
>>> http://projects.scipy.org/mailman/listinfo/scipy-user
>>>
>>>
>>>
>>>   
>>>     
>>>       
>> _______________________________________________
>> SciPy-user mailing list
>> SciPy-user at scipy.org
>> http://projects.scipy.org/mailman/listinfo/scipy-user
>>
>>   
>>     
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
>
>   



From dmitrey.kroshko at scipy.org  Mon Jul 21 12:08:52 2008
From: dmitrey.kroshko at scipy.org (dmitrey)
Date: Mon, 21 Jul 2008 19:08:52 +0300
Subject: [SciPy-user] [OpenOpt] evaluation of f(x) and df(x)
In-Reply-To: <48847F98.7030600@relativita.com>
References: <48847F98.7030600@relativita.com>
Message-ID: <4884B494.9070209@scipy.org>

Sorry, I re-read your letter once again and seems now I understand the 
question.

Emanuele Olivetti wrote:
> Dear All and Dmitrey,
>
> in my code the evaluation of f(x) and df(x) shares many
> intermediate steps. I'd like to re-use what is computed
> inside f(x) to evaluate df(x) more efficiently, during f(x)
> optimization. Then is it _always_ true that, when OpenOpt
> evaluates df(x) at a certain x=x^*, f(x) too was previously
> evaluated at x=x^*?
No, this is not guarantied.
>  And in case f(x) was evaluated multiple
> times before evaluating df(x), is it true that the last x at
> which f(x) was evaluated (before computing df(x=x^*))
> was x=x^*?
>   
No, this is not guarantied.

I can only guarantee the code
f = user_obj_fun(x)
f2 = user_obj_fun(x)
will not call user-supplied objective func for twice. Same for df, c, 
dc, dh etc.

> If these assumptions holds (as it seems from preliminary
> tests on NLP using ralg), the extra code to take advantage
> of this fact is extremely simple.
>   
It is what oofun is intended for (constructing f,c,h from lots of 
pieces, some of which can be used for several times, including 
cross-cases, for example same part of code used by c and f). Currently I 
continue my work on them. See the file (committed some minutes ago, 
still require more detailed docstrings):
http://projects.scipy.org/scipy/scikits/browser/trunk/openopt/scikits/openopt/examples/oofun_input.py
Still there is much more work to be done (including recursive 1st 
derivatives, conception of oovar that I intend to add etc).

Regards, D.
> Best,
>
> Emanuele
>
> P.S.: if the previous assumptions are false in general, I'd
> like to know it they are true at least for the NLP case.
>
>   



From pgmdevlist at gmail.com  Mon Jul 21 12:26:46 2008
From: pgmdevlist at gmail.com (Pierre GM)
Date: Mon, 21 Jul 2008 12:26:46 -0400
Subject: [SciPy-user] Statistics advise with scipy
In-Reply-To: <da09baf80807201956r6f92db05vd5262a4236bc3ae2@mail.gmail.com>
References: <da09baf80807201956r6f92db05vd5262a4236bc3ae2@mail.gmail.com>
Message-ID: <200807211226.46646.pgmdevlist@gmail.com>

On Sunday 20 July 2008 22:56:14 didier rano wrote:

> I have a problem to filter only relevant data.

Didier,
That's a trick question, whose answer depends too heavily on the kind of data 
you're processing. 
* Are you expecting some kind of trend ? If yes, detrend your data first and 
work on the residuals.
* Are you interested in finding the breaks in trend and/or regimes ? If so, I 
can send you some algos that do just that.
* Are you interested into finding the outliers ? There are common approaches, 
some based on an expected normality of the data, some based on more robust 
methods... Googling 'outliers robust method' should get you started.


From didier.rano at gmail.com  Mon Jul 21 12:59:06 2008
From: didier.rano at gmail.com (didier rano)
Date: Mon, 21 Jul 2008 12:59:06 -0400
Subject: [SciPy-user] Statistics advise with scipy
In-Reply-To: <200807211226.46646.pgmdevlist@gmail.com>
References: <da09baf80807201956r6f92db05vd5262a4236bc3ae2@mail.gmail.com>
	<200807211226.46646.pgmdevlist@gmail.com>
Message-ID: <da09baf80807210959g515a9a73l386d437451370668@mail.gmail.com>

Thanks Pierre,

What do you mean by "finding the breaks in trend and/or regimes" ?

In fact, in my case, I have several objectives:
* Draw a graph without outliers data
* Compute a trend wihout outliers data
* Determine when the behavior/trend data will change. May be "finding the
breaks in trend and/or regimes " ?

Bye
Didier Rano

2008/7/21, Pierre GM <pgmdevlist at gmail.com>:
>
> On Sunday 20 July 2008 22:56:14 didier rano wrote:
>
> > I have a problem to filter only relevant data.
>
> Didier,
> That's a trick question, whose answer depends too heavily on the kind of
> data
> you're processing.
> * Are you expecting some kind of trend ? If yes, detrend your data first
> and
> work on the residuals.
> * Are you interested in finding the breaks in trend and/or regimes ? If so,
> I
> can send you some algos that do just that.
> * Are you interested into finding the outliers ? There are common
> approaches,
> some based on an expected normality of the data, some based on more robust
> methods... Googling 'outliers robust method' should get you started.
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Didier Rano
didier.rano at gmail.com
http://www.jaxtr.com/didierrano
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080721/39997b1c/attachment.html>

From emanuele at relativita.com  Mon Jul 21 13:03:47 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Mon, 21 Jul 2008 19:03:47 +0200
Subject: [SciPy-user] [OpenOpt] evaluation of f(x) and df(x)
In-Reply-To: <4884B494.9070209@scipy.org>
References: <48847F98.7030600@relativita.com> <4884B494.9070209@scipy.org>
Message-ID: <4884C173.8040501@relativita.com>

OK you got it now. Thanks for the answer. In the meanwhile I made a little
change to my code -  just a sub-optimal solution - but allows to avoid lots
of useless computations (until now I have no example for which the
assumptions I described are false, and when they will be my algorithm
will just become slower and compute everything).

Oofun seems very interesting I'll spend more time on it.

Let me express again my appreciation for OpenOpt!

Emanuele


dmitrey wrote:
> Sorry, I re-read your letter once again and seems now I understand the 
> question.
>
> Emanuele Olivetti wrote:
>   
>> Dear All and Dmitrey,
>>
>> in my code the evaluation of f(x) and df(x) shares many
>> intermediate steps. I'd like to re-use what is computed
>> inside f(x) to evaluate df(x) more efficiently, during f(x)
>> optimization. Then is it _always_ true that, when OpenOpt
>> evaluates df(x) at a certain x=x^*, f(x) too was previously
>> evaluated at x=x^*?
>>     
> No, this is not guarantied.
>   
>>  And in case f(x) was evaluated multiple
>> times before evaluating df(x), is it true that the last x at
>> which f(x) was evaluated (before computing df(x=x^*))
>> was x=x^*?
>>   
>>     
> No, this is not guarantied.
>
> I can only guarantee the code
> f = user_obj_fun(x)
> f2 = user_obj_fun(x)
> will not call user-supplied objective func for twice. Same for df, c, 
> dc, dh etc.
>
>   
>> If these assumptions holds (as it seems from preliminary
>> tests on NLP using ralg), the extra code to take advantage
>> of this fact is extremely simple.
>>   
>>     
> It is what oofun is intended for (constructing f,c,h from lots of 
> pieces, some of which can be used for several times, including 
> cross-cases, for example same part of code used by c and f). Currently I 
> continue my work on them. See the file (committed some minutes ago, 
> still require more detailed docstrings):
> http://projects.scipy.org/scipy/scikits/browser/trunk/openopt/scikits/openopt/examples/oofun_input.py
> Still there is much more work to be done (including recursive 1st 
> derivatives, conception of oovar that I intend to add etc).
>
> Regards, D.
>   
>> Best,
>>
>> Emanuele
>>
>> P.S.: if the previous assumptions are false in general, I'd
>> like to know it they are true at least for the NLP case.
>>
>>   
>>     
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>   



From pgmdevlist at gmail.com  Mon Jul 21 13:12:39 2008
From: pgmdevlist at gmail.com (Pierre GM)
Date: Mon, 21 Jul 2008 13:12:39 -0400
Subject: [SciPy-user] Statistics advise with scipy
In-Reply-To: <da09baf80807210959g515a9a73l386d437451370668@mail.gmail.com>
References: <da09baf80807201956r6f92db05vd5262a4236bc3ae2@mail.gmail.com>
	<200807211226.46646.pgmdevlist@gmail.com>
	<da09baf80807210959g515a9a73l386d437451370668@mail.gmail.com>
Message-ID: <200807211312.40202.pgmdevlist@gmail.com>

On Monday 21 July 2008 12:59:06 didier rano wrote:
> Thanks Pierre,
>
> What do you mean by "finding the breaks in trend and/or regimes" ?

There's a lot of literature on the detection of change-points. Roughly, a 
change-point can be a step change-point (eg., switching from one mean to 
another, as you observe in your data), or a trend change-points (the slope of 
a linear model changes at some point), or both. 

www.beringclimate.noaa.gov/regimes/Regime_shift_methods_list.htm
http://ams.allenpress.com/perlserv/?request=get-pdf&doi=10.1175%2F2008JCLI1956.1

> In fact, in my case, I have several objectives:
> * Draw a graph without outliers data

Well, first you need to define what an outlier is in your problem: assuming 
normal data, how far away from the mean should a point be to be an outlier ? 
1,2, 3 standard deviation ? What if your data is not normal ? In that case, 
robust methods can give good result.

> * Compute a trend wihout outliers data
> * Determine when the behavior/trend data will change. May be "finding the
> breaks in trend and/or regimes " ?


From S.Collis at bom.gov.au  Mon Jul 21 18:00:43 2008
From: S.Collis at bom.gov.au (Scott Collis)
Date: Tue, 22 Jul 2008 08:00:43 +1000
Subject: [SciPy-user] Matplotlib Help with colorbar [SEC=UNCLASSIFIED]
Message-ID: <00E49660AB21DD44A0D4A3115DFCBF61010126F0@officeho2.bom.gov.au>

Good morning all,
While not strictly a Scipy question, I know many use Matplotlib for
visualisation.
 
I have a 2D matrix of classifications (say 1 to 10 in discrete steps)
and I am using pcolor to create a mesh plot (which I overlay contours
and then a quiver on...) Now I wish to put a colour bar on but instead
of 1-10 on the axis I want say "apple" to "Jake" (or in my case
"Unidentified" to "Rain + Hail")
 
Is there a way I can give colourbar or one of its children
(colorbar.ax.yaxis?) a list or dictionary to annotate the bar? ie
[(0,"unclassified"), ..., (10, "Rain + Hail")]
 
Dr Scott Collis
Postdoctoral Researcher
Centre for Australian Weather and Climate Research (CAWCR)
Atmosphere and Land Observation and Assessment Group
Desk: (+613) 96694766
MB: 0412177550
www: http://www.bom.gov.au/bmrc/wefor/staff/scollis/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080722/dde0131e/attachment.html>

From robert.kern at gmail.com  Mon Jul 21 18:46:38 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Mon, 21 Jul 2008 17:46:38 -0500
Subject: [SciPy-user] Matplotlib Help with colorbar [SEC=UNCLASSIFIED]
In-Reply-To: <00E49660AB21DD44A0D4A3115DFCBF61010126F0@officeho2.bom.gov.au>
References: <00E49660AB21DD44A0D4A3115DFCBF61010126F0@officeho2.bom.gov.au>
Message-ID: <3d375d730807211546v1bbea7f0wcff2b7f70d898a3f@mail.gmail.com>

On Mon, Jul 21, 2008 at 17:00, Scott Collis <S.Collis at bom.gov.au> wrote:
> Good morning all,
> While not strictly a Scipy question, I know many use Matplotlib for
> visualisation.

I suspect you will have better luck on the matplotlib-users mailing list.

  https://lists.sourceforge.net/lists/listinfo/matplotlib-users

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From stefan at sun.ac.za  Mon Jul 21 18:56:26 2008
From: stefan at sun.ac.za (=?ISO-8859-1?Q?St=E9fan_van_der_Walt?=)
Date: Tue, 22 Jul 2008 00:56:26 +0200
Subject: [SciPy-user] optimize.fmin_tnc - how to catch its output?
In-Reply-To: <g61jl9$hq0$1@ger.gmane.org>
References: <g61jl9$hq0$1@ger.gmane.org>
Message-ID: <9457e7c80807211556x5ae7d4e1m9079534f518e0cb9@mail.gmail.com>

Hi Marek

2008/7/21 Marek Wojciechowski <mwojc at p.lodz.pl>:
> Hallo!
> I tried to catch fmin_tnc output to a file in the following way:
>
>    file = open('messages', 'w')
>    stdout = sys.stdout
>    stderr = sys.stderr
>    sys.stdout = file
>    sys.stderr = file
>    res = optimize.fmin_tnc(..., messages=1)
>    file.close()
>    sys.stdout = stdout
>    sys.stderr = stderr
>
> But this does not work. I realized that fmin_tnc calls C routine which
> probably uses its own stdout, stderr. Is there a way to catch its output?

I'm afraid you are right.  The quickest way I can think of is to grab
it using standard unix pipes:

python optimise_script.py 2>&1 | cat > out.txt

Otherwise, the C-code will have to be modified.  Does anybody know
what the best way is to write to sys.stdout and sys.stderr (not printf
that goes to /dev/stdout) via the Python C API?

Regards
St?fan


From robert.kern at gmail.com  Mon Jul 21 19:22:29 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Mon, 21 Jul 2008 18:22:29 -0500
Subject: [SciPy-user] optimize.fmin_tnc - how to catch its output?
In-Reply-To: <9457e7c80807211556x5ae7d4e1m9079534f518e0cb9@mail.gmail.com>
References: <g61jl9$hq0$1@ger.gmane.org>
	<9457e7c80807211556x5ae7d4e1m9079534f518e0cb9@mail.gmail.com>
Message-ID: <3d375d730807211622p4b73cd4bo2fe69b2ebf4d8a61@mail.gmail.com>

On Mon, Jul 21, 2008 at 17:56, St?fan van der Walt <stefan at sun.ac.za> wrote:
> Hi Marek
>
> 2008/7/21 Marek Wojciechowski <mwojc at p.lodz.pl>:
>> Hallo!
>> I tried to catch fmin_tnc output to a file in the following way:
>>
>>    file = open('messages', 'w')
>>    stdout = sys.stdout
>>    stderr = sys.stderr
>>    sys.stdout = file
>>    sys.stderr = file
>>    res = optimize.fmin_tnc(..., messages=1)
>>    file.close()
>>    sys.stdout = stdout
>>    sys.stderr = stderr
>>
>> But this does not work. I realized that fmin_tnc calls C routine which
>> probably uses its own stdout, stderr. Is there a way to catch its output?
>
> I'm afraid you are right.  The quickest way I can think of is to grab
> it using standard unix pipes:
>
> python optimise_script.py 2>&1 | cat > out.txt

There is a way to redirect it to a (temporary) file from within
Python. You can then read this file. I have an old and
not-recently-tested implementation over here:

  http://www.enthought.com/~rkern/cgi-bin/hgwebdir.cgi/redir/

> Otherwise, the C-code will have to be modified.  Does anybody know
> what the best way is to write to sys.stdout and sys.stderr (not printf
> that goes to /dev/stdout) via the Python C API?

PySys_WriteStdout() and PySys_WriteStderr() as long as the output is
coming from the C extension modules themselves. Modifying the
library's C or (hah!) FORTRAN code is less palatable.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco

From didier.rano at gmail.com  Mon Jul 21 22:49:51 2008
From: didier.rano at gmail.com (didier rano)
Date: Mon, 21 Jul 2008 22:49:51 -0400
Subject: [SciPy-user] Statistics advise with scipy
In-Reply-To: <200807211312.40202.pgmdevlist@gmail.com>
References: <da09baf80807201956r6f92db05vd5262a4236bc3ae2@mail.gmail.com>
	<200807211226.46646.pgmdevlist@gmail.com>
	<da09baf80807210959g515a9a73l386d437451370668@mail.gmail.com>
	<200807211312.40202.pgmdevlist@gmail.com>
Message-ID: <da09baf80807211949i7d075d22o277b7145f31ab0e5@mail.gmail.com>

2008/7/21 Pierre GM <pgmdevlist at gmail.com>:

> On Monday 21 July 2008 12:59:06 didier rano wrote:
> > Thanks Pierre,
> >
> > What do you mean by "finding the breaks in trend and/or regimes" ?
>
> There's a lot of literature on the detection of change-points. Roughly, a
> change-point can be a step change-point (eg., switching from one mean to
> another, as you observe in your data), or a trend change-points (the slope
> of
> a linear model changes at some point), or both.
>
> www.beringclimate.noaa.gov/regimes/Regime_shift_methods_list.htm
>
> http://ams.allenpress.com/perlserv/?request=get-pdf&doi=10.1175%2F2008JCLI1956.1
>
> > In fact, in my case, I have several objectives:
> > * Draw a graph without outliers data
>
> Well, first you need to define what an outlier is in your problem: assuming
> normal data, how far away from the mean should a point be to be an outlier
> ?
> 1,2, 3 standard deviation ? What if your data is not normal ? In that case,
> robust methods can give good result.
>

My data is not normal. Do you know robusts method in scipy ? Or maybe in an
other python module ?


>
> > * Compute a trend wihout outliers data
> > * Determine when the behavior/trend data will change. May be "finding the
> > breaks in trend and/or regimes " ?
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Didier Rano
didier.rano at gmail.com
http://www.jaxtr.com/didierrano
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080721/714cb938/attachment.html>

From jan at aims.ac.za  Mon Jul 21 23:10:54 2008
From: jan at aims.ac.za (Jan Groenewald)
Date: Tue, 22 Jul 2008 05:10:54 +0200
Subject: [SciPy-user] optimize.fmin_tnc - how to catch its output?
In-Reply-To: <9457e7c80807211556x5ae7d4e1m9079534f518e0cb9@mail.gmail.com>
References: <g61jl9$hq0$1@ger.gmane.org>
	<9457e7c80807211556x5ae7d4e1m9079534f518e0cb9@mail.gmail.com>
Message-ID: <3c1fa8220807212010k1d4a12b3scc138002424f5783@mail.gmail.com>

Hi St?fan,

On Tue, Jul 22, 2008 at 12:56 AM, St?fan van der Walt <stefan at sun.ac.za> wrote:
> python optimise_script.py 2>&1 | cat > out.txt

http://en.wikipedia.org/wiki/Cat_(Unix)#cat_and_UUOC ?
You don' t need the "| cat".

Never mind the pipecat, there is a shortcut for "2>&1 >" and it is "&>".

> Otherwise, the C-code will have to be modified.  Does anybody know
> what the best way is to write to sys.stdout and sys.stderr (not printf
> that goes to /dev/stdout) via the Python C API?

I don't know but found this interesting so googled it:
http://www.ragestorm.net/tutorial?id=21#9
or
http://docs.python.org/lib/module-subprocess.html

regards,
Jan


From pgmdevlist at gmail.com  Mon Jul 21 23:07:31 2008
From: pgmdevlist at gmail.com (Pierre GM)
Date: Mon, 21 Jul 2008 23:07:31 -0400
Subject: [SciPy-user] Statistics advise with scipy
In-Reply-To: <da09baf80807211949i7d075d22o277b7145f31ab0e5@mail.gmail.com>
References: <da09baf80807201956r6f92db05vd5262a4236bc3ae2@mail.gmail.com>
	<200807211312.40202.pgmdevlist@gmail.com>
	<da09baf80807211949i7d075d22o277b7145f31ab0e5@mail.gmail.com>
Message-ID: <200807212307.31208.pgmdevlist@gmail.com>

On Monday 21 July 2008 22:49:51 didier rano wrote:

> My data is not normal. Do you know robusts method in scipy ? Or maybe in an
> other python module ?

Mmh, I'm sure you could implement some yourself. That way, we could start 
another scikits. There are already some winsorization and trimming functions 
in scipy.stats.
Alternatively, you can try to use R and numpy through rpy: 
http://rpy.sourceforge.net/


From nwagner at iam.uni-stuttgart.de  Tue Jul 22 02:56:41 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Tue, 22 Jul 2008 08:56:41 +0200
Subject: [SciPy-user] Frequency content of a transient signal
Message-ID: <web-105021035@uni-stuttgart.de>

Hi all,
  
How can I obtain the frequency content of a
transient signal (signal.png) ?

Any pointer would be appreciated.

Nils
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signal.png
Type: image/png
Size: 21303 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080722/d4b45f32/attachment.png>

From matthieu.brucher at gmail.com  Tue Jul 22 03:15:11 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Tue, 22 Jul 2008 09:15:11 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <web-105021035@uni-stuttgart.de>
References: <web-105021035@uni-stuttgart.de>
Message-ID: <e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>

Hi !

I think you can use a sliding DFT on your signal. This way, you will
get a moving estimation of the spectral information (it's a basic
time-frequency transform, before you use wavelets ;)).

Matthieu

2008/7/22 Nils Wagner <nwagner at iam.uni-stuttgart.de>:
> Hi all,
>  How can I obtain the frequency content of a
> transient signal (signal.png) ?
>
> Any pointer would be appreciated.
>
> Nils
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>



-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From nwagner at iam.uni-stuttgart.de  Tue Jul 22 03:31:33 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Tue, 22 Jul 2008 09:31:33 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
Message-ID: <web-105022673@uni-stuttgart.de>

On Tue, 22 Jul 2008 09:15:11 +0200
  "Matthieu Brucher" <matthieu.brucher at gmail.com> wrote:
> Hi !
> 
> I think you can use a sliding DFT on your signal. This 
>way, you will
> get a moving estimation of the spectral information 
>(it's a basic
> time-frequency transform, before you use wavelets ;)).
> 
> Matthieu
> 
  
Matthieu,

I am not very familiar with signal processing. Please can
you provide a short example ?
I have attached the data file. The first column 
corresponds
with the time, the second column represents the signal.
  
Thanks in advance

                       Nils
  
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mysignal.dat.gz
Type: application/x-gzip
Size: 5668 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080722/66b8c067/attachment.bin>

From matthieu.brucher at gmail.com  Tue Jul 22 03:59:13 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Tue, 22 Jul 2008 09:59:13 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <web-105022673@uni-stuttgart.de>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
	<web-105022673@uni-stuttgart.de>
Message-ID: <e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>

2008/7/22 Nils Wagner <nwagner at iam.uni-stuttgart.de>:
> On Tue, 22 Jul 2008 09:15:11 +0200
>  "Matthieu Brucher" <matthieu.brucher at gmail.com> wrote:
>>
>> Hi !
>>
>> I think you can use a sliding DFT on your signal. This way, you will
>> get a moving estimation of the spectral information (it's a basic
>> time-frequency transform, before you use wavelets ;)).
>>
>> Matthieu
>>
>  Matthieu,
>
> I am not very familiar with signal processing. Please can
> you provide a short example ?
> I have attached the data file. The first column corresponds
> with the time, the second column represents the signal.
>  Thanks in advance
>
>                      Nils

A wikipedia article states it better than me, so here is a link :
http://en.wikipedia.org/wiki/Short-time_Fourier_transform
Feel free to ask any question ;)

In fact, what you will do is :

my_ffts = []
fo i in range(n):
  my_ffts.append(fft(data[i*size:(i+1)*size]))

with size the size of the sliding window you will use. This will give
you non overlapping FFTs, but you can use overlapping FFT, such as :
  my_ffts.append(fft(data[(i-1/2.)*size:(i+3/2.)*size]))

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From nwagner at iam.uni-stuttgart.de  Tue Jul 22 04:52:31 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Tue, 22 Jul 2008 10:52:31 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
	<web-105022673@uni-stuttgart.de>
	<e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>
Message-ID: <web-105027338@uni-stuttgart.de>

On Tue, 22 Jul 2008 09:59:13 +0200
  "Matthieu Brucher" <matthieu.brucher at gmail.com> wrote:
> 2008/7/22 Nils Wagner <nwagner at iam.uni-stuttgart.de>:
>> On Tue, 22 Jul 2008 09:15:11 +0200
>>  "Matthieu Brucher" <matthieu.brucher at gmail.com> wrote:
>>>
>>> Hi !
>>>
>>> I think you can use a sliding DFT on your signal. This 
>>>way, you will
>>> get a moving estimation of the spectral information 
>>>(it's a basic
>>> time-frequency transform, before you use wavelets ;)).
>>>
>>> Matthieu
>>>
>>  Matthieu,
>>
>> I am not very familiar with signal processing. Please 
>>can
>> you provide a short example ?
>> I have attached the data file. The first column 
>>corresponds
>> with the time, the second column represents the signal.
>>  Thanks in advance
>>
>>                      Nils
> 
> A wikipedia article states it better than me, so here is 
>a link :
> http://en.wikipedia.org/wiki/Short-time_Fourier_transform
>Feel free to ask any question ;)
>
Fine. Let's start with the example from the link

How do I produce the nice spectrograms

from scipy import *
from pylab import plot, show
#
# Example taken from 
http://en.wikipedia.org/wiki/Short-time_Fourier_transform
#
def x(t):
     if t < 5:
        return cos(2*pi*10*t)
     if t >= 5. and t < 10:
        return cos(2*pi*25*t)
     if t >=10. and t< 15:
        return cos(2*pi*50*t)
     if t >=15. and t<= 20:
        return cos(2*pi*100*t)

t = linspace(0.,20.,8001) # sampled at 400 Hz
x_vec = vectorize(x)
signal = x_vec(t)
plot(t,signal)
#
# How can I obtain the nice spectrograms ?
#
show()
...
to be continued


Nils

> In fact, what you will do is :
> 
> my_ffts = []
> fo i in range(n):
>  my_ffts.append(fft(data[i*size:(i+1)*size]))
> 
> with size the size of the sliding window you will use. 
>This will give
> you non overlapping FFTs, but you can use overlapping 
>FFT, such as :
>  my_ffts.append(fft(data[(i-1/2.)*size:(i+3/2.)*size]))
> 
  


From matthieu.brucher at gmail.com  Tue Jul 22 05:17:08 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Tue, 22 Jul 2008 11:17:08 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <web-105027338@uni-stuttgart.de>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
	<web-105022673@uni-stuttgart.de>
	<e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>
	<web-105027338@uni-stuttgart.de>
Message-ID: <e76aa17f0807220217m1525f4c3m4f51f5e96fbfb939@mail.gmail.com>

> Fine. Let's start with the example from the link
>
> How do I produce the nice spectrograms
>
> from scipy import *
> from pylab import plot, show
> #
> # Example taken from
> http://en.wikipedia.org/wiki/Short-time_Fourier_transform
> #
> def x(t):
>     if t < 5:
>        return cos(2*pi*10*t)
>     if t >= 5. and t < 10:
>        return cos(2*pi*25*t)
>     if t >=10. and t< 15:
>        return cos(2*pi*50*t)
>     if t >=15. and t<= 20:
>        return cos(2*pi*100*t)
>
> t = linspace(0.,20.,8001) # sampled at 400 Hz
> x_vec = vectorize(x)
> signal = x_vec(t)
> plot(t,signal)
> #
> # How can I obtain the nice spectrograms ?
> #
> show()
> ...
> to be continued
>
>
> Nils

I've found a matlab script, it'snot very clear, but it will be a start
: http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=7463

The easiest thing to do is to use an array, so you will do something like :

>>> spectro = n.zeros((nb_chunks, precision))

nb_chunks is the number of time slices you want and precision is the
precision of the FFT you want.
sampling_f is your sampling frequency, so you can compute the size of a slice ;

>>> slice_size = time / sampling_f # time being the time in one slice

Then, something like :

>>> for i in nb_chunks:
  spectro[i] = n.abs(fft(x[i * slice_size:(i+1)*slice_size])) # you
may need to scale data, as shown in the matlab script

And then :

>>> imshow(spectro)

This is a very very crude example. I might have some time this evening
to write a full example if you still need it.

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From nwagner at iam.uni-stuttgart.de  Tue Jul 22 05:29:42 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Tue, 22 Jul 2008 11:29:42 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <e76aa17f0807220217m1525f4c3m4f51f5e96fbfb939@mail.gmail.com>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
	<web-105022673@uni-stuttgart.de>
	<e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>
	<web-105027338@uni-stuttgart.de>
	<e76aa17f0807220217m1525f4c3m4f51f5e96fbfb939@mail.gmail.com>
Message-ID: <web-105029661@uni-stuttgart.de>

On Tue, 22 Jul 2008 11:17:08 +0200
  "Matthieu Brucher" <matthieu.brucher at gmail.com> wrote:
>> Fine. Let's start with the example from the link
>>
>> How do I produce the nice spectrograms
>>
>> from scipy import *
>> from pylab import plot, show
>> #
>> # Example taken from
>> http://en.wikipedia.org/wiki/Short-time_Fourier_transform
>> #
>> def x(t):
>>     if t < 5:
>>        return cos(2*pi*10*t)
>>     if t >= 5. and t < 10:
>>        return cos(2*pi*25*t)
>>     if t >=10. and t< 15:
>>        return cos(2*pi*50*t)
>>     if t >=15. and t<= 20:
>>        return cos(2*pi*100*t)
>>
>> t = linspace(0.,20.,8001) # sampled at 400 Hz
>> x_vec = vectorize(x)
>> signal = x_vec(t)
>> plot(t,signal)
>> #
>> # How can I obtain the nice spectrograms ?
>> #
>> show()
>> ...
>> to be continued
>>
>>
>> Nils
> 
> I've found a matlab script, it'snot very clear, but it 
>will be a start
> : 
>http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=7463
> 
> The easiest thing to do is to use an array, so you will 
>do something like :
> 
>>>> spectro = n.zeros((nb_chunks, precision))
> 
> nb_chunks is the number of time slices you want and 
>precision is the
> precision of the FFT you want.
> sampling_f is your sampling frequency, so you can 
>compute the size of a slice ;
> 
>>>> slice_size = time / sampling_f # time being the time in 
>>>>one slice
> 
> Then, something like :
> 
>>>> for i in nb_chunks:
>  spectro[i] = n.abs(fft(x[i * 
>slice_size:(i+1)*slice_size])) # you
> may need to scale data, as shown in the matlab script
> 
> And then :
> 
>>>> imshow(spectro)
> 
> This is a very very crude example. I might have some 
>time this evening
> to write a full example if you still need it.
>
That would be great ! It could be added to the Cookbook 
for other users :-).

Thank you very much !

Cheers,
              Nils
  


From matthieu.brucher at gmail.com  Tue Jul 22 05:36:48 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Tue, 22 Jul 2008 11:36:48 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <web-105029661@uni-stuttgart.de>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
	<web-105022673@uni-stuttgart.de>
	<e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>
	<web-105027338@uni-stuttgart.de>
	<e76aa17f0807220217m1525f4c3m4f51f5e96fbfb939@mail.gmail.com>
	<web-105029661@uni-stuttgart.de>
Message-ID: <e76aa17f0807220236q5f0d5fb9ya854a78e6d14218c@mail.gmail.com>

>> This is a very very crude example. I might have some
>>time this evening
>> to write a full example if you still need it.
>>
> That would be great ! It could be added to the Cookbook
> for other users :-).

Let's not rush, I have to stop checking my mails and start working if
I want to find time for this :D

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From nwagner at iam.uni-stuttgart.de  Tue Jul 22 06:56:36 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Tue, 22 Jul 2008 12:56:36 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <e76aa17f0807220236q5f0d5fb9ya854a78e6d14218c@mail.gmail.com>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
	<web-105022673@uni-stuttgart.de>
	<e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>
	<web-105027338@uni-stuttgart.de>
	<e76aa17f0807220217m1525f4c3m4f51f5e96fbfb939@mail.gmail.com>
	<web-105029661@uni-stuttgart.de>
	<e76aa17f0807220236q5f0d5fb9ya854a78e6d14218c@mail.gmail.com>
Message-ID: <web-105035363@uni-stuttgart.de>

On Tue, 22 Jul 2008 11:36:48 +0200
  "Matthieu Brucher" <matthieu.brucher at gmail.com> wrote:
>>> This is a very very crude example. I might have some
>>>time this evening
>>> to write a full example if you still need it.
>>>
>> That would be great ! It could be added to the Cookbook
>> for other users :-).
> 
> Let's not rush, I have to stop checking my mails and 
>start working if
> I want to find time for this :D
> 
> Matthieu
> -- 
>French PhD student
> Website : http://matthieu-brucher.developpez.com/
> Blogs : http://matt.eifelle.com and 
>http://blog.developpez.com/?blog=92
> LinkedIn : http://www.linkedin.com/in/matthieubrucher
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
  
Hi Matthieu,

I made some progress (See test_stft.py).
But, how do I choose proper values for the parameter
NFFT,noverlap ?

Cheers,
            Nils
  
-------------- next part --------------
A non-text attachment was scrubbed...
Name: test_stft.py
Type: text/x-python
Size: 784 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080722/100fc326/attachment.py>

From stefan at sun.ac.za  Tue Jul 22 07:59:39 2008
From: stefan at sun.ac.za (=?ISO-8859-1?Q?St=E9fan_van_der_Walt?=)
Date: Tue, 22 Jul 2008 13:59:39 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
Message-ID: <9457e7c80807220459o2680f370ge1265b717579ebb7@mail.gmail.com>

2008/7/22 Matthieu Brucher <matthieu.brucher at gmail.com>:
> I think you can use a sliding DFT on your signal. This way, you will
> get a moving estimation of the spectral information (it's a basic
> time-frequency transform, before you use wavelets ;)).

Wavelets always sound like so much fun.  Attached, a script to
decompose the signal using pywt (http://wavelets.scipy.org).

Regards
St?fan
-------------- next part --------------
A non-text attachment was scrubbed...
Name: growchip_wt.py
Type: application/octet-stream
Size: 883 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080722/4e2ea495/attachment.obj>

From nwagner at iam.uni-stuttgart.de  Tue Jul 22 08:14:43 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Tue, 22 Jul 2008 14:14:43 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <9457e7c80807220459o2680f370ge1265b717579ebb7@mail.gmail.com>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
	<9457e7c80807220459o2680f370ge1265b717579ebb7@mail.gmail.com>
Message-ID: <web-105040184@uni-stuttgart.de>

On Tue, 22 Jul 2008 13:59:39 +0200
  "St?fan van der Walt" <stefan at sun.ac.za> wrote:
> 2008/7/22 Matthieu Brucher <matthieu.brucher at gmail.com>:
>> I think you can use a sliding DFT on your signal. This 
>>way, you will
>> get a moving estimation of the spectral information 
>>(it's a basic
>> time-frequency transform, before you use wavelets ;)).
> 
> Wavelets always sound like so much fun.  Attached, a 
>script to
> decompose the signal using pywt 
>(http://wavelets.scipy.org).
> 
> Regards
> St?fan
  
Hi Stefan,

Thank you for your reply.
I have installed pywt and run your script. It looks 
impressive but how can I interprete the results ?

Regards,
              Nils


From matthieu.brucher at gmail.com  Tue Jul 22 08:37:01 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Tue, 22 Jul 2008 14:37:01 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <web-105035363@uni-stuttgart.de>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
	<web-105022673@uni-stuttgart.de>
	<e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>
	<web-105027338@uni-stuttgart.de>
	<e76aa17f0807220217m1525f4c3m4f51f5e96fbfb939@mail.gmail.com>
	<web-105029661@uni-stuttgart.de>
	<e76aa17f0807220236q5f0d5fb9ya854a78e6d14218c@mail.gmail.com>
	<web-105035363@uni-stuttgart.de>
Message-ID: <e76aa17f0807220537m13f62b42p1f5e10c8ef638b07@mail.gmail.com>

>  Hi Matthieu,
>
> I made some progress (See test_stft.py).
> But, how do I choose proper values for the parameter
> NFFT,noverlap ?
>
> Cheers,
>           Nils

Of course, if pylab has everything, it's better :D

This is what I think, I didn't try it much :
NFFT is the size of one slice, so if you want to compute the FFT over
25 ms, NFFT = 24. e-3 *400
noverlap depends on the size of the slice and the number of FFTs you
want per second. If you want 50 FFTs per second (50 colmuns per
second), that means one every actual 400/50 sample, noverlap is NFFT -
8

If someone knows better this, he can show himself ;)

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From matthieu.brucher at gmail.com  Tue Jul 22 08:43:19 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Tue, 22 Jul 2008 14:43:19 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <9457e7c80807220459o2680f370ge1265b717579ebb7@mail.gmail.com>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
	<9457e7c80807220459o2680f370ge1265b717579ebb7@mail.gmail.com>
Message-ID: <e76aa17f0807220543l1ccb8eaas475d82e3356f61c0@mail.gmail.com>

2008/7/22 St?fan van der Walt <stefan at sun.ac.za>:
> 2008/7/22 Matthieu Brucher <matthieu.brucher at gmail.com>:
>> I think you can use a sliding DFT on your signal. This way, you will
>> get a moving estimation of the spectral information (it's a basic
>> time-frequency transform, before you use wavelets ;)).
>
> Wavelets always sound like so much fun.  Attached, a script to
> decompose the signal using pywt (http://wavelets.scipy.org).
>
> Regards
> St?fan

What is great with the PSD (estimation of the spectral density) is
that it shows the spectral density. Wavelet are more complex than the
PSD.
Let's just say that what a wavelet transform does on a signal will
generally never be a spectral density (IIRC).

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From fredmfp at gmail.com  Tue Jul 22 11:48:08 2008
From: fredmfp at gmail.com (fred)
Date: Tue, 22 Jul 2008 17:48:08 +0200
Subject: [SciPy-user] 3D bilinear interpolation...
Message-ID: <48860138.2060304@gmail.com>

Hi all,

I do know that scipy has a 2D interpolate.interp2d() method but what 
about in 3D ?

TIA.

Cheers,

-- 
Fred


From pav at iki.fi  Tue Jul 22 11:54:57 2008
From: pav at iki.fi (Pauli Virtanen)
Date: Tue, 22 Jul 2008 15:54:57 +0000 (UTC)
Subject: [SciPy-user] 3D bilinear interpolation...
References: <48860138.2060304@gmail.com>
Message-ID: <g64vsh$5ov$1@ger.gmane.org>

Tue, 22 Jul 2008 17:48:08 +0200, fred wrote:

> Hi all,
> 
> I do know that scipy has a 2D interpolate.interp2d() method but what
> about in 3D ?

If your grid is regularly spaced, you can user ndimage.map_coordinates

http://scipy.org/Cookbook/Interpolation?highlight=%28map_coordinates%29

-- 
Pauli Virtanen



From peridot.faceted at gmail.com  Tue Jul 22 11:55:44 2008
From: peridot.faceted at gmail.com (Anne Archibald)
Date: Tue, 22 Jul 2008 11:55:44 -0400
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <web-105035363@uni-stuttgart.de>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
	<web-105022673@uni-stuttgart.de>
	<e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>
	<web-105027338@uni-stuttgart.de>
	<e76aa17f0807220217m1525f4c3m4f51f5e96fbfb939@mail.gmail.com>
	<web-105029661@uni-stuttgart.de>
	<e76aa17f0807220236q5f0d5fb9ya854a78e6d14218c@mail.gmail.com>
	<web-105035363@uni-stuttgart.de>
Message-ID: <ce557a360807220855j7f2888a3m771cb94158cdc981@mail.gmail.com>

2008/7/22 Nils Wagner <nwagner at iam.uni-stuttgart.de>:

> I made some progress (See test_stft.py).
> But, how do I choose proper values for the parameter
> NFFT,noverlap ?

Suppose you have a time series with 4096 points and duration 1 s. If
you take a real FFT, you will get a complex array of length about
2048. In each bin of this array is a complex number representing the
amplitude and phase of a sinusoid at a particular frequency. The
frequencies are evenly spaced from zero (a constant, necessarily real)
up to (4096/2)/(1 s), that is, 2048 Hz. The spacing is linear, so bin
37 is 37 Hz (in this example). If you don't care about the phase and
just want to find the power, you can take the (squared) amplitude of
the complex number in each bin.

This is useful, but it interprets *everything* as a sinusoid. If you
have an array containing ten seconds of noise, one second of a tone,
and ten more seconds of noise, doing the above will produce a spectrum
in which sinusoids are very carefully chosen to cancel everywhere
except during that one second. Correct, and occasionally useful. but
hard to interpret. Instead, what you often want is something like the
human ear: changes on a short timescale are interpreted in the Fourier
domain, as tones, but changes on a longer scale remain in the time
domain (tones turning on and off). This is a slightly messy sort of
problem; be aware that you are introducing a timescale of transition,
and things get very confusing near that timescale. (For the human ear,
this frequency is somewhere between a few Hertz and maybe thirty
Hertz.) Also, wavelet transforms are designed to deal with this sort
of problem in a systematic way, but their results are more complicated
to interpret still.

So let's use a Fourier transform to compute a "dynamical power
spectrum". For illustrative purposes, let's suppose we have a data
stream at 48000 samples per second and we want to roughly mimic the
human ear. So we want a lowest frequency of about 10 Hz; let's take
chunks of 4096 samples, then. Now a first approach would simply chop
the time sttream into 4096-sample chunks, FFT each one, and take the
(squared) amplitude. This gives you a frequency spectrum with a
spectral resolution of (48 kHz/2)/4096, roughly five Hertz, and a time
resolution of a tenth of a second. If you want more time resolution
take shorter FFTs; if you want more frequency resolution take longer
FFTs. You can't have both (even changing the sample rate just changes
your highest frequency, not the resolution). This is a fundamental
mathematical limitation, the same one that produces the uncertainty
principle in quantum mechanics. But subject to that limitation, you've
got a dynamic power spectrum.

What about that overlap business? Well, suppose you have a tone that
turns on for a tenth of a second. Ideally, you'd see it in one FFT but
not in any of the others, and it'd be nice and strong. But what if its
start time falls in the middle of a chunk? Then you'll see it half as
strong in two adjacent chunks - much less visible. So what people
often do is make successive chunks overlap. Let's say you do that by a
factor of two. Then your tone, instead of falling in the middle of a
chunk, appears half as strong in one chunk, at full power in the next,
and half as strong in the third. Of course misalignments are still
possible, but the worst possible misalignment now puts three-quarters
of the tone in each block. The less power you want to risk losing, the
more overlap you should use (and, of course, the more time and space
it takes). Also, you should note that you won't get any more time
resolution this way - it's like (in fact it is) interpolating the
power spectrum in the time dimension. You get more measurements, but
the narrowest possible peak is still about a tenth of a second wide.

A similar phenomenon occurs in the frequency domain (often called
"scalloping"). If you take such an FFT, you get the coefficients
needed to express your signal in terms of sinusoids at frequencies (48
kHz/2)/4096 * i. What happens if the true frequency of your tone falls
halfway between two of these frequencies? Well, it appears as two
sinusoids, one on either side at about half the power (actually there
are more further out, but they are even weaker). If you're looking to
detect tones, this means that you're more sensitive to some
frequencies than others. You can address this with a similar
"interpolation" procedure. In this case you'd take your 4096 samples
and copy them into an array of size, say, 8192; you'd fill the rest of
the array with the mean value of your samples (or zero, but this
slightly reduces the junk at the low frequencies). Now when you do
your FFT, you get twice as many coefficients; the new ones interpolate
the old ones, reducing the "scalloping". As before, the more
interpolation you do, the less you lose in sensitivity, but the more
it costs in space and time.

How should you choose the FFT size and the overlap? Well, to pick the
FFT size, decide on the time resolution you want to have, keeping in
mind that the reciprocal of this is (roguhly) the frequency resolution
you will get. To pick the overlap, you need to think about space and
time costs, but overlapping by a factor of two is a good habit, or
four if it's not too slow. Whether you want to do interpolation in the
frequency domain is up to you, but it's fairly cheap. I'd go with a
factor of two.

Good luck,
Anne


From fredmfp at gmail.com  Tue Jul 22 12:00:50 2008
From: fredmfp at gmail.com (fred)
Date: Tue, 22 Jul 2008 18:00:50 +0200
Subject: [SciPy-user] 3D bilinear interpolation...
In-Reply-To: <g64vsh$5ov$1@ger.gmane.org>
References: <48860138.2060304@gmail.com> <g64vsh$5ov$1@ger.gmane.org>
Message-ID: <48860432.7040603@gmail.com>

Pauli Virtanen a ?crit :

> If your grid is regularly spaced, you can user ndimage.map_coordinates
Thanks !

Cheers,

-- 
Fred


From nwagner at iam.uni-stuttgart.de  Tue Jul 22 15:26:54 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Tue, 22 Jul 2008 21:26:54 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <ce557a360807220855j7f2888a3m771cb94158cdc981@mail.gmail.com>
References: <web-105021035@uni-stuttgart.de>
	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>
	<web-105022673@uni-stuttgart.de>
	<e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>
	<web-105027338@uni-stuttgart.de>
	<e76aa17f0807220217m1525f4c3m4f51f5e96fbfb939@mail.gmail.com>
	<web-105029661@uni-stuttgart.de>
	<e76aa17f0807220236q5f0d5fb9ya854a78e6d14218c@mail.gmail.com>
	<web-105035363@uni-stuttgart.de>
	<ce557a360807220855j7f2888a3m771cb94158cdc981@mail.gmail.com>
Message-ID: <web-105060297@uni-stuttgart.de>

On Tue, 22 Jul 2008 11:55:44 -0400
  "Anne Archibald" <peridot.faceted at gmail.com> wrote:
> 2008/7/22 Nils Wagner <nwagner at iam.uni-stuttgart.de>:
> 
>> I made some progress (See test_stft.py).
>> But, how do I choose proper values for the parameter
>> NFFT,noverlap ?
> 
> Suppose you have a time series with 4096 points and 
>duration 1 s. If
> you take a real FFT, you will get a complex array of 
>length about
> 2048. In each bin of this array is a complex number 
>representing the
> amplitude and phase of a sinusoid at a particular 
>frequency. The
> frequencies are evenly spaced from zero (a constant, 
>necessarily real)
> up to (4096/2)/(1 s), that is, 2048 Hz. The spacing is 
>linear, so bin
> 37 is 37 Hz (in this example). If you don't care about 
>the phase and
> just want to find the power, you can take the (squared) 
>amplitude of
> the complex number in each bin.
> 
> This is useful, but it interprets *everything* as a 
>sinusoid. If you
> have an array containing ten seconds of noise, one 
>second of a tone,
> and ten more seconds of noise, doing the above will 
>produce a spectrum
> in which sinusoids are very carefully chosen to cancel 
>everywhere
> except during that one second. Correct, and occasionally 
>useful. but
> hard to interpret. Instead, what you often want is 
>something like the
> human ear: changes on a short timescale are interpreted 
>in the Fourier
> domain, as tones, but changes on a longer scale remain 
>in the time
> domain (tones turning on and off). This is a slightly 
>messy sort of
> problem; be aware that you are introducing a timescale 
>of transition,
> and things get very confusing near that timescale. (For 
>the human ear,
> this frequency is somewhere between a few Hertz and 
>maybe thirty
> Hertz.) Also, wavelet transforms are designed to deal 
>with this sort
> of problem in a systematic way, but their results are 
>more complicated
> to interpret still.
> 
> So let's use a Fourier transform to compute a "dynamical 
>power
> spectrum". For illustrative purposes, let's suppose we 
>have a data
> stream at 48000 samples per second and we want to 
>roughly mimic the
> human ear. So we want a lowest frequency of about 10 Hz; 
>let's take
> chunks of 4096 samples, then. Now a first approach would 
>simply chop
> the time sttream into 4096-sample chunks, FFT each one, 
>and take the
> (squared) amplitude. This gives you a frequency spectrum 
>with a
> spectral resolution of (48 kHz/2)/4096, roughly five 
>Hertz, and a time
> resolution of a tenth of a second. If you want more time 
>resolution
> take shorter FFTs; if you want more frequency resolution 
>take longer
>FFTs. You can't have both (even changing the sample rate 
>just changes
> your highest frequency, not the resolution). This is a 
>fundamental
> mathematical limitation, the same one that produces the 
>uncertainty
> principle in quantum mechanics. But subject to that 
>limitation, you've
> got a dynamic power spectrum.
> 
> What about that overlap business? Well, suppose you have 
>a tone that
> turns on for a tenth of a second. Ideally, you'd see it 
>in one FFT but
> not in any of the others, and it'd be nice and strong. 
>But what if its
> start time falls in the middle of a chunk? Then you'll 
>see it half as
> strong in two adjacent chunks - much less visible. So 
>what people
> often do is make successive chunks overlap. Let's say 
>you do that by a
> factor of two. Then your tone, instead of falling in the 
>middle of a
> chunk, appears half as strong in one chunk, at full 
>power in the next,
> and half as strong in the third. Of course misalignments 
>are still
> possible, but the worst possible misalignment now puts 
>three-quarters
> of the tone in each block. The less power you want to 
>risk losing, the
> more overlap you should use (and, of course, the more 
>time and space
> it takes). Also, you should note that you won't get any 
>more time
> resolution this way - it's like (in fact it is) 
>interpolating the
> power spectrum in the time dimension. You get more 
>measurements, but
> the narrowest possible peak is still about a tenth of a 
>second wide.
> 
> A similar phenomenon occurs in the frequency domain 
>(often called
> "scalloping"). If you take such an FFT, you get the 
>coefficients
> needed to express your signal in terms of sinusoids at 
>frequencies (48
> kHz/2)/4096 * i. What happens if the true frequency of 
>your tone falls
> halfway between two of these frequencies? Well, it 
>appears as two
> sinusoids, one on either side at about half the power 
>(actually there
> are more further out, but they are even weaker). If 
>you're looking to
> detect tones, this means that you're more sensitive to 
>some
> frequencies than others. You can address this with a 
>similar
> "interpolation" procedure. In this case you'd take your 
>4096 samples
> and copy them into an array of size, say, 8192; you'd 
>fill the rest of
> the array with the mean value of your samples (or zero, 
>but this
> slightly reduces the junk at the low frequencies). Now 
>when you do
> your FFT, you get twice as many coefficients; the new 
>ones interpolate
> the old ones, reducing the "scalloping". As before, the 
>more
> interpolation you do, the less you lose in sensitivity, 
>but the more
> it costs in space and time.
> 
> How should you choose the FFT size and the overlap? 
>Well, to pick the
>FFT size, decide on the time resolution you want to have, 
>keeping in
> mind that the reciprocal of this is (roguhly) the 
>frequency resolution
> you will get. To pick the overlap, you need to think 
>about space and
> time costs, but overlapping by a factor of two is a good 
>habit, or
> four if it's not too slow. Whether you want to do 
>interpolation in the
> frequency domain is up to you, but it's fairly cheap. 
>I'd go with a
> factor of two.
> 
> Good luck,
> Anne
  
Hi Anne,

Thank you very much for your reply, which is very rich in 
content. It will take some time to "stomach" it though ;)

Are you aware of a good reference to delve into this topic 
?

Best wishes

                 Nils


From matthieu.brucher at gmail.com  Tue Jul 22 15:50:10 2008
From: matthieu.brucher at gmail.com (Matthieu Brucher)
Date: Tue, 22 Jul 2008 21:50:10 +0200
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <web-105060297@uni-stuttgart.de>
References: <web-105021035@uni-stuttgart.de> <web-105022673@uni-stuttgart.de>
	<e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>
	<web-105027338@uni-stuttgart.de>
	<e76aa17f0807220217m1525f4c3m4f51f5e96fbfb939@mail.gmail.com>
	<web-105029661@uni-stuttgart.de>
	<e76aa17f0807220236q5f0d5fb9ya854a78e6d14218c@mail.gmail.com>
	<web-105035363@uni-stuttgart.de>
	<ce557a360807220855j7f2888a3m771cb94158cdc981@mail.gmail.com>
	<web-105060297@uni-stuttgart.de>
Message-ID: <e76aa17f0807221250m10ca437dm3539f35e6bb9a147@mail.gmail.com>

> Hi Anne,
>
> Thank you very much for your reply, which is very rich in
> content. It will take some time to "stomach" it though ;)
>
> Are you aware of a good reference to delve into this topic
> ?

Any book on signal processing ;)

I have to correct something. When you do an FFT of a 4096 samples real
signal, you get 2047 meaning full complex values and 2 real ones.

Matthieu
-- 
French PhD student
Website : http://matthieu-brucher.developpez.com/
Blogs : http://matt.eifelle.com and http://blog.developpez.com/?blog=92
LinkedIn : http://www.linkedin.com/in/matthieubrucher


From timmichelsen at gmx-topmail.de  Tue Jul 22 18:46:06 2008
From: timmichelsen at gmx-topmail.de (Tim Michelsen)
Date: Wed, 23 Jul 2008 00:46:06 +0200
Subject: [SciPy-user] Statistics advise with scipy
In-Reply-To: <200807212307.31208.pgmdevlist@gmail.com>
References: <da09baf80807201956r6f92db05vd5262a4236bc3ae2@mail.gmail.com>	<200807211312.40202.pgmdevlist@gmail.com>	<da09baf80807211949i7d075d22o277b7145f31ab0e5@mail.gmail.com>
	<200807212307.31208.pgmdevlist@gmail.com>
Message-ID: <g65nve$q04$1@ger.gmane.org>

>> My data is not normal. Do you know robusts method in scipy ? Or maybe in an
>> other python module ?
> 
> Mmh, I'm sure you could implement some yourself. That way, we could start 
> another scikits. There are already some winsorization and trimming functions 
> in scipy.stats.
> Alternatively, you can try to use R and numpy through rpy: 
> http://rpy.sourceforge.net/
Dider,
may I ask you to give some feedback what method worked for you?
I am also working with the problem of removing outliners etc. from data.

Thanks in advance,
Timmie



From didier.rano at gmail.com  Tue Jul 22 18:55:32 2008
From: didier.rano at gmail.com (didier rano)
Date: Tue, 22 Jul 2008 18:55:32 -0400
Subject: [SciPy-user] Statistics advise with scipy
In-Reply-To: <g65nve$q04$1@ger.gmane.org>
References: <da09baf80807201956r6f92db05vd5262a4236bc3ae2@mail.gmail.com>
	<200807211312.40202.pgmdevlist@gmail.com>
	<da09baf80807211949i7d075d22o277b7145f31ab0e5@mail.gmail.com>
	<200807212307.31208.pgmdevlist@gmail.com> <g65nve$q04$1@ger.gmane.org>
Message-ID: <da09baf80807221555j26b532c9v8cd2a3c3ad425136@mail.gmail.com>

Hi,
I haven't found yet a solution to my problem. But I am reading a good
article about removing outliers:
http://www.lcgceurope.com/lcgceurope/data/articlestandard//lcgceurope/502001/4509/article.pdf
Now, I need to experiment methods described in this article.

Thanks
Didier Rano

2008/7/22 Tim Michelsen <timmichelsen at gmx-topmail.de>:

> >> My data is not normal. Do you know robusts method in scipy ? Or maybe in
> an
> >> other python module ?
> >
> > Mmh, I'm sure you could implement some yourself. That way, we could start
> > another scikits. There are already some winsorization and trimming
> functions
> > in scipy.stats.
> > Alternatively, you can try to use R and numpy through rpy:
> > http://rpy.sourceforge.net/
> Dider,
> may I ask you to give some feedback what method worked for you?
> I am also working with the problem of removing outliners etc. from data.
>
> Thanks in advance,
> Timmie
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Didier Rano
didier.rano at gmail.com
http://www.jaxtr.com/didierrano
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080722/7b53c83a/attachment.html>

From smcphers at stanford.edu  Wed Jul 23 03:07:58 2008
From: smcphers at stanford.edu (Selwyn-Lloyd McPherson)
Date: Wed, 23 Jul 2008 00:07:58 -0700
Subject: [SciPy-user] SciPy builds and installs but won't import due to f2c
Message-ID: <06BDD519-19A2-4BE1-B6B7-C3BAB3AD799E@stanford.edu>

Hi everyone,

I've been working on getting SciPy installed on my machine and it  
seems almost there. I've got a x86_64 system running Red Hat and more  
or less followed http://www.scipy.org/Installing_SciPy/Linux#head-1c4018a51422706809ee96a4db03ca0669f5f6d1 
  to build and install SciPy. After dealing with the whole fPIC mess  
that the 64-bit processor requires the install seems to finish but  
when I try to import, I get:

 >>> from scipy import *
Traceback (most recent call last):
   File "<stdin>", line 1, in ?
   File "/usr/lib64/python2.4/site-packages/scipy/linalg/__init__.py",  
line 8, in ?
     from basic import *
   File "/usr/lib64/python2.4/site-packages/scipy/linalg/basic.py",  
line 17, in ?
     from lapack import get_lapack_funcs
   File "/usr/lib64/python2.4/site-packages/scipy/linalg/lapack.py",  
line 17, in ?
     from scipy.linalg import flapack
ImportError: /usr/lib/libf2c.so.0: undefined symbol: MAIN__


Apparently a problem with f2c. . . Have any ideas?

Thanks so much for reading!


Selwyn-Lloyd


From massimo.sandal at unibo.it  Wed Jul 23 06:58:00 2008
From: massimo.sandal at unibo.it (massimo sandal)
Date: Wed, 23 Jul 2008 12:58:00 +0200
Subject: [SciPy-user] scipy.stats.gaussian_kde for 2d kernel density
	estimation
Message-ID: <48870EB8.8010208@unibo.it>

Hi,

I can't figure out how to do bivariate kernel density estimation with 
the scipy.stats.gaussian_kde module .1D evaluation seems working, but 2D 
evaluation escapes me.

I have two vectors representing x and y coordinates of points:

xvect=[72.11,81.52,66.46,52.34,81.12,76.83,...]
yvect=[26.91,17.39,28.84,15.05,10.21,26.42,...]

The problem is: how do I build the grid to evaluate the points? I would 
expect that he wants an x range and a y range (something like 
xrange=numpy.arange(0,100,1) ) from which he builds and evaluates the 
grid. However this does not work, and I do not understand how to create 
a proper 2d-grid that has the correct coordinate information.

Where can I look for information? The built-in documentation seems a bit 
terse on this point.

Thanks a lot,

m.
-- 
Massimo Sandal , Ph.D.
University of Bologna
Department of Biochemistry "G.Moruzzi"

snail mail:
Via Irnerio 48, 40126 Bologna, Italy

email:
massimo.sandal at unibo.it

web:
http://www.biocfarm.unibo.it/samori/people/sandal.html

tel: +39-051-2094388
fax: +39-051-2094387
-------------- next part --------------
A non-text attachment was scrubbed...
Name: massimo_sandal.vcf
Type: text/x-vcard
Size: 274 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080723/a9c2d299/attachment.vcf>

From wnbell at gmail.com  Wed Jul 23 07:14:11 2008
From: wnbell at gmail.com (Nathan Bell)
Date: Wed, 23 Jul 2008 06:14:11 -0500
Subject: [SciPy-user] sparsetools involvement? (was -lf77compat build
	failure)
In-Reply-To: <d05265cb0807180637s6307814em99b2103890de5920@mail.gmail.com>
References: <001301c8e84f$2dfdd750$89f985f0$@auburn.edu>
	<d05265cb0807180637s6307814em99b2103890de5920@mail.gmail.com>
Message-ID: <d05265cb0807230414s1c52b7bdp51ab753a131e8058@mail.gmail.com>

On Fri, Jul 18, 2008 at 8:37 AM, Nathan Bell <wnbell at gmail.com> wrote:
>
> Phil, could you try compiling SciPy from SVN?  Much of sparsetools has
> changed since the last release.
>
> I believe there was a missing #include <algorithm> in sparsetools.h in
> the release you're using.
>

Phil, I was mistaken about the missing #include <algorithm>, I believe
the fix is to add #include <functional> instead.  Here's a related
ticket:
http://scipy.org/scipy/scipy/ticket/706

I still would appreciate you testing SciPy from SVN on your system.

-- 
Nathan Bell wnbell at gmail.com
http://graphics.cs.uiuc.edu/~wnbell/


From dave.hirschfeld at gmail.com  Wed Jul 23 09:12:54 2008
From: dave.hirschfeld at gmail.com (Dave)
Date: Wed, 23 Jul 2008 13:12:54 +0000 (UTC)
Subject: [SciPy-user]
	=?utf-8?q?scipy=2Estats=2Egaussian=5Fkde_for_2d_kern?=
	=?utf-8?q?el_density=09estimation?=
References: <48870EB8.8010208@unibo.it>
Message-ID: <loom.20080723T131047-645@post.gmane.org>

massimo sandal <massimo.sandal <at> unibo.it> writes:

> 
> Hi,
> 
> I can't figure out how to do bivariate kernel density estimation with 
> the scipy.stats.gaussian_kde module .1D evaluation seems working, but 2D 
> evaluation escapes me.
> 
> I have two vectors representing x and y coordinates of points:
> 
> xvect=[72.11,81.52,66.46,52.34,81.12,76.83,...]
> yvect=[26.91,17.39,28.84,15.05,10.21,26.42,...]
> 
> The problem is: how do I build the grid to evaluate the points? 

Hopefully the example below will help...

-Dave

import numpy as np
import scipy.stats as stats
from matplotlib.pyplot import imshow

# Create some dummy data
rvs = np.append(stats.norm.rvs(loc=2,scale=1,size=(2000,1)),
                stats.norm.rvs(loc=0,scale=3,size=(2000,1)),
                axis=1)

kde = stats.kde.gaussian_kde(rvs.T)

# Regular grid to evaluate kde upon
x_flat = np.r_[rvs[:,0].min():rvs[:,0].max():128j]
y_flat = np.r_[rvs[:,1].min():rvs[:,1].max():128j]
x,y = np.meshgrid(x_flat,y_flat)
grid_coords = np.append(x.reshape(-1,1),y.reshape(-1,1),axis=1)

z = kde(grid_coords.T)
z = z.reshape(128,128)

imshow(z,aspect=x_flat.ptp()/y_flat.ptp())




From dfranci at seas.upenn.edu  Wed Jul 23 09:58:02 2008
From: dfranci at seas.upenn.edu (Frank Lagor)
Date: Wed, 23 Jul 2008 09:58:02 -0400
Subject: [SciPy-user] SciPy builds and installs but won't import due to
	f2c
In-Reply-To: <06BDD519-19A2-4BE1-B6B7-C3BAB3AD799E@stanford.edu>
References: <06BDD519-19A2-4BE1-B6B7-C3BAB3AD799E@stanford.edu>
Message-ID: <9fddf64a0807230658n64438a98y5192822ea9706d2e@mail.gmail.com>

Dear Selwyn-Lloyd,

I hope I can help some, but I am not at all an expert on these things.  I
had what seemed like a similar problem (VERY similar, if I can remember).
Here's what got me running-- my numpy installation was based on the lapack
version that was causing my problem (I don't know if it was a lapack or
scipy problem but whatever...) Anyways, I re-did the install of numpy, and I
did not specify what lapack to use in the install.  I you want to do this,
be very careful that you do not have a LAPACK environment variable which is
pointing somewhere and you don't know it.  If you do, just unset it (unset
LAPACK).  The numpy install should not find any LAPACK (you'll be able to
read that on the screen during install), and it will use some default one.
Then re-do the scipy install and it should work.  This is sort of a hack
fix, because you will be using a non-optimized version of Lapack (assuming
you wanted to use ATLAS or something), but it will get you running.  One
last thing, unfortunately I believe the default version of lapack installed
is lapack lite or something like that (not the full lapack).  This may cause
you to not have full functionality.  It has only caused me problems on one
function: interp2d for 2d interpolation.  Also, you are on Red Hat, why
don't you use some red hat package manager or builtin software to install
this stuff?

That's all I've got,
Frank

On Wed, Jul 23, 2008 at 3:07 AM, Selwyn-Lloyd McPherson <
smcphers at stanford.edu> wrote:

> Hi everyone,
>
> I've been working on getting SciPy installed on my machine and it
> seems almost there. I've got a x86_64 system running Red Hat and more
> or less followed
> http://www.scipy.org/Installing_SciPy/Linux#head-1c4018a51422706809ee96a4db03ca0669f5f6d1
>  to build and install SciPy. After dealing with the whole fPIC mess
> that the 64-bit processor requires the install seems to finish but
> when I try to import, I get:
>
>  >>> from scipy import *
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
>   File "/usr/lib64/python2.4/site-packages/scipy/linalg/__init__.py",
> line 8, in ?
>     from basic import *
>   File "/usr/lib64/python2.4/site-packages/scipy/linalg/basic.py",
> line 17, in ?
>     from lapack import get_lapack_funcs
>   File "/usr/lib64/python2.4/site-packages/scipy/linalg/lapack.py",
> line 17, in ?
>     from scipy.linalg import flapack
> ImportError: /usr/lib/libf2c.so.0: undefined symbol: MAIN__
>
>
> Apparently a problem with f2c. . . Have any ideas?
>
> Thanks so much for reading!
>
>
> Selwyn-Lloyd
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Frank Lagor
Ph.D. Candidate
Mechanical Engineering and Applied Mechanics
University of Pennsylvania
dfranci at seas.upenn.edu
215.898.3144
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080723/f5c72933/attachment.html>

From david.huard at gmail.com  Wed Jul 23 11:01:30 2008
From: david.huard at gmail.com (David Huard)
Date: Wed, 23 Jul 2008 11:01:30 -0400
Subject: [SciPy-user] Statistics advise with scipy
In-Reply-To: <da09baf80807221555j26b532c9v8cd2a3c3ad425136@mail.gmail.com>
References: <da09baf80807201956r6f92db05vd5262a4236bc3ae2@mail.gmail.com>
	<200807211312.40202.pgmdevlist@gmail.com>
	<da09baf80807211949i7d075d22o277b7145f31ab0e5@mail.gmail.com>
	<200807212307.31208.pgmdevlist@gmail.com> <g65nve$q04$1@ger.gmane.org>
	<da09baf80807221555j26b532c9v8cd2a3c3ad425136@mail.gmail.com>
Message-ID: <91cf711d0807230801o3f5b3b1ahc07dd0f53251b0a0@mail.gmail.com>

I've had some success with the following:

1. Define a simple statistical model for your data.  That is, from the
previous data, define a distribution for the probability of the next
point.
2. Define a cutoff probability separating valid data from outliers.
3. For each datum, compute its probability based on previous data, and
tag it as valid or outlier.

The advantage is that you can start with a simple statistical model (
for example a gaussian centered on the last valid entry ) and
customize it as you find cases that are not well handled.

David

2008/7/22 didier rano <didier.rano at gmail.com>:
> Hi,
> I haven't found yet a solution to my problem. But I am reading a good
> article about removing
> outliers: http://www.lcgceurope.com/lcgceurope/data/articlestandard//lcgceurope/502001/4509/article.pdf
> Now, I need to experiment methods described in this article.
> Thanks
> Didier Rano
>
> 2008/7/22 Tim Michelsen <timmichelsen at gmx-topmail.de>:
>>
>> >> My data is not normal. Do you know robusts method in scipy ? Or maybe
>> >> in an
>> >> other python module ?
>> >
>> > Mmh, I'm sure you could implement some yourself. That way, we could
>> > start
>> > another scikits. There are already some winsorization and trimming
>> > functions
>> > in scipy.stats.
>> > Alternatively, you can try to use R and numpy through rpy:
>> > http://rpy.sourceforge.net/
>> Dider,
>> may I ask you to give some feedback what method worked for you?
>> I am also working with the problem of removing outliners etc. from data.
>>
>> Thanks in advance,
>> Timmie
>>
>> _______________________________________________
>> SciPy-user mailing list
>> SciPy-user at scipy.org
>> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
>
> --
> Didier Rano
> didier.rano at gmail.com
> http://www.jaxtr.com/didierrano
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>


From stefan at sun.ac.za  Wed Jul 23 11:16:57 2008
From: stefan at sun.ac.za (=?ISO-8859-1?Q?St=E9fan_van_der_Walt?=)
Date: Wed, 23 Jul 2008 17:16:57 +0200
Subject: [SciPy-user] Statistics advise with scipy
In-Reply-To: <91cf711d0807230801o3f5b3b1ahc07dd0f53251b0a0@mail.gmail.com>
References: <da09baf80807201956r6f92db05vd5262a4236bc3ae2@mail.gmail.com>
	<200807211312.40202.pgmdevlist@gmail.com>
	<da09baf80807211949i7d075d22o277b7145f31ab0e5@mail.gmail.com>
	<200807212307.31208.pgmdevlist@gmail.com> <g65nve$q04$1@ger.gmane.org>
	<da09baf80807221555j26b532c9v8cd2a3c3ad425136@mail.gmail.com>
	<91cf711d0807230801o3f5b3b1ahc07dd0f53251b0a0@mail.gmail.com>
Message-ID: <9457e7c80807230816w3e25a20are4dc4f6a8d732643@mail.gmail.com>

2008/7/23 David Huard <david.huard at gmail.com>:
> I've had some success with the following:
>
> 1. Define a simple statistical model for your data.  That is, from the
> previous data, define a distribution for the probability of the next
> point.
> 2. Define a cutoff probability separating valid data from outliers.
> 3. For each datum, compute its probability based on previous data, and
> tag it as valid or outlier.
>
> The advantage is that you can start with a simple statistical model (
> for example a gaussian centered on the last valid entry ) and
> customize it as you find cases that are not well handled.

There's also

http://www.scipy.org/Cookbook/RANSAC

Cheers
St?fan


From christian-baehnisch at gmx.de  Wed Jul 23 16:46:14 2008
From: christian-baehnisch at gmx.de (Christian =?utf-8?q?B=C3=A4hnisch?=)
Date: Wed, 23 Jul 2008 22:46:14 +0200
Subject: [SciPy-user] spsolve
Message-ID: <200807232246.14492.christian-baehnisch@gmx.de>

Hello,

I'm using "scipy.linsolve.spsolve". Unfortunately, I have not been able to
figure out if  it is possible to get the number of iterations actually
used by the method. I would be very glad if somebody could help me on
this topic

greetings, Christian



From aisaac at american.edu  Wed Jul 23 18:24:50 2008
From: aisaac at american.edu (Alan G Isaac)
Date: Wed, 23 Jul 2008 18:24:50 -0400
Subject: [SciPy-user] Fwd: Re: permission requested
Message-ID: <Mahogany-0.67.0-948-20080723-182450.00@american.edu>

822



From smcphers at stanford.edu  Wed Jul 23 18:21:05 2008
From: smcphers at stanford.edu (Selwyn-Lloyd McPherson)
Date: Wed, 23 Jul 2008 15:21:05 -0700
Subject: [SciPy-user] SciPy builds and installs but won't import due to
	f2c
In-Reply-To: <9fddf64a0807230658n64438a98y5192822ea9706d2e@mail.gmail.com>
References: <06BDD519-19A2-4BE1-B6B7-C3BAB3AD799E@stanford.edu>
	<9fddf64a0807230658n64438a98y5192822ea9706d2e@mail.gmail.com>
Message-ID: <1AB18010-2249-4FEA-A6D3-96C001164ABF@stanford.edu>

Unfortunately, scipy doesn't seem to want to build without LAPACK --  
I've tried lots of different LAPACK distributions. I've tried using  
ATLAS with LAPACK and ATLAS without LAPACK (built LAPACK standalone)  
with lots of different flags but nothing seems to work. I've tried  
rebuilding the f2c libraries but it still comes up with the "undefined  
symbol" MAIN__ error. I've read here that inserting a

int MAIN__( )
{ return(0);
}

in the codebase somewhere might fix the problem but I'm not sure where  
it could go.

Is there any reason that scipy / lapack needs f2c? Can't it just use  
f2py?


Selwyn-Lloyd


On Jul 23, 2008, at 6:58 Antemeridian, Frank Lagor wrote:

> Dear Selwyn-Lloyd,
>
> I hope I can help some, but I am not at all an expert on these  
> things.  I had what seemed like a similar problem (VERY similar, if  
> I can remember).  Here's what got me running-- my numpy installation  
> was based on the lapack version that was causing my problem (I don't  
> know if it was a lapack or scipy problem but whatever...) Anyways, I  
> re-did the install of numpy, and I did not specify what lapack to  
> use in the install.  I you want to do this, be very careful that you  
> do not have a LAPACK environment variable which is  pointing  
> somewhere and you don't know it.  If you do, just unset it (unset  
> LAPACK).  The numpy install should not find any LAPACK (you'll be  
> able to read that on the screen during install), and it will use  
> some default one. Then re-do the scipy install and it should work.   
> This is sort of a hack fix, because you will be using a non- 
> optimized version of Lapack (assuming you wanted to use ATLAS or  
> something), but it will get you running.  One last thing,  
> unfortunately I believe the default version of lapack installed is  
> lapack lite or something like that (not the full lapack).  This may  
> cause you to not have full functionality.  It has only caused me  
> problems on one function: interp2d for 2d interpolation.  Also, you  
> are on Red Hat, why don't you use some red hat package manager or  
> builtin software to install this stuff?
>
> That's all I've got,
> Frank
>
> On Wed, Jul 23, 2008 at 3:07 AM, Selwyn-Lloyd McPherson <smcphers at stanford.edu 
> > wrote:
> Hi everyone,
>
> I've been working on getting SciPy installed on my machine and it
> seems almost there. I've got a x86_64 system running Red Hat and more
> or less followed http://www.scipy.org/Installing_SciPy/Linux#head-1c4018a51422706809ee96a4db03ca0669f5f6d1
>  to build and install SciPy. After dealing with the whole fPIC mess
> that the 64-bit processor requires the install seems to finish but
> when I try to import, I get:
>
>  >>> from scipy import *
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
>   File "/usr/lib64/python2.4/site-packages/scipy/linalg/__init__.py",
> line 8, in ?
>     from basic import *
>   File "/usr/lib64/python2.4/site-packages/scipy/linalg/basic.py",
> line 17, in ?
>     from lapack import get_lapack_funcs
>   File "/usr/lib64/python2.4/site-packages/scipy/linalg/lapack.py",
> line 17, in ?
>     from scipy.linalg import flapack
> ImportError: /usr/lib/libf2c.so.0: undefined symbol: MAIN__
>
>
> Apparently a problem with f2c. . . Have any ideas?
>
> Thanks so much for reading!
>
>
> Selwyn-Lloyd
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
>
> -- 
> Frank Lagor
> Ph.D. Candidate
> Mechanical Engineering and Applied Mechanics
> University of Pennsylvania
> dfranci at seas.upenn.edu
> 215.898.3144
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080723/d056cb4d/attachment.html>

From dfranci at seas.upenn.edu  Wed Jul 23 18:39:19 2008
From: dfranci at seas.upenn.edu (Frank Lagor)
Date: Wed, 23 Jul 2008 18:39:19 -0400
Subject: [SciPy-user] SciPy builds and installs but won't import due to
	f2c
In-Reply-To: <1AB18010-2249-4FEA-A6D3-96C001164ABF@stanford.edu>
References: <06BDD519-19A2-4BE1-B6B7-C3BAB3AD799E@stanford.edu>
	<9fddf64a0807230658n64438a98y5192822ea9706d2e@mail.gmail.com>
	<1AB18010-2249-4FEA-A6D3-96C001164ABF@stanford.edu>
Message-ID: <9fddf64a0807231539x262d38cj918926e957f90161@mail.gmail.com>

 Honestly, I don't know.  Here is a tip though that caused me a problem with
numpy and scipy installs and I just want to make sure you are aware of it.
Maybe it will help...

Sometimes when you execute the step "python setup.py build" when you have
been trying a bunch of different things, it does not rebuild the files as
you want it to.  So that when you do "python setup.py install" it is not
installing code that was built with the options you wanted.  To avoid this
pitful, EVERY time you try to install execute:
python setup.py clean
and,
rm -rf build  (This removes the old build directory)

Perhaps that was helpful... :(
Good Luck,
Frank

On Wed, Jul 23, 2008 at 6:21 PM, Selwyn-Lloyd McPherson <
smcphers at stanford.edu> wrote:

> Unfortunately, scipy doesn't seem to want to build without LAPACK -- I've
> tried lots of different LAPACK distributions. I've tried using ATLAS with
> LAPACK and ATLAS without LAPACK (built LAPACK standalone) with lots of
> different flags but nothing seems to work. I've tried rebuilding the f2c
> libraries but it still comes up with the "undefined symbol" MAIN__ error.
> I've read here<http://www.velocityreviews.com/forums/t332246-using-fortran-libraries-from-python.html> that
> inserting a
> int MAIN__( )
> { return(0);
> }
>
> in the codebase somewhere might fix the problem but I'm not sure where it
> could go.
>
> Is there any reason that scipy / lapack needs f2c? Can't it just use f2py?
>
> Selwyn-Lloyd
>
>
>
> On Jul 23, 2008, at 6:58 Antemeridian, Frank Lagor wrote:
>
> Dear Selwyn-Lloyd,
>
> I hope I can help some, but I am not at all an expert on these things.  I
> had what seemed like a similar problem (VERY similar, if I can remember).
> Here's what got me running-- my numpy installation was based on the lapack
> version that was causing my problem (I don't know if it was a lapack or
> scipy problem but whatever...) Anyways, I re-did the install of numpy, and I
> did not specify what lapack to use in the install.  I you want to do this,
> be very careful that you do not have a LAPACK environment variable which is
> pointing somewhere and you don't know it.  If you do, just unset it (unset
> LAPACK).  The numpy install should not find any LAPACK (you'll be able to
> read that on the screen during install), and it will use some default one.
> Then re-do the scipy install and it should work.  This is sort of a hack
> fix, because you will be using a non-optimized version of Lapack (assuming
> you wanted to use ATLAS or something), but it will get you running.  One
> last thing, unfortunately I believe the default version of lapack installed
> is lapack lite or something like that (not the full lapack).  This may cause
> you to not have full functionality.  It has only caused me problems on one
> function: interp2d for 2d interpolation.  Also, you are on Red Hat, why
> don't you use some red hat package manager or builtin software to install
> this stuff?
>
> That's all I've got,
> Frank
>
> On Wed, Jul 23, 2008 at 3:07 AM, Selwyn-Lloyd McPherson <
> smcphers at stanford.edu> wrote:
>
>> Hi everyone,
>>
>> I've been working on getting SciPy installed on my machine and it
>> seems almost there. I've got a x86_64 system running Red Hat and more
>> or less followed
>> http://www.scipy.org/Installing_SciPy/Linux#head-1c4018a51422706809ee96a4db03ca0669f5f6d1
>>  to build and install SciPy. After dealing with the whole fPIC mess
>> that the 64-bit processor requires the install seems to finish but
>> when I try to import, I get:
>>
>>  >>> from scipy import *
>> Traceback (most recent call last):
>>   File "<stdin>", line 1, in ?
>>   File "/usr/lib64/python2.4/site-packages/scipy/linalg/__init__.py",
>> line 8, in ?
>>     from basic import *
>>   File "/usr/lib64/python2.4/site-packages/scipy/linalg/basic.py",
>> line 17, in ?
>>     from lapack import get_lapack_funcs
>>   File "/usr/lib64/python2.4/site-packages/scipy/linalg/lapack.py",
>> line 17, in ?
>>     from scipy.linalg import flapack
>> ImportError: /usr/lib/libf2c.so.0: undefined symbol: MAIN__
>>
>>
>> Apparently a problem with f2c. . . Have any ideas?
>>
>> Thanks so much for reading!
>>
>>
>> Selwyn-Lloyd
>> _______________________________________________
>> SciPy-user mailing list
>> SciPy-user at scipy.org
>> http://projects.scipy.org/mailman/listinfo/scipy-user
>>
>
>
>
> --
> Frank Lagor
> Ph.D. Candidate
> Mechanical Engineering and Applied Mechanics
> University of Pennsylvania
> dfranci at seas.upenn.edu
> 215.898.3144
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>


-- 
Frank Lagor
Ph.D. Candidate
Mechanical Engineering and Applied Mechanics
University of Pennsylvania
dfranci at seas.upenn.edu
215.898.3144
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080723/ff10ff90/attachment.html>

From robert.kern at gmail.com  Wed Jul 23 18:45:23 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Wed, 23 Jul 2008 17:45:23 -0500
Subject: [SciPy-user] SciPy builds and installs but won't import due to
	f2c
In-Reply-To: <1AB18010-2249-4FEA-A6D3-96C001164ABF@stanford.edu>
References: <06BDD519-19A2-4BE1-B6B7-C3BAB3AD799E@stanford.edu>
	<9fddf64a0807230658n64438a98y5192822ea9706d2e@mail.gmail.com>
	<1AB18010-2249-4FEA-A6D3-96C001164ABF@stanford.edu>
Message-ID: <3d375d730807231545k2c5b0fb5m2e997750cb8c6809@mail.gmail.com>

On Wed, Jul 23, 2008 at 17:21, Selwyn-Lloyd McPherson
<smcphers at stanford.edu> wrote:
> Unfortunately, scipy doesn't seem to want to build without LAPACK -- I've
> tried lots of different LAPACK distributions. I've tried using ATLAS with
> LAPACK and ATLAS without LAPACK (built LAPACK standalone) with lots of
> different flags but nothing seems to work. I've tried rebuilding the f2c
> libraries but it still comes up with the "undefined symbol" MAIN__ error.
> I've read here that inserting a
> int MAIN__( )
> { return(0);
> }
> in the codebase somewhere might fix the problem but I'm not sure where it
> could go.
> Is there any reason that scipy / lapack needs f2c? Can't it just use f2py?

It doesn't use f2c to convert the Fortran code to C. However, some
Fortran compilers use libf2c as a runtime library. Without more
details about how you built things, I don't really know why you are
having linking problems. One thing to consider is that using the
environment variable LDFLAGS replaces all link flags when building
Fortran extension modules. So if you are using $LDFLAGS, then you may
have removed important link flags.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From wnbell at gmail.com  Wed Jul 23 22:48:41 2008
From: wnbell at gmail.com (Nathan Bell)
Date: Wed, 23 Jul 2008 21:48:41 -0500
Subject: [SciPy-user] spsolve
In-Reply-To: <200807232246.14492.christian-baehnisch@gmx.de>
References: <200807232246.14492.christian-baehnisch@gmx.de>
Message-ID: <d05265cb0807231948n5e9062daq9a70f829cb065438@mail.gmail.com>

On Wed, Jul 23, 2008 at 3:46 PM, Christian B?hnisch
<christian-baehnisch at gmx.de> wrote:
>
> I'm using "scipy.linsolve.spsolve". Unfortunately, I have not been able to
> figure out if  it is possible to get the number of iterations actually
> used by the method. I would be very glad if somebody could help me on
> this topic
>

Christian, the sparse solvers in scipy.linsolve are what is known as
"direct" methods (as opposed to "iterative" methods).  Rather than
iteratively solving a linear system (like the Conjugate Gradient
method), direct solvers just compute a sparse LU factorization of the
matrix in order to solve the system.  In order to do this efficiently,
and to keep L and U as sparse as possible, they incorporate a number
of fairly sophisticated algorithms/heuristics.

In SciPy 0.6 scipy.linsolve ultimately calls either the SuperLU or
UMFPACK packages to perform the solve.

If you're interested in iterative solvers, there are several in
scipy.linalg.iterative.

-- 
Nathan Bell wnbell at gmail.com
http://graphics.cs.uiuc.edu/~wnbell/


From dfranci at seas.upenn.edu  Wed Jul 23 23:02:05 2008
From: dfranci at seas.upenn.edu (Frank Lagor)
Date: Wed, 23 Jul 2008 23:02:05 -0400
Subject: [SciPy-user] error ellipse
Message-ID: <9fddf64a0807232002g31ba7a80nc4a26291c8590caf@mail.gmail.com>

Does anyone know if there is a function in scipy.stats for somewhere else to
plot a 2D error ellipse on a scatter plot of some data given the covariance
matrix?  I have not been able to find one.  Thanks in advance for your help,


Frank
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080723/652e631a/attachment.html>

From david at ar.media.kyoto-u.ac.jp  Wed Jul 23 23:23:55 2008
From: david at ar.media.kyoto-u.ac.jp (David Cournapeau)
Date: Thu, 24 Jul 2008 12:23:55 +0900
Subject: [SciPy-user] Frequency content of a transient signal
In-Reply-To: <web-105027338@uni-stuttgart.de>
References: <web-105021035@uni-stuttgart.de>	<e76aa17f0807220015w138ff3e7u81f54bcc45c4f4a1@mail.gmail.com>	<web-105022673@uni-stuttgart.de>	<e76aa17f0807220059q56cdbbc6q59fd6de48eed130b@mail.gmail.com>
	<web-105027338@uni-stuttgart.de>
Message-ID: <4887F5CB.6080903@ar.media.kyoto-u.ac.jp>

Nils Wagner wrote:
> Fine. Let's start with the example from the link
>
> How do I produce the nice spectrograms
>   

matplotlib has a function specgram, which mimics the matlab's one, and
should do what you want. You will need to adapt the window size to the
timescale of your transients, while keeping in mind that the frequency
resolution of your spectrogram is dependent on the window size (the
shorter the window, the cruder your frequency resolution is). If you
need more elaborate time-frequency representation, then wavelet, or
things like matching pursuit may be adapted, but they are not trivial to
use/code if you are new to signal processing. For audio signals, which I
am more familiar with, wavelets are generally not useful.

On matching pursuit:

http://www.scholarpedia.org/article/Matching_pursuit (I wanted to start
wrapping the MPTK in a scikits, but never got the time to start
something: http://mptk.irisa.fr/).

cheers,

David


From ajvogel at tuks.co.za  Thu Jul 24 01:59:03 2008
From: ajvogel at tuks.co.za (Adolph J. Vogel)
Date: Thu, 24 Jul 2008 07:59:03 +0200
Subject: [SciPy-user] Parallel linear solver.
Message-ID: <200807240759.03710.ajvogel@tuks.co.za>

Are there any bindings available for python to a Parallel linear solver 
package like ScaLapack? I couldnt find anything conclusive on google.

Thanx, Adolph


From christian-baehnisch at gmx.de  Thu Jul 24 02:57:33 2008
From: christian-baehnisch at gmx.de (Christian =?iso-8859-1?q?B=E4hnisch?=)
Date: Thu, 24 Jul 2008 08:57:33 +0200
Subject: [SciPy-user] spsolve
In-Reply-To: <d05265cb0807231948n5e9062daq9a70f829cb065438@mail.gmail.com>
References: <200807232246.14492.christian-baehnisch@gmx.de>
	<d05265cb0807231948n5e9062daq9a70f829cb065438@mail.gmail.com>
Message-ID: <200807240857.33610.christian-baehnisch@gmx.de>

Thank you for your quick reply! Somehow I believed spsolve is a conjugate 
gradient method ... my fault. If I got it right, the conjugate gradient 
method  in scipy.linalg.iterative should benefit from the sparsity of the 
matrices (symmetric and p. semi-definite) I put into it ?

> On Wed, Jul 23, 2008 at 3:46 PM, Christian B?hnisch
>
> <christian-baehnisch at gmx.de> wrote:
> > I'm using "scipy.linsolve.spsolve". Unfortunately, I have not been able
> > to figure out if  it is possible to get the number of iterations actually
> > used by the method. I would be very glad if somebody could help me on
> > this topic
>
> Christian, the sparse solvers in scipy.linsolve are what is known as
> "direct" methods (as opposed to "iterative" methods).  Rather than
> iteratively solving a linear system (like the Conjugate Gradient
> method), direct solvers just compute a sparse LU factorization of the
> matrix in order to solve the system.  In order to do this efficiently,
> and to keep L and U as sparse as possible, they incorporate a number
> of fairly sophisticated algorithms/heuristics.
>
> In SciPy 0.6 scipy.linsolve ultimately calls either the SuperLU or
> UMFPACK packages to perform the solve.
>
> If you're interested in iterative solvers, there are several in
> scipy.linalg.iterative.




From wnbell at gmail.com  Thu Jul 24 05:05:58 2008
From: wnbell at gmail.com (Nathan Bell)
Date: Thu, 24 Jul 2008 04:05:58 -0500
Subject: [SciPy-user] spsolve
In-Reply-To: <200807240857.33610.christian-baehnisch@gmx.de>
References: <200807232246.14492.christian-baehnisch@gmx.de>
	<d05265cb0807231948n5e9062daq9a70f829cb065438@mail.gmail.com>
	<200807240857.33610.christian-baehnisch@gmx.de>
Message-ID: <d05265cb0807240205h1514b5d4o12b66296b3ebf8f2@mail.gmail.com>

On Thu, Jul 24, 2008 at 1:57 AM, Christian B?hnisch
<christian-baehnisch at gmx.de> wrote:
> Thank you for your quick reply! Somehow I believed spsolve is a conjugate
> gradient method ... my fault.

No problem.  Documentation for the sparse solvers is incomplete, so
it's easy to get confused :)

> If I got it right, the conjugate gradient
> method  in scipy.linalg.iterative should benefit from the sparsity of the
> matrices (symmetric and p. semi-definite) I put into it ?
>

If your matrix is in a sparse format, such as the CSR or CSC formats
in scipy.sparse, then iterative solvers like conjugate gradient will
execute much more quickly.

-- 
Nathan Bell wnbell at gmail.com
http://graphics.cs.uiuc.edu/~wnbell/


From wesmckinn at gmail.com  Thu Jul 24 09:00:38 2008
From: wesmckinn at gmail.com (Wes McKinney)
Date: Thu, 24 Jul 2008 09:00:38 -0400
Subject: [SciPy-user] Equivalent to 'match' function in R?
Message-ID: <6c476c8a0807240600r138c80fdmdd5d87accfdad73e@mail.gmail.com>

Hi all,

I've been working with users lately who are transitioning from using R to
NumPy/Scipy. Some are accustomed to using the 'match' function, for
example:

> allData <- cbind(c(1,2,3,4,5), c(12, 19, 27, 38, 51))
> allData
     [,1] [,2]
[1,]    1   12
[2,]    2   19
[3,]    3   27
[4,]    4   38
[5,]    5   51
> subData <- cbind(c(3,5,1), c(NA, NA, NA))
> subData
     [,1] [,2]
[1,]    3   NA
[2,]    5   NA
[3,]    1   NA
> matchInd <- match(subData[,1], allData[,1])
> subData[,2] <- allData[matchInd,2]
> subData
     [,1] [,2]
[1,]    3   27
[2,]    5   51
[3,]    1   12
>
>
> allData <- cbind(c(1,2,3,4,5), rep(NA,5))
> allData
     [,1] [,2]
[1,]    1   NA
[2,]    2   NA
[3,]    3   NA
[4,]    4   NA
[5,]    5   NA
> subData <- cbind(c(3,5,1), c(27, 51, 12))
> subData
     [,1] [,2]
[1,]    3   27
[2,]    5   51
[3,]    1   12
> matchInd <- match(subData[,1], allData[,1])
> allData[matchInd,2] <- subData[,2]
> allData
     [,1] [,2]
[1,]    1   12
[2,]    2   NA
[3,]    3   27
[4,]    4   NA
[5,]    5   51

Before I reinvent the wheel, was curious if there was an equivalent function
elsewhere already. If not then some Cython/Pyrex is in order.

Thanks,
Wes
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080724/f23debe0/attachment.html>

From yosefmel at post.tau.ac.il  Thu Jul 24 09:35:53 2008
From: yosefmel at post.tau.ac.il (Yosef Meller)
Date: Thu, 24 Jul 2008 16:35:53 +0300
Subject: [SciPy-user] Equivalent to 'match' function in R?
In-Reply-To: <6c476c8a0807240600r138c80fdmdd5d87accfdad73e@mail.gmail.com>
References: <6c476c8a0807240600r138c80fdmdd5d87accfdad73e@mail.gmail.com>
Message-ID: <200807241635.53457.yosefmel@post.tau.ac.il>

On Thursday 24 July 2008 16:00:38 Wes McKinney wrote:
> Hi all,
>
> I've been working with users lately who are transitioning from using R to
> NumPy/Scipy. Some are accustomed to using the 'match' function, for
>
> example:
[snipped outputs]
> > allData <- cbind(c(1,2,3,4,5), c(12, 19, 27, 38, 51))
> > subData <- cbind(c(3,5,1), c(NA, NA, NA))
> > matchInd <- match(subData[,1], allData[,1])
> > subData[,2] <- allData[matchInd,2]
> > allData <- cbind(c(1,2,3,4,5), rep(NA,5))
> > subData <- cbind(c(3,5,1), c(27, 51, 12))
> > matchInd <- match(subData[,1], allData[,1])
> > allData[matchInd,2] <- subData[,2]
> > allData
>
>      [,1] [,2]
> [1,]    1   12
> [2,]    2   NA
> [3,]    3   27
> [4,]    4   NA
> [5,]    5   51
>
> Before I reinvent the wheel, was curious if there was an equivalent
> function elsewhere already. If not then some Cython/Pyrex is in order.
>
> Thanks,
> Wes

If I understand this correctly, this is what you want:
---
from scipy import c_, r_, equal, nan

vals = r_[12, 19, 27, 38, 51]
selector = r_[3, 5, 1]
idxs = r_[1, 2, 3, 4, 5] # I assume these don't have to be consecutive.
mask = equal.outer(idxs, selector).any(axis=1)

alldata = c_[idxs, where(mask, vals, nan]
---
Maybe looks a little less obvious, but does the job.


From arnar.flatberg at gmail.com  Thu Jul 24 09:49:32 2008
From: arnar.flatberg at gmail.com (Arnar Flatberg)
Date: Thu, 24 Jul 2008 15:49:32 +0200
Subject: [SciPy-user] Equivalent to 'match' function in R?
In-Reply-To: <6c476c8a0807240600r138c80fdmdd5d87accfdad73e@mail.gmail.com>
References: <6c476c8a0807240600r138c80fdmdd5d87accfdad73e@mail.gmail.com>
Message-ID: <5d3194020807240649m7d4d9c19he343da39f6c09181@mail.gmail.com>

On Thu, Jul 24, 2008 at 3:00 PM, Wes McKinney <wesmckinn at gmail.com> wrote:

> Hi all,
>
> I've been working with users lately who are transitioning from using R to
> NumPy/Scipy. Some are accustomed to using the 'match' function, for
> example:
>
> > allData <- cbind(c(1,2,3,4,5), c(12, 19, 27, 38, 51))
> > allData
>      [,1] [,2]
> [1,]    1   12
> [2,]    2   19
> [3,]    3   27
> [4,]    4   38
> [5,]    5   51
> > subData <- cbind(c(3,5,1), c(NA, NA, NA))
> > subData
>      [,1] [,2]
> [1,]    3   NA
> [2,]    5   NA
> [3,]    1   NA
>

What about using `intersect` combined with `where` ?

all_data = np.array([[1,2,3,4,5], [12,19,27,38,51]]).T
sub_data = np.array([[3,5,1], [nan,nan,nan]]).T
match_ind = np.where(np.intersect_1d(sub_data[:,0], all_data[:,0]))
sub_data[:,1] = all_data[match_ind,1]

It may not be pretty or the best approach for solving the above examples but
it behaves like R's match somewhat.

Arnar
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080724/ca5abbda/attachment.html>

From dfranci at seas.upenn.edu  Thu Jul 24 10:27:05 2008
From: dfranci at seas.upenn.edu (Frank Lagor)
Date: Thu, 24 Jul 2008 10:27:05 -0400
Subject: [SciPy-user] Parallel linear solver.
In-Reply-To: <200807240759.03710.ajvogel@tuks.co.za>
References: <200807240759.03710.ajvogel@tuks.co.za>
Message-ID: <9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>

Yes, there are.  What you are looking for is petsc4py.  PETSc is a very
good, well-developed package for scalable scientific computing.  It is based
on MPI and is very well-known and well-supported package (it is really
amazing actually how good their support actually is!)

Anyways, Lisandro Dalcin, a friend of mine, wrote petsc4py bindings that
allow you to do anything you want in PETSc in python.  The bindings are very
good, but not well documented.  It doesn't really matter however, because he
is so accessible, and PETSc's documentation is so good that you can browse
through it and guess what the correcponding syntax is in petsc4py.

I highly recommend that you check it out.  Let me know how it goes or if you
have any problems.

-Frank

On Thu, Jul 24, 2008 at 1:59 AM, Adolph J. Vogel <ajvogel at tuks.co.za> wrote:

> Are there any bindings available for python to a Parallel linear solver
> package like ScaLapack? I couldnt find anything conclusive on google.
>
> Thanx, Adolph
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Frank Lagor
Ph.D. Candidate
Mechanical Engineering and Applied Mechanics
University of Pennsylvania
dfranci at seas.upenn.edu
215.898.3144
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080724/968e2d56/attachment.html>

From mwojc at p.lodz.pl  Thu Jul 24 10:36:42 2008
From: mwojc at p.lodz.pl (Marek Wojciechowski)
Date: Thu, 24 Jul 2008 16:36:42 +0200
Subject: [SciPy-user] Parallel linear solver.
References: <200807240759.03710.ajvogel@tuks.co.za>
Message-ID: <g6a3mn$pes$1@ger.gmane.org>

Adolph J. Vogel wrote:

> Are there any bindings available for python to a Parallel linear solver
> package like ScaLapack? I couldnt find anything conclusive on google.
> 
> Thanx, Adolph

There exist python bindings for PETSC: http://code.google.com/p/petsc4py.
But i don't know its current state...

Greetings
-- 
Marek Wojciechowski



From mwojc at p.lodz.pl  Thu Jul 24 10:52:27 2008
From: mwojc at p.lodz.pl (Marek Wojciechowski)
Date: Thu, 24 Jul 2008 16:52:27 +0200
Subject: [SciPy-user] optimize.fmin_tnc - how to catch its output?
References: <g61jl9$hq0$1@ger.gmane.org>
	<9457e7c80807211556x5ae7d4e1m9079534f518e0cb9@mail.gmail.com>
	<3d375d730807211622p4b73cd4bo2fe69b2ebf4d8a61@mail.gmail.com>
Message-ID: <g6a4k9$tjo$1@ger.gmane.org>

Robert Kern wrote:

> On Mon, Jul 21, 2008 at 17:56, St?fan van der Walt <stefan at sun.ac.za>
> wrote:
>> Hi Marek
>>
>> 2008/7/21 Marek Wojciechowski <mwojc at p.lodz.pl>:
>>> Hallo!
>>> I tried to catch fmin_tnc output to a file in the following way:
>>>
>>>    file = open('messages', 'w')
>>>    stdout = sys.stdout
>>>    stderr = sys.stderr
>>>    sys.stdout = file
>>>    sys.stderr = file
>>>    res = optimize.fmin_tnc(..., messages=1)
>>>    file.close()
>>>    sys.stdout = stdout
>>>    sys.stderr = stderr
>>>
>>> But this does not work. I realized that fmin_tnc calls C routine which
>>> probably uses its own stdout, stderr. Is there a way to catch its
>>> output?
>>
>> I'm afraid you are right.  The quickest way I can think of is to grab
>> it using standard unix pipes:
>>
>> python optimise_script.py 2>&1 | cat > out.txt
> 
> There is a way to redirect it to a (temporary) file from within
> Python. You can then read this file. I have an old and
> not-recently-tested implementation over here:
> 
>   http://www.enthought.com/~rkern/cgi-bin/hgwebdir.cgi/redir/
> 
Hi!
Thanks for that! It works very nice. Tested with fmin_tnc:

    import redirfile as redir
    rout = redir.Redirector(redir.STDOUT)
    rerr = redir.Redirector(redir.STDERR)
    rout.start()
    rerr.start()
    res = optimize.fmin_tnc(..., messages=1)
    out = rout.stop()
    err = rerr.stop()

Greetings,
-- 
Marek Wojciechowski



From yannick.copin at laposte.net  Thu Jul 24 11:09:22 2008
From: yannick.copin at laposte.net (Yannick Copin)
Date: Thu, 24 Jul 2008 15:09:22 +0000 (UTC)
Subject: [SciPy-user] Equivalent to 'match' function in R?
References: <6c476c8a0807240600r138c80fdmdd5d87accfdad73e@mail.gmail.com>
Message-ID: <loom.20080724T150512-995@post.gmane.org>

Wes McKinney <wesmckinn <at> gmail.com> writes:
> Before I reinvent the wheel, was curious if there was an equivalent function
elsewhere already. If not then some Cython/Pyrex is in order.Thanks,Wes

For a similar purpose (with the handling of non-common items), I used:

def find_indices(big, small):
    """Find indices in big array of elements of small arrays. Set index=-1
    elements of small that cannot be found in big."""

    idx = np.array([ np.argmax(big==s) for s in small ])
    # Mark unfound value indices (beware: -1 is a valid index!)
    idx[big[idx]!=small] = -1

    return idx

In [63]: all_data = np.array([[1,2,3,4,5], [12,19,27,38,51]]).T
In [64]: sub_data = np.array([[3,5,1], [nan,nan,nan]]).T
In [67]: match_ind = find_indices(all_data[:,0],sub_data[:,0])
In [68]: sub_data[:,1] = all_data[match_ind,1]
In [69]: sub_data
Out[69]: 
array([[  3.,  27.],
       [  5.,  51.],
       [  1.,  12.]])

Cheers.




From wesmckinn at gmail.com  Thu Jul 24 11:12:12 2008
From: wesmckinn at gmail.com (Wes McKinney)
Date: Thu, 24 Jul 2008 11:12:12 -0400
Subject: [SciPy-user] Equivalent to 'match' function in R?
In-Reply-To: <5d3194020807240649m7d4d9c19he343da39f6c09181@mail.gmail.com>
References: <6c476c8a0807240600r138c80fdmdd5d87accfdad73e@mail.gmail.com>
	<5d3194020807240649m7d4d9c19he343da39f6c09181@mail.gmail.com>
Message-ID: <6c476c8a0807240812i397dcd09iee7c8d6a9076ec75@mail.gmail.com>

I did a 'super naive' version of this:

def match(a, b):
    bmap = dict([(v, i) for i, v in enumerate(b)])
    res = empty(len(a))
    for i, val in enumerate(a):
        res[i] = bmap.get(val, NaN)
    return res

Runs pretty slow for a test case, matching arange(20000) with a shuffled
version of itself

In [28]: timeit match(a, b)
10 loops, best of 3: 49.9 ms per loop

Same slightly less naive implementation done all with Cython and working
only with ndarrays:

In [30]: timeit cmatch(a, b)
100 loops, best of 3: 10.3 ms per loop

I don't know how to compare performance of this with R, assume it's pretty
comparable. The only thing that is kind of bust is that values not found in
the target array get translated to NA in R, but NaN's get translated to 0 as
numpy ints, you can't index an array with an array containing NaN's anyhow.
Hmm.

On Thu, Jul 24, 2008 at 9:49 AM, Arnar Flatberg <arnar.flatberg at gmail.com>
wrote:

>
>
> On Thu, Jul 24, 2008 at 3:00 PM, Wes McKinney <wesmckinn at gmail.com> wrote:
>
>> Hi all,
>>
>> I've been working with users lately who are transitioning from using R to
>> NumPy/Scipy. Some are accustomed to using the 'match' function, for
>> example:
>>
>> > allData <- cbind(c(1,2,3,4,5), c(12, 19, 27, 38, 51))
>> > allData
>>      [,1] [,2]
>> [1,]    1   12
>> [2,]    2   19
>> [3,]    3   27
>> [4,]    4   38
>> [5,]    5   51
>> > subData <- cbind(c(3,5,1), c(NA, NA, NA))
>> > subData
>>      [,1] [,2]
>> [1,]    3   NA
>> [2,]    5   NA
>> [3,]    1   NA
>>
>
> What about using `intersect` combined with `where` ?
>
> all_data = np.array([[1,2,3,4,5], [12,19,27,38,51]]).T
> sub_data = np.array([[3,5,1], [nan,nan,nan]]).T
> match_ind = np.where(np.intersect_1d(sub_data[:,0], all_data[:,0]))
> sub_data[:,1] = all_data[match_ind,1]
>
> It may not be pretty or the best approach for solving the above examples
> but it behaves like R's match somewhat.
>
> Arnar
>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080724/eda20beb/attachment.html>

From nwagner at iam.uni-stuttgart.de  Thu Jul 24 11:17:16 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Thu, 24 Jul 2008 17:17:16 +0200
Subject: [SciPy-user] Parallel linear solver.
In-Reply-To: <9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>
References: <200807240759.03710.ajvogel@tuks.co.za>
	<9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>
Message-ID: <web-105148834@uni-stuttgart.de>

On Thu, 24 Jul 2008 10:27:05 -0400
  "Frank Lagor" <dfranci at seas.upenn.edu> wrote:
> Yes, there are.  What you are looking for is petsc4py. 
> PETSc is a very
> good, well-developed package for scalable scientific 
>computing.  It is based
> on MPI and is very well-known and well-supported package 
>(it is really
> amazing actually how good their support actually is!)
> 
> Anyways, Lisandro Dalcin, a friend of mine, wrote 
>petsc4py bindings that
> allow you to do anything you want in PETSc in python. 
> The bindings are very
> good, but not well documented.  It doesn't really matter 
>however, because he
> is so accessible, and PETSc's documentation is so good 
>that you can browse
> through it and guess what the correcponding syntax is in 
>petsc4py.
> 
> I highly recommend that you check it out.  Let me know 
>how it goes or if you
> have any problems.
> 
> -Frank
> 
> On Thu, Jul 24, 2008 at 1:59 AM, Adolph J. Vogel 
><ajvogel at tuks.co.za> wrote:
> 
>> Are there any bindings available for python to a 
>>Parallel linear solver
>> package like ScaLapack? I couldnt find anything 
>>conclusive on google.
>>
>> Thanx, Adolph
>> _______________________________________________
>> SciPy-user mailing list
>> SciPy-user at scipy.org
>> http://projects.scipy.org/mailman/listinfo/scipy-user
>>
> 
> 
> 
> -- 
>Frank Lagor
> Ph.D. Candidate
> Mechanical Engineering and Applied Mechanics
> University of Pennsylvania
> dfranci at seas.upenn.edu
> 215.898.3144
  
The current version of PETSc is 2.3.3; released  May 23, 
2007. Are you aware of a forthcoming new release ?

Nils



From dfranci at seas.upenn.edu  Thu Jul 24 11:31:53 2008
From: dfranci at seas.upenn.edu (Frank Lagor)
Date: Thu, 24 Jul 2008 11:31:53 -0400
Subject: [SciPy-user] Parallel linear solver.
In-Reply-To: <web-105148834@uni-stuttgart.de>
References: <200807240759.03710.ajvogel@tuks.co.za>
	<9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>
	<web-105148834@uni-stuttgart.de>
Message-ID: <9fddf64a0807240831s5c56d0ccv44c10196b6daab85@mail.gmail.com>

Yes, there will be a new release of PETSc soon (I just asked one fo the
developers), but they don't know exactly how long it will be. The  2.3.3
release of PETSc is still current.  They use patches for a lot of their
small changes.  For example, just a two months ago or so, I downloaded a
version that was 2.3.3-p8.  The current patched version is 2.3.3-p13.  It is
still very active, and in my opinion, major releases are not a good thing,
because it may end up affecting your code.  Actually, I think this is
probably why you asked (so you could wait until a new release so you
wouldn't have to worry about code changes for a while).

Hope this help,
Frank

On Thu, Jul 24, 2008 at 11:17 AM, Nils Wagner <nwagner at iam.uni-stuttgart.de>
wrote:

> On Thu, 24 Jul 2008 10:27:05 -0400
>  "Frank Lagor" <dfranci at seas.upenn.edu> wrote:
> > Yes, there are.  What you are looking for is petsc4py.
> > PETSc is a very
> > good, well-developed package for scalable scientific
> >computing.  It is based
> > on MPI and is very well-known and well-supported package
> >(it is really
> > amazing actually how good their support actually is!)
> >
> > Anyways, Lisandro Dalcin, a friend of mine, wrote
> >petsc4py bindings that
> > allow you to do anything you want in PETSc in python.
> > The bindings are very
> > good, but not well documented.  It doesn't really matter
> >however, because he
> > is so accessible, and PETSc's documentation is so good
> >that you can browse
> > through it and guess what the correcponding syntax is in
> >petsc4py.
> >
> > I highly recommend that you check it out.  Let me know
> >how it goes or if you
> > have any problems.
> >
> > -Frank
> >
> > On Thu, Jul 24, 2008 at 1:59 AM, Adolph J. Vogel
> ><ajvogel at tuks.co.za> wrote:
> >
> >> Are there any bindings available for python to a
> >>Parallel linear solver
> >> package like ScaLapack? I couldnt find anything
> >>conclusive on google.
> >>
> >> Thanx, Adolph
> >> _______________________________________________
> >> SciPy-user mailing list
> >> SciPy-user at scipy.org
> >> http://projects.scipy.org/mailman/listinfo/scipy-user
> >>
> >
> >
> >
> > --
> >Frank Lagor
> > Ph.D. Candidate
> > Mechanical Engineering and Applied Mechanics
> > University of Pennsylvania
> > dfranci at seas.upenn.edu
> > 215.898.3144
>
> The current version of PETSc is 2.3.3; released  May 23,
> 2007. Are you aware of a forthcoming new release ?
>
> Nils
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Frank Lagor
Ph.D. Candidate
Mechanical Engineering and Applied Mechanics
University of Pennsylvania
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080724/e837503f/attachment.html>

From nwagner at iam.uni-stuttgart.de  Thu Jul 24 11:48:24 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Thu, 24 Jul 2008 17:48:24 +0200
Subject: [SciPy-user] Parallel linear solver.
In-Reply-To: <9fddf64a0807240831s5c56d0ccv44c10196b6daab85@mail.gmail.com>
References: <200807240759.03710.ajvogel@tuks.co.za>
	<9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>
	<web-105148834@uni-stuttgart.de>
	<9fddf64a0807240831s5c56d0ccv44c10196b6daab85@mail.gmail.com>
Message-ID: <web-105150168@uni-stuttgart.de>

On Thu, 24 Jul 2008 11:31:53 -0400
  "Frank Lagor" <dfranci at seas.upenn.edu> wrote:
> Yes, there will be a new release of PETSc soon (I just 
>asked one fo the
> developers), but they don't know exactly how long it 
>will be. The  2.3.3
> release of PETSc is still current.  They use patches for 
>a lot of their
> small changes.  For example, just a two months ago or 
>so, I downloaded a
> version that was 2.3.3-p8.  The current patched version 
>is 2.3.3-p13.  It is
> still very active, and in my opinion, major releases are 
>not a good thing,
> because it may end up affecting your code.  Actually, I 
>think this is
> probably why you asked (so you could wait until a new 
>release so you
> wouldn't have to worry about code changes for a while).
> 
> Hope this help,
>Frank
> 
Fank,

Thank you very much for your prompt. So far I have used
serial programs solely.
How do I benefit from parallel code in python ?
I mean what are the minimal requirements to run codes in 
parallel (hardware/software) ?
What software packages are needed to configure petsc in 
that context ?

Nils
  
   


From dfranci at seas.upenn.edu  Thu Jul 24 12:06:38 2008
From: dfranci at seas.upenn.edu (Frank Lagor)
Date: Thu, 24 Jul 2008 12:06:38 -0400
Subject: [SciPy-user] Parallel linear solver.
In-Reply-To: <web-105150168@uni-stuttgart.de>
References: <200807240759.03710.ajvogel@tuks.co.za>
	<9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>
	<web-105148834@uni-stuttgart.de>
	<9fddf64a0807240831s5c56d0ccv44c10196b6daab85@mail.gmail.com>
	<web-105150168@uni-stuttgart.de>
Message-ID: <9fddf64a0807240906s70d2efb5yac913625ab1066d2@mail.gmail.com>

Hi Nils,

Typically users come to find PETSc when they need it-- that is if they are
doing a lot of scientific computations and parallel processing is therefore
definitely needed. For me, I had access to a small 64 processor cluster and
sequential codes that would take days to run, so it was a natural fit.  I
could definitely see it being used  in smaller settings (and I'm sure many
people do), like on a desktop machine with a few processors, but that is not
what I use it for.  I'm not sure about your needs or your access to a
cluster, so you'll can probably be the best judge of if it is for you.

For software requirements-- PETSc uses BLAS, LAPACK, and an MPI distribution
as a mimimum.
 It can also interface with countless other packages (e.g. SCALAPACK, ATLAS,
SPRNG, etc.), but I don't bother with all this.  For me, there was an
OpenMPI implementation of MPI already installed on my cluster (as should be
the case for most clusters), so I just linked to it.  And the BLAS and
LAPACK on the cluster were not working currently, so I told PETSc to
download and install BLAS and LAPACK automatically.  It did and it works
fine. Anyways, that's my story-- I hope it helps.

-Frank



On Thu, Jul 24, 2008 at 11:48 AM, Nils Wagner <nwagner at iam.uni-stuttgart.de>
wrote:

> On Thu, 24 Jul 2008 11:31:53 -0400
>   "Frank Lagor" <dfranci at seas.upenn.edu> wrote:
> > Yes, there will be a new release of PETSc soon (I just
> >asked one fo the
> > developers), but they don't know exactly how long it
> >will be. The  2.3.3
> > release of PETSc is still current.  They use patches for
> >a lot of their
> > small changes.  For example, just a two months ago or
> >so, I downloaded a
> > version that was 2.3.3-p8.  The current patched version
> >is 2.3.3-p13.  It is
> > still very active, and in my opinion, major releases are
> >not a good thing,
> > because it may end up affecting your code.  Actually, I
> >think this is
> > probably why you asked (so you could wait until a new
> >release so you
> > wouldn't have to worry about code changes for a while).
> >
> > Hope this help,
> >Frank
> >
> Fank,
>
> Thank you very much for your prompt. So far I have used
> serial programs solely.
> How do I benefit from parallel code in python ?
> I mean what are the minimal requirements to run codes in
> parallel (hardware/software) ?
> What software packages are needed to configure petsc in
> that context ?
>
> Nils
>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Frank Lagor
Ph.D. Candidate
Mechanical Engineering and Applied Mechanics
University of Pennsylvania
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080724/e88da5dc/attachment.html>

From dfranci at seas.upenn.edu  Thu Jul 24 12:15:27 2008
From: dfranci at seas.upenn.edu (Frank Lagor)
Date: Thu, 24 Jul 2008 12:15:27 -0400
Subject: [SciPy-user] Parallel linear solver.
In-Reply-To: <9fddf64a0807240906s70d2efb5yac913625ab1066d2@mail.gmail.com>
References: <200807240759.03710.ajvogel@tuks.co.za>
	<9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>
	<web-105148834@uni-stuttgart.de>
	<9fddf64a0807240831s5c56d0ccv44c10196b6daab85@mail.gmail.com>
	<web-105150168@uni-stuttgart.de>
	<9fddf64a0807240906s70d2efb5yac913625ab1066d2@mail.gmail.com>
Message-ID: <9fddf64a0807240915i492e8dddqccec6c9cd9dbd44d@mail.gmail.com>

One last thing--

If you just want parallel interaction between your python codes, and you
think that petsc4py is overkill with all of it's solvers and such, then just
use mpi4py to exchange messages, perform reductions, etc. in parallel.  It
is written by the same author for petsc4py (but it is not a requirement for
petsc4py) and you may find it a bit easier to use if you want to run
basically sequential codes in parallel and exchange results.

-Frank

On Thu, Jul 24, 2008 at 12:06 PM, Frank Lagor <dfranci at seas.upenn.edu>
wrote:

> Hi Nils,
>
> Typically users come to find PETSc when they need it-- that is if they are
> doing a lot of scientific computations and parallel processing is therefore
> definitely needed. For me, I had access to a small 64 processor cluster and
> sequential codes that would take days to run, so it was a natural fit.  I
> could definitely see it being used  in smaller settings (and I'm sure many
> people do), like on a desktop machine with a few processors, but that is not
> what I use it for.  I'm not sure about your needs or your access to a
> cluster, so you'll can probably be the best judge of if it is for you.
>
> For software requirements-- PETSc uses BLAS, LAPACK, and an MPI
> distribution as a mimimum.
>  It can also interface with countless other packages (e.g. SCALAPACK,
> ATLAS, SPRNG, etc.), but I don't bother with all this.  For me, there was an
> OpenMPI implementation of MPI already installed on my cluster (as should be
> the case for most clusters), so I just linked to it.  And the BLAS and
> LAPACK on the cluster were not working currently, so I told PETSc to
> download and install BLAS and LAPACK automatically.  It did and it works
> fine. Anyways, that's my story-- I hope it helps.
>
> -Frank
>
>
>
>
> On Thu, Jul 24, 2008 at 11:48 AM, Nils Wagner <
> nwagner at iam.uni-stuttgart.de> wrote:
>
>> On Thu, 24 Jul 2008 11:31:53 -0400
>>   "Frank Lagor" <dfranci at seas.upenn.edu> wrote:
>> > Yes, there will be a new release of PETSc soon (I just
>> >asked one fo the
>> > developers), but they don't know exactly how long it
>> >will be. The  2.3.3
>> > release of PETSc is still current.  They use patches for
>> >a lot of their
>> > small changes.  For example, just a two months ago or
>> >so, I downloaded a
>> > version that was 2.3.3-p8.  The current patched version
>> >is 2.3.3-p13.  It is
>> > still very active, and in my opinion, major releases are
>> >not a good thing,
>> > because it may end up affecting your code.  Actually, I
>> >think this is
>> > probably why you asked (so you could wait until a new
>> >release so you
>> > wouldn't have to worry about code changes for a while).
>> >
>> > Hope this help,
>> >Frank
>> >
>> Fank,
>>
>> Thank you very much for your prompt. So far I have used
>> serial programs solely.
>> How do I benefit from parallel code in python ?
>> I mean what are the minimal requirements to run codes in
>> parallel (hardware/software) ?
>> What software packages are needed to configure petsc in
>> that context ?
>>
>> Nils
>>
>>
>> _______________________________________________
>> SciPy-user mailing list
>> SciPy-user at scipy.org
>> http://projects.scipy.org/mailman/listinfo/scipy-user
>>
>
>
>
> --
> Frank Lagor
> Ph.D. Candidate
> Mechanical Engineering and Applied Mechanics
> University of Pennsylvania
>



-- 
Frank Lagor
Ph.D. Candidate
Mechanical Engineering and Applied Mechanics
University of Pennsylvania
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080724/0dd6f24d/attachment.html>

From grs2103 at columbia.edu  Thu Jul 24 16:05:28 2008
From: grs2103 at columbia.edu (Gideon Simpson)
Date: Thu, 24 Jul 2008 16:05:28 -0400
Subject: [SciPy-user] quad question
Message-ID: <F05A195E-6B02-4C2F-9ED9-B219886C2D95@columbia.edu>

So I'm looking at the quad routine which seems to be built on  
quadpack.  Assuming I give it a finite interval with no weights, will  
it use QAGS?  My integrand has integrable singularities at one  
endpoint, which I don't really feel like sorting out the correct  
weighting for the QAW routine.
-gideon


From hep.sebastien.binet at gmail.com  Fri Jul 25 01:21:44 2008
From: hep.sebastien.binet at gmail.com (Sebastien Binet)
Date: Thu, 24 Jul 2008 22:21:44 -0700
Subject: [SciPy-user] mpi4py - MPI_Comm_join example ?
Message-ID: <200807242221.50661.binet@cern.ch>

hi there,

I am trying to parallelize an application with mpi4py.
For various reasons[*], I need to dynamically create new processes and get 
inter-communicators via MPI_Comm_join.
Documentation about that function is pretty scarce and I am still a MPI 
rookie.

I came up with that kind of code:
##
import sys,os,socket
from mpi4py import MPI

def msg (fmt, *args):
    map (sys.stdout.write,
         ('[%i] :: '%os.getpid(),
          fmt%args,
          '\n'))
    sys.stdout.flush()

def main(ntasks=5):
    msg ("creating a socket pair...")
    fd_parent, fd_child = socket.socketpair (socket.AF_UNIX)
    msg ("forking...")
    pid = os.fork()
    if pid == 0:
        fd_child.close() # not needed anymore
        msg ("child process...")
        main_child (fd_parent)
    else:
        fd_parent.close() # idem
        msg ("parent process...")
        main_parent (fd_child)
    return

def main_child (fd):
    comm = MPI.COMM_WORLD
    msg ("joining parent communicator...")
    icomm = MPI.Comm.Join (fd.fileno())
    fd.close() # not needed anymore
    msg ("icomm is [%s]", "OK" if icomm != MPI.COMM_NULL else "ERR")
    if icomm == MPI.COMM_NULL:
        return
    ### do something with icomm...
    return

def main_parent (fd):
    comm = MPI.COMM_WORLD
    msg ("mpi: rank=%r nprocs=%r", comm.rank, comm.size)
    msg ("joining child communicator...")
    icomm = MPI.Comm.Join (fd.fileno())
    fd.close() # not needed anymore
    msg ("icomm is [%s]", "OK" if icomm != MPI.COMM_NULL else "ERR")
    if icomm == MPI.COMM_NULL:
        return
    ### do something with icomm
    return
##

I'd like to fork as many as `ntasks` subprocesses and farm out some work...
But I am a bit stuck as to how to proceed...
Any help would be much appreciated.

Cheers,
Sebastien.

[*] main reason being that the application takes ~1-2 minutes to completely 
initialize and eats up to 1-2 Gb of memory. Hence, relying on os.fork to 
efficiently share memory between processes after the initialization completed 
seems to be a good idea.
I don't think I can use MPI_Win_xyz as the pieces of data I want to 
share/reuse among processes are C++ containers, so the usual problem of STL 
allocators and shm-segments kicks in...
Spawning new processes via MPI_Comm_spawn is a no go either b/c of the above 
main reason...
-- 
###################################
# Dr. Sebastien Binet             #
# Lawrence Berkeley National Lab. #
# 1 Cyclotron Road                #
# Berkeley, CA 94720              #
###################################
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080724/b2c6c380/attachment.sig>

From ajvogel at tuks.co.za  Fri Jul 25 01:58:28 2008
From: ajvogel at tuks.co.za (Adolph J. Vogel)
Date: Fri, 25 Jul 2008 07:58:28 +0200
Subject: [SciPy-user] Parallel linear solver.
In-Reply-To: <9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>
References: <200807240759.03710.ajvogel@tuks.co.za>
	<9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>
Message-ID: <200807250758.28653.ajvogel@tuks.co.za>

Thank you for your reply. I will check it out today, and see if I can get it 
working.

Adolph

ps. Frank I see we are semi colleagues. :)

On Thursday 24 July 2008 16:27:05 you wrote:
> Yes, there are.  What you are looking for is petsc4py.  PETSc is a very
> good, well-developed package for scalable scientific computing.  It is
> based on MPI and is very well-known and well-supported package (it is
> really amazing actually how good their support actually is!)
>
> Anyways, Lisandro Dalcin, a friend of mine, wrote petsc4py bindings that
> allow you to do anything you want in PETSc in python.  The bindings are
> very good, but not well documented.  It doesn't really matter however,
> because he is so accessible, and PETSc's documentation is so good that you
> can browse through it and guess what the correcponding syntax is in
> petsc4py.
>
> I highly recommend that you check it out.  Let me know how it goes or if
> you have any problems.
>
> -Frank
>
> On Thu, Jul 24, 2008 at 1:59 AM, Adolph J. Vogel <ajvogel at tuks.co.za> wrote:
> > Are there any bindings available for python to a Parallel linear solver
> > package like ScaLapack? I couldnt find anything conclusive on google.
> >
> > Thanx, Adolph
> > _______________________________________________
> > SciPy-user mailing list
> > SciPy-user at scipy.org
> > http://projects.scipy.org/mailman/listinfo/scipy-user



-- 
_/_/_/_/_/_/_/_/_/_/_/_/_/
Adolph J. Vogel
BEng(Mechanical)

University of Pretoria
Office: 012-420-2189
Email: ajvogel at tuks.co.za
_/_/_/_/_/_/_/_/_/_/_/_/_/


From massimo.sandal at unibo.it  Fri Jul 25 05:45:36 2008
From: massimo.sandal at unibo.it (massimo sandal)
Date: Fri, 25 Jul 2008 11:45:36 +0200
Subject: [SciPy-user] scipy.stats.gaussian_kde for 2d kernel density
	estimation
In-Reply-To: <loom.20080723T131047-645@post.gmane.org>
References: <48870EB8.8010208@unibo.it>
	<loom.20080723T131047-645@post.gmane.org>
Message-ID: <4889A0C0.40707@unibo.it>

Hi Dave,

> Hopefully the example below will help...

Thanks a lot - your example put me on the right way. I am a complete 
newbie on 2d graphs, and meshgrid just escaped my attention.

m.
-- 
Massimo Sandal , Ph.D.
University of Bologna
Department of Biochemistry "G.Moruzzi"

snail mail:
Via Irnerio 48, 40126 Bologna, Italy

email:
massimo.sandal at unibo.it

web:
http://www.biocfarm.unibo.it/samori/people/sandal.html

tel: +39-051-2094388
fax: +39-051-2094387
-------------- next part --------------
A non-text attachment was scrubbed...
Name: massimo_sandal.vcf
Type: text/x-vcard
Size: 274 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080725/9b4523b0/attachment.vcf>

From ajvogel at tuks.co.za  Fri Jul 25 05:48:33 2008
From: ajvogel at tuks.co.za (Adolph J. Vogel)
Date: Fri, 25 Jul 2008 11:48:33 +0200
Subject: [SciPy-user] Parallel linear solver.
In-Reply-To: <9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>
References: <200807240759.03710.ajvogel@tuks.co.za>
	<9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>
Message-ID: <200807251148.33395.ajvogel@tuks.co.za>

Hi Frank, 

Im having trouble with using petcs4py. When trying to run the helloworld.py 
example, I get the following:

adolph at adolph-laptop:~/lib/petsc4py-0.7.5/tests/sandbox$ mpirun -np 2 python 
helloworld.py
[adolph-laptop:07785] mca: base: component_find: unable to open osc pt2pt: 
file not found (ignored)
libibverbs: Fatal: couldn't read uverbs ABI version.
--------------------------------------------------------------------------
[0,0,0]: OpenIB on host adolph-laptop was unable to find any HCAs.
Another transport will be used instead, although this may result in
lower performance.
--------------------------------------------------------------------------
Hello, World!! I am process 0 of 1
[adolph-laptop:07786] mca: base: component_find: unable to open osc pt2pt: 
file not found (ignored)
libibverbs: Fatal: couldn't read uverbs ABI version.
--------------------------------------------------------------------------
[0,0,0]: OpenIB on host adolph-laptop was unable to find any HCAs.
Another transport will be used instead, although this may result in
lower performance.
--------------------------------------------------------------------------
-----------------------------------------------------------------------------
It seems that [at least] one of the processes that was started with
mpirun did not invoke MPI_INIT before quitting (it is possible that
more than one process did not invoke MPI_INIT -- mpirun was only
notified of the first one, which was on node n0).

mpirun can *only* be used with MPI programs (i.e., programs that
invoke MPI_INIT and MPI_FINALIZE).  You can use the "lamexec" program
to run non-MPI programs over the lambooted nodes.
-----------------------------------------------------------------------------
[0]PETSC 
ERROR: ------------------------------------------------------------------------
[0]PETSC ERROR: Caught signal number 15 Terminate: Somet process (or the batch 
system) has told this process to end
[0]PETSC ERROR: Try option -start_in_debugger or -on_error_attach_debugger
[0]PETSC ERROR: or see 
http://www.mcs.anl.gov/petsc/petsc-as/documentation/troubleshooting.html#Signal[0]PETSC 
ERROR: or try http://valgrind.org on linux or man libgmalloc on Apple to find 
memory corruption errors
[0]PETSC ERROR: configure using --with-debugging=yes, recompile, link, and run
[0]PETSC ERROR: to get more information on the crash.
[adolph-laptop:07786] MPI_ABORT invoked on rank 0 in communicator 
MPI_COMM_WORLD with errorcode 59


The invocation I used, works with my mpi4py programs. However nothing from 
petcs4py seems to work. Any help would be greatly appreciated.

Adolph

On Thursday 24 July 2008 16:27:05 you wrote:
> Yes, there are.  What you are looking for is petsc4py.  PETSc is a very
> good, well-developed package for scalable scientific computing.  It is
> based on MPI and is very well-known and well-supported package (it is
> really amazing actually how good their support actually is!)
>
> Anyways, Lisandro Dalcin, a friend of mine, wrote petsc4py bindings that
> allow you to do anything you want in PETSc in python.  The bindings are
> very good, but not well documented.  It doesn't really matter however,
> because he is so accessible, and PETSc's documentation is so good that you
> can browse through it and guess what the correcponding syntax is in
> petsc4py.
>
> I highly recommend that you check it out.  Let me know how it goes or if
> you have any problems.
>
> -Frank
>
> On Thu, Jul 24, 2008 at 1:59 AM, Adolph J. Vogel <ajvogel at tuks.co.za> wrote:
> > Are there any bindings available for python to a Parallel linear solver
> > package like ScaLapack? I couldnt find anything conclusive on google.
> >
> > Thanx, Adolph
> > _______________________________________________
> > SciPy-user mailing list
> > SciPy-user at scipy.org
> > http://projects.scipy.org/mailman/listinfo/scipy-user



-- 
_/_/_/_/_/_/_/_/_/_/_/_/_/
Adolph J. Vogel
BEng(Mechanical)

University of Pretoria
Office: 012-420-2189
Email: ajvogel at tuks.co.za
Mobile: 072 592 5836
_/_/_/_/_/_/_/_/_/_/_/_/_/


From mjakubik at ta3.sk  Fri Jul 25 05:51:29 2008
From: mjakubik at ta3.sk (Marian Jakubik)
Date: Fri, 25 Jul 2008 11:51:29 +0200
Subject: [SciPy-user] Parallel linear solver.
In-Reply-To: <9fddf64a0807240915i492e8dddqccec6c9cd9dbd44d@mail.gmail.com>
References: <200807240759.03710.ajvogel@tuks.co.za>
	<9fddf64a0807240727w6d6526dbp1af089257ba348e8@mail.gmail.com>
	<web-105148834@uni-stuttgart.de>
	<9fddf64a0807240831s5c56d0ccv44c10196b6daab85@mail.gmail.com>
	<web-105150168@uni-stuttgart.de>
	<9fddf64a0807240906s70d2efb5yac913625ab1066d2@mail.gmail.com>
	<9fddf64a0807240915i492e8dddqccec6c9cd9dbd44d@mail.gmail.com>
Message-ID: <20080725115129.37b80066@jakubik.ta3.sk>

Hi Frank,

is it possible to compare the speed of computation using parallelisms
(parallel codes) in C and Python? I am preparing a parallel code and
till now I've been decided to use C. I mean the "difference" in speed
for executable (C) and interpreted (Python) codes is important, isn't
it? 

Thank in advance for response...

Best,
Marian

D?a Thu, 24 Jul 2008 12:15:27 -0400
"Frank Lagor" <dfranci at seas.upenn.edu> nap?sal:

> One last thing--
> 
> If you just want parallel interaction between your python codes, and you
> think that petsc4py is overkill with all of it's solvers and such, then just
> use mpi4py to exchange messages, perform reductions, etc. in parallel.  It
> is written by the same author for petsc4py (but it is not a requirement for
> petsc4py) and you may find it a bit easier to use if you want to run
> basically sequential codes in parallel and exchange results.
> 
> -Frank
> 
> On Thu, Jul 24, 2008 at 12:06 PM, Frank Lagor <dfranci at seas.upenn.edu>
> wrote:
> 
> > Hi Nils,
> >
> > Typically users come to find PETSc when they need it-- that is if they are
> > doing a lot of scientific computations and parallel processing is therefore
> > definitely needed. For me, I had access to a small 64 processor cluster and
> > sequential codes that would take days to run, so it was a natural fit.  I
> > could definitely see it being used  in smaller settings (and I'm sure many
> > people do), like on a desktop machine with a few processors, but that is not
> > what I use it for.  I'm not sure about your needs or your access to a
> > cluster, so you'll can probably be the best judge of if it is for you.
> >
> > For software requirements-- PETSc uses BLAS, LAPACK, and an MPI
> > distribution as a mimimum.
> >  It can also interface with countless other packages (e.g. SCALAPACK,
> > ATLAS, SPRNG, etc.), but I don't bother with all this.  For me, there was an
> > OpenMPI implementation of MPI already installed on my cluster (as should be
> > the case for most clusters), so I just linked to it.  And the BLAS and
> > LAPACK on the cluster were not working currently, so I told PETSc to
> > download and install BLAS and LAPACK automatically.  It did and it works
> > fine. Anyways, that's my story-- I hope it helps.
> >
> > -Frank
> >
> >
> >
> >
> > On Thu, Jul 24, 2008 at 11:48 AM, Nils Wagner <
> > nwagner at iam.uni-stuttgart.de> wrote:
> >
> >> On Thu, 24 Jul 2008 11:31:53 -0400
> >>   "Frank Lagor" <dfranci at seas.upenn.edu> wrote:
> >> > Yes, there will be a new release of PETSc soon (I just
> >> >asked one fo the
> >> > developers), but they don't know exactly how long it
> >> >will be. The  2.3.3
> >> > release of PETSc is still current.  They use patches for
> >> >a lot of their
> >> > small changes.  For example, just a two months ago or
> >> >so, I downloaded a
> >> > version that was 2.3.3-p8.  The current patched version
> >> >is 2.3.3-p13.  It is
> >> > still very active, and in my opinion, major releases are
> >> >not a good thing,
> >> > because it may end up affecting your code.  Actually, I
> >> >think this is
> >> > probably why you asked (so you could wait until a new
> >> >release so you
> >> > wouldn't have to worry about code changes for a while).
> >> >
> >> > Hope this help,
> >> >Frank
> >> >
> >> Fank,
> >>
> >> Thank you very much for your prompt. So far I have used
> >> serial programs solely.
> >> How do I benefit from parallel code in python ?
> >> I mean what are the minimal requirements to run codes in
> >> parallel (hardware/software) ?
> >> What software packages are needed to configure petsc in
> >> that context ?
> >>
> >> Nils
> >>
> >>
> >> _______________________________________________
> >> SciPy-user mailing list
> >> SciPy-user at scipy.org
> >> http://projects.scipy.org/mailman/listinfo/scipy-user
> >>
> >
> >
> >
> > --
> > Frank Lagor
> > Ph.D. Candidate
> > Mechanical Engineering and Applied Mechanics
> > University of Pennsylvania
> >
> 
> 
> 


From dineshbvadhia at hotmail.com  Sat Jul 26 07:08:06 2008
From: dineshbvadhia at hotmail.com (Dinesh B Vadhia)
Date: Sat, 26 Jul 2008 04:08:06 -0700
Subject: [SciPy-user] sparse matrix: getrow()
Message-ID: <BAY109-DS3B04277F013ECA54B8C26A3810@phx.gbl>

Has the sparse.csr_matrix.getrow(,) function been fixed in the latest svn?  Thanks!

Dinesh
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080726/ceda8807/attachment.html>

From wnbell at gmail.com  Sat Jul 26 07:14:30 2008
From: wnbell at gmail.com (Nathan Bell)
Date: Sat, 26 Jul 2008 06:14:30 -0500
Subject: [SciPy-user] sparse matrix: getrow()
In-Reply-To: <BAY109-DS3B04277F013ECA54B8C26A3810@phx.gbl>
References: <BAY109-DS3B04277F013ECA54B8C26A3810@phx.gbl>
Message-ID: <d05265cb0807260414t3d5834a9w964e078fd4e38bb5@mail.gmail.com>

On Sat, Jul 26, 2008 at 6:08 AM, Dinesh B Vadhia
<dineshbvadhia at hotmail.com> wrote:
> Has the sparse.csr_matrix.getrow(,) function been fixed in the latest svn?
>

It appears to work.  You should use A[i,:] instead, since getrow()
will eventually be deprecated.  CSR and CSC matrices support most
forms of slicing and indexing now.

-- 
Nathan Bell wnbell at gmail.com
http://graphics.cs.uiuc.edu/~wnbell/


From lists at vrbka.net  Sat Jul 26 10:43:13 2008
From: lists at vrbka.net (Lubos Vrbka)
Date: Sat, 26 Jul 2008 16:43:13 +0200
Subject: [SciPy-user] sine transform prefactor
Message-ID: <488B3801.6010602@vrbka.net>

hi guys,

since i am not a mathematician, i'd like to ask one question related to 
discrete sine transform and its implementation in scipy (sandbox at the 
moment).

from definition, the continuous sine transform comes with the constant 
prefactor of sqrt(2/pi), i.e.,

FST(f(r)) = f(k) = sqrt(2/pi) int_0^{\infty} f(r) sin(kr) dr
iFST(f(k)) = f(r) = sqrt(2/pi) int_0^{\infty} f(k) sin(kr) dk

how is it in case of a discrete transform? does this sqrt(2/pi) factor 
come into play also for discrete transforms, or is it applicable only 
for continuous transforms? if it is applicable, is it already included 
inside the routines? how is it then distributed among the forward and 
reverse transform?

i think the situation here might be complicated by the fact, that the 
code in sandbox uses the fast fourier transform for the evaluation of 
the sine transform - and this might probably change the constants 
involved as well (since the fourier transform should be 'normalized' 
with a factor of 1/2pi).

i would be very grateful for any help on this issue. my program seems to 
run, but it is dependent on the choice of the constant factors involved 
in the transforms. so far i've found the correct combination (it seems 
so), but i'd like to make this more rigorous and to really understand _why_.

thanks a lot in advance. best,

-- 
Lubos _ at _"
http://www.lubos.vrbka.net


From dfranci at seas.upenn.edu  Sat Jul 26 11:46:32 2008
From: dfranci at seas.upenn.edu (Frank Lagor)
Date: Sat, 26 Jul 2008 11:46:32 -0400
Subject: [SciPy-user] prefactor response
Message-ID: <9fddf64a0807260846r68c24f97k97fb964069bacfa4@mail.gmail.com>

Hi Lubos,

I don't know a whole lot about transforms however, I do remember one bit of
information that may be useful to you:  The prefactors in the sine transform
(which is of course related to the fourier transform) are actually matters
of convention.  The sqrt(2/pi) is not always the prefactor!  The prefactor
is usually chosen to be sqrt(2/pi) so that the inverse transform will also
have a prefactor of sqrt(2/pi) and there will be some symmetry in the two
formulas so that they look alike and are easier to remember.  However,
SOMETIMES people chose the transform or inverse transform (I can't recall
which) to have a 2/pi prefactor, and the other to have no prefactor.  I
comes down to a matter of convention and it doesn't matter which one you
chose, as long as you are consistent.

Please careful and make sure you a certain of the convention chosen and you
a consistent in what you code and you will be fine.

Hope this helps,
Frank

PS> Dear Everyone, I am trying to get my email mailing list style fixed so
that I don't break mailing list etiquette.   Please email  me personal to
check me if this email sucked.  Sorry for my previous mistakes!


-- 
Frank Lagor
Ph.D. Candidate
Mechanical Engineering and Applied Mechanics
University of Pennsylvania
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080726/c27d5be5/attachment.html>

From gael.varoquaux at normalesup.org  Sat Jul 26 11:49:16 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Sat, 26 Jul 2008 17:49:16 +0200
Subject: [SciPy-user] error ellipse
In-Reply-To: <9fddf64a0807232002g31ba7a80nc4a26291c8590caf@mail.gmail.com>
References: <9fddf64a0807232002g31ba7a80nc4a26291c8590caf@mail.gmail.com>
Message-ID: <20080726154916.GB3248@phare.normalesup.org>

On Wed, Jul 23, 2008 at 11:02:05PM -0400, Frank Lagor wrote:
>    Does anyone know if there is a function in scipy.stats for somewhere else
>    to plot a 2D error ellipse on a scatter plot of some data given the
>    covariance matrix?  I have not been able to find one.  Thanks in advance
>    for your help,

For plotting related questions, you can try your luck on the matplotlib
mailing list:
https://lists.sourceforge.net/lists/listinfo/matplotlib-users

You might get more answers.

Cheers,

Ga?l


From gael.varoquaux at normalesup.org  Sat Jul 26 11:55:19 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Sat, 26 Jul 2008 17:55:19 +0200
Subject: [SciPy-user] scipy.stats.gaussian_kde for 2d kernel
	density	estimation
In-Reply-To: <4889A0C0.40707@unibo.it>
References: <48870EB8.8010208@unibo.it>
	<loom.20080723T131047-645@post.gmane.org> <4889A0C0.40707@unibo.it>
Message-ID: <20080726155519.GC3248@phare.normalesup.org>

On Fri, Jul 25, 2008 at 11:45:36AM +0200, massimo sandal wrote:
> Thanks a lot - your example put me on the right way. I am a complete newbie 
> on 2d graphs, and meshgrid just escaped my attention.

Prefer mgrid
(http://www.scipy.org/Numpy_Example_List_With_Doc#head-c2338cdfd293a3c60c3eb65d6df9a2d8e1803fc5 )
or ogrid (http://www.scipy.org/Numpy_Example_List_With_Doc#ogrid )to
meshgrid. Meshgrid is for matlab compatibility, and is less powerful.

Ga?l


From lists at vrbka.net  Sat Jul 26 12:24:48 2008
From: lists at vrbka.net (Lubos Vrbka)
Date: Sat, 26 Jul 2008 18:24:48 +0200
Subject: [SciPy-user] sine transform prefactor
In-Reply-To: <9fddf64a0807260846r68c24f97k97fb964069bacfa4@mail.gmail.com>
References: <9fddf64a0807260846r68c24f97k97fb964069bacfa4@mail.gmail.com>
Message-ID: <488B4FD0.3070804@vrbka.net>

hi,

> I don't know a whole lot about transforms however, I do remember one bit of
> information that may be useful to you:  The prefactors in the sine transform
> (which is of course related to the fourier transform) are actually matters
> of convention.  The sqrt(2/pi) is not always the prefactor!  The prefactor
> is usually chosen to be sqrt(2/pi) so that the inverse transform will also
> have a prefactor of sqrt(2/pi) and there will be some symmetry in the two
> formulas so that they look alike and are easier to remember.  However,
> SOMETIMES people chose the transform or inverse transform (I can't recall
> which) to have a 2/pi prefactor, and the other to have no prefactor.  I
> comes down to a matter of convention and it doesn't matter which one you
> chose, as long as you are consistent.
well, all this is clear to me.

if i do
iDST(DST(f(r))
on discrete data, i get the original data back (expected result). this 
would indicate that
1) either the sqrt(2/pi) norm shouldn't be used for discrete data (why?)
2) the sqrt(2/pi) factor is taken care of somewhere inside the DST 
function (how?)

this is the thing i'd really love to know - is the answer 1 or answer 2 
correct?

in my opinion, it shouldn't matter what combination of norms i use, as 
soon as they are consistent - but at the moment, i have the following 
problem. i need to iteratively solve
H(r) = C(r) + C(r)*H(r)
where the star denotes convolution and H and C are functions of 
interest. the easiest way to solve this should be fourier transforming 
this, solving
H(k) = C(k) + C(k)H(k)
by
H(k) = C(k)/(1-C(k))
and then transforming H(k) back for further processing. in practice, i 
calculate C, transorm it into fourier space, solve for H and transform 
it back, where it serves as a 'parameter' for new C.

due to the nature of my functions, the 3D-FT can be replaced by 1D 
fourier-bessel transformation AND some constants. depending on how one 
distributes the normalization of FT, one arrives at, e.g.,
FB(f(r)) = f(k) = 4pi/k int_0^infty f(r) r sin(kr) dr
iFB(f(k)) = f(r) = 1/2rpi^2 int_0^infty f(k) k sin(kr) dk
this can be calculated using fourier-sine transform of a new function 
F(r)=f(r)*r and F(k)=f(k)*k, respectively
FB(f(r)) = 4pi/k sqrt(pi/2) int_0^infty F(r) sin(kr) dr
iFB(f(r)) = 1/2rpi^2 sqrt(pi/2) int_0^infty F(k) sin(kr) dk
so one would expect that the story is clear - do fourier-sinus, multiply 
with the respective normalization constant... but the problem is, that 
this doesn't work.

at this point i asked the question - *should* i actually use these 
factors (coming from continuous transforms) also in the case of discrete 
transforms, which are self normalized to a number of discretization points?

> PS> Dear Everyone, I am trying to get my email mailing list style fixed so
> that I don't break mailing list etiquette.   Please email  me personal to
> check me if this email sucked.  Sorry for my previous mistakes!
just a comment - don't change the subject line, since it breaks the thread

best,

-- 
Lubos _ at _"
http://www.lubos.vrbka.net


From David.Kaplan at ird.fr  Sat Jul 26 14:06:51 2008
From: David.Kaplan at ird.fr (David M. Kaplan)
Date: Sat, 26 Jul 2008 20:06:51 +0200
Subject: [SciPy-user] unique, sort, sortrows
Message-ID: <1217095611.7172.42.camel@localhost>

Hi,

I am fairly new to numpy/scipy (switching from matlab).  I have noticed
a few things that left me confused.  I am wondering if I could get some
input from others.

1) On the webpage: http://www.scipy.org/NumPy_for_Matlab_Users , where
it says matlab sortrows(a,i) is equivalent to a[argsort(a[:,0],i)], I
believe this should be a[argsort(a[:,i]),:].  Or better yet:

matlab: [b,I] = sortrows(a,i)
numpy: I = argsort(a[:,i]), b = a[I,:]

I have already changed this on the webpage, but I want to make sure I am
not missing something.

2) Is there a simple equivalent of sortrows(a) (i.e., sorting by entire
rows)?  Similarly, is there a simple equivalent of the matlab Y =
unique(X,'rows')?  I looked online and there appears to have been
previous discussion of these, but nothing simple and general seemed to
come out.

3) Is there an equivalent of [Y,I,J] = unique(X)?  In this case, I is
the indices of the unique elements and J is an index from Y to X (i.e.,
where the unique elements appear in X.  I can get I with:

I,Y = unique1d( X, return_index=True )

but J, which is very useful, is not available.  I suppose I could do:

J = array([])
for i,y in enumerate(Y):
	J[ nonzero( y == X ) ] = i

But it seems like J is useful enough that there should be an easier way
(or perhaps it should be integrated into unique1d, at the risk of adding
more keyword arguments).

4) What is the best equivalent of meshgrid(x,y,z,...) or
ndgrid(x,y,z...)?  I defined my own function, but is there a built in
way:

def ndgrid( *args ):
	s = []
	for a in args:
		s.append(len(a))
	I = indices(s)
	res = []
	for a,i in zip(args,I):
		res.append(a[i])
	return res

This method will also be painful for large matrices as one has to create
twice as much data (the index matrix and the resulting matrices).

5) I was trying to do matrix multiplication of matrices with more than 2
axes.  Normally matrix multiplication of an [M,N,P] matrix with a
[P,Q,R] matrix should produce something with dimensions [M,N,Q,R], but
this produces an error.  What is the appropriate idiom for this?

Thanks for the help.

Cheers,
David

-- 
**********************************
David M. Kaplan
Charge de Recherche 1
Institut de Recherche pour le Developpement
Centre de Recherche Halieutique Mediterraneenne et Tropicale
av. Jean Monnet
B.P. 171
34203 Sete cedex
France

Phone: +33 (0)4 99 57 32 27
Fax: +33 (0)4 99 57 32 95
http://www.ur097.ird.fr/team/dkaplan/index.html
**********************************




From dfranci at seas.upenn.edu  Sat Jul 26 14:15:11 2008
From: dfranci at seas.upenn.edu (Frank Lagor)
Date: Sat, 26 Jul 2008 14:15:11 -0400
Subject: [SciPy-user] sine transform prefactor
In-Reply-To: <488B4FD0.3070804@vrbka.net>
References: <9fddf64a0807260846r68c24f97k97fb964069bacfa4@mail.gmail.com>
	<488B4FD0.3070804@vrbka.net>
Message-ID: <9fddf64a0807261115x426853e6wf888f7801472b539@mail.gmail.com>

Ludos,

First let me clarify: the factor in front of the continuous forms is
1/sqrt(2pi), not sqrt(2/pi) like I said previously (I was confused).


> if i do
> iDST(DST(f(r))
> on discrete data, i get the original data back (expected result). this
> would indicate that
> 1) either the sqrt(2/pi) norm shouldn't be used for discrete data (why?)


I don't think the factors should be added on at all.  The mere transform and
invtransform functions should handle all factors.  This is related to what I
want to type later on.


> 2) the sqrt(2/pi) factor is taken care of somewhere inside the DST function
> (how?)
>
Yes. It should be.  Like I said before, the convention chosen doesn't really
matter as long as the you do the FT and then the IFT, because the factors
from the two operations must multiple together to give 1/(2pi) in the end
(this is built into the answer.  The problem with convention only comes when
you want to compare numbers of transformed data (still in fourier space)
with someone else.  Then you need to know if your conventions are the same.


> this is the thing i'd really love to know - is the answer 1 or answer 2
> correct?
>

I think the actual answer lies in the fact that the fourier transform is
actually derived from the fourier series.  The fourier series is a sum of
some coefficients times an exponential. There is an equation for calculating
these coefficients and it is something like 1/(2pi) times an integral, and
thats where the 1/(2pi) comes in (actually it comes in when the formula for
the coefficients is derived).  See the book Functions of a complex variable
by Carrier, Krook, and Pearson for the details of this derivation.  I think
that if you see this, all your concerns about the factors will be taken care
of.

-Frank


-- 
Frank Lagor
Ph.D. Candidate
Mechanical Engineering and Applied Mechanics
University of Pennsylvania
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080726/0ef079f3/attachment.html>

From gael.varoquaux at normalesup.org  Sat Jul 26 16:24:00 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Sat, 26 Jul 2008 22:24:00 +0200
Subject: [SciPy-user] unique, sort, sortrows
In-Reply-To: <1217095611.7172.42.camel@localhost>
References: <1217095611.7172.42.camel@localhost>
Message-ID: <20080726202400.GA683@phare.normalesup.org>

Hi David,

It's been a long time since I used Matlab, so I can't really answer all
your questions, however I think I can help you with one:

On Sat, Jul 26, 2008 at 08:06:51PM +0200, David M. Kaplan wrote:
> 4) What is the best equivalent of meshgrid(x,y,z,...) or
> ndgrid(x,y,z...)?  I defined my own function, but is there a built in
> way:

I think you should look at the mgrid and ogrid functions. (See
http://www.scipy.org/Numpy_Example_List ). For many operations, prefer
ogrid, it saves memory and tends to works as easily as mgrid thanks to
broadcasting.

HTH,

Ga?l


From cimrman3 at ntc.zcu.cz  Sat Jul 26 17:35:24 2008
From: cimrman3 at ntc.zcu.cz (Robert Cimrman)
Date: Sat, 26 Jul 2008 23:35:24 +0200
Subject: [SciPy-user] unique, sort, sortrows
In-Reply-To: <1217095611.7172.42.camel@localhost>
References: <1217095611.7172.42.camel@localhost>
Message-ID: <20080726233524.vcdq55buowgwgcso@webmail.zcu.cz>

Hi David,

I can comment on unique1d, as I am the culprit. I am cc'ing to  
numpy-discussion as this is a numpy function.

Quoting "David M. Kaplan" <David.Kaplan at ird.fr>:
> 2) Is there a simple equivalent of sortrows(a) (i.e., sorting by entire
> rows)?  Similarly, is there a simple equivalent of the matlab Y =

have you looked at lexsort?

> 3) Is there an equivalent of [Y,I,J] = unique(X)?  In this case, I is
> the indices of the unique elements and J is an index from Y to X (i.e.,
> where the unique elements appear in X.  I can get I with:
>
> I,Y = unique1d( X, return_index=True )
>
> but J, which is very useful, is not available.  I suppose I could do:
>
> J = array([])
> for i,y in enumerate(Y):
> 	J[ nonzero( y == X ) ] = i
>
> But it seems like J is useful enough that there should be an easier way
> (or perhaps it should be integrated into unique1d, at the risk of adding
> more keyword arguments).

So basically Y = X[I] and X = Y[J], right? I do not recall matlab that  
well to know for sure. It certainly could be done, I could look at it  
after I return from Euroscipy (i.e. after Monday).

I would replace return_index argument by two arguments: return_direct  
(->I) and return_inverse (->J), ok? Does anyone propose better names?

Actually most of the methods in arraysetops could return optionally  
some index arrays. Would anyone use it? (I do not need it personally :)

cheers,
r.




From David.Kaplan at ird.fr  Sun Jul 27 08:43:09 2008
From: David.Kaplan at ird.fr (David M. Kaplan)
Date: Sun, 27 Jul 2008 14:43:09 +0200
Subject: [SciPy-user]  unique, sort, sortrows
In-Reply-To: 1217095611.7172.42.camel@localhost
Message-ID: <1217162589.7128.36.camel@localhost>

Hi,

Thanks for the very helpful comments.

Regarding Gael's comment, the problem with mgrid and ogrid (at least in
the version of numpy I use: 1.1.0-3) is that it currently only accepts
standard indexing so you can't use it with non-uniform values.  I use
this a lot when I have a model I want to run with a series of parameter
values, not all of which are uniformly spaced.  For example, the
following fails:

mgrid[:5,[1,7,8]]

I would like this to give something equivalent to
meshgrid(arange(5),[1,7,8]) (up to a transpose operation).  Similarly
for more input arguments.  I haven't looked at the numpy source code,
but it would seem that it shouldn't be too hard to add this
functionality as it already exists for r_ and c_:

r_[:5,[1,7,8]] # no problem here

In response to Robert's comments, I looked a bit at lexsort and didn't
immediately see how it could fix my problem of sorting rows of a matrix
because I didn't really understand it.  Finally I figured out that the
following appears to do the trick:

I = lexsort(a[:,-1::-1].T)
b = a[I,:]

As this is a bit tricky for someone to figure out, perhaps a helper
function called sortrows would be useful in numpy?  Also, an equivalent
of unique(a,'rows') would still be very useful.

As for [Y,I,J] = unique(X), yes Y=X[I] and X=Y[J].  Your fix would help
me out, though it would be nice to specify the call signature in the
help of the new version of unique1d (it took me a while to figure out
that it was I,Y and not Y,I).  I think it would also be useful to
propogate these changes to the other arraysetops functions.  In
particular, I use indexes returned by matlab's intersect command often:

[C,IA,IB] = intersect(A,B)

A use case for this is suppose you have a sparse dataset at x,y points
that is "on a grid", but lacking some of the points (e.g., instrument
only returns points that had valid data).  A simple way to solve this
would be (using a few suggested changes to numpy/scipy):

[X,Y] = mgrid[ unique(pts[:,0]), unique(pts[:,1]) ]
s = X.shape
newData = tile( NaN, s ).flatten()

p,IA,IB = intersect( c_[X.flatten(),Y.flatten()], pts, rows=True ) 
newData[IA] = Data[IB]
newData.reshape(s)

There may be other concise ways to solve this, but this one seems fairly
efficient.

Thanks again.

Cheers,
David

-- 
**********************************
David M. Kaplan
Charge de Recherche 1
Institut de Recherche pour le Developpement
Centre de Recherche Halieutique Mediterraneenne et Tropicale
av. Jean Monnet
B.P. 171
34203 Sete cedex
France

Phone: +33 (0)4 99 57 32 27
Fax: +33 (0)4 99 57 32 95
http://www.ur097.ird.fr/team/dkaplan/index.html
**********************************




From gael.varoquaux at normalesup.org  Sun Jul 27 11:55:10 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Sun, 27 Jul 2008 17:55:10 +0200
Subject: [SciPy-user] unique, sort, sortrows
In-Reply-To: <1217162589.7128.36.camel@localhost>
References: <1217162589.7128.36.camel@localhost>
Message-ID: <20080727155510.GA12519@phare.normalesup.org>

On Sun, Jul 27, 2008 at 02:43:09PM +0200, David M. Kaplan wrote:
> I haven't looked at the numpy source code,
> but it would seem that it shouldn't be too hard to add this
> functionality as it already exists for r_ and c_:

> r_[:5,[1,7,8]] # no problem here

Well, if you have time to have a look at that, there certainly is value
in adding this functionnality.

Cheers,

Ga?l


From lorenzo.isella at gmail.com  Sun Jul 27 12:17:47 2008
From: lorenzo.isella at gmail.com (Lorenzo Isella)
Date: Sun, 27 Jul 2008 18:17:47 +0200
Subject: [SciPy-user] Ordering and Counting the Repetitions of the Rows of a
	Matrix
Message-ID: <a2b3004b0807270917j47b64c40i9dc578d5dddd38ac@mail.gmail.com>

Dear All,
Consider an Nx2 matrix of the kind:

A=   1 2
       3 13
       1  2
       6  8
       3 13
       2  9
       1  1


The first entry in each row is always smaller or equal than the second
entry in the same row.
Now there are two things I would like to do with this A matrix:
(1) With a sort of n.unique1d (but have not been very successful yet),
let each row of A appear only once (i.e. get rid of the repetitions).
Therefore one should obtain the matrix:
B=   1 2
       3 13
       6  8
       2  9
       1  1

(2) At the same time, efficiently count how many times each row of B
appeared in A. I would like to get a C vector counting them as:

C=   2
       2
       1
       1
       1


Any suggestions about an efficient way of achieving this?
Many thanks

Lorenzo


From David.Kaplan at ird.fr  Sun Jul 27 13:31:01 2008
From: David.Kaplan at ird.fr (David M. Kaplan)
Date: Sun, 27 Jul 2008 19:31:01 +0200
Subject: [SciPy-user] Ordering and Counting the Repetitions of the Rows of	a
	Matrix
Message-ID: <1217179861.7128.51.camel@localhost>

Hi,

This is related to the discussion of "unique, sort, rows" that is just
before this email.  Unfortunately, unique1d doesn't have a rows option
(yet).  But I saw an email post that had a good suggestion that may work
for you: hash your array so that you can get the almost certainly unique
rows (essentially infinitesimal chance of getting a false match):

B = rand(A.shape[1],1)
C = dot(A,B)

I,C2 = unique1d(C,return_index=True)

Aunique = A[I,:]
Anum=[]
for a in Aunique:
	Anum.append( sum(all(A==a,axis=1)) )

This will do basically what you want.  Aunique won't be sorted
correctly, but you can use lexsort to fix that (see me recent post).

Cheers,
David

-- 
**********************************
David M. Kaplan
Charge de Recherche 1
Institut de Recherche pour le Developpement
Centre de Recherche Halieutique Mediterraneenne et Tropicale
av. Jean Monnet
B.P. 171
34203 Sete cedex
France

Phone: +33 (0)4 99 57 32 27
Fax: +33 (0)4 99 57 32 95
http://www.ur097.ird.fr/team/dkaplan/index.html
**********************************




From warren.weckesser at gmail.com  Sun Jul 27 15:46:29 2008
From: warren.weckesser at gmail.com (Warren Weckesser)
Date: Sun, 27 Jul 2008 15:46:29 -0400
Subject: [SciPy-user] Ordering and Counting the Repetitions of the Rows
	of a Matrix
In-Reply-To: <a2b3004b0807270917j47b64c40i9dc578d5dddd38ac@mail.gmail.com>
References: <a2b3004b0807270917j47b64c40i9dc578d5dddd38ac@mail.gmail.com>
Message-ID: <114880320807271246x1c922e7cg9539684fbad7bed9@mail.gmail.com>

Lorenzo,

Given a matrix A like you showed, here is one way to find (and count) the
unique rows:

----------
d = {}
for r in A:
    t = tuple(r)
    d[t] = d.get(t,0) + 1

# The dict d now has the counts of the unique rows of A.

B = numpy.array(d.keys())    # The unique rows of A
C = numpy.array(d.values())  # The counts of the unique rows
----------

For a large number of rows (e.g. 10000), this appears to be significantly
faster than the code that David Kaplan suggested in his email earlier today.

Regards,

Warren


On Sun, Jul 27, 2008 at 12:17 PM, Lorenzo Isella
<lorenzo.isella at gmail.com>wrote:

> Dear All,
> Consider an Nx2 matrix of the kind:
>
> A=   1 2
>       3 13
>       1  2
>       6  8
>       3 13
>       2  9
>       1  1
>
>
> The first entry in each row is always smaller or equal than the second
> entry in the same row.
> Now there are two things I would like to do with this A matrix:
> (1) With a sort of n.unique1d (but have not been very successful yet),
> let each row of A appear only once (i.e. get rid of the repetitions).
> Therefore one should obtain the matrix:
> B=   1 2
>       3 13
>       6  8
>       2  9
>       1  1
>
> (2) At the same time, efficiently count how many times each row of B
> appeared in A. I would like to get a C vector counting them as:
>
> C=   2
>       2
>       1
>       1
>       1
>
>
> Any suggestions about an efficient way of achieving this?
> Many thanks
>
> Lorenzo
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080727/c0f97b35/attachment.html>

From stefan at sun.ac.za  Sun Jul 27 17:37:24 2008
From: stefan at sun.ac.za (=?ISO-8859-1?Q?St=E9fan_van_der_Walt?=)
Date: Sun, 27 Jul 2008 23:37:24 +0200
Subject: [SciPy-user] Ordering and Counting the Repetitions of the Rows
	of a Matrix
In-Reply-To: <114880320807271246x1c922e7cg9539684fbad7bed9@mail.gmail.com>
References: <a2b3004b0807270917j47b64c40i9dc578d5dddd38ac@mail.gmail.com>
	<114880320807271246x1c922e7cg9539684fbad7bed9@mail.gmail.com>
Message-ID: <9457e7c80807271437t664acc9fu706c7a029e57cc57@mail.gmail.com>

2008/7/27 Warren Weckesser <warren.weckesser at gmail.com>:
> Lorenzo,
>
> Given a matrix A like you showed, here is one way to find (and count) the
> unique rows:
>
> ----------
> d = {}
> for r in A:
>     t = tuple(r)
>     d[t] = d.get(t,0) + 1
>
> # The dict d now has the counts of the unique rows of A.
>
> B = numpy.array(d.keys())    # The unique rows of A
> C = numpy.array(d.values())  # The counts of the unique rows
> ----------

And here's an evil one-liner:

x[np.sum(np.triu(np.all(x == x[:,None], axis=2)), axis=1)

that requires a lot of memory (given M rows of N columns, a temporary
array of MxMxN is used).  It is blazingly fast, though :)

Cheers
St?fan


From schut at sarvision.nl  Mon Jul 28 07:09:57 2008
From: schut at sarvision.nl (Vincent Schut)
Date: Mon, 28 Jul 2008 13:09:57 +0200
Subject: [SciPy-user] [python(x,y)] New Release: 2.0.0
In-Reply-To: <629b08a40807131019o36a7c35cw19ee640bbbcc173@mail.gmail.com>
References: <629b08a40807131019o36a7c35cw19ee640bbbcc173@mail.gmail.com>
Message-ID: <g6k9e5$a32$1@ger.gmane.org>

Pierre Raybaut wrote:
> Hi all,
> 
> Python(x,y) 2.0.0 is now available on http://www.pythonxy.com.
> 
> Changes history
> 07 -12 -2008 - Version 2.0.0 :
> 
>     * Added:
>         o ITK 3.6 (WrapITK) - Open-source software system for image
> processing (leading-edge segmentation and registration algorithms)
>         o Windows installer: new component manager system (updating is
> easier and more flexible, install and uninstall process are much
> cleaner, ...)
>         o Windows installer: Python(x,y) may now be installed silently
> (/S option), and the non-silent installation now allows the user to
> keep working on the machine during the process
>     * Updated:
>         o Matplotlib 0.98.1 (see release notes)
>         o SymPy 0.6.0
>         o GDAL 1.5.2
>         o PySerial 2.4
> 
> Regards,
> Pierre Raybaut

Hi Pierre,

as you include gdal and ITK in python(x,y), you might also be interested 
to include OTB (Orfeo ToolBox) in the future. OTB is a toolbox built on 
ITK which enhances/extends ITK with geospatial-focussed processing nodes 
and some extra glue (e.g. to read and propagate geospatial info during 
processing, etc). For more info, see http://otb.cnes.fr/

Cheers,
Vincent.



From elcorto at gmx.net  Mon Jul 28 07:41:59 2008
From: elcorto at gmx.net (Steve Schmerler)
Date: Mon, 28 Jul 2008 13:41:59 +0200
Subject: [SciPy-user] temporary copies for in-place array modification?
Message-ID: <20080728114159.GA5776@ramrod.de>

Hi all

Say I do this:

    >>> a
    array([1, 2, 3, 4, 5, 6])

    >>> a[1:] = a[1:] - a[:-1]

    >>> a
    array([1, 1, 1, 1, 1, 1])

    >>> b = a[1:]
    >>> ...

I.e. I want to compute the difference of a's elements and store them in place
in b = a[1:] to work with that (b is only a view so that's ok, no copy).

Are there any temp copies of `a` involved? I ask b/c `a` will be large.
Thanks!

steve


From David.Kaplan at ird.fr  Mon Jul 28 08:55:22 2008
From: David.Kaplan at ird.fr (David M. Kaplan)
Date: Mon, 28 Jul 2008 14:55:22 +0200
Subject: [SciPy-user]  unique, sort, sortrows
In-Reply-To: 1217162589.7128.36.camel@localhost
Message-ID: <1217249722.7230.36.camel@localhost>

Hi,

Well, as usual there are compromises in everything and the mgrid/ogrid
functionality is the way it currently is for some good reasons.  The
first reason is that python appears to be fairly sloppy about how it
passes indexing arguments to the __getitem__ method.  It passes a tuple
containing the arguments in all cases except when it has one argument,
in which case it just passes that argument.  This means that it is hard
to tell a tuple argument from several non-tuple arguments.  For example,
the following two produce exactly the same call to __getitem__ :

mgrid[1,2,3]
mgrid[(1,2,3)]

(__getitem__ receives a single tuple (1,2,3)), but different from:

mgrid[[1,2,3]]

(__getitem__ receives a single list = [1,2,3]).  This seems like a bug
to me, but is probably considered a feature by somebody.  In any case,
this is workable, but a bit annoying in that tuple arguments just aren't
going to work well.

The second problem is that the current implementation is fairly
efficient because it forces all arguments to the same type so as to
avoid some unnecessary copies (I think).  Once you allow non-slice
arguments, this is hard to maintain.

That being said, attached is a replacement for index_tricks.py that
should implement a reasonable set of features, while only very slightly
altering performance.  I have only touched nd_grid.  I haven't fixed the
documentation string yet, nor have I added tests to
test_index_tricks.py, but will do that if the changes will be accepted
into numpy.  With the new version, old stuff should work as usual,
except that mgrid now returns a list of arrays instead of an array of
arrays (note that this change will cause current test_index_tricks.py to
fail).  With the new changes, you can now do:

mgrid[-2:5:10j,[4.5,6,7.1],12,['abc','def']]

The following will work as expected:

mgrid[:5,(1,2,3)]

But this will not:

mgrid[(1,2,3)]  # same as mgrid[1,2,3], but different from mgrid[[1,2,3]]

Given these limitations, this seems like a fairly useful addition.  If
this looks usable, I will clean up and add tests if desired.  If not, I
recommend adding a ndgrid function to numpy that does the equivalent of
matlab [X,Y,Z,...] = ndgrid(x,y,z,...) and then making the current
meshgrid just call that changing the order of the first two arguments.

Cheers,
David



-- 
**********************************
David M. Kaplan
Charge de Recherche 1
Institut de Recherche pour le Developpement
Centre de Recherche Halieutique Mediterraneenne et Tropicale
av. Jean Monnet
B.P. 171
34203 Sete cedex
France

Phone: +33 (0)4 99 57 32 27
Fax: +33 (0)4 99 57 32 95
http://www.ur097.ird.fr/team/dkaplan/index.html
**********************************

-------------- next part --------------
A non-text attachment was scrubbed...
Name: index_tricks.py
Type: text/x-python
Size: 15089 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080728/f0a3f970/attachment.py>

From aisaac at american.edu  Mon Jul 28 09:51:47 2008
From: aisaac at american.edu (Alan G Isaac)
Date: Mon, 28 Jul 2008 09:51:47 -0400
Subject: [SciPy-user] unique, sort, sortrows
In-Reply-To: <1217249722.7230.36.camel@localhost>
References: <1217249722.7230.36.camel@localhost>
Message-ID: <488DCEF3.3010702@american.edu>

David M. Kaplan wrote:
> python appears to be fairly sloppy about how it passes 
> indexing arguments to the __getitem__ method.

I do not generally find the word 'sloppy' to be descriptive
of Python.

> It passes a tuple containing the arguments in all cases 
> except when it has one argument, in which case it just 
> passes that argument.

Well, not quite.  The bracket syntax is for passing a key
(a single object) to __getitem__.

> For example, the following two produce exactly the same 
> call to __getitem__ :
> mgrid[1,2,3] 
> mgrid[(1,2,3)] 

Well, yes.  Note::

     >>> x = 1,2,3
     >>> type(x)
     <type 'tuple'>

In Python it is the commas, not the paretheses,
that is determining the tuple type.

So perhaps the question you raise could be rephrased to
"why does an ndarray (not Python) handle treat
a list 'index' differently than a tuple 'index'?"
I do not know the history of that decision,
but it has been used to provide some additional functionality.

Cheers,
Alan Isaac



From ivo.maljevic at gmail.com  Mon Jul 28 10:07:33 2008
From: ivo.maljevic at gmail.com (Ivo Maljevic)
Date: Mon, 28 Jul 2008 10:07:33 -0400
Subject: [SciPy-user] sine transform prefactor
In-Reply-To: <9fddf64a0807261115x426853e6wf888f7801472b539@mail.gmail.com>
References: <9fddf64a0807260846r68c24f97k97fb964069bacfa4@mail.gmail.com>
	<488B4FD0.3070804@vrbka.net>
	<9fddf64a0807261115x426853e6wf888f7801472b539@mail.gmail.com>
Message-ID: <826c64da0807280707m3c340acv1b1d3413b340e0d2@mail.gmail.com>

I do not know about cosine transform much as I do not use it, but the
coefficient 1/(2pi) in
the continuous time Fourier transform is not something that is quite
randomly selected (even though, I admit, in calculating discrete FT
you have more freedom at choosing the scaling factor, and sometimes
for reasons of symmetry there is a 1/sqrt(N) scaling factor in both
directions).

Using latex, you either have completely symmetric expressions if one
uses 'f' frequency:

X(f) = \int_{-\infty}^{\infty} x(t)\ e^{- i 2\pi f t}\,dt,
x(t) = \int_{-\infty}^{\infty} X(f)\ e^{i 2 \pi f t}\,df,

or if you use angular frequency  ?=2 pi f, than you have:

    X( ?) = \int _{-\infty}^\infty x(t)\ e^{- i ? t}\,dt

    x(t) = \frac{1}{2\pi} \int _{-\infty}^{\infty} X( ?)\ e^{ i ? t}\,d ?,

Not that this will be of much help to Lubos. Lubos, you may want to
look at the following Wiki page, or if you are still not happy, I will
dig up my literature and find out more about DCT. You are probably
interested in DCT II type from this article:

http://en.wikipedia.org/wiki/Discrete_cosine_transform

and you might want to look also at:

http://en.wikipedia.org/wiki/Discrete_Fourier_transform

Ivo


2008/7/26 Frank Lagor <dfranci at seas.upenn.edu>:
> Ludos,
>
> First let me clarify: the factor in front of the continuous forms is
> 1/sqrt(2pi), not sqrt(2/pi) like I said previously (I was confused).
>
>>
>> if i do
>> iDST(DST(f(r))
>> on discrete data, i get the original data back (expected result). this
>> would indicate that
>> 1) either the sqrt(2/pi) norm shouldn't be used for discrete data (why?)
>
>
> I don't think the factors should be added on at all.  The mere transform and
> invtransform functions should handle all factors.  This is related to what I
> want to type later on.
>
>>
>> 2) the sqrt(2/pi) factor is taken care of somewhere inside the DST
>> function (how?)
>
> Yes. It should be.  Like I said before, the convention chosen doesn't really
> matter as long as the you do the FT and then the IFT, because the factors
> from the two operations must multiple together to give 1/(2pi) in the end
> (this is built into the answer.  The problem with convention only comes when
> you want to compare numbers of transformed data (still in fourier space)
> with someone else.  Then you need to know if your conventions are the same.
>
>>
>> this is the thing i'd really love to know - is the answer 1 or answer 2
>> correct?
>
> I think the actual answer lies in the fact that the fourier transform is
> actually derived from the fourier series.  The fourier series is a sum of
> some coefficients times an exponential. There is an equation for calculating
> these coefficients and it is something like 1/(2pi) times an integral, and
> thats where the 1/(2pi) comes in (actually it comes in when the formula for
> the coefficients is derived).  See the book Functions of a complex variable
> by Carrier, Krook, and Pearson for the details of this derivation.  I think
> that if you see this, all your concerns about the factors will be taken care
> of.
>
> -Frank
>
>
> --
> Frank Lagor
> Ph.D. Candidate
> Mechanical Engineering and Applied Mechanics
> University of Pennsylvania
>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>

From zachary.pincus at yale.edu  Mon Jul 28 10:25:56 2008
From: zachary.pincus at yale.edu (Zachary Pincus)
Date: Mon, 28 Jul 2008 10:25:56 -0400
Subject: [SciPy-user] Best way to save ND-arrays for Matlab?
Message-ID: <790A19AE-EF29-491D-BA9E-EBBCA4398A51@yale.edu>

Hi all,

I'm trying to use the scipy.io.mio tools to write out some arrays for  
matlab users. This works quite well, with the exception of some 3D  
arrays that I have, which (as per the documentation) get flattened to  
2D during the process. (I'm saving to version 4 .mat files, which I  
need for compatibility with some folk's versions of Matlab.)

Is there any good way to save this data in a less-awkward format for  
the eventual users? Or is this a limitation of the format?

Thanks,

Zachary Pincus
Postdoctoral Fellow, lab of Frank Slack
Molecular, Cellular and Developmental Biology
Yale University







From David.Kaplan at ird.fr  Mon Jul 28 10:56:00 2008
From: David.Kaplan at ird.fr (David M. Kaplan)
Date: Mon, 28 Jul 2008 16:56:00 +0200
Subject: [SciPy-user]  unique, sort, sortrows
In-Reply-To: 1217249722.7230.36.camel@localhost
Message-ID: <1217256960.7230.47.camel@localhost>

Hi,

>From Alan Isaac:

Well, yes.  Note::

     >>> x = 1,2,3
     >>> type(x)
     <type 'tuple'>

In Python it is the commas, not the paretheses,
that is determining the tuple type.

----

Good point.  In any case, this is a workable problem.  Adding a comma
after the last argument to mgrid[] assures that it behaves "as
expected" (e.g. mgrid[(1,2),]).

Attached is a newer much cleaner version of my replacement for
index_tricks.py.  I did some optimisation and it looks like this new
version beats the old on similar operations for large meshed matrices.
For single return arrays or small matrices it looses to the old version,
but only slightly.  As large matrices are probably the bottleneck, this
seems like a reasonable tradeoff.

Cheers,
David


-- 
**********************************
David M. Kaplan
Charge de Recherche 1
Institut de Recherche pour le Developpement
Centre de Recherche Halieutique Mediterraneenne et Tropicale
av. Jean Monnet
B.P. 171
34203 Sete cedex
France

Phone: +33 (0)4 99 57 32 27
Fax: +33 (0)4 99 57 32 95
http://www.ur097.ird.fr/team/dkaplan/index.html
**********************************

-------------- next part --------------
A non-text attachment was scrubbed...
Name: index_tricks.py
Type: text/x-python
Size: 14286 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080728/3a7742bc/attachment.py>

From zachary.pincus at yale.edu  Mon Jul 28 12:00:17 2008
From: zachary.pincus at yale.edu (Zachary Pincus)
Date: Mon, 28 Jul 2008 12:00:17 -0400
Subject: [SciPy-user] Matlab IO problem? (was: Best way to save ND-arrays
	for Matlab?)
In-Reply-To: <790A19AE-EF29-491D-BA9E-EBBCA4398A51@yale.edu>
References: <790A19AE-EF29-491D-BA9E-EBBCA4398A51@yale.edu>
Message-ID: <D09FA21A-3DD6-4737-A5A2-F9BCA582325E@yale.edu>

Hello all,

I have investigated my issue with matlab IO a little more. Clearly, I  
need to be using matlab version-5 .mat files, which support  
multidimensional arrays.

I had thought I needed to use version-4 files, because a user said  
that they couldn't read the version-5 ones. However, on closer  
investigation, they are using Version 6.5.0.180913a Release 13 of  
Matlab -- which should be able to read version-5 .mat files fine.

It appears that scipy.io.matlab saves version-5 .mat files which  
cannot always be opened by this version of matlab. I can't quite  
figure out when things work.

scipy.io.matlab.mio.savemat('test', {'five':5}, format='5')
or
scipy.io.matlab.mio.savemat('test', {'five':5*numpy.ones((5,5))},  
format='5')

both fail as below:
 >> load test
??? Error using ==> load
Can't read file.

However, some more complex .mat files, with multiple variables, seem  
to work. Other more-complex ones fail.

Anyone have any thoughts?

Zach




On Jul 28, 2008, at 10:25 AM, Zachary Pincus wrote:

> Hi all,
>
> I'm trying to use the scipy.io.mio tools to write out some arrays for
> matlab users. This works quite well, with the exception of some 3D
> arrays that I have, which (as per the documentation) get flattened to
> 2D during the process. (I'm saving to version 4 .mat files, which I
> need for compatibility with some folk's versions of Matlab.)
>
> Is there any good way to save this data in a less-awkward format for
> the eventual users? Or is this a limitation of the format?
>
> Thanks,
>
> Zachary Pincus
> Postdoctoral Fellow, lab of Frank Slack
> Molecular, Cellular and Developmental Biology
> Yale University
>
>
>
>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user



From robert.kern at gmail.com  Mon Jul 28 12:13:24 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Mon, 28 Jul 2008 11:13:24 -0500
Subject: [SciPy-user] temporary copies for in-place array modification?
In-Reply-To: <20080728114159.GA5776@ramrod.de>
References: <20080728114159.GA5776@ramrod.de>
Message-ID: <3d375d730807280913y49f580b7yc051384781933e0e@mail.gmail.com>

On Mon, Jul 28, 2008 at 06:41, Steve Schmerler <elcorto at gmx.net> wrote:
> Hi all
>
> Say I do this:
>
>    >>> a
>    array([1, 2, 3, 4, 5, 6])
>
>    >>> a[1:] = a[1:] - a[:-1]

The right-hand side of this will be a temporary array.

>    >>> a
>    array([1, 1, 1, 1, 1, 1])
>
>    >>> b = a[1:]
>    >>> ...
>
> I.e. I want to compute the difference of a's elements and store them in place
> in b = a[1:] to work with that (b is only a view so that's ok, no copy).
>
> Are there any temp copies of `a` involved? I ask b/c `a` will be large.
> Thanks!

Unfortunately, a temporary is unavoidable. If you modify `a` in-place,
you will mess up the computation. For example, we could try using the
third argument to the subtract() ufunc to place the results back into
a[1:]:

In [1]: import numpy

In [2]: a = numpy.arange(1,7)

In [3]: numpy.subtract(a[1:], a[:-1], a[1:])
Out[3]: array([1, 2, 2, 3, 3])

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From aisaac at american.edu  Mon Jul 28 12:48:39 2008
From: aisaac at american.edu (Alan G Isaac)
Date: Mon, 28 Jul 2008 12:48:39 -0400
Subject: [SciPy-user] temporary copies for in-place array modification?
In-Reply-To: <3d375d730807280913y49f580b7yc051384781933e0e@mail.gmail.com>
References: <20080728114159.GA5776@ramrod.de>
	<3d375d730807280913y49f580b7yc051384781933e0e@mail.gmail.com>
Message-ID: <Mahogany-0.67.0-1664-20080728-124839.00@american.edu>

On Mon, 28 Jul 2008, Robert Kern apparently wrote:
> Unfortunately, a temporary is unavoidable. If you modify `a` in-place, 
> you will mess up the computation. For example, we could try using the 
> third argument to the subtract() ufunc to place the results back into 
> a[1:]:
> In [1]: import numpy 
> In [2]: a = numpy.arange(1,7) 
> In [3]: numpy.subtract(a[1:], a[:-1], a[1:])
> Out[3]: array([1, 2, 2, 3, 3])


But he could work from the other end::

    >>> np.subtract(a[1:],a[:-1],a[:-1])
    array([1, 1, 1, 1, 1])

As long as he can then use a[:-1] instead of a[1:].
Right?

Alan





From robert.kern at gmail.com  Mon Jul 28 12:47:58 2008
From: robert.kern at gmail.com (Robert Kern)
Date: Mon, 28 Jul 2008 11:47:58 -0500
Subject: [SciPy-user] temporary copies for in-place array modification?
In-Reply-To: <Mahogany-0.67.0-1664-20080728-124839.00@american.edu>
References: <20080728114159.GA5776@ramrod.de>
	<3d375d730807280913y49f580b7yc051384781933e0e@mail.gmail.com>
	<Mahogany-0.67.0-1664-20080728-124839.00@american.edu>
Message-ID: <3d375d730807280947l2096f199p2f90d9d8cea57015@mail.gmail.com>

On Mon, Jul 28, 2008 at 11:48, Alan G Isaac <aisaac at american.edu> wrote:
> On Mon, 28 Jul 2008, Robert Kern apparently wrote:
>> Unfortunately, a temporary is unavoidable. If you modify `a` in-place,
>> you will mess up the computation. For example, we could try using the
>> third argument to the subtract() ufunc to place the results back into
>> a[1:]:
>> In [1]: import numpy
>> In [2]: a = numpy.arange(1,7)
>> In [3]: numpy.subtract(a[1:], a[:-1], a[1:])
>> Out[3]: array([1, 2, 2, 3, 3])
>
>
> But he could work from the other end::
>
>    >>> np.subtract(a[1:],a[:-1],a[:-1])
>    array([1, 1, 1, 1, 1])
>
> As long as he can then use a[:-1] instead of a[1:].
> Right?

Yup.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
 -- Umberto Eco


From peridot.faceted at gmail.com  Mon Jul 28 13:41:37 2008
From: peridot.faceted at gmail.com (Anne Archibald)
Date: Mon, 28 Jul 2008 13:41:37 -0400
Subject: [SciPy-user] temporary copies for in-place array modification?
In-Reply-To: <Mahogany-0.67.0-1664-20080728-124839.00@american.edu>
References: <20080728114159.GA5776@ramrod.de>
	<3d375d730807280913y49f580b7yc051384781933e0e@mail.gmail.com>
	<Mahogany-0.67.0-1664-20080728-124839.00@american.edu>
Message-ID: <ce557a360807281041h6297d0e1jb26040294ff29c97@mail.gmail.com>

2008/7/28 Alan G Isaac <aisaac at american.edu>:
> On Mon, 28 Jul 2008, Robert Kern apparently wrote:
>> Unfortunately, a temporary is unavoidable. If you modify `a` in-place,
>> you will mess up the computation. For example, we could try using the
>> third argument to the subtract() ufunc to place the results back into
>> a[1:]:
>> In [1]: import numpy
>> In [2]: a = numpy.arange(1,7)
>> In [3]: numpy.subtract(a[1:], a[:-1], a[1:])
>> Out[3]: array([1, 2, 2, 3, 3])
>
>
> But he could work from the other end::
>
>    >>> np.subtract(a[1:],a[:-1],a[:-1])
>    array([1, 1, 1, 1, 1])
>
> As long as he can then use a[:-1] instead of a[1:].
> Right?


Maybe. The result is dependent on the memory layout of a.

Anne


From jj20047 at gmail.com  Mon Jul 28 13:47:31 2008
From: jj20047 at gmail.com (jb)
Date: Mon, 28 Jul 2008 10:47:31 -0700
Subject: [SciPy-user] CPU you selected does not support x86-64 instruction
	set
Message-ID: <b22c3e260807281047n4d8a468dt1bb754b21a16185e@mail.gmail.com>

Hello:
I just installed a new version of numpy from svn on my 64-bit, Fedora
8 machine, but cannot get scipy to install.  I get the error: "CPU you
selected does not support x86-64 instruction set", which others have
had on this list.  Suggestions have been to install a new version of
numpy (which I did) and to use a gcc version > 3.4 (which I have).  So
I am stuck.  Any suggestions?

Here is a bit of information on my machine:
[root at quad1 scipy]# g77 -v
Reading specs from /usr/lib/gcc/x86_64-redhat-linux/3.4.6/specs
Configured with: ../configure --prefix=/usr --mandir=/usr/share/man
--infodir=/usr/share/info --enable-shared --enable-threads=posix
--disable-checking --with-system-zlib --enable-__cxa_atexit
--disable-libunwind-exceptions --enable-languages=c,c++,f77
--disable-libgcj --host=x86_64-redhat-linux
Thread model: posix
gcc version 3.4.6 20060404 (Red Hat 3.4.6-8)

[root at quad1 scipy]# uname -a
Linux quad1 2.6.25.10-47.fc8 #1 SMP Mon Jul 7 18:31:41 EDT 2008 x86_64
x86_64 x86_64 GNU/Linux

[john at quad1 site-packages]$ python numpy/distutils/cpuinfo.py
CPU information: CPUInfoBase__get_nbits=64 getNCPUs=4 has_3dnow
has_3dnowext has_mmx has_sse has_sse2 has_sse3 is_64bit is_AMD
is_AthlonK6_2

Here are the error messages during compile:

creating build/temp.linux-x86_64-2.5/scipy/fftpack/dfftpack
compile options: '-c'
g77:f77: scipy/fftpack/dfftpack/dsinti.f
scipy/fftpack/dfftpack/dsinti.f:0: error: CPU you selected does not
support x86-64 instruction set
scipy/fftpack/dfftpack/dsinti.f:0: error: CPU you selected does not
support x86-64 instruction set
scipy/fftpack/dfftpack/dsinti.f:0: error: CPU you selected does not
support x86-64 instruction set
scipy/fftpack/dfftpack/dsinti.f:0: error: CPU you selected does not
support x86-64 instruction set
error: Command "/usr/bin/g77 -g -Wall -fno-second-underscore -fPIC -O3
-funroll-loops -march=k6-2 -mmmx -m3dnow -msse2 -msse -msse3 -c -c
scipy/fftpack/dfftpack/dsinti.f -o
build/temp.linux-x86_64-2.5/scipy/fftpack/dfftpack/dsinti.o" failed
with exit status 1


From wnbell at gmail.com  Mon Jul 28 14:08:57 2008
From: wnbell at gmail.com (Nathan Bell)
Date: Mon, 28 Jul 2008 13:08:57 -0500
Subject: [SciPy-user] CPU you selected does not support x86-64
	instruction set
In-Reply-To: <b22c3e260807281047n4d8a468dt1bb754b21a16185e@mail.gmail.com>
References: <b22c3e260807281047n4d8a468dt1bb754b21a16185e@mail.gmail.com>
Message-ID: <d05265cb0807281108o146d7e15lbba2f13fd4a70ade@mail.gmail.com>

On Mon, Jul 28, 2008 at 12:47 PM, jb <jj20047 at gmail.com> wrote:
> Hello:
> I just installed a new version of numpy from svn on my 64-bit, Fedora
> 8 machine, but cannot get scipy to install.  I get the error: "CPU you
> selected does not support x86-64 instruction set", which others have
> had on this list.  Suggestions have been to install a new version of
> numpy (which I did) and to use a gcc version > 3.4 (which I have).  So
> I am stuck.  Any suggestions?
>

This ticket may be related:
http://projects.scipy.org/scipy/scipy/ticket/607

-- 
Nathan Bell wnbell at gmail.com
http://graphics.cs.uiuc.edu/~wnbell/


From robfelty at gmail.com  Mon Jul 28 14:27:20 2008
From: robfelty at gmail.com (Robert Felty)
Date: Mon, 28 Jul 2008 14:27:20 -0400
Subject: [SciPy-user] trouble installing audiolab
Message-ID: <200807281427.20767.robfelty@gmail.com>

I am having trouble installing the audiolab scipy toolkit. I have scipy 
installed just fine, but I am getting a library not found error when I try to 
install audiolab
This is the error I get
sudo python setup.py install
sndfile_info:
  libraries libsndfile.so.1.0.17 not found in /usr/lib64
  libraries libsndfile.so.1.0.17 not found in /usr/local/lib
  libraries libsndfile.so.1.0.17 not found in /usr/lib
Traceback (most recent call last):
  File "setup.py", line 221, in <module>
    'Topic :: Scientific/Engineering']
  File "/usr/lib64/python2.5/site-packages/numpy/distutils/core.py", line 142, 
in setup
    config = configuration()
  File "setup.py", line 144, in configuration
    sf_config   = sf_info.get_info(2)
  File "/usr/lib64/python2.5/site-packages/numpy/distutils/system_info.py", 
line 399, in get_info
    self.calc_info()
  File "setup.py", line 96, in calc_info
    raise SndfileNotFoundError("sndfile library not found")
__main__.SndfileNotFoundError: sndfile library not found

I am running on linux - Fedora 7 x86_64, with Pyton 2.5. I know that I have 
the libsndfile installed:
the 64bit version is here:
ls -l /usr/lib64/libsndfile*
-rw-r--r-- 1 root root 610500 2007-09-20 07:33 /usr/lib64/libsndfile.a
lrwxrwxrwx 1 root root     20 2008-07-28 12:45 /usr/lib64/libsndfile.so -> 
libsndfile.so.1.0.17
lrwxrwxrwx 1 root root     20 2007-09-27 14:01 /usr/lib64/libsndfile.so.1 -> 
libsndfile.so.1.0.17
-rwxr-xr-x 1 root root 328832 2007-09-20 07:33 /usr/lib64/libsndfile.so.1.0.17
. 

and the 32bit here:
ls -l /usr/lib/libsndfile*
-rw-r--r-- 1 root root 501314 2007-09-20 07:31 /usr/lib/libsndfile.a
lrwxrwxrwx 1 root root     20 2008-07-28 12:45 /usr/lib/libsndfile.so -> 
libsndfile.so.1.0.17
lrwxrwxrwx 1 root root     20 2007-09-27 14:01 /usr/lib/libsndfile.so.1 -> 
libsndfile.so.1.0.17
-rwxr-xr-x 1 root root 369612 2007-09-20 07:31 /usr/lib/libsndfile.so.1.0.17

I changed the default value of "libname" in the setup.py script from "sndfile" 
to "libsndfile" (and also tried several other variants, 
like "libsndfile.so", "libsndfile.so.1")

What I am missing here? Any help would be much appreciated.

Rob
--
Robert Felty
http://robfelty.com

"Verbing weirds language." -- Calvin (of Calvin and Hobbes)



From jj20047 at gmail.com  Mon Jul 28 17:06:04 2008
From: jj20047 at gmail.com (John)
Date: Mon, 28 Jul 2008 21:06:04 +0000 (UTC)
Subject: [SciPy-user]
	=?utf-8?q?CPU_you_selected_does_not_support_x86-64?=
	=?utf-8?q?=09instruction_set?=
References: <b22c3e260807281047n4d8a468dt1bb754b21a16185e@mail.gmail.com>
	<d05265cb0807281108o146d7e15lbba2f13fd4a70ade@mail.gmail.com>
Message-ID: <loom.20080728T210433-957@post.gmane.org>

Nathan Bell <wnbell <at> gmail.com> writes:

> This ticket may be related:
> http://projects.scipy.org/scipy/scipy/ticket/607
> 

That did it, thanks!




From fperez.net at gmail.com  Mon Jul 28 18:33:45 2008
From: fperez.net at gmail.com (Fernando Perez)
Date: Mon, 28 Jul 2008 15:33:45 -0700
Subject: [SciPy-user] Python tools at the annual SIAM meeting
Message-ID: <db6b5ecc0807281533xe151991gccfcc2275cb42a5d@mail.gmail.com>

Hi all,

for those interested, here's a brief report on the recent SIAM meeting
where a number of Python-based tools for scientific computing
(including ipython, numpy, scipy, sage, and more) were discussed:

http://fdoperez.blogspot.com/2008/07/python-tools-for-science-go-to-siam.html

The punch line is that we got selected for the annual highlights of
the conference:

http://www.ams.org/ams/siam-2008.html#python

Thanks again to all who contributed talks and attended!

Cheers,

f


From dominique.orban at gmail.com  Mon Jul 28 19:09:09 2008
From: dominique.orban at gmail.com (Dominique Orban)
Date: Mon, 28 Jul 2008 19:09:09 -0400
Subject: [SciPy-user] Python tools at the annual SIAM meeting
In-Reply-To: <db6b5ecc0807281533xe151991gccfcc2275cb42a5d@mail.gmail.com>
References: <db6b5ecc0807281533xe151991gccfcc2275cb42a5d@mail.gmail.com>
Message-ID: <8793ae6e0807281609i4075e048sbaefb22b3b211450@mail.gmail.com>

Fernando,
Congratulations on being selected for the highlights of the meeting. For
those of us who were not in San Diego, is there any chance to see the slides
of the talks in the three sessions you co-organized? That would be awesome.

Cheers,
Dominique

On Mon, Jul 28, 2008 at 6:33 PM, Fernando Perez <fperez.net at gmail.com>wrote:

> Hi all,
>
> for those interested, here's a brief report on the recent SIAM meeting
> where a number of Python-based tools for scientific computing
> (including ipython, numpy, scipy, sage, and more) were discussed:
>
>
> http://fdoperez.blogspot.com/2008/07/python-tools-for-science-go-to-siam.html
>
> The punch line is that we got selected for the annual highlights of
> the conference:
>
> http://www.ams.org/ams/siam-2008.html#python
>
> Thanks again to all who contributed talks and attended!
>
> Cheers,
>
> f
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080728/97b5ab1a/attachment.html>

From fperez.net at gmail.com  Mon Jul 28 19:16:40 2008
From: fperez.net at gmail.com (Fernando Perez)
Date: Mon, 28 Jul 2008 16:16:40 -0700
Subject: [SciPy-user] Python tools at the annual SIAM meeting
In-Reply-To: <8793ae6e0807281609i4075e048sbaefb22b3b211450@mail.gmail.com>
References: <db6b5ecc0807281533xe151991gccfcc2275cb42a5d@mail.gmail.com>
	<8793ae6e0807281609i4075e048sbaefb22b3b211450@mail.gmail.com>
Message-ID: <db6b5ecc0807281616g3700278fwbcf3f869b63ba462@mail.gmail.com>

Hi Dominique,

On Mon, Jul 28, 2008 at 4:09 PM, Dominique Orban
<dominique.orban at gmail.com> wrote:
> Fernando,
> Congratulations on being selected for the highlights of the meeting. For
> those of us who were not in San Diego, is there any chance to see the slides
> of the talks in the three sessions you co-organized? That would be awesome.

I should have thought of doing that before, sorry.  Here are my slides:

https://cirl.berkeley.edu/fperez/talks/0807_siam_intro_python_scicomp.pdf

and I'm in the process of contacting the others for theirs.  I'll
report back in a couple of days with the appropriate links as I get
them.

Cheers,


f


From zachary.pincus at yale.edu  Mon Jul 28 23:04:24 2008
From: zachary.pincus at yale.edu (Zachary Pincus)
Date: Mon, 28 Jul 2008 23:04:24 -0400
Subject: [SciPy-user] Matlab IO -- can others with matlab test this bug?
Message-ID: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>

Hi,

I have found that scipy's .mat-file IO creates version-5 files that  
cannot be read by the version of matlab I have access to, if a  
variable with less than five letters in the name is saved to that file.

The version of matlab I have is 6.5.0.180913a (R13). Here's an example  
to show the issue:

python:
 >>> import scipy.io
 >>> scipy.io.savemat('works.mat', {'abcde':1}, format='5')
 >>> scipy.io.savemat('fails.mat', {'abcd':1}, format='5')

matlab:
 >> load works
 >> load fails
??? Error using ==> load
Can't read file.

Could others please test this code out with the version(s) of Matlab  
that they have access to?

The issue is that it appears that (at least my version of) matlab  
assumes that for variables with <= 4-byte names, the "compressed data  
element" format will be used in the file. Scipy.io.savemat does not do  
this... However, the mat-file specification clearly states that this  
format is *optional*, and so the scipy routines are technically  
correct -- this is a matlab bug.

If this bug with matlab's file IO is widespread, I will make a patch  
for scipy so that short elements are written in the compressed format;  
otherwise I'll just work around the issue by ensuring that I never use  
any short names.

I would be very grateful if others would test this.

Zach



From david at ar.media.kyoto-u.ac.jp  Mon Jul 28 23:14:44 2008
From: david at ar.media.kyoto-u.ac.jp (David Cournapeau)
Date: Tue, 29 Jul 2008 12:14:44 +0900
Subject: [SciPy-user] trouble installing audiolab
In-Reply-To: <200807281427.20767.robfelty@gmail.com>
References: <200807281427.20767.robfelty@gmail.com>
Message-ID: <488E8B24.8010209@ar.media.kyoto-u.ac.jp>

Robert Felty wrote:
> I am having trouble installing the audiolab scipy toolkit. I have scipy 
> installed just fine, but I am getting a library not found error when I try to 
> install audiolab
>   

Hi Robert,

    I am sorry this is not working for you. I am not sure what the
problem is, though. You should not change libname to libsndfile, because
distutils adds it automatically for you (it will look for liblibsndfile
otherwise...).

Could you add this at line 92 of setup.py (print statement) ?

 for i in lib_dirs:
            tmp = self.check_libs(i, sndfile_libs)
            print tmp, i
            if tmp is not None:
                info    = tmp
                break

thanks,

David


From jjh at 42quarks.com  Mon Jul 28 23:43:41 2008
From: jjh at 42quarks.com (Jonathan Hunt)
Date: Tue, 29 Jul 2008 13:43:41 +1000
Subject: [SciPy-user] Matlab IO -- can others with matlab test this bug?
In-Reply-To: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>
References: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>
Message-ID: <e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>

Which version of SciPy are you using? I have 0.6.0 and MATLAB
7.6.0.324 (R2008a).

>> import scipy.io
>> scipy.io.savemat('works.mat', {'abcde':1}) # My version of SciPy doesn't support a format argument
>> scipy.io.savemat('fails.mat', {'abcd':1})

works fine.

Hope this helps.
Jonny

On Tue, Jul 29, 2008 at 1:04 PM, Zachary Pincus <zachary.pincus at yale.edu> wrote:
> Hi,
>
> I have found that scipy's .mat-file IO creates version-5 files that
> cannot be read by the version of matlab I have access to, if a
> variable with less than five letters in the name is saved to that file.
>
> The version of matlab I have is 6.5.0.180913a (R13). Here's an example
> to show the issue:
>
> python:
>  >>> import scipy.io
>  >>> scipy.io.savemat('works.mat', {'abcde':1}, format='5')
>  >>> scipy.io.savemat('fails.mat', {'abcd':1}, format='5')
>
> matlab:
>  >> load works
>  >> load fails
> ??? Error using ==> load
> Can't read file.
>
> Could others please test this code out with the version(s) of Matlab
> that they have access to?
>
> The issue is that it appears that (at least my version of) matlab
> assumes that for variables with <= 4-byte names, the "compressed data
> element" format will be used in the file. Scipy.io.savemat does not do
> this... However, the mat-file specification clearly states that this
> format is *optional*, and so the scipy routines are technically
> correct -- this is a matlab bug.
>
> If this bug with matlab's file IO is widespread, I will make a patch
> for scipy so that short elements are written in the compressed format;
> otherwise I'll just work around the issue by ensuring that I never use
> any short names.
>
> I would be very grateful if others would test this.
>
> Zach
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Jonathan J Hunt <jjh at 42quarks.com>
Homepage: http://www.42quarks.net.nz/wiki/JJH
(Further contact details there)
"Physics isn't the most important thing. Love is." Richard Feynman


From zachary.pincus at yale.edu  Tue Jul 29 00:04:09 2008
From: zachary.pincus at yale.edu (Zachary Pincus)
Date: Tue, 29 Jul 2008 00:04:09 -0400
Subject: [SciPy-user] Matlab IO -- can others with matlab test this bug?
In-Reply-To: <e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>
References: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>
	<e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>
Message-ID: <67E4404D-CC00-4775-BE98-5C0BF81B53B5@yale.edu>

Hi Johnny,

> Which version of SciPy are you using? I have 0.6.0 and MATLAB
> 7.6.0.324 (R2008a).

Foolish me for not specifying! I'm using a recent-ish SVN checkout of  
scipy: 0.7.0.dev4393.

It seems that the version-5 format writing came into the svn early  
this year, after the 0.6 release. The version-4 mat-file IO code seems  
to deal fine with short names.

Thanks again,

Zach


>
>
>>> import scipy.io
>>> scipy.io.savemat('works.mat', {'abcde':1}) # My version of SciPy  
>>> doesn't support a format argument
>>> scipy.io.savemat('fails.mat', {'abcd':1})
>
> works fine.
>
> Hope this helps.
> Jonny
>
> On Tue, Jul 29, 2008 at 1:04 PM, Zachary Pincus <zachary.pincus at yale.edu 
> > wrote:
>> Hi,
>>
>> I have found that scipy's .mat-file IO creates version-5 files that
>> cannot be read by the version of matlab I have access to, if a
>> variable with less than five letters in the name is saved to that  
>> file.
>>
>> The version of matlab I have is 6.5.0.180913a (R13). Here's an  
>> example
>> to show the issue:
>>
>> python:
>>>>> import scipy.io
>>>>> scipy.io.savemat('works.mat', {'abcde':1}, format='5')
>>>>> scipy.io.savemat('fails.mat', {'abcd':1}, format='5')
>>
>> matlab:
>>>> load works
>>>> load fails
>> ??? Error using ==> load
>> Can't read file.
>>
>> Could others please test this code out with the version(s) of Matlab
>> that they have access to?
>>
>> The issue is that it appears that (at least my version of) matlab
>> assumes that for variables with <= 4-byte names, the "compressed data
>> element" format will be used in the file. Scipy.io.savemat does not  
>> do
>> this... However, the mat-file specification clearly states that this
>> format is *optional*, and so the scipy routines are technically
>> correct -- this is a matlab bug.
>>
>> If this bug with matlab's file IO is widespread, I will make a patch
>> for scipy so that short elements are written in the compressed  
>> format;
>> otherwise I'll just work around the issue by ensuring that I never  
>> use
>> any short names.
>>
>> I would be very grateful if others would test this.
>>
>> Zach
>>
>> _______________________________________________
>> SciPy-user mailing list
>> SciPy-user at scipy.org
>> http://projects.scipy.org/mailman/listinfo/scipy-user
>>
>
>
>
> -- 
> Jonathan J Hunt <jjh at 42quarks.com>
> Homepage: http://www.42quarks.net.nz/wiki/JJH
> (Further contact details there)
> "Physics isn't the most important thing. Love is." Richard Feynman
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user



From aisaac at american.edu  Tue Jul 29 01:31:16 2008
From: aisaac at american.edu (Alan G Isaac)
Date: Tue, 29 Jul 2008 01:31:16 -0400
Subject: [SciPy-user] temporary copies for in-place array modification?
Message-ID: <488EAB24.4080009@american.edu>

>> On Mon, 28 Jul 2008, Robert Kern apparently wrote:
>>> Unfortunately, a temporary is unavoidable. If you modify `a` in-place, 
>>> you will mess up the computation. For example, we could try using the 
>>> third argument to the subtract() ufunc to place the results back into 
>>> a[1:]:
>>> In [1]: import numpy 
>>> In [2]: a = numpy.arange(1,7)
>>> In [3]: numpy.subtract(a[1:], a[:-1], a[1:])
>>> Out[3]: array([1, 2, 2, 3, 3])

> 2008/7/28 Alan G Isaac <aisaac at american.edu>:
>> But he could work from the other end::
>>    >>> np.subtract(a[1:],a[:-1],a[:-1])
>>    array([1, 1, 1, 1, 1])
>> As long as he can then use a[:-1] instead of a[1:]. 
>> Right? 


Anne Archibald wrote:
> Maybe. The result is dependent on the memory layout of a. 


What would be the right test for this being ok?
And would you give an example where this will fail?
I assumed the implicit loop would always be over the
elements in the order they appear in the array.
E.g., ::

     >>> a = np.arange(20)
     >>> a2 = a[::-2]
     >>> a2.flags
       C_CONTIGUOUS : False
       F_CONTIGUOUS : False
       OWNDATA : False
       WRITEABLE : True
       ALIGNED : True
       UPDATEIFCOPY : False
     >>> np.subtract(a2[1:],a2[:-1],a2[:-1])
     array([-2, -2, -2, -2, -2, -2, -2, -2, -2])
     >>> a2
     array([-2, -2, -2, -2, -2, -2, -2, -2, -2,  1])

Thanks,
Alan




From jjh at 42quarks.com  Tue Jul 29 01:37:29 2008
From: jjh at 42quarks.com (Jonathan Hunt)
Date: Tue, 29 Jul 2008 15:37:29 +1000
Subject: [SciPy-user] Matlab IO -- can others with matlab test this bug?
In-Reply-To: <67E4404D-CC00-4775-BE98-5C0BF81B53B5@yale.edu>
References: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>
	<e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>
	<67E4404D-CC00-4775-BE98-5C0BF81B53B5@yale.edu>
Message-ID: <e26778a60807282237i67a0294bxbf586073ec26ef2b@mail.gmail.com>

Hi Zach (and others),

I'd be happy to test the latest SVN checkout if someone could tell me
how I might build this in place (when I try and import from the scipy
directory directly it tells me that it needs to be installed which I
don't want to do as I want to leave the existing release version
installed).

Is there an easy way to go about using/modifying/testing the SVN
checkout without installing it each time one makes changes?

Thanks,
Jonny

On Tue, Jul 29, 2008 at 2:04 PM, Zachary Pincus <zachary.pincus at yale.edu> wrote:
> Hi Johnny,
>
>> Which version of SciPy are you using? I have 0.6.0 and MATLAB
>> 7.6.0.324 (R2008a).
>
> Foolish me for not specifying! I'm using a recent-ish SVN checkout of
> scipy: 0.7.0.dev4393.
>
> It seems that the version-5 format writing came into the svn early
> this year, after the 0.6 release. The version-4 mat-file IO code seems
> to deal fine with short names.
>
> Thanks again,
>
> Zach

-- 
Jonathan J Hunt <jjh at 42quarks.com>
Homepage: http://www.42quarks.net.nz/wiki/JJH
(Further contact details there)
"Physics isn't the most important thing. Love is." Richard Feynman


From wnbell at gmail.com  Tue Jul 29 03:13:48 2008
From: wnbell at gmail.com (Nathan Bell)
Date: Tue, 29 Jul 2008 02:13:48 -0500
Subject: [SciPy-user] temporary copies for in-place array modification?
In-Reply-To: <488EAB24.4080009@american.edu>
References: <488EAB24.4080009@american.edu>
Message-ID: <d05265cb0807290013g6c7fe2c1n716521a8fdf001a9@mail.gmail.com>

On Tue, Jul 29, 2008 at 12:31 AM, Alan G Isaac <aisaac at american.edu> wrote:
>> Maybe. The result is dependent on the memory layout of a.
>
> What would be the right test for this being ok?
> And would you give an example where this will fail?
> I assumed the implicit loop would always be over the
> elements in the order they appear in the array.

I wouldn't write code that depended on such details.  Doing so means
that any attempt to parallelize the implementation will break your
code in subtle, potentially non-deterministic ways.

-- 
Nathan Bell wnbell at gmail.com
http://graphics.cs.uiuc.edu/~wnbell/


From elcorto at gmx.net  Tue Jul 29 04:39:53 2008
From: elcorto at gmx.net (Steve Schmerler)
Date: Tue, 29 Jul 2008 10:39:53 +0200
Subject: [SciPy-user] temporary copies for in-place array modification?
In-Reply-To: <20080728114159.GA5776@ramrod.de>
References: <20080728114159.GA5776@ramrod.de>
Message-ID: <20080729083953.GA4340@ramrod.de>

On Jul 28 13:41, Steve Schmerler wrote:
> Hi all
> 
> Say I do this:
> 
>     >>> a
>     array([1, 2, 3, 4, 5, 6])
> 
>     >>> a[1:] = a[1:] - a[:-1]
> 
>     >>> a
>     array([1, 1, 1, 1, 1, 1])
> 
>     >>> b = a[1:]
>     >>> ...
> 
> I.e. I want to compute the difference of a's elements and store them in place
> in b = a[1:] to work with that (b is only a view so that's ok, no copy).
> 
> Are there any temp copies of `a` involved? I ask b/c `a` will be large.
> Thanks!
> 

Thanks for all answers. 

Actually, `a` will be a 3D array which I'm filling with 2D arrays in
a loop
    
    a = empty(x,y,z)
    first_2d_array(a[:,0,:]) # in-place modification
    for j in xrange(1, N+1):
        new_2d_array(a[:,j,:])

and I get 2D arrays of differences.

    a[:,1:,:] = a[:,1:,:] a[:,:-1,:]
    b = [:,1:,:]

ATM a's size is in the order of several MB, so the cost of the temporary is
nothing.

But for the future, since the temporary is in general unavoidable (or it's
safe to consider one), I'll use smaller temp 2D arrays in the loop and compute
the diffs right there (and b.shape[1] = a.shape[1]-1). 
    
    old = first_2d_array(...)
    for j in xrange(0, N)
        # view only, operate on `w`
        w = b[:,j,:]
        new_2d_array(w)
        tmp = w.copy()
        w -= old
        old = tmp


steve


From zachary.pincus at yale.edu  Tue Jul 29 08:42:51 2008
From: zachary.pincus at yale.edu (Zachary Pincus)
Date: Tue, 29 Jul 2008 08:42:51 -0400
Subject: [SciPy-user] Matlab IO -- can others with matlab test this bug?
In-Reply-To: <e26778a60807282237i67a0294bxbf586073ec26ef2b@mail.gmail.com>
References: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>
	<e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>
	<67E4404D-CC00-4775-BE98-5C0BF81B53B5@yale.edu>
	<e26778a60807282237i67a0294bxbf586073ec26ef2b@mail.gmail.com>
Message-ID: <8D21E970-E9A5-470C-A8B7-97960BB3C0FA@yale.edu>

Hi Jonny,

> I'd be happy to test the latest SVN checkout if someone could tell me
> how I might build this in place (when I try and import from the scipy
> directory directly it tells me that it needs to be installed which I
> don't want to do as I want to leave the existing release version
> installed).
>
> Is there an easy way to go about using/modifying/testing the SVN
> checkout without installing it each time one makes changes?

Perhaps some others can chime in with more "official" methods, but I  
typically do one of several things to test SVN versions:

(1) I cd to the build/lib-whatever subdirectory, after 'python  
setup.py build'. This dir contains a mostly-complete scipy install.  
After starting python and making sure that '.' is first on the  
sys.path (so that the scipy dir in that directory shadows the system  
one), 'import scipy' should get the new version. Alternately, you can  
start python from anywhere, and then do 'import sys; sys.path.insert('/ 
path/to/build/lib...', 0); import scipy' to get the same effect.

(2) The above method fails for certain things like using f2py, which  
needs certain source files that get installed, but not copied over  
into build/lib. In those cases, you can install scipy into an  
alternated location somewhere, and then add that to sys.path (or to  
the PYTHONPATH environment variable). Use the '--install-base' or '-- 
prefix' switch to 'python setup.py install' to control where the files  
get installed. I'm not really sure which is most appropriate here.

(3) In this particular case, scipy.io.matlab doesn't involve any  
compiled modules. You could just cd to scipy/io/matlab in the SVN  
checkout, and then 'import mio' should work. The desired function is  
mio.savemat....

Zach








From david at ar.media.kyoto-u.ac.jp  Tue Jul 29 08:37:30 2008
From: david at ar.media.kyoto-u.ac.jp (David Cournapeau)
Date: Tue, 29 Jul 2008 21:37:30 +0900
Subject: [SciPy-user] Matlab IO -- can others with matlab test this bug?
In-Reply-To: <8D21E970-E9A5-470C-A8B7-97960BB3C0FA@yale.edu>
References: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>	<e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>	<67E4404D-CC00-4775-BE98-5C0BF81B53B5@yale.edu>	<e26778a60807282237i67a0294bxbf586073ec26ef2b@mail.gmail.com>
	<8D21E970-E9A5-470C-A8B7-97960BB3C0FA@yale.edu>
Message-ID: <488F0F0A.70105@ar.media.kyoto-u.ac.jp>

Zachary Pincus wrote:
>
> Perhaps some others can chime in with more "official" methods, but I  
> typically do one of several things to test SVN versions:
>   

On Unix, a nice and easy way to easily switch between different numpy
installs without touching at all the environment is stow. Basically what
you do once stow is installed is:

python setup.py install --prefix=SOMEPATH/stow/numpy1
python setup.py install --prefix=SOMEPATH/stow/numpy2

And then stow makes link between the 'real' numpy and the one you want:

stow numpy1 -> numpy1 is used
stow -R numpy1 -> numpy1 is disabled
stow numpy2 -> numpy2 is enabled

I switch between many numpy configurations (one which only uses released
versions for 'real' work, many other with various compilers combination
to test the build system). It also enables true uninstallation, since
any installed version is self-contained in one directory (you can rm -rf
the directory without any chance of removing something unrelated).

cheers,

David


From jjh at 42quarks.com  Tue Jul 29 09:31:32 2008
From: jjh at 42quarks.com (Jonathan Hunt)
Date: Tue, 29 Jul 2008 23:31:32 +1000
Subject: [SciPy-user] Running SciPy SVN without installing Was: Matlab IO --
	can others with matlab test this bug?
Message-ID: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>

Hi,

Thanks for the advice. But all of these seem like hacks. Is there any
clean way to use SVN SciPy without installing it. This is particularly
important since I'm hoping to modify SciPy (I want to add a mio
wrapper for MATLAB 7/HDF5 files) and obviously I would like to avoid
having the build/install everytime I make a modification to the
library that I want to test.

Thanks,
Jonny

On Tue, Jul 29, 2008 at 10:37 PM, David Cournapeau
<david at ar.media.kyoto-u.ac.jp> wrote:
> Zachary Pincus wrote:
>>
>> Perhaps some others can chime in with more "official" methods, but I
>> typically do one of several things to test SVN versions:
>>
>
> On Unix, a nice and easy way to easily switch between different numpy
> installs without touching at all the environment is stow. Basically what
> you do once stow is installed is:
>
> python setup.py install --prefix=SOMEPATH/stow/numpy1
> python setup.py install --prefix=SOMEPATH/stow/numpy2
>
> And then stow makes link between the 'real' numpy and the one you want:
>
> stow numpy1 -> numpy1 is used
> stow -R numpy1 -> numpy1 is disabled
> stow numpy2 -> numpy2 is enabled
>
> I switch between many numpy configurations (one which only uses released
> versions for 'real' work, many other with various compilers combination
> to test the build system). It also enables true uninstallation, since
> any installed version is self-contained in one directory (you can rm -rf
> the directory without any chance of removing something unrelated).
>
> cheers,
>
> David

-- 
Jonathan J Hunt <jjh at 42quarks.com>
Homepage: http://www.42quarks.net.nz/wiki/JJH
(Further contact details there)
"Physics isn't the most important thing. Love is." Richard Feynman


From david at ar.media.kyoto-u.ac.jp  Tue Jul 29 09:23:56 2008
From: david at ar.media.kyoto-u.ac.jp (David Cournapeau)
Date: Tue, 29 Jul 2008 22:23:56 +0900
Subject: [SciPy-user] Running SciPy SVN without installing Was: Matlab
 IO --	can others with matlab test this bug?
In-Reply-To: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
Message-ID: <488F19EC.2060506@ar.media.kyoto-u.ac.jp>

Jonathan Hunt wrote:
> Hi,
>
> Thanks for the advice. But all of these seem like hacks.

stow is certainly not an hack, particularly compared to not installing
scipy. There is the possibility to use the develop mode of setuptools,
or varisous "sandboxed" python, but they are likely to be more work, and
I don't know anything about them, so I can't really comment on them (I
know other scipy developers do, though).

>  Is there any
> clean way to use SVN SciPy without installing it. This is particularly
> important since I'm hoping to modify SciPy (I want to add a mio
> wrapper for MATLAB 7/HDF5 files) and obviously I would like to avoid
> having the build/install everytime I make a modification to the
> library that I want to test.
>   

This is not obvious. For most softwares, you cannot test them without
'installing' them, why should it be different for python packages ? Any
method enabling the test of the software without installing it (besides
sandbox) is an hack by definition.

cheers,

David


From emanuele at relativita.com  Tue Jul 29 10:04:03 2008
From: emanuele at relativita.com (Emanuele Olivetti)
Date: Tue, 29 Jul 2008 16:04:03 +0200
Subject: [SciPy-user] Running SciPy SVN without installing Was: Matlab
 IO --	can others with matlab test this bug?
In-Reply-To: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
Message-ID: <488F2353.6040106@relativita.com>

Insted of stow why not using just PYTHONPATH?
(in bash, continuing the Zachary's example on numpy1)
export PYTHONPATH="SOMEPATH/stow/numpy1"
then run python/emacs/ipython/whatever and use numpy1.

Emanuele

Jonathan Hunt wrote:
> Hi,
>
> Thanks for the advice. But all of these seem like hacks. Is there any
> clean way to use SVN SciPy without installing it. This is particularly
> important since I'm hoping to modify SciPy (I want to add a mio
> wrapper for MATLAB 7/HDF5 files) and obviously I would like to avoid
> having the build/install everytime I make a modification to the
> library that I want to test.
>
> Thanks,
> Jonny
>
> On Tue, Jul 29, 2008 at 10:37 PM, David Cournapeau
> <david at ar.media.kyoto-u.ac.jp> wrote:
>> Zachary Pincus wrote:
>>> Perhaps some others can chime in with more "official" methods, but I
>>> typically do one of several things to test SVN versions:
>>>
>> On Unix, a nice and easy way to easily switch between different numpy
>> installs without touching at all the environment is stow. Basically what
>> you do once stow is installed is:
>>
>> python setup.py install --prefix=SOMEPATH/stow/numpy1
>> python setup.py install --prefix=SOMEPATH/stow/numpy2
>>
>> And then stow makes link between the 'real' numpy and the one you want:
>>
>> stow numpy1 -> numpy1 is used
>> stow -R numpy1 -> numpy1 is disabled
>> stow numpy2 -> numpy2 is enabled
>>
>> I switch between many numpy configurations (one which only uses released
>> versions for 'real' work, many other with various compilers combination
>> to test the build system). It also enables true uninstallation, since
>> any installed version is self-contained in one directory (you can rm -rf
>> the directory without any chance of removing something unrelated).
>>
>> cheers,
>>
>> David
>



From david at ar.media.kyoto-u.ac.jp  Tue Jul 29 09:56:17 2008
From: david at ar.media.kyoto-u.ac.jp (David Cournapeau)
Date: Tue, 29 Jul 2008 22:56:17 +0900
Subject: [SciPy-user] Running SciPy SVN without installing Was: Matlab
 IO --	can others with matlab test this bug?
In-Reply-To: <488F2353.6040106@relativita.com>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
	<488F2353.6040106@relativita.com>
Message-ID: <488F2181.3040307@ar.media.kyoto-u.ac.jp>

Emanuele Olivetti wrote:
> Insted of stow why not using just PYTHONPATH?
> (in bash, continuing the Zachary's example on numpy1)
> export PYTHONPATH="SOMEPATH/stow/numpy1"
> then run python/emacs/ipython/whatever and use numpy1.
>   

it breaks (or more exactly becomes complicated) once you depend on
several interdependent packages (say numpy and scipy, for example).
Granted, it may not be a problem for the problem encountered by John,
but I am always surprised by the number of people who use - buggy -
hacks to do something for which stow was invented.

cheers,

David


From jjh at 42quarks.com  Tue Jul 29 10:15:38 2008
From: jjh at 42quarks.com (Jonathan Hunt)
Date: Wed, 30 Jul 2008 00:15:38 +1000
Subject: [SciPy-user] Running SciPy SVN without installing Was: Matlab
	IO -- can others with matlab test this bug?
In-Reply-To: <488F2181.3040307@ar.media.kyoto-u.ac.jp>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
	<488F2353.6040106@relativita.com>
	<488F2181.3040307@ar.media.kyoto-u.ac.jp>
Message-ID: <e26778a60807290715s76b479c7if30f46dca8fe2477@mail.gmail.com>

Hi,

Firstly David, hack was not intended as an insult. Thanks for pointing
me to stow I haven't heard of it before.

However. Could someone tell me how most SciPy developers work on the
library. Because stow still doesn't seem to get around the need to do
a build/install everytime one makes a modification to the library. I
look at online documentation for developers but I can't find any
clues. What's the best way to modify/test SciPy SVN?

Thanks,
Jonny

On Tue, Jul 29, 2008 at 11:56 PM, David Cournapeau
<david at ar.media.kyoto-u.ac.jp> wrote:
> Emanuele Olivetti wrote:
>> Insted of stow why not using just PYTHONPATH?
>> (in bash, continuing the Zachary's example on numpy1)
>> export PYTHONPATH="SOMEPATH/stow/numpy1"
>> then run python/emacs/ipython/whatever and use numpy1.
>>
>
> it breaks (or more exactly becomes complicated) once you depend on
> several interdependent packages (say numpy and scipy, for example).
> Granted, it may not be a problem for the problem encountered by John,
> but I am always surprised by the number of people who use - buggy -
> hacks to do something for which stow was invented.
>
> cheers,
>
> David
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Jonathan J Hunt <jjh at 42quarks.com>
Homepage: http://www.42quarks.net.nz/wiki/JJH
(Further contact details there)
"Physics isn't the most important thing. Love is." Richard Feynman


From pgmdevlist at gmail.com  Tue Jul 29 10:15:24 2008
From: pgmdevlist at gmail.com (Pierre GM)
Date: Tue, 29 Jul 2008 10:15:24 -0400
Subject: [SciPy-user] Running SciPy SVN without installing Was: Matlab
	IO -- can others with matlab test this bug?
In-Reply-To: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
Message-ID: <200807291015.24171.pgmdevlist@gmail.com>

Jonathan,

Hopefully you won't need to modify a lot of files all over Scipy, but just a 
few of them ? Then you may want to install the SVN version and create 
symbolic links pointing to the files/directories you'll be actually 
modifying.
That works well, may come to bite you if you don't delete an old installation 
of Scipy before installing a new one, and is a bit problematic from times to 
times, but (goto 1)... 


From david at ar.media.kyoto-u.ac.jp  Tue Jul 29 10:07:52 2008
From: david at ar.media.kyoto-u.ac.jp (David Cournapeau)
Date: Tue, 29 Jul 2008 23:07:52 +0900
Subject: [SciPy-user] Running SciPy SVN without installing Was: Matlab
 IO -- can others with matlab test this bug?
In-Reply-To: <e26778a60807290715s76b479c7if30f46dca8fe2477@mail.gmail.com>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>	<488F2353.6040106@relativita.com>	<488F2181.3040307@ar.media.kyoto-u.ac.jp>
	<e26778a60807290715s76b479c7if30f46dca8fe2477@mail.gmail.com>
Message-ID: <488F2438.9040300@ar.media.kyoto-u.ac.jp>

Jonathan Hunt wrote:
> Hi,
>
> Firstly David, hack was not intended as an insult.

I did not take it as an insult, don't worry. I just pointed out that any
supposedly non hackish thing is more hackish than stow.

>
> However. Could someone tell me how most SciPy developers work on the
> library. 

Personally, that's how I develop on scipy and numpy.

> Because stow still doesn't seem to get around the need to do
> a build/install everytime one makes a modification to the library.

I think that getting around this need brings more problems than it
solves. I always install and test from the installed version; in most
cases, I even clean the installed version first because this often leads
to subtle problems. Installing numpy/scipy like that takes a few
seconds, and is easy to automate by make, rake files if I need
portability on Windows, etc...

cheers,

David


From david at ar.media.kyoto-u.ac.jp  Tue Jul 29 10:08:57 2008
From: david at ar.media.kyoto-u.ac.jp (David Cournapeau)
Date: Tue, 29 Jul 2008 23:08:57 +0900
Subject: [SciPy-user] Running SciPy SVN without installing Was: Matlab
 IO -- can others with matlab test this bug?
In-Reply-To: <200807291015.24171.pgmdevlist@gmail.com>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
	<200807291015.24171.pgmdevlist@gmail.com>
Message-ID: <488F2479.8020706@ar.media.kyoto-u.ac.jp>

Pierre GM wrote:
> Jonathan,
>
> Hopefully you won't need to modify a lot of files all over Scipy, but just a 
> few of them ? Then you may want to install the SVN version and create 
> symbolic links pointing to the files/directories you'll be actually 
> modifying.
>   

That's precisely how stow works ...

> That works well, may come to bite you if you don't delete an old installation 
> of Scipy before installing a new one, and is a bit problematic from times to 
> times, but (goto 1)... 
>   

... except that it solves this problem too :) Really, stow is nice. The
only problem is that it does not work on windows.

cheers,

David


From jjh at 42quarks.com  Tue Jul 29 10:32:08 2008
From: jjh at 42quarks.com (Jonathan Hunt)
Date: Wed, 30 Jul 2008 00:32:08 +1000
Subject: [SciPy-user] Running SciPy SVN without installing Was: Matlab
	IO -- can others with matlab test this bug?
In-Reply-To: <488F2479.8020706@ar.media.kyoto-u.ac.jp>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
	<200807291015.24171.pgmdevlist@gmail.com>
	<488F2479.8020706@ar.media.kyoto-u.ac.jp>
Message-ID: <e26778a60807290732s5564ea84pac3d4737781a2c3e@mail.gmail.com>

Hi,

Thanks for all the comments. If anyone uses the develop mode of
setuptools or other methods I'd be interested to hear.

Also, might I suggest (apologies if they do exist somewhere and I
missed them) that the (summarized) contents of this thread might be
useful on the SciPy developers wiki to lower the barrier for others
getting started hacking SciPy.

Thanks for all the help so far,
Jonny


-- 
Jonathan J Hunt <jjh at 42quarks.com>
Homepage: http://www.42quarks.net.nz/wiki/JJH
(Further contact details there)
"Physics isn't the most important thing. Love is." Richard Feynman


From pgmdevlist at gmail.com  Tue Jul 29 10:54:32 2008
From: pgmdevlist at gmail.com (Pierre GM)
Date: Tue, 29 Jul 2008 10:54:32 -0400
Subject: [SciPy-user] Running SciPy SVN without installing Was: Matlab
	IO -- can others with matlab test this bug?
In-Reply-To: <488F2479.8020706@ar.media.kyoto-u.ac.jp>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
	<200807291015.24171.pgmdevlist@gmail.com>
	<488F2479.8020706@ar.media.kyoto-u.ac.jp>
Message-ID: <200807291054.32764.pgmdevlist@gmail.com>

On Tuesday 29 July 2008 10:08:57 David Cournapeau wrote:
> Pierre GM wrote:
> > Then you may want to install the SVN version and
> > create symbolic links pointing to the files/directories you'll be
> > actually modifying.
>
> That's precisely how stow works ...

OK, I just found about stow. I gonna have to try. Thanks for the info.


From strawman at astraw.com  Wed Jul 30 02:28:38 2008
From: strawman at astraw.com (Andrew Straw)
Date: Tue, 29 Jul 2008 23:28:38 -0700
Subject: [SciPy-user] Matlab IO -- can others with matlab test this bug?
In-Reply-To: <488F0F0A.70105@ar.media.kyoto-u.ac.jp>
References: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>	<e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>	<67E4404D-CC00-4775-BE98-5C0BF81B53B5@yale.edu>	<e26778a60807282237i67a0294bxbf586073ec26ef2b@mail.gmail.com>	<8D21E970-E9A5-470C-A8B7-97960BB3C0FA@yale.edu>
	<488F0F0A.70105@ar.media.kyoto-u.ac.jp>
Message-ID: <48900A16.6050401@astraw.com>

David Cournapeau wrote:
> Zachary Pincus wrote:
>   
>> Perhaps some others can chime in with more "official" methods, but I  
>> typically do one of several things to test SVN versions:
>>   
>>     
>
> On Unix, a nice and easy way to easily switch between different numpy
> installs without touching at all the environment is stow. Basically what
> you do once stow is installed is:
>
> python setup.py install --prefix=SOMEPATH/stow/numpy1
> python setup.py install --prefix=SOMEPATH/stow/numpy2
>
> And then stow makes link between the 'real' numpy and the one you want:
>
> stow numpy1 -> numpy1 is used
> stow -R numpy1 -> numpy1 is disabled
> stow numpy2 -> numpy2 is enabled
>
> I switch between many numpy configurations (one which only uses released
> versions for 'real' work, many other with various compilers combination
> to test the build system). It also enables true uninstallation, since
> any installed version is self-contained in one directory (you can rm -rf
> the directory without any chance of removing something unrelated).
>   
No time to do a similar writeup for it, but: virtualenv

http://pypi.python.org/pypi/virtualenv


From jjh at 42quarks.com  Wed Jul 30 02:42:46 2008
From: jjh at 42quarks.com (Jonathan Hunt)
Date: Wed, 30 Jul 2008 16:42:46 +1000
Subject: [SciPy-user] Matlab IO -- can others with matlab test this bug?
In-Reply-To: <48900A16.6050401@astraw.com>
References: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>
	<e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>
	<67E4404D-CC00-4775-BE98-5C0BF81B53B5@yale.edu>
	<e26778a60807282237i67a0294bxbf586073ec26ef2b@mail.gmail.com>
	<8D21E970-E9A5-470C-A8B7-97960BB3C0FA@yale.edu>
	<488F0F0A.70105@ar.media.kyoto-u.ac.jp> <48900A16.6050401@astraw.com>
Message-ID: <e26778a60807292342g74d01fa7ie7f4553b51c45201@mail.gmail.com>

This is no doubt a dumb question. But why it is not possible simply to
import the library in place (i.e no install necessary, just use the
python files as they are laid out in the SVN co directory). Is it
because of the C/Fortran files that need to be compiled?

Thanks,
Jonny

-- 
Jonathan J Hunt <jjh at 42quarks.com>
Homepage: http://www.42quarks.net.nz/wiki/JJH
(Further contact details there)
"Physics isn't the most important thing. Love is." Richard Feynman


From nwagner at iam.uni-stuttgart.de  Wed Jul 30 05:41:03 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Wed, 30 Jul 2008 11:41:03 +0200
Subject: [SciPy-user] Differential Algebraic Equation Solvers
Message-ID: <web-105374238@uni-stuttgart.de>

Hi all,

I am looking for a python tool to solve a set
of DAE's, e.g.

m \ddot{x} = 2 \lambda x
m \ddot{y} = 2 \lambda y
x^2 + y^2 - l^2 = 0

  
Any pointer ?

Nils


From jr at sun.ac.za  Wed Jul 30 06:43:12 2008
From: jr at sun.ac.za (Johann Rohwer)
Date: Wed, 30 Jul 2008 12:43:12 +0200
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <web-105374238@uni-stuttgart.de>
References: <web-105374238@uni-stuttgart.de>
Message-ID: <200807301243.12871.jr@sun.ac.za>

On Wednesday, 30 July 2008, Nils Wagner wrote:
> Hi all,
>
> I am looking for a python tool to solve a set
> of DAE's, e.g.
>
> m \ddot{x} = 2 \lambda x
> m \ddot{y} = 2 \lambda y
> x^2 + y^2 - l^2 = 0

Our group has developed PySUNDIALS 
(http://pysundials.sourceforge.net), which is a Python package 
providing bindings for the SUNDIALS suite of solvers (using ctypes). 
SUNDIALS has a DAE solver (IDA), which should be able to handle your 
kind of problem. The PySUNDIALS distributrion comes with examples for 
each of the wrapped solvers.

Johann


From nwagner at iam.uni-stuttgart.de  Wed Jul 30 07:07:36 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Wed, 30 Jul 2008 13:07:36 +0200
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <200807301243.12871.jr@sun.ac.za>
References: <web-105374238@uni-stuttgart.de> <200807301243.12871.jr@sun.ac.za>
Message-ID: <web-105378668@uni-stuttgart.de>

On Wed, 30 Jul 2008 12:43:12 +0200
  Johann Rohwer <jr at sun.ac.za> wrote:
> On Wednesday, 30 July 2008, Nils Wagner wrote:
>> Hi all,
>>
>> I am looking for a python tool to solve a set
>> of DAE's, e.g.
>>
>> m \ddot{x} = 2 \lambda x
>> m \ddot{y} = 2 \lambda y
>> x^2 + y^2 - l^2 = 0
> 
> Our group has developed PySUNDIALS 
> (http://pysundials.sourceforge.net), which is a Python 
>package 
> providing bindings for the SUNDIALS suite of solvers 
>(using ctypes). 
> SUNDIALS has a DAE solver (IDA), which should be able to 
>handle your 
> kind of problem. The PySUNDIALS distributrion comes with 
>examples for 
> each of the wrapped solvers.
> 
> Johann


Hi Johann,

I have installed pysundials.
Can you give me an advice how to implement my
simple example using pysundials ?

idadenx.py seems to be a good starting point.

How do I define the set of equations, initial conditions 
etc. ?
Is it necessary to define the Jacobian ?

Thank you in advance.

      Nils


From ferrell at diablotech.com  Wed Jul 30 08:45:46 2008
From: ferrell at diablotech.com (Robert Ferrell)
Date: Wed, 30 Jul 2008 06:45:46 -0600
Subject: [SciPy-user] OS X install success! + Question
Message-ID: <5BE66D24-E4A2-4076-BF49-B77B1A1459C2@diablotech.com>

I just installed NumPy & SciPy on a new Mac Book Pro (OS X 10.5.2).   
Thanks for the great step-by-step instructions at http://www.scipy.org/Installing_SciPy/Mac_OS_X 
.

The whole process was flawless until I tried to run the tests and I  
found I don't have nose installed.  That leads to an elementary  
question about Python and OS X.

Before installing NumPy & SciPy I installed the OS X Python from  
python.org.  That's 2.5.2 (laptop shipped with 2.5.1) and was  
installed in

/Library/Frameworks/Python.framework/Versions/Current/bin/python

which is just a link to

/Library/Frameworks/Python.framework/Versions/2.5/bin/python.

numpy and scipy installed in /Library/Frameworks/Python.framework/ 
Versions/2.5/lib/python2.5/site-packages/

So far so good.  But, when I used easy_install for nose, that got  
installed into

/Library/Python/2.5/site-packages/.

What is the best way to deal with the Frameworks vs no Frameworks  
thing?  Should I just put a link from /Library/Python/2.5 to /Library/ 
Frameworks/Python.framework/Versions/2.5/lib/python2.5?

I realize this is a Python + OS X question and has nothing to do with  
SciPy per se.  But I've searched the web for an answer with no luck.   
It seems like this group is the most likely to provide useful  
information.

Thanks, and thanks for the awesome install page.  SciPy has come a  
very long way!

-robert




From zachary.pincus at yale.edu  Wed Jul 30 09:57:07 2008
From: zachary.pincus at yale.edu (Zachary Pincus)
Date: Wed, 30 Jul 2008 09:57:07 -0400
Subject: [SciPy-user] OS X install success! + Question
In-Reply-To: <5BE66D24-E4A2-4076-BF49-B77B1A1459C2@diablotech.com>
References: <5BE66D24-E4A2-4076-BF49-B77B1A1459C2@diablotech.com>
Message-ID: <5449A4F1-E90B-430B-9223-72AFC7C9DF40@yale.edu>

Hi Robert,

I'm sure others with a little deeper knowledge of easy_install and  
friends may chime in, but this should get you started.

>
> Before installing NumPy & SciPy I installed the OS X Python from
> python.org.  That's 2.5.2 (laptop shipped with 2.5.1) and was
> installed in
>
> /Library/Frameworks/Python.framework/Versions/Current/bin/python
>
> which is just a link to
>
> /Library/Frameworks/Python.framework/Versions/2.5/bin/python.
>
> numpy and scipy installed in /Library/Frameworks/Python.framework/
> Versions/2.5/lib/python2.5/site-packages/
>
> So far so good.  But, when I used easy_install for nose, that got
> installed into
>
> /Library/Python/2.5/site-packages/.
>
> What is the best way to deal with the Frameworks vs no Frameworks
> thing?  Should I just put a link from /Library/Python/2.5 to /Library/
> Frameworks/Python.framework/Versions/2.5/lib/python2.5?

Short answer: the version of easy_install on your machine is for the  
version of python that came with the operating system. So  
"easy_install nose" installed nose for that version -- not for the  
python.org python that you just installed.

You will need to install easy_install for the new version of python  
that you just installed from python.org:
http://peak.telecommunity.com/DevCenter/EasyInstall#installing-easy-install

Then run that version of easy_install that gets installed by the  
ez_setup script. That version, for better or worse, lives by default in:
/Library/Frameworks/Python.framework/Versions/2.5/bin/easy_install

(I would recommend making a symlink to /usr/local/bin for ease.  
Alternately, you could run the easy_setup.py script with the --script- 
dir=/usr/local/bin option.)


Background information:
Apple ships a version of Python with OS X, and some housekeeping  
scripts, etc., depend on this python and in some cases, the specific  
(old or broken) versions of some of the libraries it comes with (e.g.  
numpy). This is installed in:
/System/Library/Frameworks/Python.framework/Versions/2.5

This version of python is accessible as /usr/bin/python, which is a  
symlink to
/System/Library/Frameworks/Python.framework/Versions/2.5/bin/python

The basic consensus is that one really shouldn't be mucking with this  
python, so the best bet is to install the python.org python, which  
goes into /Library and /usr/local/bin.

Now, what's the deal with /Library/Python/2.5/site-packages? Well,  
Apple's basic decree is that you must not alter things in /System/...  
directories. But installing a new library for the system python would  
do just that, under the default regime (which would put them in
/System/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/ 
site-packages )
So Apple's python instead puts new libraries in /Library/Python/2.5/ 
site-packages .

This is all fine, and controlled by the default values of the sys.path  
variable (which is the list of directories Python searches when you  
try to import something). It's illuminating to do:
import sys; print sys.path
in both the system and python.org versions of python.

Now, how does one deal with multiple versions of python when  
installing things? In the pre-easy_install era, it was quite simple.  
Python things are installed with setup.py files in a directory:

python setup.py install

would install whatever library for the version of python that was used  
to run the script. So /usr/bin/python setup.py install would install  
the library for the system python, and /usr/local/bin/python setup.py  
install would do so for the python.org version. (Plain 'python' is  
the /usr/local/bin version, thanks to the python.org installer.

With easy_install, this sort of transparency is a bit lost. You need  
to look to see which version of python is invoked in the first line of  
that script to determine where the files will eventually go. If you  
have two versions of python, you need two versions of easy_install in  
different places. I personally find this a bit frustrating...

Zach









From zachary.pincus at yale.edu  Wed Jul 30 10:00:08 2008
From: zachary.pincus at yale.edu (Zachary Pincus)
Date: Wed, 30 Jul 2008 10:00:08 -0400
Subject: [SciPy-user] Matlab IO -- can others with matlab test this bug?
In-Reply-To: <e26778a60807292342g74d01fa7ie7f4553b51c45201@mail.gmail.com>
References: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>
	<e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>
	<67E4404D-CC00-4775-BE98-5C0BF81B53B5@yale.edu>
	<e26778a60807282237i67a0294bxbf586073ec26ef2b@mail.gmail.com>
	<8D21E970-E9A5-470C-A8B7-97960BB3C0FA@yale.edu>
	<488F0F0A.70105@ar.media.kyoto-u.ac.jp>
	<48900A16.6050401@astraw.com>
	<e26778a60807292342g74d01fa7ie7f4553b51c45201@mail.gmail.com>
Message-ID: <CD201218-2B0D-4978-971C-39AB6E84F0DD@yale.edu>

> This is no doubt a dumb question. But why it is not possible simply to
> import the library in place (i.e no install necessary, just use the
> python files as they are laid out in the SVN co directory). Is it
> because of the C/Fortran files that need to be compiled?

Yes, basically. Pure-python libraries can usually be imported in-place  
-- though sometimes, too-clever setup.py files can actually make it so  
the installed folder structure is different from the one in the source  
directory, and thus an in-place import will fail. (I think this is  
probably considered bad form though.)

In the specific, limited case of the matlab IO sub-sub-library, it is  
indeed pure python and can be imported in-place -- this was one of my  
hackish suggestions for testing it.

Zach


From gael.varoquaux at normalesup.org  Wed Jul 30 10:10:40 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Wed, 30 Jul 2008 16:10:40 +0200
Subject: [SciPy-user] OS X install success! + Question
In-Reply-To: <5449A4F1-E90B-430B-9223-72AFC7C9DF40@yale.edu>
References: <5BE66D24-E4A2-4076-BF49-B77B1A1459C2@diablotech.com>
	<5449A4F1-E90B-430B-9223-72AFC7C9DF40@yale.edu>
Message-ID: <20080730141040.GD21845@phare.normalesup.org>

On Wed, Jul 30, 2008 at 09:57:07AM -0400, Zachary Pincus wrote:
> With easy_install, this sort of transparency is a bit lost. You need  
> to look to see which version of python is invoked in the first line of  
> that script to determine where the files will eventually go. If you  
> have two versions of python, you need two versions of easy_install in  
> different places. I personally find this a bit frustrating...

Hum, are you sure you need two versions of easy_install? Doesn't using
the "-d" option, or the "--prefix" work? Alternatively, if you don't like
typing the options all the time, there is a configuration file to set it:
http://peak.telecommunity.com/DevCenter/EasyInstall#configuration-files

Ga?l


From jjh at 42quarks.com  Wed Jul 30 10:16:33 2008
From: jjh at 42quarks.com (Jonathan Hunt)
Date: Thu, 31 Jul 2008 00:16:33 +1000
Subject: [SciPy-user] Matlab IO -- can others with matlab test this bug?
In-Reply-To: <CD201218-2B0D-4978-971C-39AB6E84F0DD@yale.edu>
References: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>
	<e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>
	<67E4404D-CC00-4775-BE98-5C0BF81B53B5@yale.edu>
	<e26778a60807282237i67a0294bxbf586073ec26ef2b@mail.gmail.com>
	<8D21E970-E9A5-470C-A8B7-97960BB3C0FA@yale.edu>
	<488F0F0A.70105@ar.media.kyoto-u.ac.jp> <48900A16.6050401@astraw.com>
	<e26778a60807292342g74d01fa7ie7f4553b51c45201@mail.gmail.com>
	<CD201218-2B0D-4978-971C-39AB6E84F0DD@yale.edu>
Message-ID: <e26778a60807300716w50a2cb2fwaef456fa1f6cf0b0@mail.gmail.com>

Hi Zachary,

I just tested your test with yesterdays SVN SciPy and both .mat files
load with no problems in MATLAB 7.6.0.324 (R2008a) (with Python 2.5,
OS X 10.5.4). Perhaps Mathworks fixed the bug in there end if that's
where it is.

Hope that helps.
Jonny

On Thu, Jul 31, 2008 at 12:00 AM, Zachary Pincus
<zachary.pincus at yale.edu> wrote:
>> This is no doubt a dumb question. But why it is not possible simply to
>> import the library in place (i.e no install necessary, just use the
>> python files as they are laid out in the SVN co directory). Is it
>> because of the C/Fortran files that need to be compiled?
>
> Yes, basically. Pure-python libraries can usually be imported in-place
> -- though sometimes, too-clever setup.py files can actually make it so
> the installed folder structure is different from the one in the source
> directory, and thus an in-place import will fail. (I think this is
> probably considered bad form though.)
>
> In the specific, limited case of the matlab IO sub-sub-library, it is
> indeed pure python and can be imported in-place -- this was one of my
> hackish suggestions for testing it.
>
> Zach
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Jonathan J Hunt <jjh at 42quarks.com>
Homepage: http://www.42quarks.net.nz/wiki/JJH
(Further contact details there)
"Physics isn't the most important thing. Love is." Richard Feynman


From ferrell at diablotech.com  Wed Jul 30 10:17:08 2008
From: ferrell at diablotech.com (Robert Ferrell)
Date: Wed, 30 Jul 2008 08:17:08 -0600
Subject: [SciPy-user] OS X install success! + Question
In-Reply-To: <5449A4F1-E90B-430B-9223-72AFC7C9DF40@yale.edu>
References: <5BE66D24-E4A2-4076-BF49-B77B1A1459C2@diablotech.com>
	<5449A4F1-E90B-430B-9223-72AFC7C9DF40@yale.edu>
Message-ID: <3EAC1400-3C3A-4438-9DE3-2383C2F27D41@diablotech.com>

Thanks so much.  I installed a new version of easy_install, as  
suggested, that installed nose, and now everything works great.  All   
numpy tests and almost all scipy tests passed.  (Nearly) Flawless  
victory!

I really appreciate the description of what's going on with the OS X  
python directories.  You've explained it very well.

Thanks,

-robert

On Jul 30, 2008, at 7:57 AM, Zachary Pincus wrote:

> Hi Robert,
>
> I'm sure others with a little deeper knowledge of easy_install and
> friends may chime in, but this should get you started.
>
>>
>> Before installing NumPy & SciPy I installed the OS X Python from
>> python.org.  That's 2.5.2 (laptop shipped with 2.5.1) and was
>> installed in
>>
>> /Library/Frameworks/Python.framework/Versions/Current/bin/python
>>
>> which is just a link to
>>
>> /Library/Frameworks/Python.framework/Versions/2.5/bin/python.
>>
>> numpy and scipy installed in /Library/Frameworks/Python.framework/
>> Versions/2.5/lib/python2.5/site-packages/
>>
>> So far so good.  But, when I used easy_install for nose, that got
>> installed into
>>
>> /Library/Python/2.5/site-packages/.
>>
>> What is the best way to deal with the Frameworks vs no Frameworks
>> thing?  Should I just put a link from /Library/Python/2.5 to / 
>> Library/
>> Frameworks/Python.framework/Versions/2.5/lib/python2.5?
>
> Short answer: the version of easy_install on your machine is for the
> version of python that came with the operating system. So
> "easy_install nose" installed nose for that version -- not for the
> python.org python that you just installed.
>
> You will need to install easy_install for the new version of python
> that you just installed from python.org:
> http://peak.telecommunity.com/DevCenter/EasyInstall#installing-easy-install
>
> Then run that version of easy_install that gets installed by the
> ez_setup script. That version, for better or worse, lives by default  
> in:
> /Library/Frameworks/Python.framework/Versions/2.5/bin/easy_install
>
> (I would recommend making a symlink to /usr/local/bin for ease.
> Alternately, you could run the easy_setup.py script with the --script-
> dir=/usr/local/bin option.)
>
>
> Background information:
> Apple ships a version of Python with OS X, and some housekeeping
> scripts, etc., depend on this python and in some cases, the specific
> (old or broken) versions of some of the libraries it comes with (e.g.
> numpy). This is installed in:
> /System/Library/Frameworks/Python.framework/Versions/2.5
>
> This version of python is accessible as /usr/bin/python, which is a
> symlink to
> /System/Library/Frameworks/Python.framework/Versions/2.5/bin/python
>
> The basic consensus is that one really shouldn't be mucking with this
> python, so the best bet is to install the python.org python, which
> goes into /Library and /usr/local/bin.
>
> Now, what's the deal with /Library/Python/2.5/site-packages? Well,
> Apple's basic decree is that you must not alter things in /System/...
> directories. But installing a new library for the system python would
> do just that, under the default regime (which would put them in
> /System/Library/Frameworks/Python.framework/Versions/2.5/lib/ 
> python2.5/
> site-packages )
> So Apple's python instead puts new libraries in /Library/Python/2.5/
> site-packages .
>
> This is all fine, and controlled by the default values of the sys.path
> variable (which is the list of directories Python searches when you
> try to import something). It's illuminating to do:
> import sys; print sys.path
> in both the system and python.org versions of python.
>
> Now, how does one deal with multiple versions of python when
> installing things? In the pre-easy_install era, it was quite simple.
> Python things are installed with setup.py files in a directory:
>
> python setup.py install
>
> would install whatever library for the version of python that was used
> to run the script. So /usr/bin/python setup.py install would install
> the library for the system python, and /usr/local/bin/python setup.py
> install would do so for the python.org version. (Plain 'python' is
> the /usr/local/bin version, thanks to the python.org installer.
>
> With easy_install, this sort of transparency is a bit lost. You need
> to look to see which version of python is invoked in the first line of
> that script to determine where the files will eventually go. If you
> have two versions of python, you need two versions of easy_install in
> different places. I personally find this a bit frustrating...
>
> Zach
>
>
>
>
>
>
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user



From gael.varoquaux at normalesup.org  Wed Jul 30 10:21:10 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Wed, 30 Jul 2008 16:21:10 +0200
Subject: [SciPy-user] Matlab IO -- can others with matlab test this	bug?
In-Reply-To: <48900A16.6050401@astraw.com>
References: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>
	<e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>
	<67E4404D-CC00-4775-BE98-5C0BF81B53B5@yale.edu>
	<e26778a60807282237i67a0294bxbf586073ec26ef2b@mail.gmail.com>
	<8D21E970-E9A5-470C-A8B7-97960BB3C0FA@yale.edu>
	<488F0F0A.70105@ar.media.kyoto-u.ac.jp>
	<48900A16.6050401@astraw.com>
Message-ID: <20080730142110.GE21845@phare.normalesup.org>

On Tue, Jul 29, 2008 at 11:28:38PM -0700, Andrew Straw wrote:
> No time to do a similar writeup for it, but: virtualenv

> http://pypi.python.org/pypi/virtualenv

Use with caution: lots of gotcha with it (for instance the fact that
there are 90% chances that ipython or mayavi, or any program using python
other than the interpreter) will not start in your virtual environment if
not installed in your virtal env. I have shot myself in the foot many
times. This is due to how they are installed, and has been reported, and
is 'not a bug' :). I use virtualenv a lot, but do remember when you are
using it, that something is making some magic and that if things fail you
have to look closer.

Ga?l


From ferrell at diablotech.com  Wed Jul 30 10:24:09 2008
From: ferrell at diablotech.com (Robert Ferrell)
Date: Wed, 30 Jul 2008 08:24:09 -0600
Subject: [SciPy-user] OS X install success! + Question
In-Reply-To: <20080730141040.GD21845@phare.normalesup.org>
References: <5BE66D24-E4A2-4076-BF49-B77B1A1459C2@diablotech.com>
	<5449A4F1-E90B-430B-9223-72AFC7C9DF40@yale.edu>
	<20080730141040.GD21845@phare.normalesup.org>
Message-ID: <D7361000-7C72-4B59-BFD2-815F188BC41E@diablotech.com>


On Jul 30, 2008, at 8:10 AM, Gael Varoquaux wrote:

> On Wed, Jul 30, 2008 at 09:57:07AM -0400, Zachary Pincus wrote:
>> With easy_install, this sort of transparency is a bit lost. You need
>> to look to see which version of python is invoked in the first line  
>> of
>> that script to determine where the files will eventually go. If you
>> have two versions of python, you need two versions of easy_install in
>> different places. I personally find this a bit frustrating...
>
> Hum, are you sure you need two versions of easy_install? Doesn't using
> the "-d" option, or the "--prefix" work? Alternatively, if you don't  
> like
> typing the options all the time, there is a configuration file to  
> set it:
> http://peak.telecommunity.com/DevCenter/EasyInstall#configuration- 
> files

I tried the -d option.  The original (came with OS X) easy_install  
didn't like that.  Here's the complaint:

$> sudo easy_install -v -n -d/Library/Frameworks/Python.framework/ 
Versions/2.5/lib/python2.5/site-packages
TEST FAILED: /Library/Frameworks/Python.framework/Versions/2.5/lib/ 
python2.5/site-packages does NOT support .pth files
error: bad install directory or PYTHONPATH

You are attempting to install a package to a directory that is not
on PYTHONPATH and which Python does not read ".pth" files from.  The
installation directory you specified (via --install-dir, --prefix, or
the distutils default setting) was:

     /Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/ 
site-packages

and your PYTHONPATH environment variable currently contains:

     ''

That's what got me thinking there was some subtlety that I didn't  
appreciate.  Zachary's email explained it.  /Library/Python/2.5/site- 
packages is really for OS X's own python.  Now that I understand  
what's up with OS X I have greater confidence that I'll put future  
packages in the right place.

-robert


From shad0wfade at yahoo.com  Wed Jul 30 10:42:50 2008
From: shad0wfade at yahoo.com (Joshua)
Date: Wed, 30 Jul 2008 07:42:50 -0700 (PDT)
Subject: [SciPy-user] Is there a collection of useful functions/modules?
Message-ID: <267237.208.qm@web32905.mail.mud.yahoo.com>

I'm very new to Python, as in
only a week of programming, and was wondering if there is a page with a
collection of highly useful functions/modules that are not necessarily
maintained on a release-to-release basis for things like performing
FFT's on data in a file.



I wrote a function to do this using Numpy, and put in options to normalize the FFT
for me, and take the absolute value.? There are many other options I
should add, but before I add complexity, I wouldn't mind having my code
scrutinized, and made available for others to use and optimize.? That
way students and researchers don't have to reinvent the wheel, unless
they want to or are using some odd formatting scheme.? Just have the amplitude data in the list column wise.

exempli gratia:
0.0
0.1
0.0
-0.1
... et cetera.


This code makes no particular frequency sampling assumptions (other than the sampling was done correctly), and only deals with simple FFT processing.


I've attached the code and welcome scrutiny.



There are few things that I know I should to add/restructure to my code: better
error handling, printing out the phase, multi-dimensional analysis, and |Magnitude| in 20dB.



Thanks for your help
Joshua





      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080730/6ccbced3/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fftfromfile.py
Type: text/x-python
Size: 3099 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080730/6ccbced3/attachment.py>

From zachary.pincus at yale.edu  Wed Jul 30 10:44:01 2008
From: zachary.pincus at yale.edu (Zachary Pincus)
Date: Wed, 30 Jul 2008 10:44:01 -0400
Subject: [SciPy-user] OS X install success! + Question
In-Reply-To: <20080730141040.GD21845@phare.normalesup.org>
References: <5BE66D24-E4A2-4076-BF49-B77B1A1459C2@diablotech.com>
	<5449A4F1-E90B-430B-9223-72AFC7C9DF40@yale.edu>
	<20080730141040.GD21845@phare.normalesup.org>
Message-ID: <2D851907-2E03-4843-9CB1-B02A05F3FA18@yale.edu>

> On Wed, Jul 30, 2008 at 09:57:07AM -0400, Zachary Pincus wrote:
>> With easy_install, this sort of transparency is a bit lost. You need
>> to look to see which version of python is invoked in the first line  
>> of
>> that script to determine where the files will eventually go. If you
>> have two versions of python, you need two versions of easy_install in
>> different places. I personally find this a bit frustrating...
>
> Hum, are you sure you need two versions of easy_install? Doesn't using
> the "-d" option, or the "--prefix" work? Alternatively, if you don't  
> like
> typing the options all the time, there is a configuration file to  
> set it:
> http://peak.telecommunity.com/DevCenter/EasyInstall#configuration- 
> files

To the best of my limited knowledge, that won't work too well, since  
you're trying to use version A of Python to install a package for  
version B, but the installation process often (but not always?)  
gathers information from the currently-running version of Python --  
version A -- to make decisions about the installation. (Information  
which may not be correct for version B.)

Zach


From zachary.pincus at yale.edu  Wed Jul 30 10:45:46 2008
From: zachary.pincus at yale.edu (Zachary Pincus)
Date: Wed, 30 Jul 2008 10:45:46 -0400
Subject: [SciPy-user] Matlab IO -- can others with matlab test this bug?
In-Reply-To: <e26778a60807300716w50a2cb2fwaef456fa1f6cf0b0@mail.gmail.com>
References: <9595C26D-C67A-404F-AB42-9D7EDB8860F8@yale.edu>
	<e26778a60807282043m3cffe9f7m27093e1b2e4a2788@mail.gmail.com>
	<67E4404D-CC00-4775-BE98-5C0BF81B53B5@yale.edu>
	<e26778a60807282237i67a0294bxbf586073ec26ef2b@mail.gmail.com>
	<8D21E970-E9A5-470C-A8B7-97960BB3C0FA@yale.edu>
	<488F0F0A.70105@ar.media.kyoto-u.ac.jp>
	<48900A16.6050401@astraw.com>
	<e26778a60807292342g74d01fa7ie7f4553b51c45201@mail.gmail.com>
	<CD201218-2B0D-4978-971C-39AB6E84F0DD@yale.edu>
	<e26778a60807300716w50a2cb2fwaef456fa1f6cf0b0@mail.gmail.com>
Message-ID: <E988EBBE-C9D8-4CBB-9F4A-646D34FDD6EF@yale.edu>

Thanks! It looks like it's a limited matlab bug...

I'll still be making a patch to fix this, but that will definitely  
need some testing to make sure I didn't inadvertently break other  
stuff. I'll post more information after I have a patch.

Zach

On Jul 30, 2008, at 10:16 AM, Jonathan Hunt wrote:

> Hi Zachary,
>
> I just tested your test with yesterdays SVN SciPy and both .mat files
> load with no problems in MATLAB 7.6.0.324 (R2008a) (with Python 2.5,
> OS X 10.5.4). Perhaps Mathworks fixed the bug in there end if that's
> where it is.
>
> Hope that helps.
> Jonny
>
> On Thu, Jul 31, 2008 at 12:00 AM, Zachary Pincus
> <zachary.pincus at yale.edu> wrote:
>>> This is no doubt a dumb question. But why it is not possible  
>>> simply to
>>> import the library in place (i.e no install necessary, just use the
>>> python files as they are laid out in the SVN co directory). Is it
>>> because of the C/Fortran files that need to be compiled?
>>
>> Yes, basically. Pure-python libraries can usually be imported in- 
>> place
>> -- though sometimes, too-clever setup.py files can actually make it  
>> so
>> the installed folder structure is different from the one in the  
>> source
>> directory, and thus an in-place import will fail. (I think this is
>> probably considered bad form though.)
>>
>> In the specific, limited case of the matlab IO sub-sub-library, it is
>> indeed pure python and can be imported in-place -- this was one of my
>> hackish suggestions for testing it.
>>
>> Zach
>> _______________________________________________
>> SciPy-user mailing list
>> SciPy-user at scipy.org
>> http://projects.scipy.org/mailman/listinfo/scipy-user
>>
>
>
>
> -- 
> Jonathan J Hunt <jjh at 42quarks.com>
> Homepage: http://www.42quarks.net.nz/wiki/JJH
> (Further contact details there)
> "Physics isn't the most important thing. Love is." Richard Feynman
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user



From pkienzle at nist.gov  Wed Jul 30 10:50:04 2008
From: pkienzle at nist.gov (Paul Kienzle)
Date: Wed, 30 Jul 2008 10:50:04 -0400
Subject: [SciPy-user] OS X install success! + Question
In-Reply-To: <20080730141040.GD21845@phare.normalesup.org>;
	from gael.varoquaux@normalesup.org on Wed, Jul 30, 2008 at
	04:10:40PM +0200
References: <5BE66D24-E4A2-4076-BF49-B77B1A1459C2@diablotech.com>
	<5449A4F1-E90B-430B-9223-72AFC7C9DF40@yale.edu>
	<20080730141040.GD21845@phare.normalesup.org>
Message-ID: <20080730105004.C780476@jazz.ncnr.nist.gov>

On Wed, Jul 30, 2008 at 04:10:40PM +0200, Gael Varoquaux wrote:
> On Wed, Jul 30, 2008 at 09:57:07AM -0400, Zachary Pincus wrote:
> > With easy_install, this sort of transparency is a bit lost. You need  
> > to look to see which version of python is invoked in the first line of  
> > that script to determine where the files will eventually go. If you  
> > have two versions of python, you need two versions of easy_install in  
> > different places. I personally find this a bit frustrating...

For python 2.4 and up you can use:

    $ python -m easy_install --help

This uses the easy_install for the particular python interpreter used to
invoke it.

	- Paul


From gael.varoquaux at normalesup.org  Wed Jul 30 10:58:17 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Wed, 30 Jul 2008 16:58:17 +0200
Subject: [SciPy-user] OS X install success! + Question
In-Reply-To: <2D851907-2E03-4843-9CB1-B02A05F3FA18@yale.edu>
References: <5BE66D24-E4A2-4076-BF49-B77B1A1459C2@diablotech.com>
	<5449A4F1-E90B-430B-9223-72AFC7C9DF40@yale.edu>
	<20080730141040.GD21845@phare.normalesup.org>
	<2D851907-2E03-4843-9CB1-B02A05F3FA18@yale.edu>
Message-ID: <20080730145817.GF21845@phare.normalesup.org>

On Wed, Jul 30, 2008 at 10:44:01AM -0400, Zachary Pincus wrote:
> To the best of my limited knowledge, that won't work too well, since  
> you're trying to use version A of Python to install a package for  
> version B, but the installation process often (but not always?)  
> gathers information from the currently-running version of Python --  
> version A -- to make decisions about the installation. (Information  
> which may not be correct for version B.)

It seems you are right. Sorry for the noise.

Ga?l


From m.starnes05 at imperial.ac.uk  Wed Jul 30 11:11:30 2008
From: m.starnes05 at imperial.ac.uk (mark starnes)
Date: Wed, 30 Jul 2008 16:11:30 +0100
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
Message-ID: <489084A2.3000408@imperial.ac.uk>

Hi everyone,

I've looked through the list here and in Numpy-users, and checked the
'net but can't find an answer to this problem (with luck, I've missed
something obvious!).

I've an array of velocities at 80,000 points, irregularly spaced (from a
CFD analysis).  I'd like to generate the interpolated velocity at any
position in the domain, to map the data to an acoustics analysis on a
different mesh.

I tried a least squares approach but the errors are too large using
polynomials and trigonometric functions.  My conclusion is that I need a
nearest-neighbour type interpolation routine.

Is there such a routine in Scipy or Numpy?  From inspection of Matlab's
functions 'interp3' may be similar to what I would like, but I can't test.

If there's nothing available, I'll end up converting one of the
submitted scripts at the url,

http://www.mathworks.com/matlabcentral/fileexchange/loadCategory.do?objectId=14&objectType=Category


Thanks in advance,

Mark.



From josemaria.alkala at gmail.com  Wed Jul 30 11:18:02 2008
From: josemaria.alkala at gmail.com (=?ISO-8859-1?Q?Jos=E9_Mar=EDa_Garc=EDa_P=E9rez?=)
Date: Wed, 30 Jul 2008 17:18:02 +0200
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
In-Reply-To: <489084A2.3000408@imperial.ac.uk>
References: <489084A2.3000408@imperial.ac.uk>
Message-ID: <a1d6dd760807300818n199b825fk4c642c1202ede439@mail.gmail.com>

Hi Mark,
Try Radial Basis Functions. But I'm not sure if it's implemented at
SciPy/Numpy. I implemented myself long time ago, but better if there is
something already done within the project.
Regards,
Jos? M.

2008/7/30 mark starnes <m.starnes05 at imperial.ac.uk>

> Hi everyone,
>
> I've looked through the list here and in Numpy-users, and checked the
> 'net but can't find an answer to this problem (with luck, I've missed
> something obvious!).
>
> I've an array of velocities at 80,000 points, irregularly spaced (from a
> CFD analysis).  I'd like to generate the interpolated velocity at any
> position in the domain, to map the data to an acoustics analysis on a
> different mesh.
>
> I tried a least squares approach but the errors are too large using
> polynomials and trigonometric functions.  My conclusion is that I need a
> nearest-neighbour type interpolation routine.
>
> Is there such a routine in Scipy or Numpy?  From inspection of Matlab's
> functions 'interp3' may be similar to what I would like, but I can't test.
>
> If there's nothing available, I'll end up converting one of the
> submitted scripts at the url,
>
>
> http://www.mathworks.com/matlabcentral/fileexchange/loadCategory.do?objectId=14&objectType=Category
>
>
> Thanks in advance,
>
> Mark.
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080730/2c9d9346/attachment.html>

From rob.clewley at gmail.com  Wed Jul 30 11:28:50 2008
From: rob.clewley at gmail.com (Rob Clewley)
Date: Wed, 30 Jul 2008 11:28:50 -0400
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <web-105378668@uni-stuttgart.de>
References: <web-105374238@uni-stuttgart.de> <200807301243.12871.jr@sun.ac.za>
	<web-105378668@uni-stuttgart.de>
Message-ID: <a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>

Hi,

PyDSTool wraps Hairer and Wanner's Radau 5 code using SWIG. This
supports DAEs. An example is in PyDSTool/tests/DAE_example.py, and you
get the advantage of the higher level Pointset and Trajectory
structures for your solutions, with named coordinates (rather than
plain arrays with integer indices), if that appeals to you.

-Rob


From nwagner at iam.uni-stuttgart.de  Wed Jul 30 11:46:21 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Wed, 30 Jul 2008 17:46:21 +0200
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
References: <web-105374238@uni-stuttgart.de> <200807301243.12871.jr@sun.ac.za>
	<web-105378668@uni-stuttgart.de>
	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
Message-ID: <web-105392498@uni-stuttgart.de>

On Wed, 30 Jul 2008 11:28:50 -0400
  "Rob Clewley" <rob.clewley at gmail.com> wrote:
> Hi,
> 
> PyDSTool wraps Hairer and Wanner's Radau 5 code using 
>SWIG. This
> supports DAEs. An example is in 
>PyDSTool/tests/DAE_example.py, and you
> get the advantage of the higher level Pointset and 
>Trajectory
> structures for your solutions, with named coordinates 
>(rather than
> plain arrays with integer indices), if that appeals to 
>you.
> 
> -Rob
> 
Hi Rob,

Thank you for your reply.
How do I install PyDSTool ?
As opposed to other projects like numpy/scipy/matplotlib I 
can't find a setup.py.

/usr/bin/python -i src/PyDSTool/tests/DAE_example.py
Traceback (most recent call last):
   File "src/PyDSTool/tests/DAE_example.py", line 12, in ?
     from PyDSTool import *
ImportError: No module named PyDSTool

Nils


From zachary.pincus at yale.edu  Wed Jul 30 11:55:09 2008
From: zachary.pincus at yale.edu (Zachary Pincus)
Date: Wed, 30 Jul 2008 11:55:09 -0400
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
In-Reply-To: <a1d6dd760807300818n199b825fk4c642c1202ede439@mail.gmail.com>
References: <489084A2.3000408@imperial.ac.uk>
	<a1d6dd760807300818n199b825fk4c642c1202ede439@mail.gmail.com>
Message-ID: <45918DC2-F3E5-4F7E-A031-539A659491B1@yale.edu>

Mark:
>  My conclusion is that I need a
> nearest-neighbour type interpolation routine.

Jos?:
> Try Radial Basis Functions.

This was my thought as well.

Scipy has a handy implementation:
http://www.scipy.org/Cookbook/RadialBasisFunctions

Zach

From rob.clewley at gmail.com  Wed Jul 30 11:59:25 2008
From: rob.clewley at gmail.com (Rob Clewley)
Date: Wed, 30 Jul 2008 11:59:25 -0400
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <web-105392498@uni-stuttgart.de>
References: <web-105374238@uni-stuttgart.de> <200807301243.12871.jr@sun.ac.za>
	<web-105378668@uni-stuttgart.de>
	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
	<web-105392498@uni-stuttgart.de>
Message-ID: <a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>

Erm... yes, there isn't one! I know it's a bit clunky but I have been
waiting to add a setup.py until the next major release because I
wanted to get some details right about pre-compiling the C and Fortran
extensions. Those issues have been ironed out so it shouldn't be long
now...

Anyway, it's still very easy. The instructions at
http://www.cam.cornell.edu/~rclewley/cgi-bin/moin.cgi/GettingStarted
will fill you in but you basically just put the unzipped directory
wherever you like and add it to your pythonpath. There is currently no
pre-compiling of the extensions, it will all be done when you first
invoke the compiler to build your Radau model. Let me know if you have
any other questions or if there's something I should add to the
documentation to make things more clear.

Cheers
Rob

On Wed, Jul 30, 2008 at 11:46 AM, Nils Wagner
<nwagner at iam.uni-stuttgart.de> wrote:
> On Wed, 30 Jul 2008 11:28:50 -0400
>  "Rob Clewley" <rob.clewley at gmail.com> wrote:
>> Hi,
>>
>> PyDSTool wraps Hairer and Wanner's Radau 5 code using
>>SWIG. This
>> supports DAEs. An example is in
>>PyDSTool/tests/DAE_example.py, and you
>> get the advantage of the higher level Pointset and
>>Trajectory
>> structures for your solutions, with named coordinates
>>(rather than
>> plain arrays with integer indices), if that appeals to
>>you.
>>
>> -Rob
>>
> Hi Rob,
>
> Thank you for your reply.
> How do I install PyDSTool ?
> As opposed to other projects like numpy/scipy/matplotlib I
> can't find a setup.py.
>
> /usr/bin/python -i src/PyDSTool/tests/DAE_example.py
> Traceback (most recent call last):
>   File "src/PyDSTool/tests/DAE_example.py", line 12, in ?
>     from PyDSTool import *
> ImportError: No module named PyDSTool
>
> Nils
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Robert H. Clewley, Ph. D.
Assistant Professor
Department of Mathematics and Statistics
Georgia State University
720 COE, 30 Pryor St
Atlanta, GA 30303, USA

tel: 404-413-6420 fax: 404-413-6403
http://www2.gsu.edu/~matrhc
http://brainsbehavior.gsu.edu/


From nwagner at iam.uni-stuttgart.de  Wed Jul 30 12:11:44 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Wed, 30 Jul 2008 18:11:44 +0200
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>
References: <web-105374238@uni-stuttgart.de> <200807301243.12871.jr@sun.ac.za>
	<web-105378668@uni-stuttgart.de>
	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
	<web-105392498@uni-stuttgart.de>
	<a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>
Message-ID: <web-105393352@uni-stuttgart.de>

On Wed, 30 Jul 2008 11:59:25 -0400
  "Rob Clewley" <rob.clewley at gmail.com> wrote:
> Erm... yes, there isn't one! I know it's a bit clunky 
>but I have been
> waiting to add a setup.py until the next major release 
>because I
> wanted to get some details right about pre-compiling the 
>C and Fortran
> extensions. Those issues have been ironed out so it 
>shouldn't be long
> now...
> 
> Anyway, it's still very easy. The instructions at
> http://www.cam.cornell.edu/~rclewley/cgi-bin/moin.cgi/GettingStarted
> will fill you in but you basically just put the unzipped 
>directory
> wherever you like and add it to your pythonpath. There 
>is currently no
> pre-compiling of the extensions, it will all be done 
>when you first
> invoke the compiler to build your Radau model. Let me 
>know if you have
> any other questions or if there's something I should add 
>to the
> documentation to make things more clear.
> 
> Cheers
> Rob
  
Hi Rob,

I think I followed all these instructions.

  echo $PYTHONPATH
/home/nwagner/:/home/nwagner/src/PyDSTool/:/home/nwagner/src/PyDSTool/tests/:/usr/lib/python2.4/site-packages/
nwagner at linux:~> /usr/bin/python -i 
src/PyDSTool/tests/DAE_example.py
Traceback (most recent call last):
   File "src/PyDSTool/tests/DAE_example.py", line 12, in ?
     from PyDSTool import *
ImportError: No module named PyDSTool

Am I missing something ?

Cheers,
           Nils



From rob.clewley at gmail.com  Wed Jul 30 12:17:17 2008
From: rob.clewley at gmail.com (Rob Clewley)
Date: Wed, 30 Jul 2008 12:17:17 -0400
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <web-105393352@uni-stuttgart.de>
References: <web-105374238@uni-stuttgart.de> <200807301243.12871.jr@sun.ac.za>
	<web-105378668@uni-stuttgart.de>
	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
	<web-105392498@uni-stuttgart.de>
	<a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>
	<web-105393352@uni-stuttgart.de>
Message-ID: <a749952d0807300917v1ae4a90dy379b5d7f22413ac9@mail.gmail.com>

Did you also add a PyDSTool.pth into python's site-packages/ directory?

It would contain the same PyDSTool-specific lines that are in your
$PYTHONPATH, one per line, and not separated by colons.


From nwagner at iam.uni-stuttgart.de  Wed Jul 30 12:25:46 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Wed, 30 Jul 2008 18:25:46 +0200
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <a749952d0807300917v1ae4a90dy379b5d7f22413ac9@mail.gmail.com>
References: <web-105374238@uni-stuttgart.de> <200807301243.12871.jr@sun.ac.za>
	<web-105378668@uni-stuttgart.de>
	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
	<web-105392498@uni-stuttgart.de>
	<a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>
	<web-105393352@uni-stuttgart.de>
	<a749952d0807300917v1ae4a90dy379b5d7f22413ac9@mail.gmail.com>
Message-ID: <web-105393808@uni-stuttgart.de>

On Wed, 30 Jul 2008 12:17:17 -0400
  "Rob Clewley" <rob.clewley at gmail.com> wrote:
> Did you also add a PyDSTool.pth into python's 
>site-packages/ directory?

The file
  
/usr/lib/python2.4/site-packages/PyDSTool.pth

consists of two lines

/home/nwagner/src/PyDSTool/
/home/nwagner/src/PyDSTool/tests/

Nils


From rob.clewley at gmail.com  Wed Jul 30 12:36:05 2008
From: rob.clewley at gmail.com (Rob Clewley)
Date: Wed, 30 Jul 2008 12:36:05 -0400
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <web-105393808@uni-stuttgart.de>
References: <web-105374238@uni-stuttgart.de> <200807301243.12871.jr@sun.ac.za>
	<web-105378668@uni-stuttgart.de>
	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
	<web-105392498@uni-stuttgart.de>
	<a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>
	<web-105393352@uni-stuttgart.de>
	<a749952d0807300917v1ae4a90dy379b5d7f22413ac9@mail.gmail.com>
	<web-105393808@uni-stuttgart.de>
Message-ID: <a749952d0807300936q7b3eeb2cvee2638b1a5c4012e@mail.gmail.com>

As I say in the wiki in the Windows install section (I must add it to
the *nix section!) of GettingStarted, my experience is that you'll
need to add

/home
/home/wagner
/home/wagner/src

before these other lines. I don't understand why this is necessary
from a technical perspective. It seems daft to me.

On Wed, Jul 30, 2008 at 12:25 PM, Nils Wagner
<nwagner at iam.uni-stuttgart.de> wrote:
> On Wed, 30 Jul 2008 12:17:17 -0400
>  "Rob Clewley" <rob.clewley at gmail.com> wrote:
>> Did you also add a PyDSTool.pth into python's
>>site-packages/ directory?
>
> The file
>
> /usr/lib/python2.4/site-packages/PyDSTool.pth
>
> consists of two lines
>
> /home/nwagner/src/PyDSTool/
> /home/nwagner/src/PyDSTool/tests/
>
> Nils
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Robert H. Clewley, Ph. D.
Assistant Professor
Department of Mathematics and Statistics
Georgia State University
720 COE, 30 Pryor St
Atlanta, GA 30303, USA

tel: 404-413-6420 fax: 404-413-6403
http://www2.gsu.edu/~matrhc
http://brainsbehavior.gsu.edu/


From nwagner at iam.uni-stuttgart.de  Wed Jul 30 12:47:41 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Wed, 30 Jul 2008 18:47:41 +0200
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <a749952d0807300936q7b3eeb2cvee2638b1a5c4012e@mail.gmail.com>
References: <web-105374238@uni-stuttgart.de> <200807301243.12871.jr@sun.ac.za>
	<web-105378668@uni-stuttgart.de>
	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
	<web-105392498@uni-stuttgart.de>
	<a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>
	<web-105393352@uni-stuttgart.de>
	<a749952d0807300917v1ae4a90dy379b5d7f22413ac9@mail.gmail.com>
	<web-105393808@uni-stuttgart.de>
	<a749952d0807300936q7b3eeb2cvee2638b1a5c4012e@mail.gmail.com>
Message-ID: <web-105394546@uni-stuttgart.de>

On Wed, 30 Jul 2008 12:36:05 -0400
  "Rob Clewley" <rob.clewley at gmail.com> wrote:
> As I say in the wiki in the Windows install section (I 
>must add it to
> the *nix section!) of GettingStarted, my experience is 
>that you'll
> need to add
> 
> /home
> /home/wagner
> /home/wagner/src
> 
> before these other lines. I don't understand why this is 
>necessary
> from a technical perspective. It seems daft to me.
  
Strange, Now it works fine for me, but it's very slow...

Thank you again

                    Nils


From rob.clewley at gmail.com  Wed Jul 30 13:03:24 2008
From: rob.clewley at gmail.com (Rob Clewley)
Date: Wed, 30 Jul 2008 13:03:24 -0400
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <web-105394546@uni-stuttgart.de>
References: <web-105374238@uni-stuttgart.de> <web-105378668@uni-stuttgart.de>
	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
	<web-105392498@uni-stuttgart.de>
	<a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>
	<web-105393352@uni-stuttgart.de>
	<a749952d0807300917v1ae4a90dy379b5d7f22413ac9@mail.gmail.com>
	<web-105393808@uni-stuttgart.de>
	<a749952d0807300936q7b3eeb2cvee2638b1a5c4012e@mail.gmail.com>
	<web-105394546@uni-stuttgart.de>
Message-ID: <a749952d0807301003h40838f36l1d5a343f960647c6@mail.gmail.com>

>
> Strange, Now it works fine for me, but it's very slow...

I assume you mean the time to be able to run the radau test script? If
so, that's because the current version of the systems needs to
recompile everything afresh each time, even though 90% of the code is
the same. That's a distutils issue, but the version I'll be releasing
in the next few months side-steps that problem. Then, only your vector
field will need to be re-compiled if you change it, and the core radau
and dopri parts will be compiled once only at the original
installation time. Hence the complicated issues getting setup.py
working.

I think you'll find that once the model is compiled, it integrates
solutions *very* quickly. Radau is very efficient, and there's only a
small overhead in simulating models with it using the additional
structures from PyDSTool. If this isn't what you find then there's
something else wrong!

-Rob


From nwagner at iam.uni-stuttgart.de  Wed Jul 30 13:12:09 2008
From: nwagner at iam.uni-stuttgart.de (Nils Wagner)
Date: Wed, 30 Jul 2008 19:12:09 +0200
Subject: [SciPy-user] Differential Algebraic Equation Solvers
In-Reply-To: <a749952d0807301003h40838f36l1d5a343f960647c6@mail.gmail.com>
References: <web-105374238@uni-stuttgart.de> <web-105378668@uni-stuttgart.de>
	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
	<web-105392498@uni-stuttgart.de>
	<a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>
	<web-105393352@uni-stuttgart.de>
	<a749952d0807300917v1ae4a90dy379b5d7f22413ac9@mail.gmail.com>
	<web-105393808@uni-stuttgart.de>
	<a749952d0807300936q7b3eeb2cvee2638b1a5c4012e@mail.gmail.com>
	<web-105394546@uni-stuttgart.de>
	<a749952d0807301003h40838f36l1d5a343f960647c6@mail.gmail.com>
Message-ID: <web-105395109@uni-stuttgart.de>

On Wed, 30 Jul 2008 13:03:24 -0400
  "Rob Clewley" <rob.clewley at gmail.com> wrote:
>>
>> Strange, Now it works fine for me, but it's very slow...
> 
> I assume you mean the time to be able to run the radau 
>test script? 

Exactly,

>If
> so, that's because the current version of the systems 
>needs to
> recompile everything afresh each time, even though 90% 
>of the code is
> the same. That's a distutils issue, but the version I'll 
>be releasing
> in the next few months side-steps that problem. 

Will you post an announcement to this list ?
I am looking forward to the enhanced version version !

Is PyDSTool available via svn ?

>Then, 
>only your vector
> field will need to be re-compiled if you change it, and 
>the core radau
> and dopri parts will be compiled once only at the 
>original
> installation time. Hence the complicated issues getting 
>setup.py
> working.
> 
> I think you'll find that once the model is compiled, it 
>integrates
> solutions *very* quickly. Radau is very efficient, and 
>there's only a
> small overhead in simulating models with it using the 
>additional
> structures from PyDSTool. If this isn't what you find 
>then there's
> something else wrong!
> 
> -Rob

Nils
  


From bnuttall at uky.edu  Wed Jul 30 13:41:15 2008
From: bnuttall at uky.edu (Nuttall, Brandon C)
Date: Wed, 30 Jul 2008 13:41:15 -0400
Subject: [SciPy-user] Robust Statistics
In-Reply-To: <web-105394546@uni-stuttgart.de>
References: <web-105374238@uni-stuttgart.de>
	<200807301243.12871.jr@sun.ac.za>	<web-105378668@uni-stuttgart.de>
	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
	<web-105392498@uni-stuttgart.de>
	<a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>
	<web-105393352@uni-stuttgart.de>
	<a749952d0807300917v1ae4a90dy379b5d7f22413ac9@mail.gmail.com>
	<web-105393808@uni-stuttgart.de>
	<a749952d0807300936q7b3eeb2cvee2638b1a5c4012e@mail.gmail.com>
	<web-105394546@uni-stuttgart.de>
Message-ID: <F5BC85B8534D524DB972D2B2731CA53F29CAEFDFD4@EX7FM01.ad.uky.edu>

Folks,

I have some rate/time data I'm in the process of analyzing. The data represent the natural gas produced from a well over time. The data are fit best by either an exponential model (least squares best fit of log of rate data, just qi and di see below) or a "hyperbolic" model:

qt = qi *(1-b*di*t)**(-1.0/b)

        qt = rate at time t
        qi = initial rate (calculated)
        b = decline exponent (rate of change of the decline rate, calculated)
        di = initial decline rate (calculated)
        t = time

The calculated parameters are found using the optimization routine scipy.optimize.fmin_tnc(). I'm already rejecting some best fit results because either the correlation coefficient is not statistically significant or the di (slope) is not significantly different from 0.

The image I've attached shows two things. An exponential best fit is shown by the solid line and a hyperbolic best fit by the dotted line. It is common practice in analyzing this type of data to begin the best fit analysis at an early time point where the data actually start to "decline"; thus, the hyperbolic best fit started at time 2 and the first rate point (rate = about 1350) is ignored. The other thing the graph shows is that there appear to be outliers.

My question is, what Python resources are there for testing for outliers and robust statistics?

Is RANSAC appropriate for this? Note that when I run ransack.py from the cookbook, I get:

Traceback (most recent call last):

--Snip other traceback info---
  File "C:\Documents and Settings\bnuttall\Desktop\ransac.py", line 132, in fit
    A = numpy.vstack([data[:,i] for i in self.input_columns]).T
AttributeError: 'numpy.ndarray' object has no attribute 'T'

I presume the T attribute is supposed to mean the transpose operator and have changed the offending statement(s, there are 4 total) to the form:

A = numpy.vstack([data[:,i] for i in self.input_columns]).transpose()

Thanks.

Brandon Nuttall
-------------- next part --------------
A non-text attachment was scrubbed...
Name: R0114442_raw.png
Type: image/png
Size: 55799 bytes
Desc: R0114442_raw.png
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080730/d680f206/attachment.png>

From stefan at sun.ac.za  Wed Jul 30 14:17:47 2008
From: stefan at sun.ac.za (=?ISO-8859-1?Q?St=E9fan_van_der_Walt?=)
Date: Wed, 30 Jul 2008 20:17:47 +0200
Subject: [SciPy-user] Robust Statistics
In-Reply-To: <F5BC85B8534D524DB972D2B2731CA53F29CAEFDFD4@EX7FM01.ad.uky.edu>
References: <web-105374238@uni-stuttgart.de>
	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>
	<web-105392498@uni-stuttgart.de>
	<a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>
	<web-105393352@uni-stuttgart.de>
	<a749952d0807300917v1ae4a90dy379b5d7f22413ac9@mail.gmail.com>
	<web-105393808@uni-stuttgart.de>
	<a749952d0807300936q7b3eeb2cvee2638b1a5c4012e@mail.gmail.com>
	<web-105394546@uni-stuttgart.de>
	<F5BC85B8534D524DB972D2B2731CA53F29CAEFDFD4@EX7FM01.ad.uky.edu>
Message-ID: <9457e7c80807301117x7070364bx925a52afb01c1e77@mail.gmail.com>

2008/7/30 Nuttall, Brandon C <bnuttall at uky.edu>:
> I presume the T attribute is supposed to mean the transpose operator and have changed the > offending statement(s, there are 4 total) to the form:

Which version of NumPy do you have?  This was changed very long ago.

Cheers
St?fan


From gael.varoquaux at normalesup.org  Wed Jul 30 14:36:52 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Wed, 30 Jul 2008 20:36:52 +0200
Subject: [SciPy-user] Running SciPy SVN without installing Was:
	Matlab	IO -- can others with matlab test this bug?
In-Reply-To: <e26778a60807290715s76b479c7if30f46dca8fe2477@mail.gmail.com>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
	<488F2353.6040106@relativita.com>
	<488F2181.3040307@ar.media.kyoto-u.ac.jp>
	<e26778a60807290715s76b479c7if30f46dca8fe2477@mail.gmail.com>
Message-ID: <20080730183652.GJ32733@phare.normalesup.org>

On Wed, Jul 30, 2008 at 12:15:38AM +1000, Jonathan Hunt wrote:
> However. Could someone tell me how most SciPy developers work on the
> library. 

I don't work on scipy, but on other project I use a combination of
virtualenv and setuptools develop feature, giving it the right prefix. It
adds some overheard indeed.

I rerun python setup.py develop when there are any changes to the C code.

I do this because I need to be able to run different branches of the of
the same project in the same time to check for differences.

Prabhu has explained the way he uses virtualenv here:
http://prabhuramachandran.blogspot.com/2008/03/using-virtualenv-under-linux.html

I do things a bit different. One of the big difference is the instead of
python setup.py install, I use python setup.py develop.

All these approach tend to have there own quirks so you have to get to
know them.

Ga?l


From gael.varoquaux at normalesup.org  Wed Jul 30 14:38:37 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Wed, 30 Jul 2008 20:38:37 +0200
Subject: [SciPy-user] Running SciPy SVN without installing Was:
	Matlab	IO -- can others with matlab test this bug?
In-Reply-To: <488F2479.8020706@ar.media.kyoto-u.ac.jp>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
	<200807291015.24171.pgmdevlist@gmail.com>
	<488F2479.8020706@ar.media.kyoto-u.ac.jp>
Message-ID: <20080730183837.GK32733@phare.normalesup.org>

On Tue, Jul 29, 2008 at 11:08:57PM +0900, David Cournapeau wrote:
> Really, stow is nice. The
> only problem is that it does not work on windows.

Can I have two shells, one running one environment, and the other
another, say to compare results with stow?

Ga?l


From gael.varoquaux at normalesup.org  Wed Jul 30 14:41:02 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Wed, 30 Jul 2008 20:41:02 +0200
Subject: [SciPy-user] Running SciPy SVN without installing Was:
	Matlab	IO -- can others with matlab test this bug?
In-Reply-To: <e26778a60807290732s5564ea84pac3d4737781a2c3e@mail.gmail.com>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
	<200807291015.24171.pgmdevlist@gmail.com>
	<488F2479.8020706@ar.media.kyoto-u.ac.jp>
	<e26778a60807290732s5564ea84pac3d4737781a2c3e@mail.gmail.com>
Message-ID: <20080730184102.GL32733@phare.normalesup.org>

On Wed, Jul 30, 2008 at 12:32:08AM +1000, Jonathan Hunt wrote:
> Thanks for all the comments. If anyone uses the develop mode of
> setuptools or other methods I'd be interested to hear.

I do. I am not sure what to tell you about it. Just be aware that when
you use setuptools, your import path gets messed up (print sys.path to
reaize that), and as a result import priorities get scrambled. If you
know that and keep watch on what is important, when you have several
version of different packages installed, you can usual understand what is
going on.

Ga?l


From jturner at gemini.edu  Wed Jul 30 14:46:25 2008
From: jturner at gemini.edu (James Turner)
Date: Wed, 30 Jul 2008 14:46:25 -0400
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
In-Reply-To: <489084A2.3000408@imperial.ac.uk>
References: <489084A2.3000408@imperial.ac.uk>
Message-ID: <4890B701.40402@gemini.edu>

Hi Mark,

> I've an array of velocities at 80,000 points, irregularly spaced (from a
> CFD analysis).  I'd like to generate the interpolated velocity at any
> position in the domain, to map the data to an acoustics analysis on a
> different mesh.
> 
> I tried a least squares approach but the errors are too large using
> polynomials and trigonometric functions.  My conclusion is that I need a
> nearest-neighbour type interpolation routine.

If you don't have any joy with the other suggestions, there are some
useful routines by Renka in Netlib, at least one of which is 3D,
though I believe they are also piecewise polynomials (and in ForTran,
of course, so need wrapping).

   http://www.netlib.org/toms/

I also have a program implementing the Hubble Telescope's "Drizzle"
algorithm in 3D, which is similar to nearest neighbour or linear,
but unfortunately it's written in an obscure language that would
make it difficult to use from Python unless you are prepared to
install STScI's PyRAF. I also suspect it will only work well for
data that are mildly irregular and it may be fiddly to set up for
your application, but anyway, just so you have the option...
The paper is at http://xxx.lanl.gov/abs/astro-ph/9708242.

Are the points really irregular in 3D, or just in 1-2 of the
dimensions? If any dimension is regularly spaced, you might be able
to simplify the problem by resampling in 1-2D at a time?

Stefan might also have suggestions ;-). But maybe the radial basis
functions will do the trick... I'll have to look at those.

Cheers,

James.



From m.starnes05 at imperial.ac.uk  Wed Jul 30 16:15:32 2008
From: m.starnes05 at imperial.ac.uk (mark starnes)
Date: Wed, 30 Jul 2008 21:15:32 +0100
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
In-Reply-To: <45918DC2-F3E5-4F7E-A031-539A659491B1@yale.edu>
References: <489084A2.3000408@imperial.ac.uk>	<a1d6dd760807300818n199b825fk4c642c1202ede439@mail.gmail.com>
	<45918DC2-F3E5-4F7E-A031-539A659491B1@yale.edu>
Message-ID: <4890CBE4.6000908@imperial.ac.uk>

Jos?, Zachary,

I seem to be having some success running the example from the link
posted!  Thanks for your rapid and helpful responses.

Best regards,

Mark.

Zachary Pincus wrote:
> Mark:
>>  My conclusion is that I need a
>> nearest-neighbour type interpolation routine.
> 
> Jos?:
>> Try Radial Basis Functions.
> 
> This was my thought as well.
> 
> Scipy has a handy implementation:
> http://www.scipy.org/Cookbook/RadialBasisFunctions
> 
> Zach
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
> 


From m.starnes05 at imperial.ac.uk  Wed Jul 30 16:32:31 2008
From: m.starnes05 at imperial.ac.uk (mark starnes)
Date: Wed, 30 Jul 2008 21:32:31 +0100
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
In-Reply-To: <4890B701.40402@gemini.edu>
References: <489084A2.3000408@imperial.ac.uk> <4890B701.40402@gemini.edu>
Message-ID: <4890CFDF.2060304@imperial.ac.uk>

Hi James,

I'm having a go with the radial basis functions, as they seem to be
working with my Scipy (surprisingly!).  Thanks for your suggestions;
I'll move onto them if I fail with this approach.

The points are partly on a structured grid, though not regular, and
partly an unstructured grid.  I was hoping not to treat the regions
differently.

Thanks for your response, though.  I'll be sure to write back if the
radial functions don't work.

Best regards,

Mark.



James Turner wrote:
> Hi Mark,
> 
>> I've an array of velocities at 80,000 points, irregularly spaced (from a
>> CFD analysis).  I'd like to generate the interpolated velocity at any
>> position in the domain, to map the data to an acoustics analysis on a
>> different mesh.
>>
>> I tried a least squares approach but the errors are too large using
>> polynomials and trigonometric functions.  My conclusion is that I need a
>> nearest-neighbour type interpolation routine.
> 
> If you don't have any joy with the other suggestions, there are some
> useful routines by Renka in Netlib, at least one of which is 3D,
> though I believe they are also piecewise polynomials (and in ForTran,
> of course, so need wrapping).
> 
>    http://www.netlib.org/toms/
> 
> I also have a program implementing the Hubble Telescope's "Drizzle"
> algorithm in 3D, which is similar to nearest neighbour or linear,
> but unfortunately it's written in an obscure language that would
> make it difficult to use from Python unless you are prepared to
> install STScI's PyRAF. I also suspect it will only work well for
> data that are mildly irregular and it may be fiddly to set up for
> your application, but anyway, just so you have the option...
> The paper is at http://xxx.lanl.gov/abs/astro-ph/9708242.
> 
> Are the points really irregular in 3D, or just in 1-2 of the
> dimensions? If any dimension is regularly spaced, you might be able
> to simplify the problem by resampling in 1-2D at a time?
> 
> Stefan might also have suggestions ;-). But maybe the radial basis
> functions will do the trick... I'll have to look at those.
> 
> Cheers,
> 
> James.
> 
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
> 


From stefan at sun.ac.za  Wed Jul 30 17:11:36 2008
From: stefan at sun.ac.za (=?ISO-8859-1?Q?St=E9fan_van_der_Walt?=)
Date: Wed, 30 Jul 2008 23:11:36 +0200
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
In-Reply-To: <489084A2.3000408@imperial.ac.uk>
References: <489084A2.3000408@imperial.ac.uk>
Message-ID: <9457e7c80807301411p68d04884ne0a02b6d4f30828a@mail.gmail.com>

Hi Mark

2008/7/30 mark starnes <m.starnes05 at imperial.ac.uk>:
> Hi everyone,
>
> I've looked through the list here and in Numpy-users, and checked the
> 'net but can't find an answer to this problem (with luck, I've missed
> something obvious!).
>
> I've an array of velocities at 80,000 points, irregularly spaced (from a
> CFD analysis).  I'd like to generate the interpolated velocity at any
> position in the domain, to map the data to an acoustics analysis on a
> different mesh.
>
> I tried a least squares approach but the errors are too large using
> polynomials and trigonometric functions.  My conclusion is that I need a
> nearest-neighbour type interpolation routine.

Also take a look at

http://www.scipy.org/Cookbook/Matplotlib/Gridding_irregularly_spaced_data

Robert Kern's delaunay package does natural neighbour interpolation, IIRC.

Regards
St?fan


From jturner at gemini.edu  Wed Jul 30 19:30:12 2008
From: jturner at gemini.edu (James Turner)
Date: Wed, 30 Jul 2008 19:30:12 -0400
Subject: [SciPy-user] Core dump and import errors during scipy.test()
Message-ID: <4890F984.6080101@gemini.edu>

Hot on the heels of my numpy.test() core dump, I now have one from
scipy.test() using SciPy 0.6.0, again on RedHat Enterprise 3 with
Python 2.5.1. The last line (with verbosity=2) is this:

check_cosine_weighted_infinite 
(scipy.integrate.tests.test_quadpack.test_quad)Floating exception (core dumped)

I tried running gdb on the core dump and I'm not 100% sure what
I'm looking at, but maybe this bit is useful?

Loaded symbols for 
/astro/iraf/i686/gempylocal/lib/python2.5/site-packages/scipy/cluster/_vq.so
#0  0xb61359c0 in dqawfe_ (f=0xb612a6e8 <quad_function>, a=0xbfff94a0,
     omega=0xbfff94c0, integr=0xbfff94a8, epsabs=0xbfff94d8, limlst=0xbfff94b4,
     limit=0xbfff94b8, maxp1=0xbfff94bc, result=0xbfff94d0, abserr=0xbfff94c8,
     neval=0xbfff9484, ier=0xbfff9488, lst=0xbfff948c, alist=0x88024a8,
     blist=0x8802640, rlist=0x8805698, elist=0x8805830, iord=0x8719b70,
     nnlog=0x87566b8, chebmo=0x8802f80) at scipy/integrate/quadpack/dqawfe.f:267
267        10 l = dabs(omega)

In addition, I have a load of import failures at the start of the
tests, due to a missing symbol (zfftnd_fftw) in _fftpack.so. From
a Google search, it looks like this is a known SciPy problem that
David has fixed in the SVN repository. Is there a workaround for
the public SciPy release? I don't really understand what makes
this show up on my machine or whether everyone sees it (but then
presumably there would have been a SciPy 0.6.1 release?). It looks
like one of the files is just missing from the compilation.

Thanks for any suggestions,

James.



From jjh at 42quarks.com  Wed Jul 30 19:50:52 2008
From: jjh at 42quarks.com (Jonathan Hunt)
Date: Thu, 31 Jul 2008 09:50:52 +1000
Subject: [SciPy-user] Running SciPy SVN without installing Was: Matlab
	IO -- can others with matlab test this bug?
In-Reply-To: <20080730184102.GL32733@phare.normalesup.org>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
	<200807291015.24171.pgmdevlist@gmail.com>
	<488F2479.8020706@ar.media.kyoto-u.ac.jp>
	<e26778a60807290732s5564ea84pac3d4737781a2c3e@mail.gmail.com>
	<20080730184102.GL32733@phare.normalesup.org>
Message-ID: <e26778a60807301650k59e307a5t9b9e3b0148d4dce@mail.gmail.com>

Thanks for the help. I tried using develop mode in SciPy but since
SciPy just uses distutils it doesn't have develop mode. Is there an
easy way to make it use setuptools?

Thanks,
Jonny

On Thu, Jul 31, 2008 at 4:41 AM, Gael Varoquaux
<gael.varoquaux at normalesup.org> wrote:
> On Wed, Jul 30, 2008 at 12:32:08AM +1000, Jonathan Hunt wrote:
>> Thanks for all the comments. If anyone uses the develop mode of
>> setuptools or other methods I'd be interested to hear.
>
> I do. I am not sure what to tell you about it. Just be aware that when
> you use setuptools, your import path gets messed up (print sys.path to
> reaize that), and as a result import priorities get scrambled. If you
> know that and keep watch on what is important, when you have several
> version of different packages installed, you can usual understand what is
> going on.
>
> Ga?l
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>



-- 
Jonathan J Hunt <jjh at 42quarks.com>
Homepage: http://www.42quarks.net.nz/wiki/JJH
(Further contact details there)
"Physics isn't the most important thing. Love is." Richard Feynman


From gael.varoquaux at normalesup.org  Wed Jul 30 20:09:39 2008
From: gael.varoquaux at normalesup.org (Gael Varoquaux)
Date: Thu, 31 Jul 2008 02:09:39 +0200
Subject: [SciPy-user] Running SciPy SVN without installing Was:
	Matlab	IO -- can others with matlab test this bug?
In-Reply-To: <e26778a60807301650k59e307a5t9b9e3b0148d4dce@mail.gmail.com>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>
	<200807291015.24171.pgmdevlist@gmail.com>
	<488F2479.8020706@ar.media.kyoto-u.ac.jp>
	<e26778a60807290732s5564ea84pac3d4737781a2c3e@mail.gmail.com>
	<20080730184102.GL32733@phare.normalesup.org>
	<e26778a60807301650k59e307a5t9b9e3b0148d4dce@mail.gmail.com>
Message-ID: <20080731000939.GG27172@phare.normalesup.org>

On Thu, Jul 31, 2008 at 09:50:52AM +1000, Jonathan Hunt wrote:
> Thanks for the help. I tried using develop mode in SciPy but since
> SciPy just uses distutils it doesn't have develop mode. Is there an
> easy way to make it use setuptools?

Use setupegg.py instead of setup.py.

Ga?l


From m.starnes05 at imperial.ac.uk  Thu Jul 31 04:13:26 2008
From: m.starnes05 at imperial.ac.uk (mark starnes)
Date: Thu, 31 Jul 2008 09:13:26 +0100
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
In-Reply-To: <9457e7c80807301411p68d04884ne0a02b6d4f30828a@mail.gmail.com>
References: <489084A2.3000408@imperial.ac.uk>
	<9457e7c80807301411p68d04884ne0a02b6d4f30828a@mail.gmail.com>
Message-ID: <48917426.5000901@imperial.ac.uk>

Hi St?fan, thanks for the tip.  I tried this approach before, but
couldn't get it to work, due to my installation (I broke it, got it
working again, then wasn't brave enough to tinker).  However, I got the
radial basis functions working by modifying my PYTHONPATH, rather than
with a rebuild, so may try the Delauney / natgrid approaches too.  The
radial basis function seems to have trouble with more than 5000 points
so I may need other options.

Thanks for taking the time to help!

Best regards,

Mark.



St?fan van der Walt wrote:
> Hi Mark
> 
> 2008/7/30 mark starnes <m.starnes05 at imperial.ac.uk>:
>> Hi everyone,
>>
>> I've looked through the list here and in Numpy-users, and checked the
>> 'net but can't find an answer to this problem (with luck, I've missed
>> something obvious!).
>>
>> I've an array of velocities at 80,000 points, irregularly spaced (from a
>> CFD analysis).  I'd like to generate the interpolated velocity at any
>> position in the domain, to map the data to an acoustics analysis on a
>> different mesh.
>>
>> I tried a least squares approach but the errors are too large using
>> polynomials and trigonometric functions.  My conclusion is that I need a
>> nearest-neighbour type interpolation routine.
> 
> Also take a look at
> 
> http://www.scipy.org/Cookbook/Matplotlib/Gridding_irregularly_spaced_data
> 
> Robert Kern's delaunay package does natural neighbour interpolation, IIRC.
> 
> Regards
> St?fan
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
> 


From m.starnes05 at imperial.ac.uk  Thu Jul 31 04:58:05 2008
From: m.starnes05 at imperial.ac.uk (mark starnes)
Date: Thu, 31 Jul 2008 09:58:05 +0100
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
In-Reply-To: <9457e7c80807301411p68d04884ne0a02b6d4f30828a@mail.gmail.com>
References: <489084A2.3000408@imperial.ac.uk>
	<9457e7c80807301411p68d04884ne0a02b6d4f30828a@mail.gmail.com>
Message-ID: <48917E9D.207@imperial.ac.uk>

Hi again, St?fan.

Is Robert Kern's package limited to two-dimensional data?  I've had a
look and can't see any three-dimensional options.

Best regards,

Mark.

St?fan van der Walt wrote:
> Hi Mark
> 
> 2008/7/30 mark starnes <m.starnes05 at imperial.ac.uk>:
>> Hi everyone,
>>
>> I've looked through the list here and in Numpy-users, and checked the
>> 'net but can't find an answer to this problem (with luck, I've missed
>> something obvious!).
>>
>> I've an array of velocities at 80,000 points, irregularly spaced (from a
>> CFD analysis).  I'd like to generate the interpolated velocity at any
>> position in the domain, to map the data to an acoustics analysis on a
>> different mesh.
>>
>> I tried a least squares approach but the errors are too large using
>> polynomials and trigonometric functions.  My conclusion is that I need a
>> nearest-neighbour type interpolation routine.
> 
> Also take a look at
> 
> http://www.scipy.org/Cookbook/Matplotlib/Gridding_irregularly_spaced_data
> 
> Robert Kern's delaunay package does natural neighbour interpolation, IIRC.
> 
> Regards
> St?fan
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
> 


From josemaria.alkala at gmail.com  Thu Jul 31 05:18:50 2008
From: josemaria.alkala at gmail.com (=?ISO-8859-1?Q?Jos=E9_Mar=EDa_Garc=EDa_P=E9rez?=)
Date: Thu, 31 Jul 2008 11:18:50 +0200
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
In-Reply-To: <48917E9D.207@imperial.ac.uk>
References: <489084A2.3000408@imperial.ac.uk>
	<9457e7c80807301411p68d04884ne0a02b6d4f30828a@mail.gmail.com>
	<48917E9D.207@imperial.ac.uk>
Message-ID: <a1d6dd760807310218g7e49fdbajd259f957704c4228@mail.gmail.com>

Mark,
>From my experience working with RBF, they work pretty well even when you use
few points for the interpolation. They track non linear behavior very well.
I have worked with RBF with very big FEM models (200000-500000 grid points)
and with more than 3D (in other disciplines), but I don't take all the
points at the same. What I would do is to use for example the 100 nearest
points to the geometric point where you want to interpolate (probably with
10 would be enough). That's something you can try: test using 10, 20, 50,
100 points, and you will see that the difference is small pretty soon, and
for sure smaller that the error you may expect from a CFD simulation.
Hope this tip helps!
Jos? M.


2008/7/31 mark starnes <m.starnes05 at imperial.ac.uk>

> Hi again, St?fan.
>
> Is Robert Kern's package limited to two-dimensional data?  I've had a
> look and can't see any three-dimensional options.
>
> Best regards,
>
> Mark.
>
> St?fan van der Walt wrote:
> > Hi Mark
> >
> > 2008/7/30 mark starnes <m.starnes05 at imperial.ac.uk>:
> >> Hi everyone,
> >>
> >> I've looked through the list here and in Numpy-users, and checked the
> >> 'net but can't find an answer to this problem (with luck, I've missed
> >> something obvious!).
> >>
> >> I've an array of velocities at 80,000 points, irregularly spaced (from a
> >> CFD analysis).  I'd like to generate the interpolated velocity at any
> >> position in the domain, to map the data to an acoustics analysis on a
> >> different mesh.
> >>
> >> I tried a least squares approach but the errors are too large using
> >> polynomials and trigonometric functions.  My conclusion is that I need a
> >> nearest-neighbour type interpolation routine.
> >
> > Also take a look at
> >
> >
> http://www.scipy.org/Cookbook/Matplotlib/Gridding_irregularly_spaced_data
> >
> > Robert Kern's delaunay package does natural neighbour interpolation,
> IIRC.
> >
> > Regards
> > St?fan
> > _______________________________________________
> > SciPy-user mailing list
> > SciPy-user at scipy.org
> > http://projects.scipy.org/mailman/listinfo/scipy-user
> >
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080731/b334d9fe/attachment.html>

From m.starnes05 at imperial.ac.uk  Thu Jul 31 07:12:10 2008
From: m.starnes05 at imperial.ac.uk (mark starnes)
Date: Thu, 31 Jul 2008 12:12:10 +0100
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
In-Reply-To: <a1d6dd760807310218g7e49fdbajd259f957704c4228@mail.gmail.com>
References: <489084A2.3000408@imperial.ac.uk>	<9457e7c80807301411p68d04884ne0a02b6d4f30828a@mail.gmail.com>	<48917E9D.207@imperial.ac.uk>
	<a1d6dd760807310218g7e49fdbajd259f957704c4228@mail.gmail.com>
Message-ID: <48919E0A.2020505@imperial.ac.uk>

Hi Jos?,

Thanks for your comments.  Knowing that other people split the
interpolation into regions is useful.  I'll be attempting an approach
similar to that you suggested; splitting the domain, not around each
desired point but into regions containing sets of points.

I didn't expect the effect you mention about accuracy being related to
the number of points.  Looks like I need to read up on the theory....

Thanks again,

Mark.


Jos? Mar?a Garc?a P?rez wrote:
> Mark,
> From my experience working with RBF, they work pretty well even when you
> use few points for the interpolation. They track non linear behavior
> very well.
> I have worked with RBF with very big FEM models (200000-500000 grid
> points) and with more than 3D (in other disciplines), but I don't take
> all the points at the same. What I would do is to use for example the
> 100 nearest points to the geometric point where you want to interpolate
> (probably with 10 would be enough). That's something you can try: test
> using 10, 20, 50, 100 points, and you will see that the difference is
> small pretty soon, and for sure smaller that the error you may expect
> from a CFD simulation.
> Hope this tip helps!
> Jos? M.
> 
> 
> 2008/7/31 mark starnes <m.starnes05 at imperial.ac.uk
> <mailto:m.starnes05 at imperial.ac.uk>>
> 
>     Hi again, St?fan.
> 
>     Is Robert Kern's package limited to two-dimensional data?  I've had a
>     look and can't see any three-dimensional options.
> 
>     Best regards,
> 
>     Mark.
> 
>     St?fan van der Walt wrote:
>     > Hi Mark
>     >
>     > 2008/7/30 mark starnes <m.starnes05 at imperial.ac.uk
>     <mailto:m.starnes05 at imperial.ac.uk>>:
>     >> Hi everyone,
>     >>
>     >> I've looked through the list here and in Numpy-users, and checked the
>     >> 'net but can't find an answer to this problem (with luck, I've missed
>     >> something obvious!).
>     >>
>     >> I've an array of velocities at 80,000 points, irregularly spaced
>     (from a
>     >> CFD analysis).  I'd like to generate the interpolated velocity at any
>     >> position in the domain, to map the data to an acoustics analysis on a
>     >> different mesh.
>     >>
>     >> I tried a least squares approach but the errors are too large using
>     >> polynomials and trigonometric functions.  My conclusion is that I
>     need a
>     >> nearest-neighbour type interpolation routine.
>     >
>     > Also take a look at
>     >
>     >
>     http://www.scipy.org/Cookbook/Matplotlib/Gridding_irregularly_spaced_data
>     >
>     > Robert Kern's delaunay package does natural neighbour
>     interpolation, IIRC.
>     >
>     > Regards
>     > St?fan
>     > _______________________________________________
>     > SciPy-user mailing list
>     > SciPy-user at scipy.org <mailto:SciPy-user at scipy.org>
>     > http://projects.scipy.org/mailman/listinfo/scipy-user
>     >
>     _______________________________________________
>     SciPy-user mailing list
>     SciPy-user at scipy.org <mailto:SciPy-user at scipy.org>
>     http://projects.scipy.org/mailman/listinfo/scipy-user
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user


From david at ar.media.kyoto-u.ac.jp  Thu Jul 31 08:07:20 2008
From: david at ar.media.kyoto-u.ac.jp (David Cournapeau)
Date: Thu, 31 Jul 2008 21:07:20 +0900
Subject: [SciPy-user] Running SciPy SVN without installing Was:	Matlab
 IO -- can others with matlab test this bug?
In-Reply-To: <20080730183837.GK32733@phare.normalesup.org>
References: <e26778a60807290631k1742185en6d950f2e61533ed@mail.gmail.com>	<200807291015.24171.pgmdevlist@gmail.com>	<488F2479.8020706@ar.media.kyoto-u.ac.jp>
	<20080730183837.GK32733@phare.normalesup.org>
Message-ID: <4891AAF8.9030906@ar.media.kyoto-u.ac.jp>

Gael Varoquaux wrote:
> On Tue, Jul 29, 2008 at 11:08:57PM +0900, David Cournapeau wrote:
>   
>> Really, stow is nice. The
>> only problem is that it does not work on windows.
>>     
>
> Can I have two shells, one running one environment, and the other
> another, say to compare results with stow?
>   

Sure, just get one stow environment per shell, and set PYTHONPATH
differently for each one (for example by using an alias
pythonfoo='PYTHONPATH=foo python'). The great advantage of stow is its
simplicity. If you screw up something, you just remove the directory,
there is no side effect at all. In a way, stow fixes one of the things
which I really hate with unix, the FHS.

cheers,

David


From stefan at sun.ac.za  Thu Jul 31 08:32:38 2008
From: stefan at sun.ac.za (=?ISO-8859-1?Q?St=E9fan_van_der_Walt?=)
Date: Thu, 31 Jul 2008 14:32:38 +0200
Subject: [SciPy-user] [Fwd: 3D interpolation over irregular data]
In-Reply-To: <48917E9D.207@imperial.ac.uk>
References: <489084A2.3000408@imperial.ac.uk>
	<9457e7c80807301411p68d04884ne0a02b6d4f30828a@mail.gmail.com>
	<48917E9D.207@imperial.ac.uk>
Message-ID: <9457e7c80807310532j4ae692d8qf038fbf8162f3004@mail.gmail.com>

2008/7/31 mark starnes <m.starnes05 at imperial.ac.uk>:
> Is Robert Kern's package limited to two-dimensional data?  I've had a
> look and can't see any three-dimensional options.

Sorry, Mark, my mistake, it doesn't.

The only Python wrapper for 3-dimensional Delaunay I'm aware of is

from enthought.tvtk.api import tvtk
d = tvtk.Delaunay3D()

Unfortunately, I don't have much more detail on how to use it.

Regards
St?fan


From bsouthey at gmail.com  Thu Jul 31 09:47:14 2008
From: bsouthey at gmail.com (Bruce Southey)
Date: Thu, 31 Jul 2008 08:47:14 -0500
Subject: [SciPy-user] Robust Statistics
In-Reply-To: <F5BC85B8534D524DB972D2B2731CA53F29CAEFDFD4@EX7FM01.ad.uky.edu>
References: <web-105374238@uni-stuttgart.de>	<200807301243.12871.jr@sun.ac.za>	<web-105378668@uni-stuttgart.de>	<a749952d0807300828h5a1939c7kdbc89afc7c64160d@mail.gmail.com>	<web-105392498@uni-stuttgart.de>	<a749952d0807300859q4507c8cl7c840701c963a949@mail.gmail.com>	<web-105393352@uni-stuttgart.de>	<a749952d0807300917v1ae4a90dy379b5d7f22413ac9@mail.gmail.com>	<web-105393808@uni-stuttgart.de>	<a749952d0807300936q7b3eeb2cvee2638b1a5c4012e@mail.gmail.com>	<web-105394546@uni-stuttgart.de>
	<F5BC85B8534D524DB972D2B2731CA53F29CAEFDFD4@EX7FM01.ad.uky.edu>
Message-ID: <4891C262.9020002@gmail.com>

Nuttall, Brandon C wrote:
> Folks,
>
> I have some rate/time data I'm in the process of analyzing. The data represent the natural gas produced from a well over time. The data are fit best by either an exponential model (least squares best fit of log of rate data, just qi and di see below) or a "hyperbolic" model:
>
> qt = qi *(1-b*di*t)**(-1.0/b)
>
>         qt = rate at time t
>         qi = initial rate (calculated)
>         b = decline exponent (rate of change of the decline rate, calculated)
>         di = initial decline rate (calculated)
>         t = time
>
> The calculated parameters are found using the optimization routine scipy.optimize.fmin_tnc(). I'm already rejecting some best fit results because either the correlation coefficient is not statistically significant or the di (slope) is not significantly different from 0.
>
> The image I've attached shows two things. An exponential best fit is shown by the solid line and a hyperbolic best fit by the dotted line. It is common practice in analyzing this type of data to begin the best fit analysis at an early time point where the data actually start to "decline"; thus, the hyperbolic best fit started at time 2 and the first rate point (rate = about 1350) is ignored. The other thing the graph shows is that there appear to be outliers.
>
> My question is, what Python resources are there for testing for outliers and robust statistics?
>
> Is RANSAC appropriate for this? Note that when I run ransack.py from the cookbook, I get:
>
> Traceback (most recent call last):
>
> --Snip other traceback info---
>   File "C:\Documents and Settings\bnuttall\Desktop\ransac.py", line 132, in fit
>     A = numpy.vstack([data[:,i] for i in self.input_columns]).T
> AttributeError: 'numpy.ndarray' object has no attribute 'T'
>
> I presume the T attribute is supposed to mean the transpose operator and have changed the offending statement(s, there are 4 total) to the form:
>
> A = numpy.vstack([data[:,i] for i in self.input_columns]).transpose()
>
> Thanks.
>
> Brandon Nuttall
>   
>
> ------------------------------------------------------------------------
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>   
Hi,
First, what do you mean by 'outlier'? Please see, for example, 
http://en.wikipedia.org/wiki/Outlier because outliers fully depend on 
the (implicit or explicit) model fitted. Note this is really defined for 
linear models not nonlinear models that you are fitting.

Second, NumPy provides the tools to do fit various models but you need 
to implement your own statistics especially for nonlinear models. (In 
particular I would strongly suggest looking at likelihood ratio test and 
Bayesian information criterion for model selection.) Really your best 
bet is to use to R directly or interface to R with Rpy. That way you 
also have access to many nonlinear statistical routines that can fit 
that type of data.

If you do not need a parametric model, try fitting splines. These may be 
more suited to handle the variability near t=60 although that may be 
more related to data collection and time scale.

Regards
Bruce


From lists at vrbka.net  Thu Jul 31 10:43:22 2008
From: lists at vrbka.net (Lubos Vrbka)
Date: Thu, 31 Jul 2008 16:43:22 +0200
Subject: [SciPy-user] sine transformation weirdness
Message-ID: <4891CF8A.2040505@vrbka.net>

hi guys,

while using the sine transformation from sandbox
http://projects.scipy.org/scipy/scipy/browser/trunk/scipy/sandbox/image/transforms.py
i encountered some confusion with respect to the constants involved in 
the transformation process. to test it, i tried to do discrete sine 
transformation of
f(r) = exp(-r)
analytically, the result is
f(k) = sqrt(2/pi) * k/(1+k^2)

for the discrete transformation i used dr=0.1, 1000 points, dk = 
pi/(npoints*dr) since the dst uses effectively the double number of 
sampling points. you can see the result in the attached file (hopefully 
it gets it through). the blue line is the analytical result (without the 
sqrt(2/pi) factor), green is the numerical discrete sine transform and 
the red line is f(k,numerical)/f(k,analytical) ratio. in case the figure 
doesn't make it through, the red line looks like one half of reverted 
parabola. for r=0, its value is ~10. it reaches 0 at the distance 
corresponding to the last sampling point.

apparently, there is some non-constant factor present here - i just 
don't know what factor it might be. it has to come from the FFT used 
inside the DST routine... i'd be very glad for any hint in this respect.

thanks in advance! with best regards,
lubos

-- 
Lubos _ at _"
http://www.lubos.vrbka.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fst_exp_r.png
Type: image/png
Size: 14188 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20080731/6eafc5e1/attachment.png>

From jturner at gemini.edu  Thu Jul 31 18:06:50 2008
From: jturner at gemini.edu (James Turner)
Date: Thu, 31 Jul 2008 18:06:50 -0400
Subject: [SciPy-user] Core dump and import errors during scipy.test()
In-Reply-To: <4890F984.6080101@gemini.edu>
References: <4890F984.6080101@gemini.edu>
Message-ID: <4892377A.3040803@gemini.edu>

> #0  0xb61359c0 in dqawfe_ (f=0xb612a6e8 <quad_function>, a=0xbfff94a0,
>     omega=0xbfff94c0, integr=0xbfff94a8, epsabs=0xbfff94d8, 
> limlst=0xbfff94b4,
>     limit=0xbfff94b8, maxp1=0xbfff94bc, result=0xbfff94d0, 
> abserr=0xbfff94c8,
>     neval=0xbfff9484, ier=0xbfff9488, lst=0xbfff948c, alist=0x88024a8,
>     blist=0x8802640, rlist=0x8805698, elist=0x8805830, iord=0x8719b70,
>     nnlog=0x87566b8, chebmo=0x8802f80) at 
> scipy/integrate/quadpack/dqawfe.f:267
> 267        10 l = dabs(omega)

I had a look at the file scipy/integrate/quadpack/dqawfe.f and
realized that whilst "dabs()" is a double-precision built-in
function, it is being assigned to an integer, "l" on line 267.
I wondered if the core dump is just due to their different
lengths, so I changed "l" to be declared as double precision and
now that crash seems to have gone away, but I'm getting another
one from the same _vq.so library:

Loaded symbols for 
/astro/iraf/i686/gempylocal/lib/python2.5/site-packages/scipy/cluster/_vq.so
#0  gesdd_ (min_lwork_=0xbfff6af0, max_lwork_=0xbfff6af0, prefix=0x1, m=0x5,
     n=0xbfff6b20, compute_uv_=0x0, __g77_length_prefix=-1249493092)
     at scipy/linalg/src/calc_lwork.f:39
39      MNTHR = INT( MINMN*11.0D0 / 6.0D0 )

The last lines from test(verbosity=2) are:
Check the behavior when the inputs and outputs are multidimensional. ... ok
Make sure that appropriate exceptions are raised when invalid valuesFloating 
exception (core dumped)

(...seems like they weren't?)

Are these kinds of test crashes a common occurence? It seems like
they should be if they're declaration errors? Or does no-one else
build the official public SciPy from source?

I still have to look into my other problem with the imports, but
it sounds like that one is already fixed in the latest SVN anyway.

Thanks!

James.



From brennan.williams at visualreservoir.com  Thu Jul 31 18:12:17 2008
From: brennan.williams at visualreservoir.com (Brennan Williams)
Date: Fri, 01 Aug 2008 10:12:17 +1200
Subject: [SciPy-user] scipy.io.numpyio fwrite - appending or updating an
	array
Message-ID: <489238C1.7040902@visualreservoir.com>

I have an existing binary file containing numpy array data. It has been 
created using open,fwrite & close and I can read the data using fread.

I want to be able to either append a new array to the end of the file or 
update an existing array within the file.

I've tried opening the file with a mode of either 'ab+' or 'wb+' and 
then writing the data using something like....

     fd = open(vfname, 'ab+')
     if fd:
       filepos=(self.id-1)*self.yarray.size*4
       fd.seek(filepos)
       fwrite(fd, self.yarray.size, self.yarray,'f')
       fd.close()

When I use a mode of 'ab+' it looks like the data has been written to 
the file ok (no errors reported) but when I read it back I get my 
original data.

When I use 'wb+' then my updated data gets written and read back ok. But 
when I reload the file, everything apart from my updated data (i.e. 
everything before it in the file) is now zero.

The '+' in the mode seems to make no difference.

What am I doing wrong?

Thanks

Bren.


From brennan.williams at visualreservoir.com  Thu Jul 31 22:42:15 2008
From: brennan.williams at visualreservoir.com (Brennan Williams)
Date: Fri, 01 Aug 2008 14:42:15 +1200
Subject: [SciPy-user] scipy.io.numpyio fwrite - appending or updating
 an	array
In-Reply-To: <489238C1.7040902@visualreservoir.com>
References: <489238C1.7040902@visualreservoir.com>
Message-ID: <48927807.3080603@visualreservoir.com>

I've tried replacing numpyio with both fopen and now also npfile but I'm 
getting the same problem, i.e. if I write a numpy array to the file,
everything else before that position in the file is now zero. It is as 
if it is a new file, not an existing one.

Brennan Williams wrote:
> I have an existing binary file containing numpy array data. It has been 
> created using open,fwrite & close and I can read the data using fread.
>
> I want to be able to either append a new array to the end of the file or 
> update an existing array within the file.
>
> I've tried opening the file with a mode of either 'ab+' or 'wb+' and 
> then writing the data using something like....
>
>      fd = open(vfname, 'ab+')
>      if fd:
>        filepos=(self.id-1)*self.yarray.size*4
>        fd.seek(filepos)
>        fwrite(fd, self.yarray.size, self.yarray,'f')
>        fd.close()
>
> When I use a mode of 'ab+' it looks like the data has been written to 
> the file ok (no errors reported) but when I read it back I get my 
> original data.
>
> When I use 'wb+' then my updated data gets written and read back ok. But 
> when I reload the file, everything apart from my updated data (i.e. 
> everything before it in the file) is now zero.
>
> The '+' in the mode seems to make no difference.
>
> What am I doing wrong?
>
> Thanks
>
> Bren.
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>
>