From oanjao at yahoo.com  Tue May  1 00:17:31 2007
From: oanjao at yahoo.com (Craig Finch)
Date: Mon, 30 Apr 2007 15:17:31 -0700 (PDT)
Subject: [C++-sig] Need strategic advice designing Boost/Python program
In-Reply-To: <4636463B.8090709@sympatico.ca>
Message-ID: <62078.32342.qm@web30315.mail.mud.yahoo.com>

 
> This, together with your point further down that you want to add
> python
> methods that are only ever used inside python make it sound as if
> a C++ base class derived from in python may be the best approach.

I think I am beginning to see why this approach is good.  I think this
is a cleaner solution than using __init__.py to inject methods.  By
"cleaner," I mean that my intentions will be more obvious to someone
who looks at my code for the first time.  From the Tutorial, this is
the only way I know to inject methods, so let me know if there's
another way.

> I don't understand what you mean by "I can't access the data members
> of
> the C++ class". Of course, you have to expose them to python via
> accessors
> (say, as 'properties'), but as soon as that is done you can easily
> derive
> and extend that base class from within python. 

I keep getting error messages like this one when I define a constructor
in Python:
Boost.Python.ArgumentError: Python argument types in
    python_constructor.value(python_constructor)
did not match C++ signature:
    value(python_constructor {lvalue})
This error message is from a simple test case that I have created. 
BTW, I'm injecting the method from __init__.py.  I'm probably doing
something wrong--should I start a new discussion thread, since this is
kind of a side topic?

> I'm not sure I understand. extract<> is a C++ API, that has nothing
> to
> do with a Python interface. The only context in which you may need
> extract<>
> here is when talking to a Python object (such as an instance of the
> derived
> Python class) from within C++, which, as you stated earlier, you
> won't.

I guess that wasn't very clear.  To elaborate: when a Python programmer
uses my C++ class, he should not be concerned about C details like
array length, memory allocation, etc.  The Python user should be able
to create lists, NumPy arrays, etc.  The underlying C++ code should
then check for type errors and convert them to C arrays for efficient
processing.  My plan was to use numeric::array objects in C++, which
would provide the right Python behavior, and would require the use of
<extract> to get the correct type from each element of the array.  It
looks like Ralf is giving me a lot to think about on the array subject,
so I'm going to look at that before I say any more on this subject!
 
> The same IS-A relationship that you would expect from a pure C++
> program
> still holds in this hybrid setting. If you have exposed a C++
> function
> (or method) expecting a parent reference (or pointer), a child
> reference
> (or pointer) will be accepted as well.

Great!  Things like this make me very impressed with Boost.python.


--------------
Please reply to cfinch at ieee.org

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From seefeld at sympatico.ca  Tue May  1 00:22:38 2007
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Mon, 30 Apr 2007 18:22:38 -0400
Subject: [C++-sig] Need strategic advice designing Boost/Python program
In-Reply-To: <62078.32342.qm@web30315.mail.mud.yahoo.com>
References: <62078.32342.qm@web30315.mail.mud.yahoo.com>
Message-ID: <46366C2E.3040303@sympatico.ca>

Craig Finch wrote:

> I keep getting error messages like this one when I define a constructor
> in Python:
> Boost.Python.ArgumentError: Python argument types in
>     python_constructor.value(python_constructor)
> did not match C++ signature:
>     value(python_constructor {lvalue})
> This error message is from a simple test case that I have created. 

Since you don't show any specific code I have to guess: Are you
calling the base class constructor in your derived class constructor ?
In contrast to C++, Python doesn't do this automatically for you, so
if you forget to initialize the base class you may see conflicts such
as the above when passing a derived where a base is expected.

> BTW, I'm injecting the method from __init__.py.  I'm probably doing
> something wrong--should I start a new discussion thread, since this is
> kind of a side topic?

If you have a specific question (with code samples etc.) that may
make sense. It's up to you.

I'm not sure what __init__.py is here. So I guess a complete test case
would indeed help.

Regards,
		Stefan

-- 

      ...ich hab' noch einen Koffer in Berlin...


From seefeld at sympatico.ca  Tue May  1 01:00:59 2007
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Mon, 30 Apr 2007 19:00:59 -0400
Subject: [C++-sig] Need strategic advice designing Boost/Python program
In-Reply-To: <148457.12444.qm@web31106.mail.mud.yahoo.com>
References: <148457.12444.qm@web31106.mail.mud.yahoo.com>
Message-ID: <4636752B.90007@sympatico.ca>

Ralf W. Grosse-Kunstleve wrote:
>> This, together with your point further down that you want to add python
>> methods that are only ever used inside python make it sound as if
>> a C++ base class derived from in python may be the best approach.
> 
> FWIW: there is a very powerful alternative to this: injecting.
> For example:
> 
> class _histogram(boost.python.injector, ext.histogram):
> 
>   def show(self, f=None, prefix="", format_cutoffs="%.8g"):
>     if (f is None): f = sys.stdout
>     fmt = "%s" + format_cutoffs + " - " + format_cutoffs + ": %d"
>     for info in self.slot_infos():
>       print >> f, fmt % (prefix, info.low_cutoff, info.high_cutoff, info.n)
> 
> The boost.python.injector is a small utility class defined
> in boost_adaptbx (adaptor toolbox), boost/python.py.
> 
> The difference is that objects of ext.histogram will have the
> .show() method, even if they are created in C++.
> I.e. even if you only have a C++ constructor, you can have pure
> Python methods.

Sorry, can you elaborate on that a bit ? What is special in the code
above ? It looks like a 'normal' definition of a derived python class
('_histogram') that derives from a (presumably) C++ class (ext.histogram).
What does boost.python.injector do that wouldn't be available without it ?

(I don't understand the 'only C++ constructor' part, in fact)

Thanks,
		Stefan


-- 

      ...ich hab' noch einen Koffer in Berlin...


From rwgk at yahoo.com  Tue May  1 02:16:50 2007
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Mon, 30 Apr 2007 17:16:50 -0700 (PDT)
Subject: [C++-sig] Need strategic advice designing Boost/Python program
Message-ID: <631819.11856.qm@web31111.mail.mud.yahoo.com>

> Sorry, can you elaborate on that a bit ? What is special in the code
> above ? It looks like a 'normal' definition of a derived python class
> ('_histogram') that derives from a (presumably) C++ class (ext.histogram).

If you just derive without the boost.python.injector, the C++ class ext.histogram
doesn't know about the .show() method, only _histogram does.

Maybe I should have shown the bare-bones approach first. Here it is again:



  def histogram_show(self, f=None, prefix="", format_cutoffs="%.8g"):
    # exact same code

  ext.histogram.show = histogram_show

Why is this cool?

If you have a C++ function returning a new histogram object, you want that
to have the .show() method:

C++:
  histogram make_histogram();

  def("make_histogram", make_histogram);

Python:

  h = make_histogram()
  h.show()

It doesn't help if you have the plain inheritance _histogram, since h above is just an ext.histogram.
To get the same result without injecting the .show() method, you'd have to def a Python wrapper
function, something like:

  def make_histogram():
      h = ext.make_histogram()
      return _histogram(h)

And that for every C++ function returning a histogram object.

Ralf




__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070430/59e57fe4/attachment.htm>

From dave at boost-consulting.com  Tue May  1 13:41:34 2007
From: dave at boost-consulting.com (David Abrahams)
Date: Tue, 01 May 2007 07:41:34 -0400
Subject: [C++-sig] [Boost.Python] inconsistent use of function pointer
	typedef
References: <f0smha$gms$1@sea.gmane.org>
Message-ID: <87hcqwkbyp.fsf@grogan.peloton>


on Fri Apr 27 2007, Sebastian Ramacher <s.ramacher-AT-gmx.at> wrote:

> I just wanted to repost a patch I've submitted to the SourceForge
> tracker some days ago. The patch fixes the inconsistent use of the
> function pointer typedef 'convertible_function' at the declaration and
> definition of 'boost::python::converter::registry::insert' and
> 'boost::python::converter::registry::push_back'.

This is just a warning fix, right?

-- 
Dave Abrahams
Boost Consulting
http://www.boost-consulting.com

Don't Miss BoostCon 2007! ==> http://www.boostcon.com



From oanjao at yahoo.com  Wed May  2 00:42:38 2007
From: oanjao at yahoo.com (Craig Finch)
Date: Tue, 1 May 2007 15:42:38 -0700 (PDT)
Subject: [C++-sig] Problem injecting Python __init__ into wrapped C++ class
Message-ID: <891935.77183.qm@web30301.mail.mud.yahoo.com>

Injecting Python methods into wrapped Boost C++ classes sounds like an
excellent idea.  It is described in the Boost.python tutorial at
http://www.boost.org/libs/python/doc/tutorial/doc/html/python/techniques.html#python.extending_wrapped_objects_in_python

I have been able to get this to work for ordinary methods, but I would
like to define an __init__ method in Python for a wrapped C++ class.  I
have created a test case, following the methods in the Tutorial, that
demonstrates the runtime error that results when I try this.  Because
this involves creating a Python package with a directory structure that
would be difficult to reproduce correctly in this message, I'm posting
a link to a .tgz file that contains everything you need to build and
run the example (using make and gcc) and see the error message.

http://www.shocksolution.com/files/TestConstructors.tgz

When I build and run the code, I get the following output:
----------------------------------------------------------
Python value is 18
C++ value is 18
Python value is
Traceback (most recent call last):
  File "TestConstructors/TestConstructors.py", line 8, in ?
    print "Python value is", y.value()
Boost.Python.ArgumentError: Python argument types in
    python_constructor.value(python_constructor)
did not match C++ signature:
    value(python_constructor {lvalue})

I would appreciate it if someone could take a look at the code (it's
really very small) and tell me what I'm doing wrong here.

--------------
Please reply to cfinch at ieee.org

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From rwgk at yahoo.com  Wed May  2 02:04:37 2007
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Tue, 1 May 2007 17:04:37 -0700 (PDT)
Subject: [C++-sig] Problem injecting Python __init__ into wrapped C++
	class
Message-ID: <588216.9155.qm@web31111.mail.mud.yahoo.com>

> I have been able to get this to work for ordinary methods, but I would

> like to define an __init__ method in Python for a wrapped C++ class.  I
> have created a test case, following the methods in the Tutorial, that
> demonstrates the runtime error that results when I try this.



I'm doing it with a Python wrapper function for the constructor, e.g. sticking to my

histogram example (although the following is completely made up):



def histogram(a, b, c):

  if (c is None):

     return ext.histogram(a, b)

  return ext.histogram(a, b, c)



The disadvantage is that help(histogram) only reports the function, not the entire class

(in all our modules histogram is ext.histogram until overridden somehow).



I don't really like this solution, but I'm having difficulties imagining something more slick.



Ralf







__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070501/676165ad/attachment.htm>

From mitch at niftyneato.com  Wed May  2 07:15:34 2007
From: mitch at niftyneato.com (Mitch Jones)
Date: Tue, 01 May 2007 22:15:34 -0700
Subject: [C++-sig] Problem exposing container iterator with
 boost::iterator_adaptor
In-Reply-To: <7465b6170704271147m4c3a1crb5530c0713eece7c@mail.gmail.com>
Message-ID: <C25D6C86.3AF0C%mitch@niftyneato.com>

on 4/27/07 11:47 AM, Roman Yakovenko at roman.yakovenko at gmail.com scribbled:

> Your code looks correct, can you create small example, which reproduce
> the problem?

Well, the sample turned out to be 150 or so lines, and in the end I figured
out the problem as I was working with the sample.

> Also what happens if you replace
> .def("__iter__", range(&Container::begin, &Container::end)
> with
> .def("__iter__", range(&WidgetContainer::begin, &WidgetContainer::end)
> ?

This was a typo and should have read WidgetContainer all along.

The problem boiled down to how my iterator adaptor was declared which looked
something like 

class WidgetIterator : public iterator_fa?ade<WidgetIterator,
shared_ptr<Widget>, single_pass_traversal_tag>

What would happen is that dereference would get called and the current item
would get passed back, but if you notice, the definition of the iterator
doesn't define the reference type, it uses the default, which is absolutely
not what you want in this case.

In this case, dereference looks like

shared_ptr<Widget> & dereference() const { return current;}

The shared_ptr never gets properly copied causing premature destruction
which leads to the "pure virtual method called" error.

So, in answer to my question "Is the iterator_adaptor a red herring and the
real issue in Widget and its descendants and I've missed something obvious?"
The answer was no and yes.

-- 
Mitch Jones                                 mitch at niftyneato.com





From jonas at MIT.EDU  Wed May  2 14:25:58 2007
From: jonas at MIT.EDU (Eric Jonas)
Date: Wed, 02 May 2007 08:25:58 -0400
Subject: [C++-sig] more examples of automatic type conversion
Message-ID: <1178108758.13074.4.camel@convolution.mit.edu>

Hello! I'm using boost::python to try and interface numpy (that is, the
newest numpy, not numeric or numarray) ndarray objects to my code which
depends on boost::multi_array. I've written functions that take a
pyobject* and return a multiarray and vice versa. Now I'm at the point
of interfacing my c++ classes (which take multi_array references in
their constructors). Rather than adulterating their interfaces by adding
additional constructors with PyObject* arguments, is there some way of
telling boost::python to use my conversion functions automatically? 

Thanks, 
		...Eric Jonas




From seefeld at sympatico.ca  Wed May  2 14:33:21 2007
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Wed, 02 May 2007 08:33:21 -0400
Subject: [C++-sig] boost::python::eval
Message-ID: <46388511.40702@sympatico.ca>

David,

when looking at the Embedding Python tutorial, I realized that the
functionality provided by exec() and exec_file() isn't quite complete,
with respect to the C API's offering: Py_eval_input isn't covered.

The attached patch attempts to fix that, i.e. a new eval() function
is defined that returns the result of the evaluated expression passed
as argument.

I'd like to check this into HEAD, as well as the 1_34 branch, as soon
as the release is out (assuming that this is indeed to become the new
trunk), so it can make it into 1.35.

OK ?

Thanks,
		Stefan

-- 

      ...ich hab' noch einen Koffer in Berlin...
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070502/c7432cb2/attachment.txt>

From dave at boost-consulting.com  Wed May  2 15:02:48 2007
From: dave at boost-consulting.com (David Abrahams)
Date: Wed, 02 May 2007 09:02:48 -0400
Subject: [C++-sig] boost::python::eval
In-Reply-To: <46388511.40702@sympatico.ca> (Stefan Seefeld's message of "Wed\,
	02 May 2007 08\:33\:21 -0400")
References: <46388511.40702@sympatico.ca>
Message-ID: <87y7k7e5tz.fsf@grogan.peloton>


on Wed May 02 2007, Stefan Seefeld <seefeld-AT-sympatico.ca> wrote:

> David,
>
> when looking at the Embedding Python tutorial, I realized that the
> functionality provided by exec() and exec_file() isn't quite complete,
> with respect to the C API's offering: Py_eval_input isn't covered.
>
> The attached patch attempts to fix that, i.e. a new eval() function
> is defined that returns the result of the evaluated expression passed
> as argument.
>
> I'd like to check this into HEAD, as well as the 1_34 branch, as soon
> as the release is out (assuming that this is indeed to become the new
> trunk), so it can make it into 1.35.
>
> OK ?

Sure.

-- 
Dave Abrahams
Boost Consulting
http://www.boost-consulting.com

Don't Miss BoostCon 2007! ==> http://www.boostcon.com


From roman.yakovenko at gmail.com  Wed May  2 15:32:21 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 2 May 2007 16:32:21 +0300
Subject: [C++-sig] more examples of automatic type conversion
In-Reply-To: <1178108758.13074.4.camel@convolution.mit.edu>
References: <1178108758.13074.4.camel@convolution.mit.edu>
Message-ID: <7465b6170705020632g29b04f5eoce3608934f8265c7@mail.gmail.com>

On 5/2/07, Eric Jonas <jonas at mit.edu> wrote:
> Hello! I'm using boost::python to try and interface numpy (that is, the
> newest numpy, not numeric or numarray) ndarray objects to my code which
> depends on boost::multi_array. I've written functions that take a
> pyobject* and return a multiarray and vice versa. Now I'm at the point
> of interfacing my c++ classes (which take multi_array references in
> their constructors). Rather than adulterating their interfaces by adding
> additional constructors with PyObject* arguments, is there some way of
> telling boost::python to use my conversion functions automatically?

http://www.language-binding.net/pyplusplus/troubleshooting_guide/automatic_conversion/automatic_conversion.html

Let me know if this helped you.


-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From tim at klingt.org  Wed May  2 15:41:07 2007
From: tim at klingt.org (Tim Blechmann)
Date: Wed, 02 May 2007 15:41:07 +0200
Subject: [C++-sig] boost::python threading problem
Message-ID: <1178113267.7782.15.camel@localhost>

hi all,

trying to hunt a deadlock, i've figured out, that during the execution
of wrapped c++ functions, python global interpreter lock is not
released.
in my specific case it lead to this behaviour:
thread A (python) calls a wrapped c++ function, that locks mutex M
thread B (c++) locks mutex M before calling a python function.

from my understanding, this shouldn't be a problem, if the gil would be
released for the time of the c++ function call... while this can be done
from within the wrapped c++ function (via Py_BEGIN_ALLOW_THREADS and
Py_END_ALLOW_THREADS), i would be in favor of a solution, where i
wouldn't have to change the c++ code ...

is this a known issue? are there any plans to change this behavior?

thanks, tim

--
tim at klingt.org    ICQ: 96771783
http://tim.klingt.org

There's no such entity as "most people". These are generalities. All
generalities are meaningless. You've got to pin it down to a specific
person doing a specific thing at a specific time and space.
  William S. Burroughs
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070502/596b3bd9/attachment.pgp>

From ndbecker2 at gmail.com  Wed May  2 19:06:05 2007
From: ndbecker2 at gmail.com (Neal Becker)
Date: Wed, 02 May 2007 13:06:05 -0400
Subject: [C++-sig] more examples of automatic type conversion
References: <1178108758.13074.4.camel@convolution.mit.edu>
Message-ID: <f1agdt$mhq$1@sea.gmane.org>

Eric Jonas wrote:

> Hello! I'm using boost::python to try and interface numpy (that is, the
> newest numpy, not numeric or numarray) ndarray objects to my code which
> depends on boost::multi_array. I've written functions that take a
> pyobject* and return a multiarray and vice versa. Now I'm at the point
> of interfacing my c++ classes (which take multi_array references in
> their constructors). Rather than adulterating their interfaces by adding
> additional constructors with PyObject* arguments, is there some way of
> telling boost::python to use my conversion functions automatically?
> 
> Thanks,
> ...Eric Jonas

I'm very interested in interfacing various c++ array containers to numpy. 
Do you have any small examples?



From jonas at MIT.EDU  Wed May  2 19:31:44 2007
From: jonas at MIT.EDU (Eric Jonas)
Date: Wed, 02 May 2007 13:31:44 -0400
Subject: [C++-sig] more examples of automatic type conversion
In-Reply-To: <f1agdt$mhq$1@sea.gmane.org>
References: <1178108758.13074.4.camel@convolution.mit.edu>
	<f1agdt$mhq$1@sea.gmane.org>
Message-ID: <1178127105.13074.18.camel@convolution.mit.edu>

> I'm very interested in interfacing various c++ array containers to numpy. 
> Do you have any small examples?
> 

I think we all are, and the existing numpy/numeric interface is getting
a bit crufty, and the new numpy type module is somewhat complex :) one
thing I'm struggling with is my poor understanding of python's internal
memory management and the like. I'll hopefully have the automatic type
conversion working in a day or two, and I'll be sure to post my results
to the list. I'm not trying to do anything fancy like the
numeric/numarray stuff currently in boost; I just want functions that
take in const boost::multi_array<T, n> & to handle the conversion for
me. I want to mostly use my python as a scripting interface on top of
some number-crunching C++ monte-carlo code. 

Interesting questions include:   What sort of containers do we use?
Boost::array? Boost::multiarray? The ublas containers? all of them? Just
std::vector< std::vector (...) How do we handle the static typing
problems?

I also can't promise that I've written anything other than an app that
leaks memory like a submarine with a screen door. C++'s notion of
"ownership" is so simple compared to python. 
				...Eric





From p.jaroszynski at gmail.com  Wed May  2 22:56:51 2007
From: p.jaroszynski at gmail.com (Piotr Jaroszynski)
Date: Wed, 2 May 2007 22:56:51 +0200
Subject: [C++-sig] Policies again
Message-ID: <200705022256.52150.p.jaroszynski@gmail.com>

Hello,

I am still not the best at polcies...

Let me show you my current problem:

#include <boost/python.hpp>
namespace bp = boost::python;

struct Blah;
struct Foo {
    Blah * b;
    Foo(Blah * blah) : b(blah) { }
};
struct Blah {
    Foo * f;
    Blah() : f(new Foo(this)) { }
};

BOOST_PYTHON_MODULE(policies)
{
    bp::class_<Foo> f("Foo", bp::no_init);
    bp::class_<Blah> b("Blah");
    b.add_property("f", bp::make_getter(&Blah::f, policy??));
}

Which policy should I use in such case?

-- 
Best Regards,
Piotr Jaroszynski


From ndbecker2 at gmail.com  Thu May  3 01:51:31 2007
From: ndbecker2 at gmail.com (Neal Becker)
Date: Wed, 02 May 2007 19:51:31 -0400
Subject: [C++-sig] more examples of automatic type conversion
References: <1178108758.13074.4.camel@convolution.mit.edu>
	<f1agdt$mhq$1@sea.gmane.org>
	<1178127105.13074.18.camel@convolution.mit.edu>
Message-ID: <f1b863$o1q$1@sea.gmane.org>

Eric Jonas wrote:

>> I'm very interested in interfacing various c++ array containers to numpy.
>> Do you have any small examples?
>> 
> 
> I think we all are, and the existing numpy/numeric interface is getting
> a bit crufty, and the new numpy type module is somewhat complex :) one
> thing I'm struggling with is my poor understanding of python's internal
> memory management and the like. I'll hopefully have the automatic type
> conversion working in a day or two, and I'll be sure to post my results
> to the list. I'm not trying to do anything fancy like the
> numeric/numarray stuff currently in boost; I just want functions that
> take in const boost::multi_array<T, n> & to handle the conversion for
> me. I want to mostly use my python as a scripting interface on top of
> some number-crunching C++ monte-carlo code.
> 
> Interesting questions include:   What sort of containers do we use?
> Boost::array? Boost::multiarray? The ublas containers? all of them? Just
> std::vector< std::vector (...) How do we handle the static typing
> problems?
> 

I've used:
ublas::{vector,matrix}
boost::multi_array
blitz::array

Recently, I've been playing with blitz/python and I am starting to really
like it.  OTOH, it doesn't seem to be a very active project and my
questions have gone largely unanswered.  OTOH, the code is far easier to
comprehend (for me, at least) than the first 2 above, so I can pretty much
answer my own questions.

I like that blitz::array handles more than just 1/2 dimensions, and I am
starting to _really_ like that it does ref counting.  That means I can do:

u = array1_int ((1,2,3,4))
u2 = u[::2]

and now u2 is a proxy for u.  This behavior is (I believe) like numpy.  Or:

u = array1_complex ((1,2,3,4))
u2 = real (u)
u2[:] = 7

Again, this will write to the real part of u.  And, I haven't needed to even
think about ownership.



From ndbecker2 at gmail.com  Thu May  3 01:56:23 2007
From: ndbecker2 at gmail.com (Neal Becker)
Date: Wed, 02 May 2007 19:56:23 -0400
Subject: [C++-sig] more examples of automatic type conversion
References: <1178108758.13074.4.camel@convolution.mit.edu>
	<f1agdt$mhq$1@sea.gmane.org>
	<1178127105.13074.18.camel@convolution.mit.edu>
Message-ID: <f1b8f7$o1q$2@sea.gmane.org>

Eric Jonas wrote:

>> I'm very interested in interfacing various c++ array containers to numpy.
>> Do you have any small examples?
>> 
> 
> I think we all are, and the existing numpy/numeric interface is getting
> a bit crufty, and the new numpy type module is somewhat complex :) one
> thing I'm struggling with is my poor understanding of python's internal
> memory management and the like. I'll hopefully have the automatic type
> conversion working in a day or two, and I'll be sure to post my results
> to the list. I'm not trying to do anything fancy like the
> numeric/numarray stuff currently in boost; I just want functions that
> take in const boost::multi_array<T, n> & to handle the conversion for
> me. I want to mostly use my python as a scripting interface on top of
> some number-crunching C++ monte-carlo code.
> 
> Interesting questions include:   What sort of containers do we use?
> Boost::array? Boost::multiarray? The ublas containers? all of them? Just
> std::vector< std::vector (...) How do we handle the static typing
> problems?
> 

Also, I am quite interested in writing generic algorithms that will work
with any of the above.  I have mentioned this on boost-devel.  Some people
suggested looking at gil.  I don't really think that is what I want,
though.  Seems rather specific to image processing and image formats.

I'm still interested in a generalization of boost::range to multi-dimensions
for this purpose.



From fullung at gmail.com  Thu May  3 10:33:51 2007
From: fullung at gmail.com (Albert Strasheim)
Date: Thu, 3 May 2007 10:33:51 +0200
Subject: [C++-sig] more examples of automatic type conversion
In-Reply-To: <7465b6170705020632g29b04f5eoce3608934f8265c7@mail.gmail.com>
References: <1178108758.13074.4.camel@convolution.mit.edu>
	<7465b6170705020632g29b04f5eoce3608934f8265c7@mail.gmail.com>
Message-ID: <20070503083351.GA5134@dogbert.sdsl.sun.ac.za>

On Wed, 02 May 2007, Roman Yakovenko wrote:

> On 5/2/07, Eric Jonas <jonas at mit.edu> wrote:
> > Hello! I'm using boost::python to try and interface numpy (that is, the
> > newest numpy, not numeric or numarray) ndarray objects to my code which
> > depends on boost::multi_array. I've written functions that take a
> > pyobject* and return a multiarray and vice versa. Now I'm at the point
> > of interfacing my c++ classes (which take multi_array references in
> > their constructors). Rather than adulterating their interfaces by adding
> > additional constructors with PyObject* arguments, is there some way of
> > telling boost::python to use my conversion functions automatically?
> 
> http://www.language-binding.net/pyplusplus/troubleshooting_guide/automatic_conversion/automatic_conversion.html
> 
> Let me know if this helped you.

As luck would have it, I also started playing with this idea yesterday 
and I found the above-mentioned page to help me along. I'm mostly 
interested in wrapping an interface written in terms of uBLAS vectors, 
matrices and vector-of-vectors.

One issue I've run into with uBLAS is that there doesn't seem to be an 
array adaptor (the A in matrix<T, L, A>) that can be used to wrap 
existing memory without copying at some point. Writing an array adaptor 
like this shouldn't be too hard though.

Another issue is that you have to be very careful with the 
vector/matrix type you use, since converting a NumPy array to
matrix<T, L, adapter_that_doesnt_copy<T> > and passing that to a 
function that expects matrix<T> doesn't work (strange crashes, probably 
due to the slight difference in size between the objects, due to the 
different array adaptors being used).

Anyway, I've more or less got pass-by-value and return-by-value working 
using Roman's example, but I can't quite figure out how to make 
Boost.Python convert pointer and reference arguments or return values 
into PyObject*'s using the same scheme. Should this just work if I have 
my automatic conversion set up correctly, or is there additional magic 
that needs to be added? Again, it would be *very* useful if copying 
could be avoided here.

Any hints would be appreciated.

Cheers,

Albert


From roman.yakovenko at gmail.com  Thu May  3 13:08:41 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 3 May 2007 14:08:41 +0300
Subject: [C++-sig] more examples of automatic type conversion
In-Reply-To: <20070503083351.GA5134@dogbert.sdsl.sun.ac.za>
References: <1178108758.13074.4.camel@convolution.mit.edu>
	<7465b6170705020632g29b04f5eoce3608934f8265c7@mail.gmail.com>
	<20070503083351.GA5134@dogbert.sdsl.sun.ac.za>
Message-ID: <7465b6170705030408geea84b4ib2c4e52e4b957a07@mail.gmail.com>

On 5/3/07, Albert Strasheim <fullung at gmail.com> wrote:
>
> On Wed, 02 May 2007, Roman Yakovenko wrote:
>
> > On 5/2/07, Eric Jonas <jonas at mit.edu> wrote:
> > > Hello! I'm using boost::python to try and interface numpy (that is,
> the
> > > newest numpy, not numeric or numarray) ndarray objects to my code
> which
> > > depends on boost::multi_array. I've written functions that take a
> > > pyobject* and return a multiarray and vice versa. Now I'm at the point
> > > of interfacing my c++ classes (which take multi_array references in
> > > their constructors). Rather than adulterating their interfaces by
> adding
> > > additional constructors with PyObject* arguments, is there some way of
> > > telling boost::python to use my conversion functions automatically?
> >
> >
> http://www.language-binding.net/pyplusplus/troubleshooting_guide/automatic_conversion/automatic_conversion.html
> >
> > Let me know if this helped you.
>
> As luck would have it, I also started playing with this idea yesterday
> and I found the above-mentioned page to help me along. I'm mostly
> interested in wrapping an interface written in terms of uBLAS vectors,
> matrices and vector-of-vectors.


:-), good

One issue I've run into with uBLAS is that there doesn't seem to be an
> array adaptor (the A in matrix<T, L, A>) that can be used to wrap
> existing memory without copying at some point. Writing an array adaptor
> like this shouldn't be too hard though.


You don't have, use indexing suite v2:
http://language-binding.net/pyplusplus/documentation/containers.html
And example of usage:
http://language-binding.net/pyplusplus/documentation/functions/call_policies.html#return-range


> Another issue is that you have to be very careful with the
> vector/matrix type you use, since converting a NumPy array to
> matrix<T, L, adapter_that_doesnt_copy<T> > and passing that to a
> function that expects matrix<T> doesn't work (strange crashes, probably
> due to the slight difference in size between the objects, due to the
> different array adaptors being used).
>
> Anyway, I've more or less got pass-by-value and return-by-value working
> using Roman's example, but I can't quite figure out how to make
> Boost.Python convert pointer and reference arguments or return values
> into PyObject*'s using the same scheme. Should this just work if I have
> my automatic conversion set up correctly, or is there additional magic
> that needs to be added? Again, it would be *very* useful if copying
> could be avoided here.


Post small example, may be I will be able to help

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070503/e2a363ad/attachment.htm>

From j.reid at mail.cryst.bbk.ac.uk  Thu May  3 13:30:45 2007
From: j.reid at mail.cryst.bbk.ac.uk (John Reid)
Date: Thu, 03 May 2007 12:30:45 +0100
Subject: [C++-sig] more examples of automatic type conversion
In-Reply-To: <f1agdt$mhq$1@sea.gmane.org>
References: <1178108758.13074.4.camel@convolution.mit.edu>
	<f1agdt$mhq$1@sea.gmane.org>
Message-ID: <f1ch6l$55b$1@sea.gmane.org>

Neal Becker wrote:
> I'm very interested in interfacing various c++ array containers to numpy. 
> Do you have any small examples?

I have what works for me although it is not extensively tested. It is 
not fancy as it copies data on conversion to/from python but I don't 
need any more than that. I've just implemented conversion for 
boost::multi_array but other arrays should be straightforward. Also it 
is a shameless hack of Roman's tuple example.

No doubt it could be improved considerably but I hope it is useful.

Here is an example of using it to export a property of a simple struct:

struct numpy_test
{
	typedef boost::multi_array< double, 2 > double_array;
	double_array a;
	void print_shape() const { std::cout << a.shape()[ 0 ] << "," << 
a.shape()[ 1 ] << "\n"; }
	void print_first() const { std::cout << a[ 0 ][ 0 ] << "\n"; }
};

void
export_model()
{
	numpy_multi_array_converter< numpy_test::double_array 
 >::register_to_and_from_python();
	class_< numpy_test >( "numpy_test" )
		.def( "print_shape", &numpy_test::print_shape )
		.def( "print_first", &numpy_test::print_first )
		.add_property(
			"array",
			make_getter( &numpy_test::a, return_value_policy< return_by_value >() ),
			numpy_multi_array_converter< numpy_test::double_array 
 >::set_member_with_resize< numpy_test, &numpy_test::a > )
		;
}


and here is the code to do the conversion:



namespace detail {

template< typename T >
struct get_dtype
{
	static const char * name() { throw std::logic_error( "get_dtype not 
specialised for this type" ); }
};

#define DECLARE_DTYPE_FOR( type, dtype ) template< > struct get_dtype< 
type > { static const char * name() { return dtype; } };
DECLARE_DTYPE_FOR( double, "float64" )
DECLARE_DTYPE_FOR( float, "float32" )
DECLARE_DTYPE_FOR( int, "int32" )
DECLARE_DTYPE_FOR( unsigned, "uint32" )
DECLARE_DTYPE_FOR( long, "int64" )
DECLARE_DTYPE_FOR( unsigned long, "uint64" )
}

// namespace impl

template< typename MultiArrayType >
struct numpy_multi_array_converter
{
     typedef MultiArrayType multi_array_t;
     typedef std::vector< std::size_t > shape_t;

	static void register_to_and_from_python()
	{
		register_from_python();
		register_to_python();
	}

	static void register_to_python()
	{
		boost::python::to_python_converter< multi_array_t, 
numpy_multi_array_converter< multi_array_t > >();
	}

	static void register_from_python()
	{
		boost::python::converter::registry::push_back(
			&numpy_multi_array_converter< multi_array_t >::convertible,
			&numpy_multi_array_converter< multi_array_t >::construct,
			boost::python::type_id< multi_array_t >()
		);
	}

     static
	void *
	convertible( PyObject * obj )
	{
		using namespace boost::python;
		try
		{
			shape_t shape;
			get_shape( object( handle<>( borrowed( obj ) ) ), shape );
			if( multi_array_t::dimensionality != shape.size() ) return 0;
		}
		catch( ... )
		{
			return 0;
		}

		return obj;
     }

	template<
		typename C,
		multi_array_t C::* pm
	>
	static
	void
	set_member_with_resize( C & c, const multi_array_t & a )
	{
		std::vector< unsigned > extents;
		for( unsigned dim = 0; a.num_dimensions() != dim; ++dim ) 
extents.push_back( a.shape()[ dim ] );
		(c.*pm).resize( extents );
		(c.*pm) = a;
	}

     static
	void
     construct(
		PyObject* obj,
		boost::python::converter::rvalue_from_python_stage1_data* data )
	{
		using namespace boost::python;

		//get the storage
         typedef converter::rvalue_from_python_storage< multi_array_t > 
storage_t;
         storage_t * the_storage = reinterpret_cast< storage_t * >( data );
         void * memory_chunk = the_storage->storage.bytes;

		//new placement
		object py_obj( handle<>( borrowed( obj ) ) );
		shape_t shape;
		get_shape( py_obj, shape );
		multi_array_t * a = new (memory_chunk) multi_array_t( shape );

		//extract each element from numpy array and put in c array
		index i( a->num_dimensions(), 0 );
		do
		{
			using boost::python::list;
			using boost::python::tuple;

			list numpy_index;
			for( unsigned dim = 0; a->num_dimensions() != dim; ++dim ) { 
numpy_index.append( i[ dim ] ); }
			( *a )( i ) = extract< multi_array_t::element >( py_obj[ tuple( 
numpy_index ) ] );
		}
		while( increment_index( i, *a ) );

         data->convertible = memory_chunk;
     }

	static
	PyObject *
	convert( const multi_array_t & c_array )
	{
		using namespace boost::python;

		object numpy = object( handle<>( ::PyImport_Import( object( "numpy" 
).ptr() ) ) );
		if( ! numpy  ) throw std::logic_error( "Could not import numpy" );
		object array_function = numpy.attr( "empty" );
		if( ! array_function  ) throw std::logic_error( "Could not find array 
function" );

		//create a numpy array to put it in
		boost::python::list extents;
		for( unsigned dim = 0; c_array.num_dimensions() != dim; ++dim ) 
extents.append( c_array.shape()[ dim ] );

		object result(
			array_function(
				extents,
				numpy.attr( "dtype" )( detail::get_dtype< multi_array_t::element 
 >::name() ) ) );

		//copy the elements	
		index i( c_array.num_dimensions(), 0 );
		do
		{
			boost::python::list numpy_index;
			for( unsigned dim = 0; c_array.num_dimensions() != dim; ++dim ) { 
numpy_index.append( i[dim] ); }
			result[ tuple( numpy_index ) ] = c_array( i );
		}
		while( increment_index( i, c_array ) );

		return incref( result.ptr() );
	}

protected:
	static
	void
	get_shape( boost::python::object & obj, shape_t & shape )
	{
		using namespace boost::python;
		shape.clear();
		object py_shape = obj.attr( "shape" );
		const std::size_t N = len( py_shape );
		for( std::size_t i = 0; N != i; ++i ) shape.push_back( extract< 
std::size_t >( py_shape[ i ] ) );
	}

	typedef std::vector< typename multi_array_t::index > index; /**< To 
iterate over entries in num_dimensions independent fashion. */

	/**< Iterates over entries in num_dimensions independent fashion. */
	static
	bool
	increment_index( index & i, const multi_array_t & c_array )
	{
		for( unsigned dim = 0; i.size() != dim; ++dim )
		{
			++i[dim];
			if( i[dim] != c_array.shape()[dim] ) return true;
			else i[dim] = 0;
		}
		return false;
	}

};




From ndbecker2 at gmail.com  Thu May  3 14:43:13 2007
From: ndbecker2 at gmail.com (Neal Becker)
Date: Thu, 03 May 2007 08:43:13 -0400
Subject: [C++-sig] more examples of automatic type conversion
References: <1178108758.13074.4.camel@convolution.mit.edu>
	<f1agdt$mhq$1@sea.gmane.org>
	<1178127105.13074.18.camel@convolution.mit.edu>
Message-ID: <f1cld1$kad$1@sea.gmane.org>

I've placed some basic test code for blitz++ here:

https://nbecker.dyndns.org/blitz.hpp

Next I will add all the arithmetic operators.  That won't take long, since
I've already done this for ublas, and stlsoft::fixed_array.



From p.jaroszynski at gmail.com  Thu May  3 16:30:07 2007
From: p.jaroszynski at gmail.com (Piotr Jaroszynski)
Date: Thu, 3 May 2007 16:30:07 +0200
Subject: [C++-sig] Policies again
In-Reply-To: <200705022256.52150.p.jaroszynski@gmail.com>
References: <200705022256.52150.p.jaroszynski@gmail.com>
Message-ID: <200705031630.08286.p.jaroszynski@gmail.com>

On Wednesday 02 of May 2007 22:56:51 Piotr Jaroszynski wrote:
> Let me show you my current problem:
To be even more precise:

#include <boost/python.hpp>
namespace bp = boost::python;

struct Blah;

struct Foo {
    Blah * b;
    Foo(Blah * blah) : b(blah) { }
};

struct Blah_f {
    Foo * f;
    Blah_f(Blah * blah) : f(new Foo(blah)) { }
};

struct Blah : Blah_f {
    Blah() : Blah_f(this) { }
};

BOOST_PYTHON_MODULE(policies)
{
    bp::register_ptr_to_python<Foo*>();
    bp::class_<Foo, boost::noncopyable> f("Foo", bp::no_init);
    bp::class_<Blah> b("Blah");
    b.add_property("f", bp::make_getter(&Blah::f,
	 bp::return_internal_reference<>()));
}

You will probably ask where is the big difference... So when using the 
return_internal_reference policy, which is probably wrong anyway, I get:
Traceback (most recent call last):
  File "./test.py", line 5, in ?
    print Blah().f
Boost.Python.ArgumentError: Python argument types in
    None.None(Blah)
did not match C++ signature:
    None(Blah_f {lvalue})

I think I should use one of the with_custodian_and_ward variations, but I 
couldn't make any of them to even compile.

-- 
Best Regards,
Piotr Jaroszynski


From jwiegley at gmail.com  Fri May  4 15:39:13 2007
From: jwiegley at gmail.com (John Wiegley)
Date: Fri, 4 May 2007 07:39:13 -0600
Subject: [C++-sig] Making boost::optional<T> passable to/from Python
Message-ID: <5452682E-E1E0-4DAE-BFA1-DDC3B8C9F09A@gmail.com>

The following code will make it possible to use boost::optional<T>  
very easily with Boost.Python.  For every time you expect to be used  
optionally, you must call the following during module initialization:

   python_optional<some_type_t>();

You can now accept and return boost::optional<some_type_t>().  In  
both directions, None will reflect an uninitialized value, while data  
of the correct type represents an initialized value.  Needless to  
say, valid conversions must already exist for some_type_t, otherwise  
an error is generated claiming that the optional type cannot be  
converted.

John Wiegley

template <typename T, typename TfromPy>
struct object_from_python
{
   object_from_python() {
     boost::python::converter::registry::push_back
       (&TfromPy::convertible, &TfromPy::construct,
        boost::python::type_id<T>());
   }
};

template <typename T, typename TtoPy, typename TfromPy>
struct register_python_conversion
{
   register_python_conversion() {
     boost::python::to_python_converter<T, TtoPy>();
     object_from_python<T, TfromPy>();
   }
};

template <typename T>
struct python_optional : public boost::noncopyable
{
   struct optional_to_python
   {
     static PyObject * convert(const boost::optional<T>& value)
     {
       return (value ? boost::python::to_python_value<T>()(*value) :
	              boost::python::detail::none());
     }
   };

   struct optional_from_python
   {
     static void * convertible(PyObject * source)
     {
       using namespace boost::python::converter;

       if (source == Py_None)
	return source;

       const registration& converters(registered<T>::converters);

       if (implicit_rvalue_convertible_from_python(source,  
converters)) {
	rvalue_from_python_stage1_data data =
	  rvalue_from_python_stage1(source, converters);
	return rvalue_from_python_stage2(source, data, converters);
       }
       return NULL;
     }

     static void construct(PyObject * source,
			  boost::python::converter::rvalue_from_python_stage1_data * data)
     {
       using namespace boost::python::converter;

       void * const storage = ((rvalue_from_python_storage<T> *)  
data)->storage.bytes;

       if (data->convertible == source)	    // == None
	new (storage) boost::optional<T>(); // A Boost uninitialized value
       else
	new (storage) boost::optional<T>(*static_cast<T *>(data->convertible));

       data->convertible = storage;
     }
   };

   explicit python_optional() {
     register_python_conversion<boost::optional<T>,
       optional_to_python, optional_from_python>();
   }
};



From gbrunick at andrew.cmu.edu  Fri May  4 18:02:46 2007
From: gbrunick at andrew.cmu.edu (Gerard Brunick)
Date: Fri, 04 May 2007 12:02:46 -0400
Subject: [C++-sig] Passing wrapped objects between extensions via python.
Message-ID: <463B5926.2030302@andrew.cmu.edu>

I'd like to be able to add new C++ code "on the fly."  I have the "main" 
extension module:
*************************
#include <iostream>
#include <boost/python.hpp>
#include <boost/function.hpp>

void do_function(boost::function<double (double)> func, double arg) {
    std::cout << "Doing function: " << func(arg) << std::endl;
}

BOOST_PYTHON_MODULE(main) {
    boost::python::class_<boost::function<double (double)> 
 >("double_double");
    boost::python::def("do_function", do_function);
}
***************************

Later I decide to add functionality (during a session in which the 
python is running
and the main extension module has been loaded) so I compile a new 
extension module:

***************************
#include <boost/python.hpp>
#include <boost/function.hpp>

double triple(double d) { return d * 3; }

BOOST_PYTHON_MODULE(helper) {
    boost::python::scope().attr("triple")
        =  boost::function<double (double)>(triple);
}
****************************

This all seems to work fine.  In fact I have

*********
 >>> import main, helper
 >>> main.do_function(helper.triple, 3)
Doing function: 9
**********

but

****************
 >>> import helper, main
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: No to_python (by-value) converter found for C++ type: class 
boost::function<double __cdec
l(double),class std::allocator<void> >
*****************

which I take to be a good thing, because it means that helper actually 
finds the to_python
converter in main.  The thing is, I'm a little uneasy about all of this 
because I feel that
I should have to do something to make the helper.pyd link again the 
main.pyd.  Is it really this
easy, or am I doing something that is unsafe?

Thanks,
Gerard


From roman.yakovenko at gmail.com  Fri May  4 21:23:30 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 4 May 2007 22:23:30 +0300
Subject: [C++-sig] Policies again
In-Reply-To: <200705031630.08286.p.jaroszynski@gmail.com>
References: <200705022256.52150.p.jaroszynski@gmail.com>
	<200705031630.08286.p.jaroszynski@gmail.com>
Message-ID: <7465b6170705041223g624c4c2dp16d0905df21d5ac6@mail.gmail.com>

On 5/3/07, Piotr Jaroszynski <p.jaroszynski at gmail.com> wrote:
> On Wednesday 02 of May 2007 22:56:51 Piotr Jaroszynski wrote:
> > Let me show you my current problem:
> To be even more precise:
>
> #include <boost/python.hpp>
> namespace bp = boost::python;
>
> struct Blah;
>
> struct Foo {
>     Blah * b;
>     Foo(Blah * blah) : b(blah) { }
> };
>
> struct Blah_f {
>     Foo * f;
>     Blah_f(Blah * blah) : f(new Foo(blah)) { }
> };
>
> struct Blah : Blah_f {
>     Blah() : Blah_f(this) { }
> };
>
> BOOST_PYTHON_MODULE(policies)
> {
>     bp::register_ptr_to_python<Foo*>();
>     bp::class_<Foo, boost::noncopyable> f("Foo", bp::no_init);
>     bp::class_<Blah> b("Blah");
>     b.add_property("f", bp::make_getter(&Blah::f,
>          bp::return_internal_reference<>()));
> }
>
> You will probably ask where is the big difference... So when using the
> return_internal_reference policy, which is probably wrong anyway, I get:
> Traceback (most recent call last):
>   File "./test.py", line 5, in ?
>     print Blah().f
> Boost.Python.ArgumentError: Python argument types in
>     None.None(Blah)
> did not match C++ signature:
>     None(Blah_f {lvalue})
>
> I think I should use one of the with_custodian_and_ward variations, but I
> couldn't make any of them to even compile.

No,  return_internal_reference should be fine. I don't remember why,
but Py++ generates a wrapper class for you use-case and adds accessor
function. Take a look on member variables tester. I attached the
generated code.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: member_variables.cpp
Type: text/x-c++src
Size: 20869 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070504/6baf2626/attachment.cpp>

From p.jaroszynski at gmail.com  Sat May  5 12:30:30 2007
From: p.jaroszynski at gmail.com (Piotr Jaroszynski)
Date: Sat, 5 May 2007 12:30:30 +0200
Subject: [C++-sig] Policies again
In-Reply-To: <7465b6170705041223g624c4c2dp16d0905df21d5ac6@mail.gmail.com>
References: <200705022256.52150.p.jaroszynski@gmail.com>
	<200705031630.08286.p.jaroszynski@gmail.com>
	<7465b6170705041223g624c4c2dp16d0905df21d5ac6@mail.gmail.com>
Message-ID: <200705051230.31231.p.jaroszynski@gmail.com>

On Friday 04 of May 2007 21:23:30 Roman Yakovenko wrote:
> No,  return_internal_reference should be fine. I don't remember why,
> but Py++ generates a wrapper class for you use-case and adds accessor
> function. Take a look on member variables tester. I attached the
> generated code.

Thanks a lot!
Making accessors functions did the trick. I wonder whether it can be 
considered a bug in the make_getter as using:

> ? ? b.add_property("f", bp::make_getter(&Blah::f,
> ? ? ? ? ?bp::return_internal_reference<>()));

results in:
> Traceback (most recent call last):
> ? File "./test.py", line 5, in ?
> ? ? print Blah().f
> Boost.Python.ArgumentError: Python argument types in
> ? ? None.None(Blah)
> did not match C++ signature:
> ? ? None(Blah_f {lvalue})

-- 
Best Regards,
Piotr Jaroszynski


From s.ramacher at gmx.at  Sat May  5 15:03:26 2007
From: s.ramacher at gmx.at (Sebastian Ramacher)
Date: Sat, 05 May 2007 15:03:26 +0200
Subject: [C++-sig] [Boost.Python] inconsistent use of function pointer
	typedef
In-Reply-To: <87hcqwkbyp.fsf@grogan.peloton>
References: <f0smha$gms$1@sea.gmane.org> <87hcqwkbyp.fsf@grogan.peloton>
Message-ID: <f1hvb5$7k1$1@sea.gmane.org>

David Abrahams wrote:
> 
> This is just a warning fix, right?
> 

Actually not. At least I haven't seen a warning here (msvc-8.0).
I just stumbled accross it while using boost::function for
convertible_function and got some unresolved symbol errors.



From tgbrooks at gmail.com  Sat May  5 20:46:21 2007
From: tgbrooks at gmail.com (Tom Brooks)
Date: Sat, 5 May 2007 14:46:21 -0400
Subject: [C++-sig] Passing inherited classes back to C++
Message-ID: <112c5ffc0705051146h5c980458pf411338a3413f3be@mail.gmail.com>

I currently have these C++ classes:

class Renderable
{
public:
    Renderable();
    virtual ~Renderable();
    virtual void Draw() = 0;
};

class Quad : public Renderable
{
public:
    Quad(float x, float y, float h, float w);
    ~Quad();
    void Draw();
};

And I need to expose them to Python. Then, I need to be able to make one in
Python, and pass it back to a C++ function  as a Renderable*.

I made this wrapper class for Renderable:

struct RenderableWrap : Renderable, wrapper<Renderable>
{
public:
    void Draw()
    {
        this->get_override("Draw")();
    }
};

And I have this Boost.Python code for the Python side:

    class_<RenderableWrap, boost::noncopyable>("Renderable")
        .def("Draw",pure_virtual(&Renderable::Draw))
    ;

    class_<Quad, bases<RenderableWrap> >("Quad",init<float, float, float,
float>())
        .def("Draw",&Quad::Draw)
    ;

But when I pass in a Quad() to my C++ function, I get an error that it can't
convert from a Quad to a Renderable*. I have next to no clue how the base
class should have been done, since the tutorial and anything else I can find
online don't go over what to do for the base class (when there are virtual
functions, at least). For example, I don't know if I inherit from
Renderable, or RenderableWrap (or both?). I don't know whether the base
class needs a wrapper itself (and would it if it had a virtual function?).

I'd appreciate any help!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070505/74843b3c/attachment.htm>

From roman.yakovenko at gmail.com  Sat May  5 20:55:42 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sat, 5 May 2007 21:55:42 +0300
Subject: [C++-sig] Passing inherited classes back to C++
In-Reply-To: <112c5ffc0705051146h5c980458pf411338a3413f3be@mail.gmail.com>
References: <112c5ffc0705051146h5c980458pf411338a3413f3be@mail.gmail.com>
Message-ID: <7465b6170705051155s3c3b2b4r692875e03c610779@mail.gmail.com>

On 5/5/07, Tom Brooks <tgbrooks at gmail.com> wrote:
> I currently have these C++ classes:
>
> class Renderable
> {
> public:
>     Renderable();
>     virtual ~Renderable();
>     virtual void Draw() = 0;
> };
>
> class Quad : public Renderable
> {
> public:
>      Quad(float x, float y, float h, float w);
>     ~Quad();
>     void Draw();
> };
>
> And I need to expose them to Python. Then, I need to be able to make one in
> Python, and pass it back to a C++ function  as a Renderable*.
>
> I made this wrapper class for Renderable:
>
> struct RenderableWrap : Renderable, wrapper<Renderable>
> {
> public:
>     void Draw()
>     {
>         this->get_override("Draw")();
>     }
> };
>
> And I have this Boost.Python code for the Python side:
>
>     class_<RenderableWrap,
> boost::noncopyable>("Renderable")
>         .def("Draw",pure_virtual(&Renderable::Draw))
>     ;
>
>     class_<Quad, bases<RenderableWrap> >("Quad",init<float, float, float,

Quad derives from Renderable, not RenderableWrap

> float>())
>         .def("Draw",&Quad::Draw)
>     ;
>
> But when I pass in a Quad() to my C++ function, I get an error that it can't
> convert from a Quad to a Renderable*. I have next to no clue how the base
> class should have been done, since the tutorial and anything else I can find
> online don't go over what to do for the base class (when there are virtual
> functions, at least). For example, I don't know if I inherit from
> Renderable, or RenderableWrap (or both?). I don't know whether the base
> class needs a wrapper itself (and would it if it had a virtual function?).
>
> I'd appreciate any help!

Next code was generated by Py++ GUI
(http://language-binding.net/pyplusplus/documentation/tutorials/pyplusplus_gui.html)

#include "boost/python.hpp"

#include "1.h"

namespace bp = boost::python;

struct Renderable_wrapper : Renderable, bp::wrapper< Renderable > {

    Renderable_wrapper( )
    : Renderable( )
      , bp::wrapper< Renderable >(){
        // null constructor

    }

    virtual void Draw(  ){
        bp::override func_Draw = this->get_override( "Draw" );
        func_Draw(  );
    }

};

struct Quad_wrapper : Quad, bp::wrapper< Quad > {

    Quad_wrapper(Quad const & arg )
    : Quad( arg )
      , bp::wrapper< Quad >(){
        // copy constructor

    }

    Quad_wrapper(float x, float y, float h, float w )
    : Quad( x, y, h, w )
      , bp::wrapper< Quad >(){
        // constructor

    }

    virtual void Draw(  ) {
        if( bp::override func_Draw = this->get_override( "Draw" ) )
            func_Draw(  );
        else
            this->Quad::Draw(  );
    }


    void default_Draw(  ) {
        Quad::Draw( );
    }

};

BOOST_PYTHON_MODULE(pyplusplus){
    bp::class_< Renderable_wrapper, boost::noncopyable >( "Renderable" )
        .def( bp::init< >() )
        .def(
            "Draw"
            , bp::pure_virtual( &::Renderable::Draw ) );

    bp::class_< Quad_wrapper, bp::bases< Renderable > >( "Quad",
bp::init< float, float, float, float >(( bp::arg("x"), bp::arg("y"),
bp::arg("h"), bp::arg("w") )) )
        .def(
            "Draw"
            , &::Quad::Draw
            , &Quad_wrapper::default_Draw );
}


-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From tgbrooks at gmail.com  Sat May  5 22:15:40 2007
From: tgbrooks at gmail.com (Tom Brooks)
Date: Sat, 5 May 2007 16:15:40 -0400
Subject: [C++-sig] Passing inherited classes back to C++
In-Reply-To: <7465b6170705051155s3c3b2b4r692875e03c610779@mail.gmail.com>
References: <112c5ffc0705051146h5c980458pf411338a3413f3be@mail.gmail.com>
	<7465b6170705051155s3c3b2b4r692875e03c610779@mail.gmail.com>
Message-ID: <112c5ffc0705051315h48b41513sa8e7952c7abcdbb9@mail.gmail.com>

Thank you for the fast reply. I would have tried Py++ myself, but I'm unable
to install it. I get an error that the .NET framework SDK must be installed
in order to build Python extensions when I try to install GCC_XML. I tried
to download and install the .NET SDK, but I'm sure if I've really got all of
it. I didn't select much for the installation process. Though it is
off-topic, any help with this would be nice too. I use Visual Studios C++
express on Windows XP. I have Python 2.4 and the Windows Platform SDK.

As for the code, I haven't looked at it in a lot of detail yet, but a simple
copy-and-paste gives me a run-time error that I'm calling a pure-virtual
function when I call Draw() from the passed in Renderable*. Any more
thoughts?

On 5/5/07, Roman Yakovenko <roman.yakovenko at gmail.com> wrote:
>
> On 5/5/07, Tom Brooks <tgbrooks at gmail.com> wrote:
> > I currently have these C++ classes:
> >
> > class Renderable
> > {
> > public:
> >     Renderable();
> >     virtual ~Renderable();
> >     virtual void Draw() = 0;
> > };
> >
> > class Quad : public Renderable
> > {
> > public:
> >      Quad(float x, float y, float h, float w);
> >     ~Quad();
> >     void Draw();
> > };
> >
> > And I need to expose them to Python. Then, I need to be able to make one
> in
> > Python, and pass it back to a C++ function  as a Renderable*.
> >
> > I made this wrapper class for Renderable:
> >
> > struct RenderableWrap : Renderable, wrapper<Renderable>
> > {
> > public:
> >     void Draw()
> >     {
> >         this->get_override("Draw")();
> >     }
> > };
> >
> > And I have this Boost.Python code for the Python side:
> >
> >     class_<RenderableWrap,
> > boost::noncopyable>("Renderable")
> >         .def("Draw",pure_virtual(&Renderable::Draw))
> >     ;
> >
> >     class_<Quad, bases<RenderableWrap> >("Quad",init<float, float,
> float,
>
> Quad derives from Renderable, not RenderableWrap
>
> > float>())
> >         .def("Draw",&Quad::Draw)
> >     ;
> >
> > But when I pass in a Quad() to my C++ function, I get an error that it
> can't
> > convert from a Quad to a Renderable*. I have next to no clue how the
> base
> > class should have been done, since the tutorial and anything else I can
> find
> > online don't go over what to do for the base class (when there are
> virtual
> > functions, at least). For example, I don't know if I inherit from
> > Renderable, or RenderableWrap (or both?). I don't know whether the base
> > class needs a wrapper itself (and would it if it had a virtual
> function?).
> >
> > I'd appreciate any help!
>
> Next code was generated by Py++ GUI
> (
> http://language-binding.net/pyplusplus/documentation/tutorials/pyplusplus_gui.html
> )
>
> #include "boost/python.hpp"
>
> #include "1.h"
>
> namespace bp = boost::python;
>
> struct Renderable_wrapper : Renderable, bp::wrapper< Renderable > {
>
>     Renderable_wrapper( )
>     : Renderable( )
>       , bp::wrapper< Renderable >(){
>         // null constructor
>
>     }
>
>     virtual void Draw(  ){
>         bp::override func_Draw = this->get_override( "Draw" );
>         func_Draw(  );
>     }
>
> };
>
> struct Quad_wrapper : Quad, bp::wrapper< Quad > {
>
>     Quad_wrapper(Quad const & arg )
>     : Quad( arg )
>       , bp::wrapper< Quad >(){
>         // copy constructor
>
>     }
>
>     Quad_wrapper(float x, float y, float h, float w )
>     : Quad( x, y, h, w )
>       , bp::wrapper< Quad >(){
>         // constructor
>
>     }
>
>     virtual void Draw(  ) {
>         if( bp::override func_Draw = this->get_override( "Draw" ) )
>             func_Draw(  );
>         else
>             this->Quad::Draw(  );
>     }
>
>
>     void default_Draw(  ) {
>         Quad::Draw( );
>     }
>
> };
>
> BOOST_PYTHON_MODULE(pyplusplus){
>     bp::class_< Renderable_wrapper, boost::noncopyable >( "Renderable" )
>         .def( bp::init< >() )
>         .def(
>             "Draw"
>             , bp::pure_virtual( &::Renderable::Draw ) );
>
>     bp::class_< Quad_wrapper, bp::bases< Renderable > >( "Quad",
> bp::init< float, float, float, float >(( bp::arg("x"), bp::arg("y"),
> bp::arg("h"), bp::arg("w") )) )
>         .def(
>             "Draw"
>             , &::Quad::Draw
>             , &Quad_wrapper::default_Draw );
> }
>
>
> --
> Roman Yakovenko
> C++ Python language binding
> http://www.language-binding.net/
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070505/6067fa59/attachment.htm>

From roman.yakovenko at gmail.com  Sun May  6 07:12:47 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sun, 6 May 2007 08:12:47 +0300
Subject: [C++-sig] Passing inherited classes back to C++
In-Reply-To: <112c5ffc0705051315h48b41513sa8e7952c7abcdbb9@mail.gmail.com>
References: <112c5ffc0705051146h5c980458pf411338a3413f3be@mail.gmail.com>
	<7465b6170705051155s3c3b2b4r692875e03c610779@mail.gmail.com>
	<112c5ffc0705051315h48b41513sa8e7952c7abcdbb9@mail.gmail.com>
Message-ID: <7465b6170705052212p7ac075q4ea2c042577002a0@mail.gmail.com>

On 5/5/07, Tom Brooks <tgbrooks at gmail.com> wrote:
> Thank you for the fast reply. I would have tried Py++ myself, but I'm unable
> to install it. I get an error that the .NET framework SDK must be installed
> in order to build Python extensions when I try to install GCC_XML. I tried
> to download and install the .NET SDK, but I'm sure if I've really got all of
> it. I didn't select much for the installation process. Though it is
> off-topic, any help with this would be nice too. I use Visual Studios C++
> express on Windows XP. I have Python 2.4 and the Windows Platform SDK.
>
> As for the code, I haven't looked at it in a lot of detail yet, but a simple
> copy-and-paste gives me a run-time error that I'm calling a pure-virtual
> function when I call Draw() from the passed in Renderable*. Any more
> thoughts?

Please provide a complete example

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From tgbrooks at gmail.com  Sun May  6 11:12:43 2007
From: tgbrooks at gmail.com (Tom Brooks)
Date: Sun, 6 May 2007 05:12:43 -0400
Subject: [C++-sig] Passing inherited classes back to C++
In-Reply-To: <7465b6170705052212p7ac075q4ea2c042577002a0@mail.gmail.com>
References: <112c5ffc0705051146h5c980458pf411338a3413f3be@mail.gmail.com>
	<7465b6170705051155s3c3b2b4r692875e03c610779@mail.gmail.com>
	<112c5ffc0705051315h48b41513sa8e7952c7abcdbb9@mail.gmail.com>
	<7465b6170705052212p7ac075q4ea2c042577002a0@mail.gmail.com>
Message-ID: <112c5ffc0705060212q519a6593x665d3a2293de8a29@mail.gmail.com>

I run this python code:

quad = Renderer.Quad(250, 250, 64,64)
c = Renderer.Color()
Renderer.AddRenderable(quad) #Removing this line makes it not crash

AddRenderable does:
renderables.push_back(renderable);

And later that frame, this is done:
    for(std::list<Renderable*>::iterator itr = renderables.begin(); itr !=
renderables.end(); itr++)
    {
        Renderable* renderable = *itr;
        renderable->Draw(); //It crashes here! With error "R6025: pure
virtual function call".
    }

I use your Py++ code (only thing changed is it's namespace, which is now
Renderer), along with:

    def("AddRenderable",&AddRenderable);

in the Renderer namespace.

My C++ code runs just fine if I don't use anything from Python.

There, I think that's complete.

On 5/6/07, Roman Yakovenko <roman.yakovenko at gmail.com> wrote:
>
> On 5/5/07, Tom Brooks <tgbrooks at gmail.com> wrote:
> > Thank you for the fast reply. I would have tried Py++ myself, but I'm
> unable
> > to install it. I get an error that the .NET framework SDK must be
> installed
> > in order to build Python extensions when I try to install GCC_XML. I
> tried
> > to download and install the .NET SDK, but I'm sure if I've really got
> all of
> > it. I didn't select much for the installation process. Though it is
> > off-topic, any help with this would be nice too. I use Visual Studios
> C++
> > express on Windows XP. I have Python 2.4 and the Windows Platform SDK.
> >
> > As for the code, I haven't looked at it in a lot of detail yet, but a
> simple
> > copy-and-paste gives me a run-time error that I'm calling a pure-virtual
> > function when I call Draw() from the passed in Renderable*. Any more
> > thoughts?
>
> Please provide a complete example
>
> --
> Roman Yakovenko
> C++ Python language binding
> http://www.language-binding.net/
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070506/35206635/attachment.htm>

From roman.yakovenko at gmail.com  Sun May  6 11:22:09 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sun, 6 May 2007 12:22:09 +0300
Subject: [C++-sig] Passing inherited classes back to C++
In-Reply-To: <112c5ffc0705060212q519a6593x665d3a2293de8a29@mail.gmail.com>
References: <112c5ffc0705051146h5c980458pf411338a3413f3be@mail.gmail.com>
	<7465b6170705051155s3c3b2b4r692875e03c610779@mail.gmail.com>
	<112c5ffc0705051315h48b41513sa8e7952c7abcdbb9@mail.gmail.com>
	<7465b6170705052212p7ac075q4ea2c042577002a0@mail.gmail.com>
	<112c5ffc0705060212q519a6593x665d3a2293de8a29@mail.gmail.com>
Message-ID: <7465b6170705060222x67a4b5dvd0e525c3c0deb9f1@mail.gmail.com>

On 5/6/07, Tom Brooks <tgbrooks at gmail.com> wrote:
> I run this python code:
>
> quad = Renderer.Quad(250, 250, 64,64)
> c = Renderer.Color()
> Renderer.AddRenderable(quad) #Removing this line makes it not crash
>
> AddRenderable does:
> renderables.push_back(renderable);
>
> And later that frame, this is done:
>     for(std::list<Renderable*>::iterator itr =
> renderables.begin(); itr != renderables.end(); itr++)
>     {
>         Renderable* renderable = *itr;
>         renderable->Draw(); //It crashes here! With error "R6025: pure
> virtual function call".
>     }
>
> I use your Py++ code (only thing changed is it's namespace, which is now
> Renderer), along with:
>
>     def("AddRenderable",&AddRenderable);
>
> in the Renderer namespace.
>
> My C++ code runs just fine if I don't use anything from Python.
>
> There, I think that's complete.

This is not a complete example, but it allows to make a guess - memory
management.
To prove this, try to hold a reference to "quad" variable somewhere in
Python and run the code.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From tgbrooks at gmail.com  Sun May  6 11:37:34 2007
From: tgbrooks at gmail.com (Tom Brooks)
Date: Sun, 6 May 2007 05:37:34 -0400
Subject: [C++-sig] Passing inherited classes back to C++
In-Reply-To: <7465b6170705060222x67a4b5dvd0e525c3c0deb9f1@mail.gmail.com>
References: <112c5ffc0705051146h5c980458pf411338a3413f3be@mail.gmail.com>
	<7465b6170705051155s3c3b2b4r692875e03c610779@mail.gmail.com>
	<112c5ffc0705051315h48b41513sa8e7952c7abcdbb9@mail.gmail.com>
	<7465b6170705052212p7ac075q4ea2c042577002a0@mail.gmail.com>
	<112c5ffc0705060212q519a6593x665d3a2293de8a29@mail.gmail.com>
	<7465b6170705060222x67a4b5dvd0e525c3c0deb9f1@mail.gmail.com>
Message-ID: <112c5ffc0705060237l430a861emd259215a5cee0a62@mail.gmail.com>

Wow, that was a great guess! I kept a reference to the quad in Python, and
it all went fine. The FAQ has a part about just this situation, so I think
I'm all set now. Thanks a bunch for the help!

BTW, for future reference, what else would you have wanted to see?

On 5/6/07, Roman Yakovenko <roman.yakovenko at gmail.com> wrote:
>
> On 5/6/07, Tom Brooks <tgbrooks at gmail.com> wrote:
> > I run this python code:
> >
> > quad = Renderer.Quad(250, 250, 64,64)
> > c = Renderer.Color()
> > Renderer.AddRenderable(quad) #Removing this line makes it not crash
> >
> > AddRenderable does:
> > renderables.push_back(renderable);
> >
> > And later that frame, this is done:
> >     for(std::list<Renderable*>::iterator itr =
> > renderables.begin(); itr != renderables.end(); itr++)
> >     {
> >         Renderable* renderable = *itr;
> >         renderable->Draw(); //It crashes here! With error "R6025: pure
> > virtual function call".
> >     }
> >
> > I use your Py++ code (only thing changed is it's namespace, which is now
> > Renderer), along with:
> >
> >     def("AddRenderable",&AddRenderable);
> >
> > in the Renderer namespace.
> >
> > My C++ code runs just fine if I don't use anything from Python.
> >
> > There, I think that's complete.
>
> This is not a complete example, but it allows to make a guess - memory
> management.
> To prove this, try to hold a reference to "quad" variable somewhere in
> Python and run the code.
>
> --
> Roman Yakovenko
> C++ Python language binding
> http://www.language-binding.net/
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070506/2847fda0/attachment.htm>

From ngoodspeed at solidworks.com  Sun May  6 19:03:37 2007
From: ngoodspeed at solidworks.com (Nat Goodspeed)
Date: Sun, 6 May 2007 13:03:37 -0400
Subject: [C++-sig] Passing inherited classes back to C++
In-Reply-To: <112c5ffc0705060237l430a861emd259215a5cee0a62@mail.gmail.com>
References: <112c5ffc0705051146h5c980458pf411338a3413f3be@mail.gmail.com><7465b6170705051155s3c3b2b4r692875e03c610779@mail.gmail.com><112c5ffc0705051315h48b41513sa8e7952c7abcdbb9@mail.gmail.com><7465b6170705052212p7ac075q4ea2c042577002a0@mail.gmail.com><112c5ffc0705060212q519a6593x665d3a2293de8a29@mail.gmail.com><7465b6170705060222x67a4b5dvd0e525c3c0deb9f1@mail.gmail.com>
	<112c5ffc0705060237l430a861emd259215a5cee0a62@mail.gmail.com>
Message-ID: <94F7A8DD4408D8499C6812FE42E2D4B7011FCC7A@corp-mail4.solidworks.swk>

________________________________________
From: c++-sig-bounces at python.org [mailto:c++-sig-bounces at python.org] On
Behalf Of Tom Brooks
Sent: Sunday, May 06, 2007 5:38 AM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Passing inherited classes back to C++

BTW, for future reference, what else would you have wanted to see? 

[Nat] Roman, forgive me for putting words in your mouth.

Tom, I think he's requesting that you attach complete, minimal source
files to your mail message that he could save to some work directory,
build and execute to see (and possibly debug) the behavior you describe.

Providing snippets of the crucial code in your mail message is helpful,
and sometimes it's enough. But other times someone wants to help, can't
guess the problem from your description and doesn't have time to write
the boilerplate code into which to paste your snippets. In such cases,
attaching complete (minimal) source files encourages them to test your
program anyway.

For other kinds of questions, e.g. compilation errors, sometimes the
failing code snippet is fine by itself and the problem has to do with
missing (or badly-ordered) #include statements. Unless both parties see
the whole source file, you can get replies of the form "It works for me"
even while you're saying, "But it's still broken for me."


From roman.yakovenko at gmail.com  Sun May  6 20:09:50 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sun, 6 May 2007 21:09:50 +0300
Subject: [C++-sig] Passing inherited classes back to C++
In-Reply-To: <94F7A8DD4408D8499C6812FE42E2D4B7011FCC7A@corp-mail4.solidworks.swk>
References: <112c5ffc0705051146h5c980458pf411338a3413f3be@mail.gmail.com>
	<7465b6170705051155s3c3b2b4r692875e03c610779@mail.gmail.com>
	<112c5ffc0705051315h48b41513sa8e7952c7abcdbb9@mail.gmail.com>
	<7465b6170705052212p7ac075q4ea2c042577002a0@mail.gmail.com>
	<112c5ffc0705060212q519a6593x665d3a2293de8a29@mail.gmail.com>
	<7465b6170705060222x67a4b5dvd0e525c3c0deb9f1@mail.gmail.com>
	<112c5ffc0705060237l430a861emd259215a5cee0a62@mail.gmail.com>
	<94F7A8DD4408D8499C6812FE42E2D4B7011FCC7A@corp-mail4.solidworks.swk>
Message-ID: <7465b6170705061109ja36fec3m7717c1d92a6f72b9@mail.gmail.com>

On 5/6/07, Nat Goodspeed <ngoodspeed at solidworks.com> wrote:
> ________________________________________
> From: c++-sig-bounces at python.org [mailto:c++-sig-bounces at python.org] On
> Behalf Of Tom Brooks
> Sent: Sunday, May 06, 2007 5:38 AM
> To: Development of Python/C++ integration
> Subject: Re: [C++-sig] Passing inherited classes back to C++
>
> BTW, for future reference, what else would you have wanted to see?
>
> [Nat] Roman, forgive me for putting words in your mouth.
>
> Tom, I think he's requesting that you attach complete, minimal source
> files to your mail message that he could save to some work directory,
> build and execute to see (and possibly debug) the behavior you describe.

No, I don't like to help people, brrr :-).

 The purpose of small and complete example to force people to help
them self. In most cases we deal with complex libraries and
dependencies - extracting the exact use-case that fails, it is a half
way to solution.

Also I agree with you it is much easier to see the problem when you
see small and complete source code( C++ and Python ).

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From fullung at gmail.com  Sun May  6 20:44:55 2007
From: fullung at gmail.com (Albert Strasheim)
Date: Sun, 6 May 2007 20:44:55 +0200
Subject: [C++-sig] more examples of automatic type conversion
In-Reply-To: <7465b6170705030408geea84b4ib2c4e52e4b957a07@mail.gmail.com>
References: <1178108758.13074.4.camel@convolution.mit.edu>
	<7465b6170705020632g29b04f5eoce3608934f8265c7@mail.gmail.com>
	<20070503083351.GA5134@dogbert.sdsl.sun.ac.za>
	<7465b6170705030408geea84b4ib2c4e52e4b957a07@mail.gmail.com>
Message-ID: <20070506184455.GA15374@dogbert.sdsl.sun.ac.za>

Hello all

On Thu, 03 May 2007, Roman Yakovenko wrote:

> On 5/3/07, Albert Strasheim <fullung at gmail.com> wrote:
<snip>
> >Anyway, I've more or less got pass-by-value and return-by-value working
> >using Roman's example, but I can't quite figure out how to make
> >Boost.Python convert pointer and reference arguments or return values
> >into PyObject*'s using the same scheme. Should this just work if I have
> >my automatic conversion set up correctly, or is there additional magic
> >that needs to be added? Again, it would be *very* useful if copying
> >could be avoided here.
> 
> 
> Post small example, may be I will be able to help

To recap, the idea is to wrap classes with member functions that take 
uBLAS vectors or matrices by value, (const) reference and (const) 
pointer. 

However, on the Python side one always wants to pass in a NumPy array 
which should be converted on the fly. Using an appropriate array 
adaptor, one should be able to construct a uBLAS matrix that reused the 
NumPy array's underlying data buffer without copying (except in special 
cases which I won't go into now).

Similarly, for (const) pointer or (const) reference returns, one wants 
to wrap a new NumPy array around the return value so that the Python 
side always sees a NumPy array. Again, copying should be avoided.
For the return_internal_reference case, it would be nice if the object 
"owning" the matrix being returned could be kept alive until all NumPy 
arrays that refer to the data are deleted.

In the manage_new_object case, the array_adaptor in the array being 
returned can be told to give up the ownership of the data and then 
uBLAS matrix can be deallocated immediately (effectively only 
destroying a few bits and pieces).

I expect some complications here if you have a NumPy array taking over 
ownership of a data buffer allocated in C++ code, so I guess one would 
have to look at an array_adaptor that allocates the matrix's storage as 
a NumPy array which can be "extracted" when returning to Python.

Here's some code that shows what I'm trying to using tuples instead of 
NumPy arrays and array_t instead of uBLAS matrices.

#include <boost/python.hpp>
#include <iostream>

namespace py = boost::python;

struct array_t {
    int rows;
    int cols;
    void* data;
    array_t(int rows_, int cols_) : rows(rows_), cols(cols_), data(0){}
    array_t() : rows(0), cols(0), data(0){}
};

struct array_converter {
    static void register_to_and_from_python() {
        register_from_python();
        register_to_python();
    }

    static void register_to_python() {
        py::to_python_converter<array_t, array_converter>();
    }

    static void register_from_python() {
        py::converter::registry::push_back(
            &array_converter::convertible,
            &array_converter::construct,
            py::type_id<array_t>());
    }

    static void* convertible(PyObject* obj) {
        return PyTuple_Check(obj) != 0 ? obj : 0;
    }

    static void construct(PyObject* obj, py::converter::rvalue_from_python_stage1_data* data) {
        // get the storage
        typedef py::converter::rvalue_from_python_storage<array_t> storage_t;
        storage_t* the_storage = reinterpret_cast<storage_t*>(data);
        void* memory_chunk = the_storage->storage.bytes;
        // placement new
        int rows = PyInt_AsLong(PyTuple_GET_ITEM(obj, 0));
        int cols = PyInt_AsLong(PyTuple_GET_ITEM(obj, 1));
        // XXX get data pointer here
        array_t* a = new (memory_chunk) array_t(rows, cols);
        data->convertible = memory_chunk;
    }

    static PyObject* convert(const array_t& a) {
        return py::incref(py::make_tuple(a.rows, a.cols).ptr());
    }
};

struct array_test {
    array_test() : data_(3, 4){}
    array_t value_return() const {
        return data_;
    }
    array_t const* const_pointer_return() const {
        return &data_;
    }
    array_t* pointer_return() {
        return &data_;
    }
    array_t const& const_ref_return() const {
        return data_;
    }
    array_t& ref_return() {
        return data_;
    }
    void value_arg(array_t a) {
        print(a);
    }
    void pointer_arg(array_t* a) {
        print(*a);
    }
    void const_pointer_arg(array_t const* a){}
    void ref_arg(array_t& a){}
    void const_ref_arg(array_t const& a){}
private:
    void print(array_t const& a) {
        std::cout << "(" << a.rows << ", " << a.cols << ")" << std::endl;
    }
    array_t data_;
};

BOOST_PYTHON_MODULE(pyublas)
{
	array_converter::register_to_and_from_python();
	py::class_<array_test, boost::noncopyable>("array_test")
            .def("value_return", &array_test::value_return)
            .def("const_pointer_return", &array_test::const_pointer_return, py::return_internal_reference<>())
#if 0
            .def("pointer_return", &array_test::pointer_return)
            .def("const_ref_return", &array_test::const_ref_return)
            .def("ref_return", &array_test::ref_return)
#endif
            .def("value_arg", &array_test::value_arg)
            .def("pointer_arg", &array_test::pointer_arg)
#if 0
            .def("const_pointer_arg", &array_test::const_pointer_arg)
            .def("ref_arg", &array_test::ref_arg)
            .def("const_ref_arg", &array_test::const_ref_arg)
#endif
        ;
}

And the Python code

import pyublas
print pyublas
print dir(pyublas)
at = pyublas.array_test()
print at
print at.value_return()
# XXX this doesn't work yet
print at.const_pointer_return()
a = (10, 20)
at.value_arg(a)
# XXX this doesn't work yet
at.pointer_arg(a)

The call to const_pointer_return raises the following error:

Traceback (most recent call last):
  File "test_pyublas.py", line 24, in ?
    print at.const_pointer_return()
TypeError: No Python class registered for C++ class struct array_t

The call to pointer_arg raises the following error:

Traceback (most recent call last):
  File "test_pyublas.py", line 28, in ?
    at.pointer_arg(a)
Boost.Python.ArgumentError: Python argument types in
    array_test.pointer_arg(array_test, tuple)
did not match C++ signature:
    pointer_arg(struct array_test {lvalue}, struct array_t *)

In both cases, I would like conversions to happen from the PyObjects to 
the expected C++ data structures. Is it possible to convince 
Boost.Python to do this?

Thanks for any feedback.

Cheers,

Albert


From fullung at gmail.com  Mon May  7 12:10:35 2007
From: fullung at gmail.com (Albert Strasheim)
Date: Mon, 7 May 2007 12:10:35 +0200
Subject: [C++-sig] more examples of automatic type conversion
In-Reply-To: <20070506184455.GA15374@dogbert.sdsl.sun.ac.za>
References: <1178108758.13074.4.camel@convolution.mit.edu>
	<7465b6170705020632g29b04f5eoce3608934f8265c7@mail.gmail.com>
	<20070503083351.GA5134@dogbert.sdsl.sun.ac.za>
	<7465b6170705030408geea84b4ib2c4e52e4b957a07@mail.gmail.com>
	<20070506184455.GA15374@dogbert.sdsl.sun.ac.za>
Message-ID: <20070507101035.GA22718@dogbert.sdsl.sun.ac.za>

Hello all

On Sun, 06 May 2007, Albert Strasheim wrote:

> Hello all
> 
> On Thu, 03 May 2007, Roman Yakovenko wrote:
> 
> > On 5/3/07, Albert Strasheim <fullung at gmail.com> wrote:
> <snip>
> > >Anyway, I've more or less got pass-by-value and return-by-value working
> > >using Roman's example, but I can't quite figure out how to make
> > >Boost.Python convert pointer and reference arguments or return values
> > >into PyObject*'s using the same scheme. Should this just work if I have
> > >my automatic conversion set up correctly, or is there additional magic
> > >that needs to be added? Again, it would be *very* useful if copying
> > >could be avoided here.
> > 
> > 
> > Post small example, may be I will be able to help

> <snip>

Looks like SWIG typemaps do exactly what I want in this case.

Cheers,

Albert


From dave at boost-consulting.com  Mon May  7 21:24:22 2007
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 07 May 2007 15:24:22 -0400
Subject: [C++-sig] [Boost.Python] inconsistent use of function pointer
	typedef
References: <f0smha$gms$1@sea.gmane.org> <87hcqwkbyp.fsf@grogan.peloton>
	<f1hvb5$7k1$1@sea.gmane.org>
Message-ID: <878xc0sahl.fsf@grogan.peloton>


on Sat May 05 2007, Sebastian Ramacher <s.ramacher-AT-gmx.at> wrote:

> David Abrahams wrote:
>> 
>> This is just a warning fix, right?
>> 
>
> Actually not. At least I haven't seen a warning here (msvc-8.0).
> I just stumbled accross it while using boost::function for
> convertible_function and got some unresolved symbol errors.

I'm sorry, I don't understand.  Is this a bugfix, or a stylistic
improvement, or something else?

-- 
Dave Abrahams
Boost Consulting
http://www.boost-consulting.com

Don't Miss BoostCon 2007! ==> http://www.boostcon.com



From roman.yakovenko at gmail.com  Tue May  8 06:19:47 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Tue, 8 May 2007 07:19:47 +0300
Subject: [C++-sig] [ANN]Py++ 0.9.0
In-Reply-To: <7465b6170705072119x410cb33cq50f88eb1baba11c6@mail.gmail.com>
References: <7465b6170705072112q639a50acwcda287e7b4ffb103@mail.gmail.com>
	<7465b6170705072119x410cb33cq50f88eb1baba11c6@mail.gmail.com>
Message-ID: <7465b6170705072119u429f9e82h153b309f3912241e@mail.gmail.com>

Hello!

I'm pleased to announce the 0.9.0 release of Py++.

What is Py++?
=============

Py++ is an object-oriented framework for creating a code generator for
Boost.Python library.

Where is Py++?
==============

Site: http://language-binding.net/pyplusplus/pyplusplus.html

Download: http://language-binding.net/pyplusplus/download.html

What's new?
===========

Features
--------

* Added exposing of copy constructor, ``operator=`` and ``operator<<``.
   * ``operator=`` is exposed under "assign" name
   * ``operator<<`` is exposed under "__str__" name

* Added new call policies:
    * as_tuple
    * custom_call_policies
    * return_range

* Added an initial support for multi-module development. Now you can mark your
  declarations as ``already_exposed`` and `Py++`_ will do the rest.

* `input_c_buffer`_ - new functions transformation, which allows to
pass a Python
  sequence to C++ function, instead of pair of arguments: pointer to
buffer and size.

* Added ability to control generated "include" directives. Now you can ask Py++
  to include a header file, when it generates code for some declaration.

* Code generation improvements: system header files (Boost.Python or
Py++ defined)
  will be included from the generated files only in case the generated code
  depends on them.

* Performance: Py++ runs 1.5 - 2 times faster.

* Py++ will generate documentation for automatically constructed properties.

* Added iteration functionality to Boost.Python Indexing Suite V2 ``std::map``
  and ``std::multimap`` containers.

--
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From bruce.who.hk at gmail.com  Tue May  8 14:20:54 2007
From: bruce.who.hk at gmail.com (Bruce Who)
Date: Tue, 8 May 2007 20:20:54 +0800
Subject: [C++-sig] Why this does not work- overriding C++ virtual functions
	in python script?
Message-ID: <c8dd1d2b0705080520i3b28bb57x785b99d121624508@mail.gmail.com>

Hi, all

I have managed to get a python wrapper for my Cpp modules, but there
is something wrong.

This is my cpp files:

// cmod.h file
#ifndef __CMOD_H
#define __CMOD_H

class EventHandler
{
public:
    virtual void OnEvent(){}
    virtual ~EventHandler(){}
};

void func(void);

void setEventHandler(EventHandler* pEH);

#endif

// cmod.cpp file
#include "cmod.h"

static EventHandler* s_pEH;

void func(void)
{
    if(s_pEH != NULL)
    {
        printf("call OnEvent");
        s_pEH->OnEvent();
    }
}

void setEventHandler(EventHandler* pEH)
{
    s_pEH = pEH;
}

// cmod.i file
%module cmod
%include "cmod.h"


%inline %{
#include "cmod.h"

%}


and I compile all these files with Visual Studio 2005:

swig -c++ -python cmod.i
cl -c cmod_wrap.cxx -ID:\program\Python25\include\
cl -c cmod.cpp
link /subsystem:windows /LIBPATH:D:\program\Python25\libs\
cmod_wrap.obj cmod.obj /dll /OUT:_cmod.pyd

and then we can get cmod.py and _cmod.pyd. I write a python script:

# testcmod.py file
import cmod

class EventHandler(cmod.EventHandler):
    def OnEvent(self):
        print 'called'

eh = EventHandler()
cmod.setEventHandler(eh)
cmod.func()


Once this script runs, it is supposed to print 'called' but it does
not! I guess that if we cannot override C++ virtual functions in
python script? Hope someone can help me. Any help would be
appreciated!


Bruce


From nitro at dr-code.org  Tue May  8 14:52:48 2007
From: nitro at dr-code.org (Nitro)
Date: Tue, 08 May 2007 14:52:48 +0200
Subject: [C++-sig] [Swig-user] Why this does not work- overriding C++
	virtual functions in python script?
In-Reply-To: <c8dd1d2b0705080520i3b28bb57x785b99d121624508@mail.gmail.com>
References: <c8dd1d2b0705080520i3b28bb57x785b99d121624508@mail.gmail.com>
Message-ID: <op.tr0d2aofz3pzfc@nitrogenycs2>

Am 08.05.2007, 14:20 Uhr, schrieb Bruce Who <bruce.who.hk at gmail.com>:

> Hi, all
>
> I have managed to get a python wrapper for my Cpp modules, but there
> is something wrong.
>
> This is my cpp files:
>
> // cmod.h file
> #ifndef __CMOD_H
> #define __CMOD_H
>
> class EventHandler
> {
> public:
>     virtual void OnEvent(){}
>     virtual ~EventHandler(){}
> };
>
> void func(void);
>
> void setEventHandler(EventHandler* pEH);
>
> #endif
>
> // cmod.cpp file
> #include "cmod.h"
>
> static EventHandler* s_pEH;
>
> void func(void)
> {
>     if(s_pEH != NULL)
>     {
>         printf("call OnEvent");
>         s_pEH->OnEvent();
>     }
> }
>
> void setEventHandler(EventHandler* pEH)
> {
>     s_pEH = pEH;
> }
>
> // cmod.i file
> %module cmod
> %include "cmod.h"
>
>
> %inline %{
> #include "cmod.h"
>
> %}
>
>
> and I compile all these files with Visual Studio 2005:
>
> swig -c++ -python cmod.i
> cl -c cmod_wrap.cxx -ID:\program\Python25\include\
> cl -c cmod.cpp
> link /subsystem:windows /LIBPATH:D:\program\Python25\libs\
> cmod_wrap.obj cmod.obj /dll /OUT:_cmod.pyd
>
> and then we can get cmod.py and _cmod.pyd. I write a python script:
>
> # testcmod.py file
> import cmod
>
> class EventHandler(cmod.EventHandler):
>     def OnEvent(self):
>         print 'called'
>
> eh = EventHandler()
> cmod.setEventHandler(eh)
> cmod.func()
>
>
> Once this script runs, it is supposed to print 'called' but it does
> not! I guess that if we cannot override C++ virtual functions in
> python script? Hope someone can help me. Any help would be
> appreciated!

Read the "directors" part of the manual.

-Matthias


From fullung at gmail.com  Tue May  8 23:45:43 2007
From: fullung at gmail.com (Albert Strasheim)
Date: Tue, 8 May 2007 23:45:43 +0200
Subject: [C++-sig] Automatic type conversion again
Message-ID: <20070508214542.GA1577@dogbert.sdsl.sun.ac.za>

Hello all

Referring to Roman's excellent guide here:

http://www.language-binding.net/pyplusplus/troubleshooting_guide/automatic_conversion/automatic_conversion.html

Is there any way to allow conversion of PyObject* when passing to 
functions that expect references or pointers? In Roman's code he 
explicitly checks that doing this throws a TypeError:

http://www.language-binding.net/pyplusplus/troubleshooting_guide/automatic_conversion/test.py.html

e.g.

self.failUnlessRaises( TypeError, tuples.test_triplet_ref_110, (1,1,0) )

I can do this with SWIG typemaps, but SWIG doesn't seem to have a 
built-in equivalent for return_internal_reference or 
with_custodian_and_ward_postcall, so I'd like to stick to using 
Boost.Python if at all possible.

Any advice would be appreciated. Thanks.

Cheers,

Albert


From bruce.who.hk at gmail.com  Wed May  9 03:13:18 2007
From: bruce.who.hk at gmail.com (Bruce Who)
Date: Wed, 9 May 2007 09:13:18 +0800
Subject: [C++-sig] [Swig-user] Why this does not work- overriding C++
	virtual functions in python script?
In-Reply-To: <op.tr0d2aofz3pzfc@nitrogenycs2>
References: <c8dd1d2b0705080520i3b28bb57x785b99d121624508@mail.gmail.com>
	<op.tr0d2aofz3pzfc@nitrogenycs2>
Message-ID: <c8dd1d2b0705081813g1867e1eel4b1e53eb96c48622@mail.gmail.com>

Hi, Nitro

Thanks for your clue. I've added "directors" to my .i file:

%module(directors="1") cmod
%feature("director");

but it seems that Visual Studio2005 doesnot support this, and it fails
to compile:

f:\swigproj\cmod_wrap.cxx(2733) : error C2039: 'unexpected_handler' :
is not a member of 'std'
f:\swigproj\cmod_wrap.cxx(2734) : error C2039: 'unexpected_handler' :
is not a member of 'std'

I look at the source code and find that it has something to do with
directors. Is there any workaround?

On 5/8/07, Nitro <nitro at dr-code.org> wrote:
> Am 08.05.2007, 14:20 Uhr, schrieb Bruce Who <bruce.who.hk at gmail.com>:
>
>>
>
> Read the "directors" part of the manual.
>
> -Matthias
>


From s.somani at fincad.com  Wed May  9 03:37:19 2007
From: s.somani at fincad.com (Sohail Somani)
Date: Tue, 8 May 2007 18:37:19 -0700
Subject: [C++-sig] [Swig-user] Why this does not work- overriding C++
	virtualfunctions in python script?
In-Reply-To: <c8dd1d2b0705081813g1867e1eel4b1e53eb96c48622@mail.gmail.com>
References: <c8dd1d2b0705080520i3b28bb57x785b99d121624508@mail.gmail.com><op.tr0d2aofz3pzfc@nitrogenycs2>
	<c8dd1d2b0705081813g1867e1eel4b1e53eb96c48622@mail.gmail.com>
Message-ID: <1C1EBEF8DBACDC439D038EA051674EC78A4B60@xbox.financialcad.com>

You probably need to #include <stdexcept> or <exception> in the
generated code. 

> -----Original Message-----
> From: swig-user-bounces at lists.sourceforge.net 
> [mailto:swig-user-bounces at lists.sourceforge.net] On Behalf Of 
> Bruce Who
> Sent: Tuesday, May 08, 2007 6:13 PM
> To: Nitro
> Cc: swig-user; c++-sig of python
> Subject: Re: [Swig-user] Why this does not work- overriding 
> C++ virtualfunctions in python script?
> 
> Hi, Nitro
> 
> Thanks for your clue. I've added "directors" to my .i file:
> 
> %module(directors="1") cmod
> %feature("director");
> 
> but it seems that Visual Studio2005 doesnot support this, and it fails
> to compile:
> 
> f:\swigproj\cmod_wrap.cxx(2733) : error C2039: 'unexpected_handler' :
> is not a member of 'std'
> f:\swigproj\cmod_wrap.cxx(2734) : error C2039: 'unexpected_handler' :
> is not a member of 'std'
> 
> I look at the source code and find that it has something to do with
> directors. Is there any workaround?
> 
> On 5/8/07, Nitro <nitro at dr-code.org> wrote:
> > Am 08.05.2007, 14:20 Uhr, schrieb Bruce Who 
> <bruce.who.hk at gmail.com>:
> >
> >>
> >
> > Read the "directors" part of the manual.
> >
> > -Matthias
> >
> 
> --------------------------------------------------------------
> -----------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> _______________________________________________
> Swig-user mailing list
> Swig-user at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/swig-user
> 


From bruce.who.hk at gmail.com  Wed May  9 04:13:39 2007
From: bruce.who.hk at gmail.com (Bruce Who)
Date: Wed, 9 May 2007 10:13:39 +0800
Subject: [C++-sig] [Swig-user] Why this does not work- overriding C++
	virtualfunctions in python script?
In-Reply-To: <1C1EBEF8DBACDC439D038EA051674EC78A4B60@xbox.financialcad.com>
References: <c8dd1d2b0705080520i3b28bb57x785b99d121624508@mail.gmail.com>
	<op.tr0d2aofz3pzfc@nitrogenycs2>
	<c8dd1d2b0705081813g1867e1eel4b1e53eb96c48622@mail.gmail.com>
	<1C1EBEF8DBACDC439D038EA051674EC78A4B60@xbox.financialcad.com>
Message-ID: <c8dd1d2b0705081913j5a417a53v734e0abe8c87cab3@mail.gmail.com>

I add #include<exception> in the generated .cxx file and removed std
namespace before exceptions and it compiles. I donnot know why. But my
code still doesnot work! My OnEvent() event is still not called. Does
that mean director doesnot work for me?

On 5/9/07, Sohail Somani <s.somani at fincad.com> wrote:
> You probably need to #include <stdexcept> or <exception> in the
> generated code.
>
> > -----Original Message-----
> > From: swig-user-bounces at lists.sourceforge.net
> > [mailto:swig-user-bounces at lists.sourceforge.net] On Behalf Of
> > Bruce Who
> > Sent: Tuesday, May 08, 2007 6:13 PM
> > To: Nitro
> > Cc: swig-user; c++-sig of python
> > Subject: Re: [Swig-user] Why this does not work- overriding
> > C++ virtualfunctions in python script?
> >
> > Hi, Nitro
> >
> > Thanks for your clue. I've added "directors" to my .i file:
> >
> > %module(directors="1") cmod
> > %feature("director");
> >
> > but it seems that Visual Studio2005 doesnot support this, and it fails
> > to compile:
> >
> > f:\swigproj\cmod_wrap.cxx(2733) : error C2039: 'unexpected_handler' :
> > is not a member of 'std'
> > f:\swigproj\cmod_wrap.cxx(2734) : error C2039: 'unexpected_handler' :
> > is not a member of 'std'
> >
> > I look at the source code and find that it has something to do with
> > directors. Is there any workaround?
> >
> > On 5/8/07, Nitro <nitro at dr-code.org> wrote:
> > > Am 08.05.2007, 14:20 Uhr, schrieb Bruce Who
> > <bruce.who.hk at gmail.com>:
> > >
> > >>
> > >
> > > Read the "directors" part of the manual.
> > >
> > > -Matthias
> > >
> >
> > --------------------------------------------------------------
> > -----------
> > This SF.net email is sponsored by DB2 Express
> > Download DB2 Express C - the FREE version of DB2 express and take
> > control of your XML. No limits. Just data. Click to get it now.
> > http://sourceforge.net/powerbar/db2/
> > _______________________________________________
> > Swig-user mailing list
> > Swig-user at lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/swig-user
> >
>

Bruce


From roman.yakovenko at gmail.com  Wed May  9 05:51:04 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 9 May 2007 06:51:04 +0300
Subject: [C++-sig] [Swig-user] Why this does not work- overriding C++
	virtualfunctions in python script?
In-Reply-To: <c8dd1d2b0705081913j5a417a53v734e0abe8c87cab3@mail.gmail.com>
References: <c8dd1d2b0705080520i3b28bb57x785b99d121624508@mail.gmail.com>
	<op.tr0d2aofz3pzfc@nitrogenycs2>
	<c8dd1d2b0705081813g1867e1eel4b1e53eb96c48622@mail.gmail.com>
	<1C1EBEF8DBACDC439D038EA051674EC78A4B60@xbox.financialcad.com>
	<c8dd1d2b0705081913j5a417a53v734e0abe8c87cab3@mail.gmail.com>
Message-ID: <7465b6170705082051k6f55ff0chd79bd2729b5c4d54@mail.gmail.com>

I have nothing again SWIG, but can you move this discussion to its mailing list?
It has nothing to do with Boost.Python.

Thanks

On 5/9/07, Bruce Who <bruce.who.hk at gmail.com> wrote:
> I add #include<exception> in the generated .cxx file and removed std
> namespace before exceptions and it compiles. I donnot know why. But my
> code still doesnot work! My OnEvent() event is still not called. Does
> that mean director doesnot work for me?
>
> On 5/9/07, Sohail Somani <s.somani at fincad.com> wrote:
> > You probably need to #include <stdexcept> or <exception> in the
> > generated code.
> >
> > > -----Original Message-----
> > > From: swig-user-bounces at lists.sourceforge.net
> > > [mailto:swig-user-bounces at lists.sourceforge.net] On Behalf Of
> > > Bruce Who
> > > Sent: Tuesday, May 08, 2007 6:13 PM
> > > To: Nitro
> > > Cc: swig-user; c++-sig of python
> > > Subject: Re: [Swig-user] Why this does not work- overriding
> > > C++ virtualfunctions in python script?
> > >
> > > Hi, Nitro
> > >
> > > Thanks for your clue. I've added "directors" to my .i file:
> > >
> > > %module(directors="1") cmod
> > > %feature("director");
> > >
> > > but it seems that Visual Studio2005 doesnot support this, and it fails
> > > to compile:
> > >
> > > f:\swigproj\cmod_wrap.cxx(2733) : error C2039: 'unexpected_handler' :
> > > is not a member of 'std'
> > > f:\swigproj\cmod_wrap.cxx(2734) : error C2039: 'unexpected_handler' :
> > > is not a member of 'std'
> > >
> > > I look at the source code and find that it has something to do with
> > > directors. Is there any workaround?
> > >
> > > On 5/8/07, Nitro <nitro at dr-code.org> wrote:
> > > > Am 08.05.2007, 14:20 Uhr, schrieb Bruce Who
> > > <bruce.who.hk at gmail.com>:
> > > >
> > > >>
> > > >
> > > > Read the "directors" part of the manual.
> > > >
> > > > -Matthias
> > > >
> > >
> > > --------------------------------------------------------------
> > > -----------
> > > This SF.net email is sponsored by DB2 Express
> > > Download DB2 Express C - the FREE version of DB2 express and take
> > > control of your XML. No limits. Just data. Click to get it now.
> > > http://sourceforge.net/powerbar/db2/
> > > _______________________________________________
> > > Swig-user mailing list
> > > Swig-user at lists.sourceforge.net
> > > https://lists.sourceforge.net/lists/listinfo/swig-user
> > >
> >
>
> Bruce
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>


-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Wed May  9 06:48:52 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 9 May 2007 07:48:52 +0300
Subject: [C++-sig] Automatic type conversion again
In-Reply-To: <20070508214542.GA1577@dogbert.sdsl.sun.ac.za>
References: <20070508214542.GA1577@dogbert.sdsl.sun.ac.za>
Message-ID: <7465b6170705082148t386f92bgacfe5731c4bcae65@mail.gmail.com>

On 5/9/07, Albert Strasheim <fullung at gmail.com> wrote:
> Hello all
>
> Referring to Roman's excellent guide here:
>
> http://www.language-binding.net/pyplusplus/troubleshooting_guide/automatic_conversion/automatic_conversion.html
>
> Is there any way to allow conversion of PyObject* when passing to
> functions that expect references or pointers? In Roman's code he
> explicitly checks that doing this throws a TypeError:
>
> http://www.language-binding.net/pyplusplus/troubleshooting_guide/automatic_conversion/test.py.html
>
> e.g.
>
> self.failUnlessRaises( TypeError, tuples.test_triplet_ref_110, (1,1,0) )
>
> I can do this with SWIG typemaps, but SWIG doesn't seem to have a
> built-in equivalent for return_internal_reference or
> with_custodian_and_ward_postcall, so I'd like to stick to using
> Boost.Python if at all possible.

Memory management is pretty good reason to stick with Boost.Python.

The short answer - you cannot, the long answer - I believe it is
possible to find acceptable solution to the problem ( may be not ).

I could be wrong! You cannot, because an automatic conversion creates
temporal object, so the changes you do to it in the function cannot be
propagated to Python.

There are few possible work around.

The first on is to use decorators.

Every relevant exported function should have a decorator, which will
normalize argument types.

Explanation: Lets say your function takes YourArray*, while user gave
you an instance of ItsArray type as argument. You can create new
function that will create new YourArray instance from ItsArray. You
can define it as:

boost::python::object Its2YourArrayConverter( boost::python::object );

Also you create new YourArray instance, it doesn't necessary means
that you copy array content. You can use call policies to bind between
objects ( return and argument ) life time.

Thus you don't use "automatic conversion".

Using Python magic you can apply this decorator on every relevant function.

HTH

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From neworderofjamie at gmail.com  Wed May  9 13:23:36 2007
From: neworderofjamie at gmail.com (James Knight)
Date: Wed, 9 May 2007 12:23:36 +0100
Subject: [C++-sig] Wrapping std::map using Boost::Python
Message-ID: <9fadca7e0705090423h4fdf20fcs84eee57bb090ca64@mail.gmail.com>

Hi,

I'm trying to wrap std::map using boost::python so my embedded python code
can access maps.  I have the following wrapper class:

//Map - takes e.g std::map<std::string,thing*> as MapType
template<class MapType>
struct MapItem
{
    //Typedefine the key and value types for ease
    typedef typename MapType::key_type KeyType;
    typedef typename MapType::mapped_type ValueType;

    static ValueType& Get(MapType & x,KeyType const& i)
    {
        if( x.find(i) != x.end() )
            return x[i];
        //Some kind of error!
    }
    static void Set(MapType & x,KeyType const& i,ValueType const& v)
    {
        x[i]=v; // use map autocreation feature
    }
    static void Del(MapType & x,KeyType const& i)
    {
        if( x.find(i) != x.end() )
            x.erase(i);

    }
}

And the following Boost::Python class definition to wrap a particular map:

[source lang="cpp"]
boost::python::class_<std::map<std::string,SubGrid> >("SubGridMap")
    .def("__len__",&std::map<std::string,SubGrid>::size)
    .def("clear",&std::map<std::string,SubGrid>::clear)
    .def("__getitem__",&MapItem<std::map<std::string,SubGrid>
>::Get,boost::python::return_value_policy<boost::python::copy_non_const_reference>())
    .def("__setitem__",&MapItem<std::map<std::string,SubGrid>
>::Set,boost::python::with_custodian_and_ward<1,2>()) // to let container
keep value
    .def("__delitem__",&MapItem<std::map<std::string,SubGrid> >::Del)
;

I really need iterator support and following the Python Wiki entry on STL
containers, have written a function like this to convert the std::map to a
std::list of tuples as boost::python::iterator doesn't seem to be compatible
with std::map (which seems a bit icky but makes sense):

static std::list<boost::tuple<KeyType,ValueType> > Items(MapType const& x)
{
    std::list<boost::tuple<KeyType,ValueType> > t;
    typename MapType::const_iterator it;

    for(it=x.begin(); it!=x.end(); ++it)
        t.push_back(boost::make_tuple(it->first,it->second));

    return t;
}


But the wiki entry irritatingly stops short of telling you how the hell you
wrap it into an __iter__ and the attachement to the wiki entry seems to be
empty...Has anyone done this or have any ideas?

Cheers

Jamie
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070509/de04026c/attachment.htm>

From seefeld at sympatico.ca  Wed May  9 14:32:53 2007
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Wed, 09 May 2007 08:32:53 -0400
Subject: [C++-sig] [Swig-user] Why this does not work- overriding C++
 virtualfunctions in python script?
In-Reply-To: <7465b6170705082051k6f55ff0chd79bd2729b5c4d54@mail.gmail.com>
References: <c8dd1d2b0705080520i3b28bb57x785b99d121624508@mail.gmail.com>	<op.tr0d2aofz3pzfc@nitrogenycs2>	<c8dd1d2b0705081813g1867e1eel4b1e53eb96c48622@mail.gmail.com>	<1C1EBEF8DBACDC439D038EA051674EC78A4B60@xbox.financialcad.com>	<c8dd1d2b0705081913j5a417a53v734e0abe8c87cab3@mail.gmail.com>
	<7465b6170705082051k6f55ff0chd79bd2729b5c4d54@mail.gmail.com>
Message-ID: <4641BF75.3020608@sympatico.ca>

Roman Yakovenko wrote:
> I have nothing again SWIG, but can you move this discussion to its mailing list?
> It has nothing to do with Boost.Python.

[While I agree with your point, I'd also like to mention that the topic of this mailing
 list claims to be "Development of Python/C++ integration", and thus it is worth pointing
 out that boost.python is not the only authoritative binding.]

FWIW,
		Stefan

-- 

      ...ich hab' noch einen Koffer in Berlin...


From wsf at fultondesigns.co.uk  Wed May  9 21:58:46 2007
From: wsf at fultondesigns.co.uk (William S Fulton)
Date: Wed, 09 May 2007 20:58:46 +0100
Subject: [C++-sig] [Swig-user] Why this does not work- overriding
 C++	virtualfunctions in python script?
In-Reply-To: <c8dd1d2b0705081913j5a417a53v734e0abe8c87cab3@mail.gmail.com>
References: <c8dd1d2b0705080520i3b28bb57x785b99d121624508@mail.gmail.com>	<op.tr0d2aofz3pzfc@nitrogenycs2>	<c8dd1d2b0705081813g1867e1eel4b1e53eb96c48622@mail.gmail.com>	<1C1EBEF8DBACDC439D038EA051674EC78A4B60@xbox.financialcad.com>
	<c8dd1d2b0705081913j5a417a53v734e0abe8c87cab3@mail.gmail.com>
Message-ID: <464227F6.4070701@fultondesigns.co.uk>

Take one of the director examples as they work then modify to suite.
That way you will probably spot what you are doing wrong too.

William
Bruce Who wrote:
> I add #include<exception> in the generated .cxx file and removed std
> namespace before exceptions and it compiles. I donnot know why. But my
> code still doesnot work! My OnEvent() event is still not called. Does
> that mean director doesnot work for me?
> 



From bruce.who.hk at gmail.com  Thu May 10 04:17:36 2007
From: bruce.who.hk at gmail.com (Bruce Who)
Date: Thu, 10 May 2007 10:17:36 +0800
Subject: [C++-sig] [Swig-user] Why this does not work- overriding C++
	virtualfunctions in python script?
In-Reply-To: <464227F6.4070701@fultondesigns.co.uk>
References: <c8dd1d2b0705080520i3b28bb57x785b99d121624508@mail.gmail.com>
	<op.tr0d2aofz3pzfc@nitrogenycs2>
	<c8dd1d2b0705081813g1867e1eel4b1e53eb96c48622@mail.gmail.com>
	<1C1EBEF8DBACDC439D038EA051674EC78A4B60@xbox.financialcad.com>
	<c8dd1d2b0705081913j5a417a53v734e0abe8c87cab3@mail.gmail.com>
	<464227F6.4070701@fultondesigns.co.uk>
Message-ID: <c8dd1d2b0705091917l191a1099p2277a7dd09902df4@mail.gmail.com>

Hi,

Thanks for all of you. It finally works.

I have tried swig/testswig/python/extend, and the unexpected_handler
issue still exists. I solve this as what I said in my previous post.
The script swig/testswig/python/extend/runme.py runs with errors:

Runtime Error!
Program: D:\program\Python25\python.exe
abnormal program termination

and I find that this error is caused by this call ``e.getName()''. I
think swig is not very friendly to win32 users.

However, the only difference between my .i file and the example.i is
that the order of #include and %include, this is my original .i file:


%module(directors="1") cmod
%include "cmod.h"

%feature("director");

%inline %{
#include "cmod.h"
%}


so I change it to:

%module(directors="1") cmod

%inline %{
#include "cmod.h"
%}

%feature("director");

%include "cmod.h"


then compile all files and run my script and everything goes fine!



Bruce


From roman.yakovenko at gmail.com  Thu May 10 06:31:51 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 10 May 2007 07:31:51 +0300
Subject: [C++-sig] Wrapping std::map using Boost::Python
In-Reply-To: <9fadca7e0705090423h4fdf20fcs84eee57bb090ca64@mail.gmail.com>
References: <9fadca7e0705090423h4fdf20fcs84eee57bb090ca64@mail.gmail.com>
Message-ID: <7465b6170705092131l55fbd97dw633e8ca6d41f029@mail.gmail.com>

On 5/9/07, James Knight <neworderofjamie at gmail.com> wrote:
>
> Hi,
>
> I'm trying to wrap std::map using boost::python so my embedded python code
> can access maps.  I have the following wrapper class:


Boost.Python exposes std::map:
http://boost.cvs.sourceforge.net/*checkout*/boost/boost/libs/python/doc/v2/indexing.html#map_indexing_suite_class
http://boost.cvs.sourceforge.net/boost/boost/libs/python/test/map_indexing_suite.cpp?view=markup&sortdir=down

May be you can use them?



-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070510/3a7a6cc9/attachment.htm>

From bruce.who.hk at gmail.com  Thu May 10 10:54:04 2007
From: bruce.who.hk at gmail.com (Bruce Who)
Date: Thu, 10 May 2007 16:54:04 +0800
Subject: [C++-sig] [Swig-user] Why this does not work- overriding C++
	virtualfunctions in python script?
In-Reply-To: <c8dd1d2b0705091917l191a1099p2277a7dd09902df4@mail.gmail.com>
References: <c8dd1d2b0705080520i3b28bb57x785b99d121624508@mail.gmail.com>
	<op.tr0d2aofz3pzfc@nitrogenycs2>
	<c8dd1d2b0705081813g1867e1eel4b1e53eb96c48622@mail.gmail.com>
	<1C1EBEF8DBACDC439D038EA051674EC78A4B60@xbox.financialcad.com>
	<c8dd1d2b0705081913j5a417a53v734e0abe8c87cab3@mail.gmail.com>
	<464227F6.4070701@fultondesigns.co.uk>
	<c8dd1d2b0705091917l191a1099p2277a7dd09902df4@mail.gmail.com>
Message-ID: <c8dd1d2b0705100154t1ae97a9do1ed1b982d08612ff@mail.gmail.com>

Hi, all

I guess the Runtime Error is due to the compiler. I compiles it with
VC++6.0 and the error occurs while with VC++8.0 everything is OK.

On 5/10/07, Bruce Who <bruce.who.hk at gmail.com> wrote:
> Hi,
>
> Thanks for all of you. It finally works.
>
> I have tried swig/testswig/python/extend, and the unexpected_handler
> issue still exists. I solve this as what I said in my previous post.
> The script swig/testswig/python/extend/runme.py runs with errors:
>
> Runtime Error!
> Program: D:\program\Python25\python.exe
> abnormal program termination
>
> and I find that this error is caused by this call ``e.getName()''. I
> think swig is not very friendly to win32 users.
>
> However, the only difference between my .i file and the example.i is
> that the order of #include and %include, this is my original .i file:
>
>
> %module(directors="1") cmod
> %include "cmod.h"
>
> %feature("director");
>
> %inline %{
> #include "cmod.h"
> %}
>
>
> so I change it to:
>
> %module(directors="1") cmod
>
> %inline %{
> #include "cmod.h"
> %}
>
> %feature("director");
>
> %include "cmod.h"
>
>
> then compile all files and run my script and everything goes fine!


Bruce


From bruce.who.hk at gmail.com  Thu May 10 17:57:16 2007
From: bruce.who.hk at gmail.com (Bruce Who)
Date: Thu, 10 May 2007 23:57:16 +0800
Subject: [C++-sig] A callback function with arguments causes error
Message-ID: <c8dd1d2b0705100857l5a8db92coab3d0213a5e64ae7@mail.gmail.com>

Hi, all

I have some callback functions in my .cpp files and I want to wrap
them into python scripts with swig. So I have tried this example
first: swigwin-1.3.31/Examples/python/callback. This example compiles
and runs well. But this callback function takes no arguments, so I
changed it a little:

// example.h
class Callback {
public:
	virtual ~Callback() { std::cout << "Callback::~Callback()" << std:: endl; }
	// XXX this is the original statement
	// virtual void run() { std::cout << "Callback::run()" << std::endl; }
	virtual void run(int* pn) { std::cout << "Callback::run()" << (*pn)
<< std::endl; }
};


class Caller {
private:
	Callback *_callback;
public:
	Caller(): _callback(0) {}
	~Caller() { delCallback(); }
	void delCallback() { delete _callback; _callback = 0; }
	void setCallback(Callback *cb) { delCallback(); _callback = cb; }
	// XXX void call() { if (_callback) _callback->run(); }
	void call()
	{
		if (_callback)
        	{
        	    int i = 90;
        	    // XXX _callback->run(i);
        	    _callback->run(&i);
        	}
    	}
};


As you see, I just add a int* argument to this callback. It compiles
but it cause python.exe to crash while running. Could anybody tell me
how to solve this? Any help would be appreciated!


Bruce


From luca.sbardella at gmail.com  Thu May 10 18:32:34 2007
From: luca.sbardella at gmail.com (Luca Sbardella)
Date: Thu, 10 May 2007 18:32:34 +0200
Subject: [C++-sig] inheritance problem when using constructor
Message-ID: <8315652a0705100932y6c9a611bg8f4ba29a59250ea5@mail.gmail.com>

Hi all,
I have a problem regarding inheritance/polymorphism which I cannot get to
the bottom of it.
In C++ I have a pure virtual class, lets called it an interface, called
"its" and an implementation of it called "ts".
I exposed the two to python using the "wrapper.hpp" module.

struct ts_wrapper : its,  wrapper<its>
{
int    get_numdates() const {....}
}

typedef boost::shared_ptr<its> ITS;
register_ptr_to_python<ITS>();

class_<ts_wrapper, boost::noncopyable>("itimeserie", "Time serie interface")
        .def("get_numdates",
pure_virtual(&its::get_numdates));

class_<ts, bases<ts_wrapper>>("timeserie", "An implementation of
itimeserie", init<const list&,const dict&>());

def("ts_create",&timeserieFactory::Create,"Get a time serie");


timeserieFactory returns a smart pointer to the interface "its", ITS.

Now, in Python, when I create an instance of "timeserie" using the
"ts_create" everything works fine, i.e. I can call the method "get_numdates"
without any problem. However when I create an instance of "timeserie" using
the constructor, the object is constructed correctly but when I call
"get_numdates" this is what python think about it

Boost.Python.ArgumentError: Python argument types in
    itimeserie.get_numdates(timeserie)
did not match C++ signature:
    get_numdates(class ts_wrapper {lvalue})
    get_numdates(class its {lvalue})

somehow when I use the constructor, the instance is not recognized as
derived from the interface.
Luca
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070510/ad92c7a4/attachment.htm>

From p.jaroszynski at gmail.com  Thu May 10 19:02:19 2007
From: p.jaroszynski at gmail.com (Piotr Jaroszynski)
Date: Thu, 10 May 2007 19:02:19 +0200
Subject: [C++-sig] inheritance problem when using constructor
In-Reply-To: <8315652a0705100932y6c9a611bg8f4ba29a59250ea5@mail.gmail.com>
References: <8315652a0705100932y6c9a611bg8f4ba29a59250ea5@mail.gmail.com>
Message-ID: <200705101902.19867.p.jaroszynski@gmail.com>

On Thursday 10 of May 2007 18:32:34 Luca Sbardella wrote:
> class_<ts, bases<ts_wrapper>>("timeserie", "An implementation of
> itimeserie", init<const list&,const dict&>());

ts is not derived from ts_wrapper but from its, is it? So I think you should 
use bases<its>.

-- 
Best Regards,
Piotr Jaroszynski


From luca.sbardella at gmail.com  Thu May 10 19:38:35 2007
From: luca.sbardella at gmail.com (Luca Sbardella)
Date: Thu, 10 May 2007 19:38:35 +0200
Subject: [C++-sig] inheritance problem when using constructor
In-Reply-To: <200705101902.19867.p.jaroszynski@gmail.com>
References: <8315652a0705100932y6c9a611bg8f4ba29a59250ea5@mail.gmail.com>
	<200705101902.19867.p.jaroszynski@gmail.com>
Message-ID: <8315652a0705101038j3da6f0fdlc342e371be115237@mail.gmail.com>

obviously.
thanks

On 10/05/07, Piotr Jaroszynski <p.jaroszynski at gmail.com> wrote:
>
> On Thursday 10 of May 2007 18:32:34 Luca Sbardella wrote:
> > class_<ts, bases<ts_wrapper>>("timeserie", "An implementation of
> > itimeserie", init<const list&,const dict&>());
>
> ts is not derived from ts_wrapper but from its, is it? So I think you
> should
> use bases<its>.
>
> --
> Best Regards,
> Piotr Jaroszynski
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070510/6919af26/attachment.htm>

From pk at cs.tut.fi  Thu May 10 22:27:35 2007
From: pk at cs.tut.fi (=?ISO-8859-1?Q?Pertti_Kellom=E4ki?=)
Date: Thu, 10 May 2007 23:27:35 +0300
Subject: [C++-sig] Doing dlopen() from within a Python extension
Message-ID: <46438037.7010306@cs.tut.fi>

I am wrapping a library using Boost.Python, and the
library does a "dlopen(path,RTLD_LOCAL | RTLD_NOW)"
to load some plugin code. The dlopen() fails and
returns NULL.

There seems to have been a lot of discussion around
the issue, but I am yet to find a definite answer of
how to make the dlopen() work. Any pointers?

My system is Ubuntu 6.10, and I am building the Python
extension using distutils. Python version is 2.4.
-- 
Pertti


From pk at cs.tut.fi  Thu May 10 23:08:20 2007
From: pk at cs.tut.fi (=?ISO-8859-1?Q?Pertti_Kellom=E4ki?=)
Date: Fri, 11 May 2007 00:08:20 +0300
Subject: [C++-sig] Doing dlopen() from within a Python extension
In-Reply-To: <46438037.7010306@cs.tut.fi>
References: <46438037.7010306@cs.tut.fi>
Message-ID: <464389C4.6010103@cs.tut.fi>

Pertti Kellom?ki wrote:
> There seems to have been a lot of discussion around
> the issue, but I am yet to find a definite answer of
> how to make the dlopen() work.

It seems that setting the RTLD_GLOBAL flag with
sys.setldopenflags() before loading my extensions resolves
the issue.
-- 
Pertti


From ajay.murthy at philips.com  Fri May 11 06:33:51 2007
From: ajay.murthy at philips.com (Ajay Murthy)
Date: Fri, 11 May 2007 10:03:51 +0530
Subject: [C++-sig] Very New to Boost Python
Message-ID: <OF4F43E4BF.AAB22381-ON652572D8.00163537-652572D8.0018E29B@philips.com>

Hi All,

I am very new to Boost.Python and I am facing the following problem.

Here is my C++ class declaration:

class Matrix
{
public:
    Matrix(int rows, int cols);
    Matrix(const Matrix& m);
    inline int GetCols() { return numCols;}
    inline int GetRows() { return numRows;}
    int** Get_Buffer();
    void SetStride(int x, int y);
    Matrix& operator*(const Matrix& m);
    Vector& operator[](int index);
    Matrix& operator=(const Matrix& m);
private:
    void SetSize(int rows, int cols);
    int numCols, numRows;
    Vector* matrix;
    int ystride;

};

And the corresponding Boost.Python wrapper is

BOOST_PYTHON_MODULE(DataTypes)
{
    class_<Matrix>("Matrix", init<int, int>())
        .def("GetCols", &Matrix::GetCols)
        .def("GetRows", &Matrix::GetRows)
        .def("SetStride", &Matrix::SetStride)
        //.def("Get_Buffer", &Matrix::Get_Buffer)
        //.def(self * self)
        //.def(__getitem__(self, int)
        //.def(__setitem__(self, int, int)
    ;
}

As you might have already observed, I have commented out few lines in the 
wrapper. If i uncomment it, bjam doesnt work. 
>From the initial analysis it seems to be a problem with the methods which 
accepts/returns arguments other than the primitive types. I may be wrong 
here.

Anyways, could anyone please explain me what the problem could be and the 
solution for the same.

Also, it would help if you could point me to some good documentation. The 
documentation in the boost.python page is very helpful for basic 
understanding but not after that.

thanks in advance,
Ajay
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070511/367fdf26/attachment.htm>

From meine at informatik.uni-hamburg.de  Fri May 11 12:09:12 2007
From: meine at informatik.uni-hamburg.de (Hans Meine)
Date: Fri, 11 May 2007 12:09:12 +0200
Subject: [C++-sig] Telling boost::build V2 how to find pyconfig.h?
Message-ID: <200705111209.12723.meine@informatik.uni-hamburg.de>

Hi!

For several hours now I have been struggling with boost::build V2, trying to 
compile the 1.34.0 beta version.

Please, can anyone tell me how to tell the build system that in addition to 
the regular python include path (which is found correctly), a 
platform-specific include path that contains only pyconfig.h is needed?

Looking at boost_1_34_0/tools/build/v2/tools/python.jam, I am not getting any 
smarter.  I do not even understand how the include path is set and why it is 
correct (I see the "rule" compute-default-paths, but my python installation 
is not in $prefix at all, so that cannot apply for me).  I expect 
distutils.sysconfig to be used, but I did not find that code yet.

With older boost versions, I simply set
PYTHON_PYCONFIG="`python -c "import distutils.sysconfig; print 
distutils.sysconfig.get_config_h_filename()"`"
and copied $PYTHON_PYCONFIG into $BUILDDIR/bin/boost/libs/python/build

I would be content if I found a comparable working hack to build 1.34.0 in our 
system environment, but the only location where I could copy pyconfig.h was 
the boost's source root directory, which is included through -I"."

If you tell me how to add an additional -I commandline argument to gcc, that 
would be fine, too.

Otherwise, the preferred solution would be to ask $(interpreter-cmd) for the 
location of pyconfig.h (see above) and compare the directory with the python 
includes directory, so it works for all users automatically.

-- 
Ciao, /  /
     /--/
    /  / ANS


From roman.yakovenko at gmail.com  Fri May 11 21:14:20 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 11 May 2007 21:14:20 +0200
Subject: [C++-sig] Very New to Boost Python
In-Reply-To: <OF4F43E4BF.AAB22381-ON652572D8.00163537-652572D8.0018E29B@philips.com>
References: <OF4F43E4BF.AAB22381-ON652572D8.00163537-652572D8.0018E29B@philips.com>
Message-ID: <7465b6170705111214n1367d7a3jfab593ccd6d85deb@mail.gmail.com>

On 5/11/07, Ajay Murthy <ajay.murthy at philips.com> wrote:
>
> Hi All,
>
> I am very new to Boost.Python and I am facing the following problem.
>
> Here is my C++ class declaration:
>
> class Matrix
> {
> public:
>     Matrix(int rows, int cols);
>     Matrix(const Matrix& m);
>     inline int GetCols() { return numCols;}
>     inline int GetRows() { return numRows;}
>     int** Get_Buffer();
>     void SetStride(int x, int y);
>     Matrix& operator*(const Matrix& m);
>     Vector& operator[](int index);
>     Matrix& operator=(const Matrix& m);
> private:
>     void SetSize(int rows, int cols);
>     int numCols, numRows;
>     Vector* matrix;
>     int ystride;
>
> };
>
> And the corresponding Boost.Python wrapper is
>
> BOOST_PYTHON_MODULE(DataTypes)
> {
>     class_<Matrix>("Matrix", init<int, int>())
>         .def("GetCols", &Matrix::GetCols)
>         .def("GetRows", &Matrix::GetRows)
>         .def("SetStride", &Matrix::SetStride)
>         //.def("Get_Buffer", &Matrix::Get_Buffer)
>         //.def(self * self)
>         //.def(__getitem__(self, int)
>         //.def(__setitem__(self, int, int)
>     ;
> }
>

You have to expose class Vector too and may be to define call policies.

>
> As you might have already observed, I have commented out few lines in the
wrapper. If i uncomment it, bjam doesnt work.
> From the initial analysis it seems to be a problem with the methods which
accepts/returns arguments other than the primitive types. I may be wrong
here.
>
> Anyways, could anyone please explain me what the problem could be and the
solution for the same.

You cannot export Get_Buffer, cause Python doesn't expose its user to
low-level details such as int**, so does Boost.Python. The are few possible
solution, one of them is to use custom return policies. I needed to expose
the functionality similar to yours, so I create such call policy:
http://www.language-binding.net/pyplusplus/documentation/functions/call_policies.html#return-range

> Also, it would help if you could point me to some good documentation. The
documentation in the boost.python page is very helpful for basic
understanding but not after that.

http://www.language-binding.net/pyplusplus/links.html#help-resources

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070511/84f5e39b/attachment.htm>

From bruce.who.hk at gmail.com  Mon May 14 11:25:51 2007
From: bruce.who.hk at gmail.com (Bruce Who)
Date: Mon, 14 May 2007 17:25:51 +0800
Subject: [C++-sig] A callback function with arguments causes error
In-Reply-To: <c8dd1d2b0705100857l5a8db92coab3d0213a5e64ae7@mail.gmail.com>
References: <c8dd1d2b0705100857l5a8db92coab3d0213a5e64ae7@mail.gmail.com>
Message-ID: <c8dd1d2b0705140225s47c36911o7445c0f6b1f40bb9@mail.gmail.com>

Hi, all

Has nobody tried a callback function with arguments? I just want to
implement an event-handling by doing this, but I donnot know how to
make it work.

On 5/10/07, Bruce Who <bruce.who.hk at gmail.com> wrote:
> Hi, all
>
> I have some callback functions in my .cpp files and I want to wrap
> them into python scripts with swig. So I have tried this example
> first: swigwin-1.3.31/Examples/python/callback. This example compiles
> and runs well. But this callback function takes no arguments, so I
> changed it a little:
>
> // example.h
> class Callback {
> public:
>         virtual ~Callback() { std::cout << "Callback::~Callback()" << std:: endl; }
>         // XXX this is the original statement
>         // virtual void run() { std::cout << "Callback::run()" << std::endl; }
>         virtual void run(int* pn) { std::cout << "Callback::run()" << (*pn)
> << std::endl; }
> };
>
>
> class Caller {
> private:
>         Callback *_callback;
> public:
>         Caller(): _callback(0) {}
>         ~Caller() { delCallback(); }
>         void delCallback() { delete _callback; _callback = 0; }
>         void setCallback(Callback *cb) { delCallback(); _callback = cb; }
>         // XXX void call() { if (_callback) _callback->run(); }
>         void call()
>         {
>                 if (_callback)
>                 {
>                     int i = 90;
>                     // XXX _callback->run(i);
>                     _callback->run(&i);
>                 }
>         }
> };
>
>
> As you see, I just add a int* argument to this callback. It compiles
> but it cause python.exe to crash while running. Could anybody tell me
> how to solve this? Any help would be appreciated!


Bruce


From meine at informatik.uni-hamburg.de  Mon May 14 11:52:16 2007
From: meine at informatik.uni-hamburg.de (Hans Meine)
Date: Mon, 14 May 2007 11:52:16 +0200
Subject: [C++-sig] Better introspection/apidoc extraction support?
Message-ID: <200705141152.16771.meine@informatik.uni-hamburg.de>

Hi!

The epydoc developers would like to support boost::python function 
introspection.  In boost 1.34.0, there is support for C++ signatures appended 
to the docstrings, but it would be better AFAICS if boost::python would 
additionally provide means for extracting the argument names and default 
values via ("half-private") attributes, similar to CPython.

I would imagine that this is not even hard to do for someone with sufficient 
insight, is such a thing planned?  (Or are there good reasons why that is not 
yet supported?)

IMHO it would greatly help API documentation systems. (Especially since 
signatures in docstrings are discouraged according to PEP 257, as I just 
learned.)

-- 
Ciao, /  /
     /--/
    /  / ANS


From lucy.devlin at hotmail.co.uk  Mon May 14 14:14:43 2007
From: lucy.devlin at hotmail.co.uk (Lucy Devlin)
Date: Mon, 14 May 2007 12:14:43 +0000
Subject: [C++-sig] Deleting the C++ part of a Boost.Python exposed Python
 object from C++
Message-ID: <BLU102-W41B072924BE0024F9AEE10A33E0@phx.gbl>


Hi Everyone,Apologies for the verbose subject/title :-pI am very new to Boost.Python so I am still finding my feet...I've spent a couple of days looking at a problem that I just can't figure out how to solve. What look like similar problems and solutions don't appear to suit my needs, so if someone can point me in the right direction I would be very grateful.I'm working on interfacing with a C++ application for which the source code should not be affected (if possible) and the following situation seems like a reasonable pattern of behaviour, but I can't see how to wrap it with Boost.Python. Look at the following C++ code snippet...ProcessX* process = new ProcessX;ProcessManager* manager = ...; // Get the process managermanager->addProcess(process);if (!manager->isValid()){    // Prompt for user input and modify the process    // Note that process is still accessible at this point    process->doSomething();}// The manager has ownership of process which
// will be deleted by the manager after execution
manager->executeProcesses();Now, I was hoping to find a way to wrap the C++ classes involved (ProcessX and ProcessManager) such that a python ProcessX instance wraps a C++ ProcessX instance and knows when the C++ instance is deleted. Such as...ProcessX process = myapi.ProcessX()ProcessManager manager = ...# Get the process managermanager.addProcess(process);if not manager.isValid():    # User input, etc    process.doSomething() # process instance is still validmanager.executeProcesses()try:    process.doSomething() # The wrapped C++ object was deletedexcept DeleteCPPException:    print "Exception raised" # And rightly so!I don't really need the Python process instance to endure after the deletion of its C++ counterpart but I need to transfer ownership to the process manager and maintain temporary access to the Python instance in order to doSomething.Any suggestions? Reading material? Good advice?I apologise if this has come up a few times before! I would expect it to have but didn't find anything in the archives.Many thanks in advance,L
_________________________________________________________________
Try Live.com: where your online world comes together - with news, sports, weather, and much more.
http://www.live.com/getstarted
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070514/25f0b79e/attachment.htm>

From lucy.devlin at hotmail.co.uk  Mon May 14 16:51:51 2007
From: lucy.devlin at hotmail.co.uk (Lucy Devlin)
Date: Mon, 14 May 2007 14:51:51 +0000
Subject: [C++-sig] Auto-Linking to Boost.Python
Message-ID: <BLU106-W34DAC60D8EE012CDB623A3A33E0@phx.gbl>


Hi Everyone,I've noticed that the auto-linking in VC7.1 is trying to link me against boost_python-vc71-mt-gd-1_34.lib rather than boost_python-vc71-mt-gyd-1_34.lib (note the extra 'y') even though I have BOOST_DEBUG_PYTHON defined in my project. Is this right or am I doing something wrong? I built the actual boost_python-vc71-mt-gyd-1_34.lib (and DLL) without any problems, it's just linking to this library that's going wrong for me.Obviously I am using Boost 1.34 ;-)Regards and many thanks,L
_________________________________________________________________
Try Live.com - your fast, personalized homepage with all the things you care about in one place.
http://www.live.com/getstarted
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070514/e9fc89bc/attachment.htm>

From rwgk at yahoo.com  Mon May 14 17:59:00 2007
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Mon, 14 May 2007 08:59:00 -0700 (PDT)
Subject: [C++-sig] Better introspection/apidoc extraction support?
Message-ID: <375753.33058.qm@web31104.mail.mud.yahoo.com>

I did the implementation of the "docstring_options" feature. I'm not aware of plans to work on this, but I'm interested to at least learn what could be done to better support epydoc. If it is not too time-consuming I'd be willing to work on it a bit.

> The epydoc developers would like to support boost::python function 
> introspection.  In boost 1.34.0, there is support for C++ signatures appended 
> to the docstrings, but it would be better AFAICS if boost::python would 
> additionally provide means for extracting the argument names and default 
> values via ("half-private") attributes, similar to CPython.

How exactly does this work?  Are there web pages about this?

Thanks!

Ralf







       
____________________________________________________________________________________Be a better Heartthrob. Get better relationship answers from someone who knows. Yahoo! Answers - Check it out. 
http://answers.yahoo.com/dir/?link=list&sid=396545433
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070514/feff1926/attachment.htm>

From monty at inaugust.com  Mon May 14 20:27:07 2007
From: monty at inaugust.com (Monty Taylor)
Date: Mon, 14 May 2007 11:27:07 -0700
Subject: [C++-sig] [Swig-user] A callback function with arguments causes
	error
In-Reply-To: <c8dd1d2b0705140225s47c36911o7445c0f6b1f40bb9@mail.gmail.com>
References: <c8dd1d2b0705100857l5a8db92coab3d0213a5e64ae7@mail.gmail.com>
	<c8dd1d2b0705140225s47c36911o7445c0f6b1f40bb9@mail.gmail.com>
Message-ID: <4648A9FB.3000804@inaugust.com>

Hi Bruce,

I'm using a callback with arguments from Python using directors.

I've got this:

class BaseCallback {
  public:
  virtual ~BaseCallback() {};
  virtual void callback(int res, NdbTransaction * trans);

};


Then from python:

class PythonCallback(ndbapi.BaseCallback):

  def __init__(self, recAttr):
    self.recAttr=recAttr

  def __call__(self, *args, **kw):
    self.callback(*args,**kw)

  def callback(self, ret, myTrans):
    x=self.recAttr.get_value()

Works like a charm. Are you sure everything is working right with memory
ownership and that int *? I'd try it with just a normal int first to see
if that's the problem, then go in and figure out how to get Python to
allocate and pass the int* in such a way that it still exists when you
access it in the callback... but I could be way off base there.

Monty

Bruce Who wrote:
> Hi, all
> 
> Has nobody tried a callback function with arguments? I just want to
> implement an event-handling by doing this, but I donnot know how to
> make it work.
> 
> On 5/10/07, Bruce Who <bruce.who.hk at gmail.com> wrote:
>> Hi, all
>>
>> I have some callback functions in my .cpp files and I want to wrap
>> them into python scripts with swig. So I have tried this example
>> first: swigwin-1.3.31/Examples/python/callback. This example compiles
>> and runs well. But this callback function takes no arguments, so I
>> changed it a little:
>>
>> // example.h
>> class Callback {
>> public:
>>         virtual ~Callback() { std::cout << "Callback::~Callback()" << std:: endl; }
>>         // XXX this is the original statement
>>         // virtual void run() { std::cout << "Callback::run()" << std::endl; }
>>         virtual void run(int* pn) { std::cout << "Callback::run()" << (*pn)
>> << std::endl; }
>> };
>>
>>
>> class Caller {
>> private:
>>         Callback *_callback;
>> public:
>>         Caller(): _callback(0) {}
>>         ~Caller() { delCallback(); }
>>         void delCallback() { delete _callback; _callback = 0; }
>>         void setCallback(Callback *cb) { delCallback(); _callback = cb; }
>>         // XXX void call() { if (_callback) _callback->run(); }
>>         void call()
>>         {
>>                 if (_callback)
>>                 {
>>                     int i = 90;
>>                     // XXX _callback->run(i);
>>                     _callback->run(&i);
>>                 }
>>         }
>> };
>>
>>
>> As you see, I just add a int* argument to this callback. It compiles
>> but it cause python.exe to crash while running. Could anybody tell me
>> how to solve this? Any help would be appreciated!
> 
> 
> Bruce
> 
> -------------------------------------------------------------------------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> _______________________________________________
> Swig-user mailing list
> Swig-user at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/swig-user



From bruce.who.hk at gmail.com  Tue May 15 02:25:30 2007
From: bruce.who.hk at gmail.com (Bruce Who)
Date: Tue, 15 May 2007 08:25:30 +0800
Subject: [C++-sig] [Swig-user] A callback function with arguments causes
	error
In-Reply-To: <4648A9FB.3000804@inaugust.com>
References: <c8dd1d2b0705100857l5a8db92coab3d0213a5e64ae7@mail.gmail.com>
	<c8dd1d2b0705140225s47c36911o7445c0f6b1f40bb9@mail.gmail.com>
	<4648A9FB.3000804@inaugust.com>
Message-ID: <c8dd1d2b0705141725xe68fc29t51ab7619dca4651@mail.gmail.com>

Hi, Monty,

Thanks for your code, but I still cannot figure out why it doesnot
work. This is all of my code, it is just the callback example of swig
with a little modification, but it just doesnot work:

// ================== example.h
#include <cstdio>
#include <iostream>

class Callback {
public:
	virtual ~Callback() { std::cout << "Callback::~Callback()" << std:: endl; }
	// virtual void run() { std::cout << "Callback::run()" << std::endl; }
	virtual void run(int* pn) { std::cout << "Callback::run()" << (*pn)
<< std::endl; }
};


class Caller {
private:
	Callback *_callback;
public:
	Caller(): _callback(0) {}
	~Caller() { delCallback(); }
	void delCallback() { delete _callback; _callback = 0; }
	void setCallback(Callback *cb) { delCallback(); _callback = cb; }
	// void call() { if (_callback) _callback->run(); }
	void call()
    {
        if (_callback)
        {
            int i = 90;
            _callback->run(&i);
        }
    }
};


// ================== example.cxx
#include "example.h"


// ================== example.i
%module(directors="1") example
%{
#include "example.h"
%}

%include "std_string.i"

%feature("director") Callback;

%include "example.h"


@rem ================== build.bat
@rem I compile my code with this bat file
set FILENAME=example

call "D:\program\Microsoft Visual Studio 8\VC\bin\vcvars32"
swig -c++ -python %FILENAME%.i

cl -c %FILENAME%_wrap.cxx -ID:\program\Python25\include\

link /subsystem:windows /LIBPATH:D:\program\Python25\libs\
%FILENAME%_wrap.obj /dll /OUT:_%FILENAME%.pyd


#===================================================
# file: runme.py
# you can test the wrapper with this script

import example

class PyCallback(example.Callback):
    def __init__(self):
        example.Callback.__init__(self)
    def run(self, i):
        print "PyCallback.run()", i

# Create an Caller instance

caller = example.Caller()

# Add a simple C++ callback (caller owns the callback, so
# we disown it first by clearing the .thisown flag).

print "Adding and calling a normal C++ callback"
print "----------------------------------------"

callback = example.Callback()
callback.thisown = 0
caller.setCallback(callback)
caller.call()
caller.delCallback();

print
print "Adding and calling a Python callback"
print "------------------------------------"

# Add a Python callback (caller owns the callback, so we
# disown it first by calling __disown__).

caller.setCallback(PyCallback().__disown__())
caller.call()
caller.delCallback()

print
print "Adding and calling another Python callback"
print "------------------------------------------"

# Lets do the same but use the weak reference this time.

callback = PyCallback().__disown__()
caller.setCallback(callback)
caller.call()
caller.delCallback()

# All done.

print
print "python exit"


After compiling all code, I run it:
E:\test>runme.py
Adding and calling a normal C++ callback
----------------------------------------
Callback::run()90
Callback::~Callback()

Adding and calling a Python callback
------------------------------------
PyCallback.run()

and then a windows message box pops up saying that there is something
wrong with python.exe.


On 5/15/07, Monty Taylor <monty at inaugust.com> wrote:
> Hi Bruce,
>
> I'm using a callback with arguments from Python using directors.
>
> I've got this:
>
> class BaseCallback {
>   public:
>   virtual ~BaseCallback() {};
>   virtual void callback(int res, NdbTransaction * trans);
>
> };
>
>
> Then from python:
>
> class PythonCallback(ndbapi.BaseCallback):
>
>   def __init__(self, recAttr):
>     self.recAttr=recAttr
>
>   def __call__(self, *args, **kw):
>     self.callback(*args,**kw)
>
>   def callback(self, ret, myTrans):
>     x=self.recAttr.get_value()
>
> Works like a charm. Are you sure everything is working right with memory
> ownership and that int *? I'd try it with just a normal int first to see
> if that's the problem, then go in and figure out how to get Python to
> allocate and pass the int* in such a way that it still exists when you
> access it in the callback... but I could be way off base there.
>
> Monty


From j.reid at mail.cryst.bbk.ac.uk  Tue May 15 12:49:41 2007
From: j.reid at mail.cryst.bbk.ac.uk (John Reid)
Date: Tue, 15 May 2007 11:49:41 +0100
Subject: [C++-sig] Better introspection/apidoc extraction support?
In-Reply-To: <375753.33058.qm@web31104.mail.mud.yahoo.com>
References: <375753.33058.qm@web31104.mail.mud.yahoo.com>
Message-ID: <f2c39r$j06$1@sea.gmane.org>

Would there be any chance of improving the formatting of mismatched 
arguments in python to C++ function call errors at the same time as this 
work? I'm not sure how related it is though. See
http://tinyurl.com/2wa9pe

I think it would be a major improvement in useability.

Best,
John.



Ralf W. Grosse-Kunstleve wrote:
> I did the implementation of the "docstring_options" feature. I'm not 
> aware of plans to work on this, but I'm interested to at least learn 
> what could be done to better support epydoc. If it is not too 
> time-consuming I'd be willing to work on it a bit.
> 
>  > The epydoc developers would like to support boost::python function
>  > introspection.  In boost 1.34.0, there is support for C++ signatures 
> appended
>  > to the docstrings, but it would be better AFAICS if boost::python would
>  > additionally provide means for extracting the argument na mes and 
> default
>  > values via ("half-private") attributes, similar to CPython.
> 
> How exactly does this work?  Are there web pages about this?
> 
> Thanks!
> 
> Ralf
> 
> 
> ------------------------------------------------------------------------
> Moody friends. Drama queens. Your life? Nope! - their life, your story.
> Play Sims Stories at Yahoo! Games. 
> <http://us.rd.yahoo.com/evt=48224/*http://sims.yahoo.com/>
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig



From meine at informatik.uni-hamburg.de  Tue May 15 13:37:24 2007
From: meine at informatik.uni-hamburg.de (Hans Meine)
Date: Tue, 15 May 2007 13:37:24 +0200
Subject: [C++-sig] Better introspection/apidoc extraction support?
In-Reply-To: <375753.33058.qm@web31104.mail.mud.yahoo.com>
References: <375753.33058.qm@web31104.mail.mud.yahoo.com>
Message-ID: <200705151337.25198.meine@informatik.uni-hamburg.de>

Hi!

Am Montag, 14. Mai 2007 17:59:00 schrieb Ralf W. Grosse-Kunstleve:
> I did the implementation of the "docstring_options" feature. I'm not aware
> of plans to work on this, but I'm interested to at least learn what could
> be done to better support epydoc. If it is not too time-consuming I'd be
> willing to work on it a bit.

Great!

> > The epydoc developers would like to support boost::python function
> > introspection.  In boost 1.34.0, there is support for C++ signatures
> > appended to the docstrings, but it would be better AFAICS if
> > boost::python would additionally provide means for extracting the
> > argument names and default values via ("half-private") attributes,
> > similar to CPython.
>
> How exactly does this work?  Are there web pages about this?

Not that I know of.  AFAICS, there is no public API for that (that's why I 
wrote "half-private"), but if you look at a functions func_code.co_XXX and 
func_defaults properties, you see that there is a lot of internal information 
available in CPython.

Then, there is the inspect module (implemented in python), which uses the 
above internal stuff to offer a more convenient API.  E.g. 
inspect.getargspec() uses func_code.co_varnames[:func_code.co_argcount] to 
query the argument names, func_code.co_flags to determine whether the 
function has *args or **kwargs parameters (4/8 bitflags), ...

I would think that basically *any* API within boost::python functions would be 
OK, since there is no official way beyond inspect.  (Of course, it would be 
best if the inspect module would become compatible with boost::python some 
day, so apidoc projects do not have to duplicate code..)

-- 
Ciao, /  /
     /--/
    /  / ANS


From Mark.English at rbccm.com  Tue May 15 14:12:07 2007
From: Mark.English at rbccm.com (English, Mark)
Date: Tue, 15 May 2007 13:12:07 +0100
Subject: [C++-sig] Better introspection/apidoc extraction support?
Message-ID: <EE7D40281B517A41AF5A90847A61F27930CAE47A@SEW39051.oak.fg.rbc.com>

> Not that I know of.  AFAICS, there is no public API for that (that's why I 
> wrote "half-private"), but if you look at a functions func_code.co_XXX and 
> func_defaults properties, you see that there is a lot of internal information 
> available in CPython.

> Then, there is the inspect module (implemented in python), which uses the 
> above internal stuff to offer a more convenient API.  E.g. 
> inspect.getargspec() uses func_code.co_varnames[:func_code.co_argcount] to 
> query the argument names, func_code.co_flags to determine whether the 
> function has *args or **kwargs parameters (4/8 bitflags), ...

> I would think that basically *any* API within boost::python functions would be 
> OK, since there is no official way beyond inspect.  (Of course, it would be 
> best if the inspect module would become compatible with boost::python some 
> day, so apidoc projects do not have to duplicate code..)

(Apologies for the ridiculous trailer)

For an introspection example, there was a signature.py drifting around which I'm having trouble tracking down, written I believe by Neel Krishnaswami. There's a copy here:
http://funformkit.sourceforge.net/FunFormKit/Signature.py

Speaking of signatures, have you looked at the draft Pep 362 ?
http://www.python.org/dev/peps/pep-0362/

There's a reference implementation:
http://svn.python.org/view/sandbox/trunk/pep362/

Hope that helps,
Mark
________________________________________

This E-Mail (including any attachments) may contain privileged or confidential information.  It is intended only for the addressee(s) indicated above.

The sender does not waive any of its rights, privileges or other protections respecting this information.  

Any distribution, copying or other use of this E-Mail or the information it contains, by other than an intended recipient, is not sanctioned and is prohibited.

If you received this E-Mail in error, please delete it and advise the sender (by return E-Mail or otherwise) immediately. 

This E-Mail (including any attachments) has been scanned for viruses. 

It is believed to be free of any virus or other defect that might affect any computer system into which it is received and opened. 

However, it is the responsibility of the recipient to ensure that it is virus free. 

The sender accepts no responsibility for any loss or damage arising in any way from its use.

E-Mail received by or sent from RBC Capital Markets is subject to review by Supervisory personnel. 

Such communications are retained and may be produced to regulatory authorities or others with legal rights to the information.

IRS CIRCULAR 230 NOTICE:  TO COMPLY WITH U.S. TREASURY REGULATIONS, WE ADVISE YOU THAT ANY U.S. FEDERAL TAX ADVISE INCLUDED IN THIS COMMUNICATION IS NOT INTENDED OR WRITTEN TO BE USED, AND CANNOT BE USED, TO AVOID ANY U.S. FEDERAL TAX PENALTIES OR TO PROMOTE, MARKET, OR RECOMMEND TO ANOTHER PARTY ANY TRANSACTION OR MATTER.


From meine at informatik.uni-hamburg.de  Tue May 15 15:01:41 2007
From: meine at informatik.uni-hamburg.de (Hans Meine)
Date: Tue, 15 May 2007 15:01:41 +0200
Subject: [C++-sig] Better introspection/apidoc extraction support?
In-Reply-To: <EE7D40281B517A41AF5A90847A61F27930CAE47A@SEW39051.oak.fg.rbc.com>
References: <EE7D40281B517A41AF5A90847A61F27930CAE47A@SEW39051.oak.fg.rbc.com>
Message-ID: <200705151501.42122.meine@informatik.uni-hamburg.de>

Hi!

Am Dienstag, 15. Mai 2007 14:12:07 schrieb English, Mark:
> For an introspection example, there was a signature.py drifting around
> which I'm having trouble tracking down, written I believe by Neel
> Krishnaswami. There's a copy here:
> http://funformkit.sourceforge.net/FunFormKit/Signature.py

That looks like a predecessor / alternative to the std. 'inspect' module.
(And does not yet work with boost::python or course.)

> Speaking of signatures, have you looked at the draft Pep 362 ?
> http://www.python.org/dev/peps/pep-0362/

That PEP is very interesting indeed.  Since it describes a complete API, it 
looks like a perfect implementation goal for BPL.  (Eager/lazy creation does 
not really matter here, as long as BPL functions get a proper __signature__ 
attribute.)  Ralf, what do you think?

-- 
Ciao, /  /
     /--/
    /  / ANS


From bruce.who.hk at gmail.com  Tue May 15 16:49:24 2007
From: bruce.who.hk at gmail.com (Bruce Who)
Date: Tue, 15 May 2007 22:49:24 +0800
Subject: [C++-sig] [Swig-user] A callback function with arguments causes
	error
In-Reply-To: <op.tscvrq2lz3pzfc@nitrogenycs2>
References: <c8dd1d2b0705100857l5a8db92coab3d0213a5e64ae7@mail.gmail.com>
	<c8dd1d2b0705140225s47c36911o7445c0f6b1f40bb9@mail.gmail.com>
	<4648A9FB.3000804@inaugust.com>
	<c8dd1d2b0705141725xe68fc29t51ab7619dca4651@mail.gmail.com>
	<op.tscvrq2lz3pzfc@nitrogenycs2>
Message-ID: <c8dd1d2b0705150749q4471ff5dmfb9edeadee20eaa8@mail.gmail.com>

hi, Nitro

Thanks for your hint, and I tried a similar method, I add __asm {int
3} before the call, but the unassembld code are really hard to read
and I donnot think I can figure it out even it's translated into C++
source code.

I just wonder if this kind of usage is not supported by swig or maybe
I miss something very basic, since I just made a little modification
to the swig example.

On 5/15/07, Nitro <nitro at dr-code.org> wrote:
> Am 15.05.2007, 02:25 Uhr, schrieb Bruce Who <bruce.who.hk at gmail.com>:
>
> <snip>
>
> Why don't you compile a debug version and invoke the debugger when it
> crashes (or before it crashes)? My favourite trick for debugging an
> extension module is like this:
>
> test.py:
>
> import myModule
> doSomeNonCrashingPythonThingsHere()
> myModule.InvokeDebugger()  # Debugger gets involved here
> this_line_will_make_it_crash()
>
>
> myModule.i:
>
> ....
>
> %inline
> %{
> void InvokeDebugger()
> {
>      int* x = 0;
>      y = *x;
> }
> %}
>
>
> When it crashes on the invoke debugger line, windows gives me that error
> box where I can choose "Debug". When I do this I end up in the y = *x
> line, then set the "next instruction" to the end of the InvokeDebugger
> function, set all breakpoints I want to set (for example in the
> myModuke_wrap.cpp) and hit F5 to contiue debugging. The debugger will then
> either break on the next breakpoint or break when it crashes.
> This should give you enough pointers to see what you are doing wrong.
>
> -Matthias
>
> P.S.: I think this only works in debug mode and probably only on msvc. In
> release mode the InvokeDebugger function is sometimes either not executed
> or it doesn't care whether it dereferences a 0 pointer.
>

Bruce


From bruce.who.hk at gmail.com  Wed May 16 05:10:27 2007
From: bruce.who.hk at gmail.com (Bruce Who)
Date: Wed, 16 May 2007 11:10:27 +0800
Subject: [C++-sig] Any win32 project using swig?
Message-ID: <c8dd1d2b0705152010y131b3735sc6aded6a63b603cf@mail.gmail.com>

Hi,all

I'm trying to use swig to wrap a C++ library to python, but there's so
many problems that I cannot solve. I am now looking for a existing
project which uses swig for python binding, is a open-source win32
project and not too complex. If there is such a project I think I can
learn something by looking through its source code. Could anybody
recommend one or more (the more the better) for me? Thanks in advance.

Bruce


From jlenz2 at math.uiuc.edu  Wed May 16 20:39:42 2007
From: jlenz2 at math.uiuc.edu (John Lenz)
Date: Wed, 16 May 2007 13:39:42 -0500
Subject: [C++-sig] [Swig-user] Any win32 project using swig?
In-Reply-To: <c8dd1d2b0705152010y131b3735sc6aded6a63b603cf@mail.gmail.com>
References: <c8dd1d2b0705152010y131b3735sc6aded6a63b603cf@mail.gmail.com>
Message-ID: <464B4FEE.7010902@math.uiuc.edu>

On 05/15/2007 10:10 PM, Bruce Who wrote:
> Hi,all
> 
> I'm trying to use swig to wrap a C++ library to python, but there's so
> many problems that I cannot solve. I am now looking for a existing
> project which uses swig for python binding, is a open-source win32
> project and not too complex. If there is such a project I think I can
> learn something by looking through its source code. Could anybody
> recommend one or more (the more the better) for me? Thanks in advance.
> 
> Bruce


Subversion is the first one that comes to mind.  cross platform and
python bindings (as well as several other languages).

http://subversion.tigris.org/


From dave at boost-consulting.com  Fri May 18 01:09:09 2007
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 17 May 2007 17:09:09 -0600
Subject: [C++-sig] Auto-Linking to Boost.Python
References: <BLU106-W34DAC60D8EE012CDB623A3A33E0@phx.gbl>
Message-ID: <871whf9hey.fsf@grogan.peloton>


on Mon May 14 2007, Lucy Devlin <lucy.devlin-AT-hotmail.co.uk> wrote:

> Hi Everyone,
>
> I've noticed that the auto-linking in VC7.1 is trying to link me
> against boost_python-vc71-mt-gd-1_34.lib rather than
> boost_python-vc71-mt-gyd-1_34.lib (note the extra 'y') even though I
> have BOOST_DEBUG_PYTHON defined in my project. Is this right or am I
> doing something wrong? I built the actual
> boost_python-vc71-mt-gyd-1_34.lib (and DLL) without any problems, it's
> just linking to this library that's going wrong for me.

Are you certain?  I was pretty sure I never did anything to support
auto linking in Boost.Python.


-- 
Dave Abrahams
Boost Consulting
http://www.boost-consulting.com

Don't Miss BoostCon 2007! ==> http://www.boostcon.com



From t_spens at yahoo.com  Fri May 18 02:02:21 2007
From: t_spens at yahoo.com (t_spens)
Date: Thu, 17 May 2007 17:02:21 -0700 (PDT)
Subject: [C++-sig] Problems importing a c++ shared library
Message-ID: <10674432.post@talk.nabble.com>


ImportError: dynamic module does not define init function (initlibhello) - I
have no idea why I'm getting this error.

I'm using gcc to create my library libhello.so from the .cpp file below
Here are my gcc arguments:
gcc -g -Wall -I ../../../usr/include/boost -I ../../../usr/include/python2.5
-c hello.cpp -o hello.o
gcc -shared -fPIC -g -Wall -I ../../../usr/include/boost -I
../../../usr/include/python2.5 -L /usr/lib -lstdc++ -lboost_python -o
libhello.so hello.o

#include <iostream>
using namespace std;
void print(char * stringToPrint){cout << stringToPrint << endl;}
int main ()
{
        print("hello from c++");
        return 0;
}
#include <boost/python.hpp>
using namespace boost::python;
BOOST_PYTHON_MODULE(hello)
{
        def("print", print);
}

~/pythonScripts$ python
Python 2.5.1c1 (release25-maint, Apr 12 2007, 21:00:25) 
[GCC 4.1.2 (Ubuntu 4.1.2-0ubuntu4)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import libhello
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ImportError: dynamic module does not define init function (initlibhello)
>>>

Any ideas?
-- 
View this message in context: http://www.nabble.com/Problems-importing-a-c%2B%2B-shared-library-tf3775017.html#a10674432
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From p.jaroszynski at gmail.com  Fri May 18 02:37:21 2007
From: p.jaroszynski at gmail.com (Piotr Jaroszynski)
Date: Fri, 18 May 2007 02:37:21 +0200
Subject: [C++-sig] Problems importing a c++ shared library
In-Reply-To: <10674432.post@talk.nabble.com>
References: <10674432.post@talk.nabble.com>
Message-ID: <200705180237.21804.p.jaroszynski@gmail.com>

On Friday 18 of May 2007 02:02:21 t_spens wrote:
> BOOST_PYTHON_MODULE(hello)
> >>> import libhello
these two must match, so try BOOST_PYTHON_MODULE(libhello).

-- 
Best Regards,
Piotr Jaroszynski


From j.reid at mail.cryst.bbk.ac.uk  Fri May 18 10:15:32 2007
From: j.reid at mail.cryst.bbk.ac.uk (John Reid)
Date: Fri, 18 May 2007 09:15:32 +0100
Subject: [C++-sig] Auto-Linking to Boost.Python
In-Reply-To: <871whf9hey.fsf@grogan.peloton>
References: <BLU106-W34DAC60D8EE012CDB623A3A33E0@phx.gbl>
	<871whf9hey.fsf@grogan.peloton>
Message-ID: <f2jncs$sfm$1@sea.gmane.org>

David Abrahams wrote:
> Are you certain?  I was pretty sure I never did anything to support
> auto linking in Boost.Python.
> 
> 

I noticed this at the end of python/detail/config.hpp:

//  enable automatic library variant selection 
------------------------------//

#if !defined(BOOST_PYTHON_SOURCE) && !defined(BOOST_ALL_NO_LIB) && 
!defined(BOOST_PYTHON_NO_LIB)
//
// Set the name of our library, this will get undef'ed by auto_link.hpp
// once it's done with it:
//
#define BOOST_LIB_NAME boost_python
//
// If we're importing code from a dll, then tell auto_link.hpp about it:
//
#ifdef BOOST_PYTHON_DYNAMIC_LIB
#  define BOOST_DYN_LINK
#endif
//
// And include the header that does the work:
//
#include <boost/config/auto_link.hpp>
#endif  // auto-linking disabled



From iladyko at odesk.com  Fri May 18 10:17:56 2007
From: iladyko at odesk.com (Irene Ladyko)
Date: Fri, 18 May 2007 15:17:56 +0700
Subject: [C++-sig] embeded python and winsound
Message-ID: <1453152813.20070518151756@odesk.com>

Hello, c++-sig.

I use Python version 2.4.4 under win32.
I built python using pcbuild.sln. There is winsound module project in
it, so I built it too. After that I can run python.exe interpeter
directly and load winsound module in it. All be OK.
    D:\....>python.exe
    Python 2.4.4 (#71, May 17 2007, 16:53:33) [MSC v.1310 32 bit (Intel)] on win32
    Type "help", "copyright", "credits" or "license" for more information.
    >>>
    >>>import winsound
    >>>winsound.PlaySound("SystemExit", winsound.SND_ALIAS)

But I am interested in embeded python. I use this interpreter in
C++ code (MSVC7.1). And when I load winsound programmatically, the following error appears:
"No module named winsound". But such module as 'array' is imported
correctly:
          import array

What is wrong. Please, point me on my error.


Code:
#include <string>

#include "Python.h"


int main(int /*argc*/, char** /*argv*/)
{
        PyThreadState *m_threadState;
        
        // Initialize python
        Py_NoSiteFlag = 1;
        Py_Initialize();
        
        // Enable thread support.
        PyEval_InitThreads();

        // Delete default thread state.
        PyThreadState *l_default = PyThreadState_Get();
        Py_EndInterpreter(l_default);

        // Release global interpreter lock.
        PyEval_ReleaseLock();

        //start python
        PyEval_AcquireLock();
        m_threadState = Py_NewInterpreter();
        PyEval_ReleaseThread(m_threadState);

        //-------Load winsound module----------!!!!!!!!!!!!!!!!!!!!!!!!
        PyEval_AcquireThread(m_threadState);

        PyObject *l_module;
        PyObject *l_dict;
        PyObject *l_retValue;

        // we will use __main__ module as our global and local namespace.
        l_module = PyImport_AddModule("__main__");
        if (!l_module)
        {
                PyEval_ReleaseThread(m_threadState);
                return -1;
        }

        std::string l_str = "import winsound";
        l_dict = PyModule_GetDict(l_module);
        l_retValue = PyRun_StringFlags(
                        (char *)l_str.c_str(),
                        Py_file_input,
                        l_dict,
                        l_dict,
                        NULL);
        if (!l_retValue)
        {
//!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
                // Enter into this block!!!!!!!!!!!!!!!!!!!
                PyEval_ReleaseThread(m_threadState);
                return -1;
        }
        Py_DECREF(l_retValue);

        PyEval_ReleaseThread(m_threadState);

        //End
        PyEval_AcquireThread(m_threadState);
        Py_EndInterpreter(m_threadState);
        PyEval_ReleaseLock();
        PyEval_AcquireLock();
        // Create default interpreter back.
        Py_NewInterpreter();
        Py_Finalize();

        return 0;
}

-- 
Best Regards, Irene Ladyko.



From pk at cs.tut.fi  Fri May 18 11:30:10 2007
From: pk at cs.tut.fi (=?ISO-8859-1?Q?Pertti_Kellom=E4ki?=)
Date: Fri, 18 May 2007 12:30:10 +0300
Subject: [C++-sig] Exposing a template class
Message-ID: <464D7222.4020509@cs.tut.fi>

Are there examples around of exposing instances of
template classes? I'm assuming it should be as easy
as enclosing a  BOOST_PYTHON_MODULE inside a template
and instantiating the template, but it would be reassuring
to have some hand holding in case there are issues I have
not thought about.
-- 
Pertti


From roman.yakovenko at gmail.com  Fri May 18 20:26:19 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 18 May 2007 20:26:19 +0200
Subject: [C++-sig] embeded python and winsound
In-Reply-To: <1453152813.20070518151756@odesk.com>
References: <1453152813.20070518151756@odesk.com>
Message-ID: <7465b6170705181126p517f5942q2bbaadfd592e6011@mail.gmail.com>

On 5/18/07, Irene Ladyko <iladyko at odesk.com> wrote:
> Hello, c++-sig.
>
> I use Python version 2.4.4 under win32.
> I built python using pcbuild.sln. There is winsound module project in
> it, so I built it too. After that I can run python.exe interpeter
> directly and load winsound module in it. All be OK.
>     D:\....>python.exe
>     Python 2.4.4 (#71, May 17 2007, 16:53:33) [MSC v.1310 32 bit (Intel)] on win32
>     Type "help", "copyright", "credits" or "license" for more information.
>     >>>
>     >>>import winsound
>     >>>winsound.PlaySound("SystemExit", winsound.SND_ALIAS)
>
> But I am interested in embeded python. I use this interpreter in
> C++ code (MSVC7.1). And when I load winsound programmatically, the following error appears:
> "No module named winsound". But such module as 'array' is imported
> correctly:
>           import array
>
> What is wrong. Please, point me on my error.

I guess the problem is related to location of winsound module. May be
you can move the module to one of the directories Python is looking at
or adjust sys.path variable.

By the way you can use Boost.Python functionality to simplify a lot your code:
http://boost.org/libs/python/doc/tutorial/doc/html/python/embedding.html

>
>
> Code:
> #include <string>
>
> #include "Python.h"
>
>
> int main(int /*argc*/, char** /*argv*/)
> {
>         PyThreadState *m_threadState;
>
>         // Initialize python
>         Py_NoSiteFlag = 1;
>         Py_Initialize();
>
>         // Enable thread support.
>         PyEval_InitThreads();
>
>         // Delete default thread state.
>         PyThreadState *l_default = PyThreadState_Get();
>         Py_EndInterpreter(l_default);
>
>         // Release global interpreter lock.
>         PyEval_ReleaseLock();
>
>         //start python
>         PyEval_AcquireLock();
>         m_threadState = Py_NewInterpreter();
>         PyEval_ReleaseThread(m_threadState);
>
>         //-------Load winsound module----------!!!!!!!!!!!!!!!!!!!!!!!!
>         PyEval_AcquireThread(m_threadState);
>
>         PyObject *l_module;
>         PyObject *l_dict;
>         PyObject *l_retValue;
>
>         // we will use __main__ module as our global and local namespace.
>         l_module = PyImport_AddModule("__main__");
>         if (!l_module)
>         {
>                 PyEval_ReleaseThread(m_threadState);
>                 return -1;
>         }
>
>         std::string l_str = "import winsound";
>         l_dict = PyModule_GetDict(l_module);
>         l_retValue = PyRun_StringFlags(
>                         (char *)l_str.c_str(),
>                         Py_file_input,
>                         l_dict,
>                         l_dict,
>                         NULL);
>         if (!l_retValue)
>         {
> //!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
>                 // Enter into this block!!!!!!!!!!!!!!!!!!!
>                 PyEval_ReleaseThread(m_threadState);
>                 return -1;
>         }
>         Py_DECREF(l_retValue);
>
>         PyEval_ReleaseThread(m_threadState);
>
>         //End
>         PyEval_AcquireThread(m_threadState);
>         Py_EndInterpreter(m_threadState);
>         PyEval_ReleaseLock();
>         PyEval_AcquireLock();
>         // Create default interpreter back.
>         Py_NewInterpreter();
>         Py_Finalize();
>
>         return 0;
> }
>
> --
> Best Regards, Irene Ladyko.
>
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>


-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Fri May 18 20:27:55 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 18 May 2007 20:27:55 +0200
Subject: [C++-sig] Exposing a template class
In-Reply-To: <464D7222.4020509@cs.tut.fi>
References: <464D7222.4020509@cs.tut.fi>
Message-ID: <7465b6170705181127u63e13c0fue99b57e175f4ab2a@mail.gmail.com>

On 5/18/07, Pertti Kellom?ki <pk at cs.tut.fi> wrote:
> Are there examples around of exposing instances of
> template classes? I'm assuming it should be as easy
> as enclosing a  BOOST_PYTHON_MODULE inside a template
> and instantiating the template, but it would be reassuring
> to have some hand holding in case there are issues I have
> not thought about.

It is not that clear what you want to do. Exposing instantiation of
template class is not different from exposing non-template class.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From pk at cs.tut.fi  Fri May 18 20:42:50 2007
From: pk at cs.tut.fi (=?ISO-8859-1?Q?Pertti_Kellom=E4ki?=)
Date: Fri, 18 May 2007 21:42:50 +0300
Subject: [C++-sig] Exposing a template class
In-Reply-To: <7465b6170705181127u63e13c0fue99b57e175f4ab2a@mail.gmail.com>
References: <464D7222.4020509@cs.tut.fi>
	<7465b6170705181127u63e13c0fue99b57e175f4ab2a@mail.gmail.com>
Message-ID: <464DF3AA.80701@cs.tut.fi>

Roman Yakovenko wrote:
> It is not that clear what you want to do. Exposing instantiation of
> template class is not different from exposing non-template class.

I have a template class, whose three different instantiations
are used as base classes for some concrete classes. I was
wondering whether I could write a template that would expand
into the appropriate declarations for exposing a particular
instantiation. Just writing the wrapper struct as a template
would of course already go a long way.
-- 
Pertti


From roman.yakovenko at gmail.com  Fri May 18 20:44:26 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 18 May 2007 20:44:26 +0200
Subject: [C++-sig] Exposing a template class
In-Reply-To: <464DF3AA.80701@cs.tut.fi>
References: <464D7222.4020509@cs.tut.fi>
	<7465b6170705181127u63e13c0fue99b57e175f4ab2a@mail.gmail.com>
	<464DF3AA.80701@cs.tut.fi>
Message-ID: <7465b6170705181144l7d23e3dft2084942ace05e6d0@mail.gmail.com>

On 5/18/07, Pertti Kellom?ki <pk at cs.tut.fi> wrote:
> Roman Yakovenko wrote:
> > It is not that clear what you want to do. Exposing instantiation of
> > template class is not different from exposing non-template class.
>
> I have a template class, whose three different instantiations
> are used as base classes for some concrete classes. I was
> wondering whether I could write a template that would expand
> into the appropriate declarations for exposing a particular
> instantiation. Just writing the wrapper struct as a template
> would of course already go a long way.


Do you mean something like this:

template< class T >
void export_xx(const sd::string& name){
    boost::python::class_< T, ... >( name )
         ....
    ;
}

export_xx<X1>( "X1" );
export_xx<X2>( "X2" );
export_xx<X3>( "X3" );

Than yes, you can do this.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From seefeld at sympatico.ca  Fri May 18 20:55:04 2007
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Fri, 18 May 2007 14:55:04 -0400
Subject: [C++-sig] Exposing a template class
In-Reply-To: <7465b6170705181144l7d23e3dft2084942ace05e6d0@mail.gmail.com>
References: <464D7222.4020509@cs.tut.fi>	<7465b6170705181127u63e13c0fue99b57e175f4ab2a@mail.gmail.com>	<464DF3AA.80701@cs.tut.fi>
	<7465b6170705181144l7d23e3dft2084942ace05e6d0@mail.gmail.com>
Message-ID: <464DF688.3040700@sympatico.ca>

Roman Yakovenko wrote:

> template< class T >
> void export_xx(const sd::string& name){
>     boost::python::class_< T, ... >( name )
>          ....
>     ;
> }
> 
> export_xx<X1>( "X1" );
> export_xx<X2>( "X2" );
> export_xx<X3>( "X3" );

or even:

template <class T>
void export_instance(std::string const &name)
{
  boost::python::class_<your_class_template<T>, ...>(name.c_str())
  ...
}

export_instance<int>("int_container");
export_instance<float>("float_container");

> Than yes, you can do this.
> 

By the way, what you are referring to as 'template class' is actually
a class template.

HTH,
		Stefan

-- 

      ...ich hab' noch einen Koffer in Berlin...


From roman.yakovenko at gmail.com  Fri May 18 21:02:02 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 18 May 2007 21:02:02 +0200
Subject: [C++-sig] Exposing a template class
In-Reply-To: <464DF688.3040700@sympatico.ca>
References: <464D7222.4020509@cs.tut.fi>
	<7465b6170705181127u63e13c0fue99b57e175f4ab2a@mail.gmail.com>
	<464DF3AA.80701@cs.tut.fi>
	<7465b6170705181144l7d23e3dft2084942ace05e6d0@mail.gmail.com>
	<464DF688.3040700@sympatico.ca>
Message-ID: <7465b6170705181202w611f6a50wfed90035da40e631@mail.gmail.com>

On 5/18/07, Stefan Seefeld <seefeld at sympatico.ca> wrote:
> By the way, what you are referring to as 'template class' is actually
> a class template.

Thanks.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From pk at cs.tut.fi  Sat May 19 08:33:29 2007
From: pk at cs.tut.fi (=?ISO-8859-1?Q?Pertti_Kellom=E4ki?=)
Date: Sat, 19 May 2007 09:33:29 +0300
Subject: [C++-sig] Exposing a template class
In-Reply-To: <7465b6170705181144l7d23e3dft2084942ace05e6d0@mail.gmail.com>
References: <464D7222.4020509@cs.tut.fi>	<7465b6170705181127u63e13c0fue99b57e175f4ab2a@mail.gmail.com>	<464DF3AA.80701@cs.tut.fi>
	<7465b6170705181144l7d23e3dft2084942ace05e6d0@mail.gmail.com>
Message-ID: <464E9A39.4090708@cs.tut.fi>

Roman Yakovenko wrote:
> Do you mean something like this:
[...]
> Than yes, you can do this.

That's exactly what I was looking for. Thanks!
-- 
Pertti


From srossross at gmail.com  Sat May 19 21:38:12 2007
From: srossross at gmail.com (Sean Ross-Ross)
Date: Sat, 19 May 2007 13:38:12 -0600
Subject: [C++-sig] Boost "From Python" Conversion and Template Classes
Message-ID: <1ABEC499-AD7B-4753-B2F2-0DC027881F4C@gmail.com>

I have a project where there is an internal reference count pointer  
to the objects passed through. And all of class methods are templated  
similar to:

	template<class ScalarType, class MV, class OP>
	class BasicEigenproblem : public virtual Eigenproblem<ScalarType,  
MV, OP> {
	....
		setOperator( const Teuchos::RefCountPtr<OP>& Op );
	....
	}

In the C++ code to call this method you must do:
	
	RefCountPtr< Matrix > A = rcp( new Matrix( ... ) );
	eigenproblem. setOperator(A);


I would like to have the methods in python be just:
	
	A = Matrix()
	eigenproblem. setOperator(A)

where I have templated BasicEigenproblem<double,object,object>.
And without dealing with the RefCountPtr, and have some method that  
calls in the background:

	RefCountPtr< object > RCPA = rcp(  A );


I have tried experimenting with the extraction methods with no luck:

8:	using namespace boost::python;

10:	typedef Teuchos::RefCountPtr<object> RCP;

12:	void* extract_rcp(PyObject* obj)
13:	{
14:	    RCP *r = new RCP( new object(handle<>(obj)), false );
15:	    return r;
16:	}

18:	converter::registry::insert ( &extract_rcp,  type_id<RCP>() );

I get a few errors with this at compile time:

	extraction.cpp:18: error: expected constructor, destructor, or type  
conversion before '(' token

I can not see to find a concrete example of how to do this. Am I one  
the right track, any Sugestions?

Thanks, and sorry for the long winded message.

~Sean




From tim at klingt.org  Sat May 19 23:34:21 2007
From: tim at klingt.org (Tim Blechmann)
Date: Sat, 19 May 2007 21:34:21 +0000
Subject: [C++-sig] call objects with argument-list / keyword
Message-ID: <1179610461.9762.32.camel@localhost>

hi all,

is it possible to call a python object with argument-list as tuple and
keyword-arguments as dictionary?

something like:

object callme;
tuple args;
dict keywords;
callme(args, keywords);

or do i need to call PyObject_Call directly?

thanks, tim

--
tim at klingt.org    ICQ: 96771783
http://tim.klingt.org

You can play a shoestring if you're sincere
  John Coltrane
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070519/12f6c7ca/attachment.pgp>

From fullung at gmail.com  Mon May 21 15:14:05 2007
From: fullung at gmail.com (Albert Strasheim)
Date: Mon, 21 May 2007 15:14:05 +0200
Subject: [C++-sig] Moving "decorator" into C++ part of module
Message-ID: <20070521131405.GA12892@dogbert.sdsl.sun.ac.za>

Hello all

I have a class with various methods that return uBLAS matrices. I've 
wrapped various instantiations of these matrices with:

template<typename T>
struct matrix_wrapper
{
    static void wrap(const char* python_name)
    {
        py::class_<T, boost::noncopyable>(python_name, py::no_init)
            .add_property("__array_struct__", array_struct__)
            ;
    }

    static void cleanup(void* obj) {
        PyArrayInterface* ai = (PyArrayInterface*) obj;
        delete [] ai->shape;
        delete [] ai->strides;
        delete ai;
    }

    static PyObject* array_struct__(T const& self)
    {
        // http://numpy.scipy.org/array_interface.shtml
        PyArrayInterface* ai = new PyArrayInterface;
        // ... more code here ...
        ai->data = (void*) self.data().begin();
        return PyCObject_FromVoidPtr(ai, cleanup);
    }
}

In Python this allows me to do:

foo = WrappedClass()
cmatrix = foo.somefunc()
numpyarr = numpy.asarray(cmatrix)

without a copy of cmatrix's contents having to be made. Now I'd just 
like to get rid of the asarray bit so that it looks like my wrapped 
functions return NumPy arrays. In Python I can "decorate" this function 
like so:

def return_asarray(func):
    from numpy import asarray
    def call(*args, **kwargs):
        return asarray(func(*args, **kwargs))
    return call
WrappedClass.somefunc = return_asarray(WrappedClass.somefunc)

but I would like to move this "decoration" into the C++ code, so that 
my module doesn't need to have the extra Python bits.

Any thoughts on how this could be done? I looked at the documentation 
for class_, but nothing jumped out at me.

Regards,

Albert


From roman.yakovenko at gmail.com  Mon May 21 15:26:45 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Mon, 21 May 2007 16:26:45 +0300
Subject: [C++-sig] Moving "decorator" into C++ part of module
In-Reply-To: <20070521131405.GA12892@dogbert.sdsl.sun.ac.za>
References: <20070521131405.GA12892@dogbert.sdsl.sun.ac.za>
Message-ID: <7465b6170705210626m2f7a5577r72eaa29af9bbe657@mail.gmail.com>

On 5/21/07, Albert Strasheim <fullung at gmail.com> wrote:
>
> Hello all
> ...
> without a copy of cmatrix's contents having to be made. Now I'd just
> like to get rid of the asarray bit so that it looks like my wrapped
> functions return NumPy arrays. In Python I can "decorate" this function
> like so:
>
> def return_asarray(func):
>     from numpy import asarray
>     def call(*args, **kwargs):
>         return asarray(func(*args, **kwargs))
>     return call
> WrappedClass.somefunc = return_asarray(WrappedClass.somefunc)
>
> but I would like to move this "decoration" into the C++ code, so that
> my module doesn't need to have the extra Python bits.
>
> Any thoughts on how this could be done? I looked at the documentation
> for class_, but nothing jumped out at me.
>

Even call policies? If I understand right, you can create new call policy,
implement postcall only member function and that's all.

Am I missing something?

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070521/521ee3d7/attachment.htm>

From ndbecker2 at gmail.com  Mon May 21 15:45:26 2007
From: ndbecker2 at gmail.com (Neal Becker)
Date: Mon, 21 May 2007 09:45:26 -0400
Subject: [C++-sig] Moving "decorator" into C++ part of module
References: <20070521131405.GA12892@dogbert.sdsl.sun.ac.za>
Message-ID: <f2s7pn$lsj$1@sea.gmane.org>

Albert Strasheim wrote:

> Hello all
> 
> I have a class with various methods that return uBLAS matrices. I've
> wrapped various instantiations of these matrices with:
> 
[...]

I'm very interested in this.

I thought array_interface was only a proposal at this point.  You are using
it?  If I have python2.5, do I need something to try this out?  (Like,
where is the array_interface include file?)

Could you post a complete code that shows how you convert ublas::matrix to
numpy?  Do you have converters in the other direction?



From fullung at gmail.com  Mon May 21 15:56:34 2007
From: fullung at gmail.com (Albert Strasheim)
Date: Mon, 21 May 2007 15:56:34 +0200
Subject: [C++-sig] Moving "decorator" into C++ part of module
In-Reply-To: <7465b6170705210626m2f7a5577r72eaa29af9bbe657@mail.gmail.com>
References: <20070521131405.GA12892@dogbert.sdsl.sun.ac.za>
	<7465b6170705210626m2f7a5577r72eaa29af9bbe657@mail.gmail.com>
Message-ID: <20070521135634.GB12892@dogbert.sdsl.sun.ac.za>

Hello

On Mon, 21 May 2007, Roman Yakovenko wrote:

> Even call policies? If I understand right, you can create new call policy,
> implement postcall only member function and that's all.
> 
> Am I missing something?

Thanks, got it working.

Cheers,

Albert


From fullung at gmail.com  Mon May 21 16:02:07 2007
From: fullung at gmail.com (Albert Strasheim)
Date: Mon, 21 May 2007 16:02:07 +0200
Subject: [C++-sig] Moving "decorator" into C++ part of module
In-Reply-To: <f2s7pn$lsj$1@sea.gmane.org>
References: <20070521131405.GA12892@dogbert.sdsl.sun.ac.za>
	<f2s7pn$lsj$1@sea.gmane.org>
Message-ID: <20070521140207.GC12892@dogbert.sdsl.sun.ac.za>

Howdy

On Mon, 21 May 2007, Neal Becker wrote:

> Albert Strasheim wrote:
> 
> > Hello all
> > 
> > I have a class with various methods that return uBLAS matrices. I've
> > wrapped various instantiations of these matrices with:
> > 
> [...]
> 
> I'm very interested in this.
> 
> I thought array_interface was only a proposal at this point.  You are using
> it?  If I have python2.5, do I need something to try this out?  (Like,
> where is the array_interface include file?)

NumPy's array interface is ready to go. I think you might be thinking 
about the enhanced buffer protocol PEP thing.

> Could you post a complete code that shows how you convert ublas::matrix to
> numpy?  Do you have converters in the other direction?

Code attached. This is still very much a work in progress. Test it with:

import numpy as N
import os

from numpy.testing import *
# directory where you compiled the module
set_local_path(os.path.join('..', 'win_build', 'Debug'))
import pyublas
restore_path()

at = pyublas.array_test()
print at
print at.ref_return()
print at.const_ref_return()
print at.pointer_return()
print at.const_pointer_return()

As you can see, I can now convert return types (just haven't implemented 
return by-value yet). Now that I understand call policies properly, I'm 
going to look at converting arguments.

Cheers,

Albert
-------------- next part --------------
#include <boost/numeric/ublas/matrix.hpp>

#include <boost/python.hpp>



#define PYUBLAS_FILE_WITH_INIT

#ifndef PYUBLAS_FILE_WITH_INIT

#define NO_IMPORT_ARRAY

#endif

#include <numpy/arrayobject.h>



// XXX for debugging purposes... remove later

#include <boost/numeric/ublas/io.hpp>

#include <iostream>



namespace ublas = boost::numeric::ublas;

namespace py = boost::python;



template<typename T>

struct matrix_wrapper

{

    static void wrap(const char* python_name)

    {

        py::class_<T, boost::noncopyable>(python_name, py::no_init)

            // This will have to copy for things like symmetric matrices

            .add_property("__array_struct__", array_struct__)

            ;

        // register conversion for value arguments

        py::to_python_converter<T, matrix_wrapper<T> >();

        // return conversion for return by-value

        py::converter::registry::push_back(convertible, construct, py::type_id<T>());

    }



    static void cleanup(void* obj)

    {

        PyArrayInterface* ai = (PyArrayInterface*) obj;

        delete [] ai->shape;

        delete [] ai->strides;

        delete ai;

    }



    static PyObject* array_struct__(T const& self)

    {

        // http://numpy.scipy.org/array_interface.shtml

        PyArrayInterface* ai = new PyArrayInterface;



        // contains the integer 2 as a sanity check

        ai->two = 2;

        // number of dimensions: two for a matrix

        ai->nd = 2;

        // kind in array --- character code of typestr

        ai->typekind = 'f';

        // size of each element

        ai->itemsize = sizeof(T::value_type);



        // how should be data interpreted

        ai->flags = NPY_CONTIGUOUS | NPY_ALIGNED | NPY_NOTSWAPPED | NPY_WRITEABLE;



        // A length-nd array of shape information

        ai->shape = new npy_intp[2];

        ai->shape[0] = self.size1();

        ai->shape[1] = self.size2();



        // A length-nd array of stride information

        ai->strides = new npy_intp[2];

        ai->strides[0] = ai->shape[1] * ai->itemsize;

        ai->strides[1] = ai->itemsize;



        // A pointer to the first element of the array

        ai->data = (void*) self.data().begin();



        ai->descr = NULL;



        return PyCObject_FromVoidPtr(ai, cleanup);

    }



    static void* convertible(PyObject* obj) {

        return PyArray_Check(obj) ? obj : NULL;

    }



    static void construct(PyObject* obj, py::converter::rvalue_from_python_stage1_data* data)

    {

        // get the storage

        typedef py::converter::rvalue_from_python_storage<T> storage_t;

        storage_t* the_storage = reinterpret_cast<storage_t*>(data);

        void* memory_chunk = the_storage->storage.bytes;

        // placement new

        T* a = new (memory_chunk) T(0, 0);   // TODO need a special array adaptor here...

        data->convertible = memory_chunk;

    }



    static PyObject* convert(const T& a) {

        // TODO still need to do some work here

        return py::incref(Py_None);

    }

};



typedef ublas::matrix<double, ublas::row_major, ublas::array_adaptor<double> > matrix_type;



struct array_test

{

    typedef matrix_type array_t;

    array_test() : data_(3, 4)

    {

        for (unsigned i = 0; i < data_.size1(); ++i) {

            for (unsigned j = 0; j < data_.size2(); ++j) {

                data_(i, j) = 3 * i + j;

            }

        }

    }

    array_t value_return() const

    {

        return data_;

    }

    array_t const* const_pointer_return() const

    {

        return &data_;

    }

    array_t* pointer_return()

    {

        return &data_;

    }

    array_t const& const_ref_return() const

    {

        return data_;

    }

    array_t& ref_return()

    {

        return data_;

    }

    void value_arg(array_t a)

    {

        print(a);

    }

    void pointer_arg(array_t* a)

    {

        print(*a);

    }

    void const_pointer_arg(array_t const* a)

    {

    }

    void ref_arg(array_t& a)

    {

    }

    void const_ref_arg(array_t const& a)

    {

    }

private:

    void print(array_t const& a)

    {

        std::cout << a << std::endl;

    }



    array_t data_;

};



template<class BasePolicy_ = default_call_policies>

struct return_asarray : BasePolicy_

{

    template <class ArgumentPackage>

    static PyObject* postcall(ArgumentPackage const& args_, PyObject* result)

    {

        result = BasePolicy_::postcall(args_, result);

        if (result == 0) {

            return 0;

        }

        return PyArray_FromStructInterface(result);

    }

};



BOOST_PYTHON_MODULE(pyublas)

{

    import_array();



    matrix_wrapper<matrix_type>::wrap("matrix_double");



    py::class_<array_test, boost::noncopyable>("array_test")

        .def("value_return", &array_test::value_return)

        .def("const_pointer_return", &array_test::const_pointer_return, return_asarray<py::return_internal_reference<> >())

        .def("pointer_return", &array_test::pointer_return, return_asarray<py::return_internal_reference<> >())

        .def("const_ref_return", &array_test::const_ref_return, return_asarray<py::return_internal_reference<> >())

        .def("ref_return", &array_test::ref_return, return_asarray<py::return_internal_reference<> >())

#if 0

        .def("value_arg", &array_test::value_arg)

        .def("pointer_arg", &array_test::pointer_arg)

        .def("const_pointer_arg", &array_test::const_pointer_arg)

        .def("ref_arg", &array_test::ref_arg)

        .def("const_ref_arg", &array_test::const_ref_arg)

#endif

        ;

}


From nick.lee at volition-inc.com  Mon May 21 18:07:17 2007
From: nick.lee at volition-inc.com (Nick Lee)
Date: Mon, 21 May 2007 11:07:17 -0500
Subject: [C++-sig] building static boost.python libraries in 1.34?
Message-ID: <f2sg3l$tok$1@sea.gmane.org>

I just downloaded Boost 1.34 as well as Boost.Jam 3.1.14. I'm trying to 
manually build the Boost.Python libraries from the command line, but it 
looks like things have changed from 1.33.1. I switched to the 
libs\python\build\ directory and entered the following command:

   bjam --toolset=msvc release

It looks like it built the dynamic-link library 
("boost_python-vc80-mt-1_34.dll") and its corresponding import library 
("boost_python-vc80-mt-1_34.lib"), but the static-link library is nowhere to 
be found. According to the new documentation, the file should be named 
something like "libboost_python-vc80-xxx.lib", but I can't find it anywhere. 
Under 1.33.1, the file would have been named "libboost_python.lib", and the 
command-line call to bjam would have automatically created it.

How do I build the Boost.Python static-link library in 1.34? Is there a 
command-line parameter I'm missing for bjam? 





From ndbecker2 at gmail.com  Mon May 21 20:21:17 2007
From: ndbecker2 at gmail.com (Neal Becker)
Date: Mon, 21 May 2007 14:21:17 -0400
Subject: [C++-sig] Moving "decorator" into C++ part of module
References: <20070521131405.GA12892@dogbert.sdsl.sun.ac.za>
	<f2s7pn$lsj$1@sea.gmane.org>
	<20070521140207.GC12892@dogbert.sdsl.sun.ac.za>
Message-ID: <f2snut$bbk$1@sea.gmane.org>

This looks very good so far.  I see there is c++ ublas -> python numpy
conversion.

Have you tried the other direction?  python numpy -> c++ ublas?



From fullung at gmail.com  Mon May 21 20:54:33 2007
From: fullung at gmail.com (Albert Strasheim)
Date: Mon, 21 May 2007 20:54:33 +0200
Subject: [C++-sig] Moving "decorator" into C++ part of module
In-Reply-To: <f2snut$bbk$1@sea.gmane.org>
References: <20070521131405.GA12892@dogbert.sdsl.sun.ac.za>
	<f2s7pn$lsj$1@sea.gmane.org>
	<20070521140207.GC12892@dogbert.sdsl.sun.ac.za>
	<f2snut$bbk$1@sea.gmane.org>
Message-ID: <20070521185433.GA19431@dogbert.sdsl.sun.ac.za>

Hello

On Mon, 21 May 2007, Neal Becker wrote:

> This looks very good so far.  I see there is c++ ublas -> python numpy
> conversion.

With a few more templates and whatnot one can take care of things 
figuring out that for a matrix<int> the typecode should be 'i', etc. For 
now I'm just focusing on getting double working. But soon...

> Have you tried the other direction?  python numpy -> c++ ublas?

I'm trying to sort this out now. Here there are a few more 
complications:

1. Some array adaptor complications (nothing too major) to make a uBLAS 
matrix use the array's data without a copy

2. Conversion of NumPy array arguments to uBLAS

I'm currently stuck here. I was hoping to do something like:

template <
    class T,
    std::size_t arg,
    class BasePolicy_ = py::default_call_policies
>
struct convert_array : BasePolicy_
{
    BOOST_STATIC_ASSERT(arg > 0);

    template <class ArgumentPackage>
    static bool precall(ArgumentPackage const& args_)
    {
        unsigned int arity_ = PyTuple_GET_SIZE(args_);
        if (arg > arity_) {
            PyErr_SetString(
                PyExc_IndexError,
                "convert_array: argument index out of range");
            return false;
        }
        PyObject* obj = PyTuple_GetItem(args_, arg);
        if (obj == NULL) {
            return false;
        }
        if (!PyArray_Check(obj)) {
            PyErr_SetString(
                PyExc_TypeError,
                "convert_array: ndarray argument expected");
            return NULL;
        }

        // XXX magic happens here see below

        return BasePolicy_::precall(args_);
    }
};

What I wanted to attempt at XXX is to create an instance of a uBLAS 
vector/matrix/whatever inside its associated PyObject, with the 
uBLAS matrix using the memory of the NumPy array. I was hoping I could 
then do a little bait and switch on the tuple item to put this new 
uBLAS-PyObject-using-a-Numpy-array into the arguments before the 
function gets called.

You could then wrap the method something like this:

py::class_<array_test, boost::noncopyable>("array_test")
        .def("pointer_arg",
             &array_test::pointer_arg, 
             convert_array<matrix<float>, 1>())

Unfortunately, it seems checking of argument types happens before the 
the convert_array precall (not much of a *pre*call is it? ;-)), so this 
trick doesn't work (you get an ArgumentError).

Any ideas would help at this point. :-)

Regards,

Albert


From nick.lee at volition-inc.com  Mon May 21 20:55:40 2007
From: nick.lee at volition-inc.com (Nick Lee)
Date: Mon, 21 May 2007 13:55:40 -0500
Subject: [C++-sig] building static boost.python libraries in 1.34?
References: <f2sg3l$tok$1@sea.gmane.org>
Message-ID: <f2spvc$k9j$1@sea.gmane.org>

Um, nevermind. I think I figured it out.

I just ran bjam from the boost root directory rather than from the 
libs\python\build\ directory. Built everything properly. Funny how things 
work better when you follow directions. :)

Niq 





From ndbecker2 at gmail.com  Mon May 21 21:11:43 2007
From: ndbecker2 at gmail.com (Neal Becker)
Date: Mon, 21 May 2007 15:11:43 -0400
Subject: [C++-sig] Moving "decorator" into C++ part of module
References: <20070521131405.GA12892@dogbert.sdsl.sun.ac.za>
	<f2s7pn$lsj$1@sea.gmane.org>
	<20070521140207.GC12892@dogbert.sdsl.sun.ac.za>
	<f2snut$bbk$1@sea.gmane.org>
	<20070521185433.GA19431@dogbert.sdsl.sun.ac.za>
Message-ID: <f2sqtg$o00$1@sea.gmane.org>

Albert Strasheim wrote:

> Hello
> 
> On Mon, 21 May 2007, Neal Becker wrote:
> 
>> This looks very good so far.  I see there is c++ ublas -> python numpy
>> conversion.
> 
> With a few more templates and whatnot one can take care of things
> figuring out that for a matrix<int> the typecode should be 'i', etc. For
> now I'm just focusing on getting double working. But soon...
> 
>> Have you tried the other direction?  python numpy -> c++ ublas?
> 
> I'm trying to sort this out now. Here there are a few more
> complications:
> 
> 1. Some array adaptor complications (nothing too major) to make a uBLAS
> matrix use the array's data without a copy
> 
> 2. Conversion of NumPy array arguments to uBLAS
> 
> I'm currently stuck here. I was hoping to do something like:
> 
> template <
>     class T,
>     std::size_t arg,
>     class BasePolicy_ = py::default_call_policies
>>
> struct convert_array : BasePolicy_
> {
>     BOOST_STATIC_ASSERT(arg > 0);
> 
>     template <class ArgumentPackage>
>     static bool precall(ArgumentPackage const& args_)
>     {
>         unsigned int arity_ = PyTuple_GET_SIZE(args_);
>         if (arg > arity_) {
>             PyErr_SetString(
>                 PyExc_IndexError,
>                 "convert_array: argument index out of range");
>             return false;
>         }
>         PyObject* obj = PyTuple_GetItem(args_, arg);
>         if (obj == NULL) {
>             return false;
>         }
>         if (!PyArray_Check(obj)) {
>             PyErr_SetString(
>                 PyExc_TypeError,
>                 "convert_array: ndarray argument expected");
>             return NULL;
>         }
> 
>         // XXX magic happens here see below
> 
>         return BasePolicy_::precall(args_);
>     }
> };
> 
> What I wanted to attempt at XXX is to create an instance of a uBLAS
> vector/matrix/whatever inside its associated PyObject, with the
> uBLAS matrix using the memory of the NumPy array. I was hoping I could
> then do a little bait and switch on the tuple item to put this new
> uBLAS-PyObject-using-a-Numpy-array into the arguments before the
> function gets called.
> 
> You could then wrap the method something like this:
> 
> py::class_<array_test, boost::noncopyable>("array_test")
>         .def("pointer_arg",
>              &array_test::pointer_arg,
>              convert_array<matrix<float>, 1>())
> 
> Unfortunately, it seems checking of argument types happens before the
> the convert_array precall (not much of a *pre*call is it? ;-)), so this
> trick doesn't work (you get an ArgumentError).
> 
> Any ideas would help at this point. :-)
> 

Mostly I'd be interested in the following.  My use case is I have lots of
c++ algorithms written for ublas interface (or maybe a more generic
superset of that).  I want to use them from python.  I want to call them
with numpy arrays.

Someone has to convert the numpy array to a ublas interface, then call the
c++ algorithm, and maybe convert the ublas object back to numpy to python.  



From jmoorkan at uci.edu  Tue May 22 04:56:25 2007
From: jmoorkan at uci.edu (Jayram)
Date: Mon, 21 May 2007 19:56:25 -0700
Subject: [C++-sig] boost python installation problem
Message-ID: <f2tm4p$q49$1@sea.gmane.org>

Hi all,

My configuration is the following:

MS-Visual Studio 95 - "vc-8_0", Boost = 1.33.1, Python = 2.5

I was trying to build python boost using the command:

>> bjam -sTOOLS=vc-8_0 -sPYTHON_VERSION=2.5 -sPYTHON_ROOT=C:\Python25"

A part of build output is the following which starts immediately after 
compiling all the files.
---------------------------
vc-C++ 
..\..\..\bin\boost\libs\python\build\boost_python.dll\vc-8_0\debug\threading-multi\object_operators.obj
object_operators.cpp
vc-C++ 
..\..\..\bin\boost\libs\python\build\boost_python.dll\vc-8_0\debug\threading-multi\wrapper.obj
wrapper.cpp
vc-Link 
..\..\..\bin\boost\libs\python\build\boost_python.dll\vc-8_0\debug\threading-multi\boost_python_debug.dll 
..\..\..\bin\boost\libs\python\build\boost_python.dll\vc-8_0\debug\threading-multi\boost_python_debug.lib
   Creating library 
..\..\..\bin\boost\libs\python\build\boost_python.dll\vc-8_0\debug\threading-multi\boost_python_debug.lib 
and object 
..\..\..\bin\boost\libs\python\build\boost_python.dll\vc-8_0\debug\threading-multi\boost_python_debug.exp
function.obj : error LNK2019: unresolved external symbol __imp__PyErr_Format 
referenced in function "public: virtual void * __thiscall 
boost::python::error_already_set::`vector deleting destructor'(unsigned 
int)" (??_Eerror_already_set at python@boost@@UAEPAXI at Z)
numeric.obj : error LNK2001: unresolved external symbol __imp__PyErr_Format
from_python.obj : error LNK2001: unresolved external symbol 
__imp__PyErr_Format
registry.obj : error LNK2001: unresolved external symbol __imp__PyErr_Format
-------------------------
and the unresolved external symbol error continues and 
boost_python_debug.dll is not generated.

I have been working only in GNU/Linux build environment, BUT M$ Windows is 
totally new to me.
So can anyone point out how to fix this error ?

What I understood is that vc-Link does not find some important libraries 
during linking which is associated with Python and Boost ?
But I am unable to proceed from here to determine what is the problem and 
how to fix it ?

Thanks and any suggestion to me will be helpful.

- Jay







From luca.sbardella at gmail.com  Tue May 22 12:53:28 2007
From: luca.sbardella at gmail.com (Luca Sbardella)
Date: Tue, 22 May 2007 12:53:28 +0200
Subject: [C++-sig] exposing iterator for the values of std::map
Message-ID: <8315652a0705220353m3e0e41d9k89200115951c9641@mail.gmail.com>

Hi,
I'm trying to expose to python the iterator of a specific class with this
skeleton

class bla;
typedef boost::shared_ptr<bla> BLA;

class bla  {
public:
  typedef std::map<std:string,BLA> container;
  typedef container::iterator  iterator;

  static BLA  get(const std::string& code);

  static iterator  begin() {return s_db.begin();}
  static iterator  end()   {return s_db.end();}

  std::string code() const {return ...}

private:
  static container s_db;
};

the bla class is exposed to Python in the usual way. my question is how do I
expose the iterator such that in Python I can iterate through the elements
in s_db as it was a list (I want to see the BLA elements not the keys of the
map).
In other words I would like to be able to do this

b = bla(...)
for  i in b:
  print i.code()

Naively, I tried to use .def("__iter__", range(&bla::begin,&bla::end)) but,
not surprisingly, it doesn't work (something to do with the implementation
of NextPolicies I assume).

Best
Luca
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070522/e51738f7/attachment.htm>

From rwgk at yahoo.com  Tue May 22 19:14:42 2007
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Tue, 22 May 2007 10:14:42 -0700 (PDT)
Subject: [C++-sig] Better introspection/apidoc extraction support?
Message-ID: <566772.7647.qm@web31107.mail.mud.yahoo.com>

> > How exactly does this work?  Are there web pages about this?
>
> Not that I know of.  AFAICS, there is no public API for that (that's why I 
> wrote "half-private"), but if you look at a functions func_code.co_XXX and 
> func_defaults properties, you see that there is a lot of internal information 
> available in CPython.

Sorry, this looks like more work than I'm able to squeeze in. I was hoping for
a well-established, time-tested API, but it looks more like work in progress.

If someone contributes patches including docs I'd be happy to try them out and check
them in.

The current docstring support isn't all that complicated. Look in boost/libs/python/src/object/function.cpp,
mainly function::signature().

Ralf







      ____________________________________________________________________________________
Park yourself in front of a world of choices in alternative vehicles. Visit the Yahoo! Auto Green Center.
http://autos.yahoo.com/green_center/ 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070522/fc843d8c/attachment.htm>

From pk at cs.tut.fi  Wed May 23 18:17:31 2007
From: pk at cs.tut.fi (=?ISO-8859-1?Q?Pertti_Kellom=E4ki?=)
Date: Wed, 23 May 2007 19:17:31 +0300
Subject: [C++-sig] Transfer of ownership, constructors and Py++
Message-ID: <4654691B.7020906@cs.tut.fi>

I am trying out Py++, and so far it looks very nice.
I figured how to do transfer of ownership for ordinary
member functions using function_transformers.transfer_ownership(),
but is there a simple way to do the same for constructors?
-- 
Pertti


From roman.yakovenko at gmail.com  Wed May 23 19:39:41 2007
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 23 May 2007 20:39:41 +0300
Subject: [C++-sig] Transfer of ownership, constructors and Py++
In-Reply-To: <4654691B.7020906@cs.tut.fi>
References: <4654691B.7020906@cs.tut.fi>
Message-ID: <7465b6170705231039i19f894e9q23ebf8f20c95387e@mail.gmail.com>

On 5/23/07, Pertti Kellom?ki <pk at cs.tut.fi> wrote:
> I am trying out Py++, and so far it looks very nice.

Thanks

> I figured how to do transfer of ownership for ordinary
> member functions using function_transformers.transfer_ownership(),
> but is there a simple way to do the same for constructors?

No, this functionality is missing, but it should be pretty simple to add it.
Please open "Feature request": http://sourceforge.net/tracker/?group_id=118209
Or if you need it badly we can work together on this feature.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From nick.lee at volition-inc.com  Thu May 24 17:24:12 2007
From: nick.lee at volition-inc.com (Nick Lee)
Date: Thu, 24 May 2007 10:24:12 -0500
Subject: [C++-sig] newbie question: returning a tuple of strings
Message-ID: <f34ams$ana$1@sea.gmane.org>

How do I write a C++ function that returns a tuple of strings to Python 
using Boost.Python? Is std::string the only string format that Boost.Python 
recognizes?

Niq 





From seefeld at sympatico.ca  Thu May 24 17:36:17 2007
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Thu, 24 May 2007 11:36:17 -0400
Subject: [C++-sig] newbie question: returning a tuple of strings
In-Reply-To: <f34ams$ana$1@sea.gmane.org>
References: <f34ams$ana$1@sea.gmane.org>
Message-ID: <4655B0F1.2050707@sympatico.ca>

Nick Lee wrote:
> How do I write a C++ function that returns a tuple of strings to Python 
> using Boost.Python? Is std::string the only string format that Boost.Python 
> recognizes?

char const * should work just fine:

object result = make_tuple("first", "second", "third");
...

HTH,
		Stefan

-- 

      ...ich hab' noch einen Koffer in Berlin...


From meine at informatik.uni-hamburg.de  Fri May 25 14:40:54 2007
From: meine at informatik.uni-hamburg.de (Hans Meine)
Date: Fri, 25 May 2007 14:40:54 +0200
Subject: [C++-sig] Better introspection/apidoc extraction support?
In-Reply-To: <566772.7647.qm@web31107.mail.mud.yahoo.com>
References: <566772.7647.qm@web31107.mail.mud.yahoo.com>
Message-ID: <200705251440.54910.meine@informatik.uni-hamburg.de>

Hi Ralf!

Am Dienstag, 22. Mai 2007 19:14:42 schrieb Ralf W. Grosse-Kunstleve:
> Sorry, this looks like more work than I'm able to squeeze in. I was hoping
> for a well-established, time-tested API, but it looks more like work in
> progress.

That's a pity, but I can understand it.  OTOH, I don't think there is much 
progress ATM, as long as PEP 362 is not implemented.  (Is there a reason you 
did not reply to my later post where I mentioned 
http://www.python.org/dev/peps/pep-0362/ - I just want to make sure you saw 
it.)

I indeed wrote a parser for the C++ signatures for epydoc, but it is quite 
complicated since I have to
- parse arbitrary C++ types (templates with argument literals,
  namespaces, ...), which made me hack together a rather largish
  pyparsing-based parser (and as you know, writing a C++ parser is not what
  you wanna do..)
- parse arbitrary Python __repr__ output for the default arguments.
  Again, that is not doable with just regular expressions, also the __repr__
  can be basically anything, so it cannot be 100% safe.

It would be much better if there was an API to extract argument names and 
default values.

> The current docstring support isn't all that complicated. Look in
> boost/libs/python/src/object/function.cpp, mainly function::signature().

When using epydoc, I noticed that the resulting docstrings are no valid reST 
anymore if the original docstring does not end with a \n:

"""foo bar
C++ signature:
   test(void) -> void"""

However, this would be a valid description list:

"""foo bar

C++ signature:
   test(void) -> void"""

Would it make sense to just add two \n instead of one in
libs/python/src/object/function.cpp:510?

Or look at the attached patch, which also cares about docstrings to which a \n 
has manually been appended for boost 1.34.0beta + epydoc support.

That does not yet lead to a very pretty output, but it is a minimal change 
that makes epydoc w/ reST work at least, without further modifications.

-- 
Ciao, /  /
     /--/
    /  / ANS
-------------- next part --------------
A non-text attachment was scrubbed...
Name: boost_separated_cppsignatures.diff
Type: text/x-diff
Size: 602 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070525/93a94b4f/attachment.diff>

From srossross at gmail.com  Fri May 25 17:18:53 2007
From: srossross at gmail.com (Sean Ross-Ross)
Date: Fri, 25 May 2007 09:18:53 -0600
Subject: [C++-sig] exposing Template methods and classes
Message-ID: <69E14FC7-CB04-49E8-BBBF-E62625042FBD@gmail.com>


Is there a way that I can smartly template classes or functions  
depending on the type of python Object passed to them?

for example, say you expose a vector class and as a constructor pass  
a extra argument "dtype", like numpy, that templates the vector to  
that  type.

or any what at all that does not involve templating it on a PyObject.  
So I don't have to template every type that I could possibly want?

~Sean





From Lawrence.Spector at CanfieldSci.com  Fri May 25 20:35:52 2007
From: Lawrence.Spector at CanfieldSci.com (Lawrence Spector)
Date: Fri, 25 May 2007 14:35:52 -0400
Subject: [C++-sig] [Boost.Python] Issues Building Boost.Python Extensions
	with Boost.Build v2 using Boost 1.34.0
In-Reply-To: <69E14FC7-CB04-49E8-BBBF-E62625042FBD@gmail.com>
References: <69E14FC7-CB04-49E8-BBBF-E62625042FBD@gmail.com>
Message-ID: <EC0FE44F31A3714F96D382F93BCB6003929C6A@Telesto.domain-01>

I successfully was building Boost.Python extensions using bjam version 1
in Boost 1.33.1.  Since the upgrade, I converted my Jamfile to
Jamfile.v2 and am getting the following errors when I try to build:

bjam -sTOOLS=vc-8_0 -sBOOST_ROOT=C:\3rdparty\boost_1_34_0 -sBUILD=debug
error: Could not find parent for project at '.'
error: Did not find Jamfile or project-root.jam in any parent directory.

Here's my Jamfile:

# This is the top of our own project tree
project-root ;

import python ;

extension MyLib  # Declare a Python extension
:   MyLib.cpp                    # source
    # requirements and dependencies for Boost.Python extensions
    <template>@boost/libs/python/build/extension
    ## Requirements ##
    : <include>$(MyDir)/Include
      <define>_ATL_DLL <define>_AFXDLL
      <define>_UNICODE <define>UNICODE
      <runtime-link>dynamic 
    ;

--------------------------------

Here's my Jamrules:

path-global BOOST_ROOT : C:/3rdparty/boost_1_34_0 ;

project boost : $(BOOST_ROOT) ;

--------------------------------

One more note, I tried the quickstart example as suggested in the Python
tutorial and got this:

C:\3rdparty\boost_1_34_0\libs\python\example\quickstart>bjam
toolset=msvc --verbose-test test --v2
Jamroot:17: in modules.load 
rule python-extension unknown in module
Jamfile</C:/3rdparty/boost_1_34_0/libs/python/example/quickstart>.
C:/3rdparty/boost_1_34_0/tools/build/v2/build\project.jam:312: in
load-jamfile
C:/3rdparty/boost_1_34_0/tools/build/v2/build\project.jam:68: in load
C:/3rdparty/boost_1_34_0/tools/build/v2/build\project.jam:170: in
project.find
C:/3rdparty/boost_1_34_0/tools/build/v2\build-system.jam:237: in load
C:\3rdparty\boost_1_34_0\tools\build\v2/kernel\modules.jam:261: in
import
C:\3rdparty\boost_1_34_0\tools\build\v2/kernel/bootstrap.jam:132: in
boost-build

C:\3rdparty\boost_1_34_0\libs\python\example\quickstart\boost-build.jam:
7: in module scope

When using the following command:

bjam --toolset=msvc --verbose-test test

Any ideas?

Thanks in advance,

Lawrence


From fullung at gmail.com  Fri May 25 18:27:47 2007
From: fullung at gmail.com (Albert Strasheim)
Date: Fri, 25 May 2007 18:27:47 +0200
Subject: [C++-sig] exposing Template methods and classes
References: <69E14FC7-CB04-49E8-BBBF-E62625042FBD@gmail.com>
Message-ID: <004f01c79ee9$a1454af0$0100a8c0@sun.ac.za>

Hello

----- Original Message ----- 
From: "Sean Ross-Ross" <srossross at gmail.com>
To: "Development of Python/C++ integration" <c++-sig at python.org>
Sent: Friday, May 25, 2007 5:18 PM
Subject: [C++-sig] exposing Template methods and classes


> Is there a way that I can smartly template classes or functions
> depending on the type of python Object passed to them?
>
> for example, say you expose a vector class and as a constructor pass
> a extra argument "dtype", like numpy, that templates the vector to
> that  type.
>
> or any what at all that does not involve templating it on a PyObject.
> So I don't have to template every type that I could possibly want?

You're going to have to template on every type you possibly want, or the 
code that has to deal with that type isn't going to be in your extension 
module.

My code is still a bit messy, but I'm playing with the idea of using 
Boost.Python's def_visitor to make this a bit easier. Take a look here:

http://pyspkrec.googlecode.com/svn/numpycpp/src/multi_array_test.cpp
http://pyspkrec.googlecode.com/svn/numpycpp/src/ublas_matrix_test.cpp

I have a struct called generic_visitor which basically contains all the 
repetitive type stuff, and then when I wrap a specific method, I write 
something much smaller (a copy_visitor in those files) to wrap a specific 
method. Then the def part only says: def(generic_visitor<copy_visitor>()).

Hope this helps.

Cheers,

Albert 



From gazihan at cse.wustl.edu  Sat May 26 01:10:02 2007
From: gazihan at cse.wustl.edu (Gazihan Alankus)
Date: Fri, 25 May 2007 18:10:02 -0500
Subject: [C++-sig] Boost.Python std::cerr segfault when built with -lGL
Message-ID: <f37qch$p5q$1@sea.gmane.org>

Hello, 

I'm using Boost.Python to create a python library. When I use -lGL in my
linker flags, the first string I send to std::cerr results in an exception.
When I leave -lGL out, everything is fine and std::cerr works with no
problems. I get this problem when I add -lGL to the simplest Boost.Python
sample program that has nothing to do with OpenGL. Any ideas? Does anybody
else see this behavior? 

I'm on Gentoo Linux, using gcc 4.1.1, Python 2.4 and Boost 1.33. I tried
Python 2.3 too and it's the same. 

-Gazi 





From seefeld at sympatico.ca  Sat May 26 01:31:51 2007
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Fri, 25 May 2007 19:31:51 -0400
Subject: [C++-sig] Boost.Python std::cerr segfault when built with -lGL
In-Reply-To: <f37qch$p5q$1@sea.gmane.org>
References: <f37qch$p5q$1@sea.gmane.org>
Message-ID: <465771E7.2060104@sympatico.ca>

Gazihan Alankus wrote:
> Hello, 
> 
> I'm using Boost.Python to create a python library. When I use -lGL in my
> linker flags, the first string I send to std::cerr results in an exception.
> When I leave -lGL out, everything is fine and std::cerr works with no
> problems. I get this problem when I add -lGL to the simplest Boost.Python
> sample program that has nothing to do with OpenGL. Any ideas? Does anybody
> else see this behavior? 
> 
> I'm on Gentoo Linux, using gcc 4.1.1, Python 2.4 and Boost 1.33. I tried
> Python 2.3 too and it's the same. 

(Is this 'GL' name specific to Gentoo ? On my distribution I would name it 'gl',
 i.e. all minor case.)

Have you tried reducing your test case to the point where boost.python isn't
part of the setup any more ? It doesn't sound as if your problem is related to
neither Python nor boost, in any case.

Regards,
		Stefan


-- 

      ...ich hab' noch einen Koffer in Berlin...


From gazihan at cse.wustl.edu  Sat May 26 02:15:35 2007
From: gazihan at cse.wustl.edu (Gazihan Alankus)
Date: Fri, 25 May 2007 19:15:35 -0500
Subject: [C++-sig] Boost.Python std::cerr segfault when built with -lGL
References: <f37qch$p5q$1@sea.gmane.org> <465771E7.2060104@sympatico.ca>
Message-ID: <f37u7e$4nq$1@sea.gmane.org>

Stefan Seefeld wrote:
> (Is this 'GL' name specific to Gentoo ? On my distribution I would name it
> 'gl',
>  i.e. all minor case.)

Yes, it's libGL.so in my system and the file was created by mesa. I remember
having it as libgl.so in another system.

> 
> Have you tried reducing your test case to the point where boost.python
> isn't part of the setup any more ? It doesn't sound as if your problem is
> related to neither Python nor boost, in any case.

Yes, apparently I wasn't right about that -lGL causing that, althought it
looked like that for the last hour before posting that message. Now it's
almost randomly based on my build parameters. I'll post later when I figure
things out. Thanks for the reply. 

-Gazi 




From ajay.murthy at philips.com  Tue May 29 11:32:00 2007
From: ajay.murthy at philips.com (Ajay Murthy)
Date: Tue, 29 May 2007 15:02:00 +0530
Subject: [C++-sig] Exposing Class to Python
Message-ID: <OF749430DA.15731EEB-ON652572EA.0033AE67-652572EA.00342F61@philips.com>

Hi,
I have been trying extending python with C++.
Now I am aware of exposing a list of methods to the python side using the 
method table(PyMethodDef) and initialising the module with this method 
table ( Py_InitModule)
Even though this serves the purpose. the methods now belong to the module 
and not to the class.

I would like to call the methods on the object to which it belongs and not 
by using the module.

So, what do i do to expose the class also to the python side?  And how do 
i bring in the relationship b/w the Class and its methods.

thanks in advance,
Ajay
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070529/722fe8e9/attachment.htm>

From gabriel.becedillas at corest.com  Tue May 29 20:06:25 2007
From: gabriel.becedillas at corest.com (Gabriel Becedillas)
Date: Tue, 29 May 2007 15:06:25 -0300
Subject: [C++-sig] sp_counted_impl_pd with shared_ptr_deleter
Message-ID: <465C6BA1.7050306@corest.com>

I have the following problem:
1_) A boost::shared_ptr<T> is converted to python.
2_) The PyObject holding the boost::shared_ptr<T> is converted back from 
python, (apparently) yielding the same boost::shared_ptr<T>.
3_) The boost::shared_ptr<T> hits refcount 0 somewhere else in C++Land 
without holding the Python GIL.
4_) Python's API gets called without holding the GIL because a 
sp_counted_impl_pd with a shared_ptr_deleter modifies the PyObject's 
refcount and Py_Dealloc gets called (check shared_ptr_deleter::operator()).

This is wrong because you shouldn't use Python's API without holding the 
GIL. In my particular case I don't know when or where a 
boost::shared_ptr<T> will hit refcount 0 (thats the funny part of using 
smart pointers).

I took a look around and found that:
1_) shared_ptr_from_python is putting the shared_ptr_deleter.
2_) shared_ptr_to_python is checking if the boost::shared_ptr<T> holds a 
shared_ptr_deleter to get the PyObject inside.

Is there a way to make shared_ptr_from_python return the original 
boost::shared_ptr<T> without changing the 
sp_counted_impl_pd/shared_ptr_deleter ?

Thanks.


From gazihan at cse.wustl.edu  Wed May 30 10:28:28 2007
From: gazihan at cse.wustl.edu (Gazihan Alankus)
Date: Wed, 30 May 2007 03:28:28 -0500
Subject: [C++-sig] Boost.Python std::cerr segfault when built with -lGL
References: <f37qch$p5q$1@sea.gmane.org> <465771E7.2060104@sympatico.ca>
	<f37u7e$4nq$1@sea.gmane.org>
Message-ID: <f3jcjf$vu6$1@sea.gmane.org>

Gazihan Alankus wrote:

> Stefan Seefeld wrote:
>> 
>> Have you tried reducing your test case to the point where boost.python
>> isn't part of the setup any more ? It doesn't sound as if your problem is
>> related to neither Python nor boost, in any case.
> 
> Yes, apparently I wasn't right about that -lGL causing that, althought it
> looked like that for the last hour before posting that message. Now it's
> almost randomly based on my build parameters. I'll post later when I
> figure things out. Thanks for the reply.

My doubts were on that I'm not using bjam and I thought my build parameters
could be the cause, but I just disproved it. I edited the sample
libs/python/example/quickstart/extending.cpp to write something to
std::cerr and compiled it using what bjam -o tells me and it ran fine. When
I added -lGL to that and it created the segfault again. So I'm pretty sure
that when I link a Boost.Python library to OpenGL, usage of std::cerr gives
a segfault. gdb tells me it's being encountered in operator<<().
fprintf(stderr, "something") works fine however. And yes, I have no
problems with cerr and OpenGL when Boost.Python is not around. Btw, I
recompiled my gcc (libstdc++ and everything), python, boost and mesa in
case there was some kind of corruption before. 

So I'm not sure how to track this down right now. I have no other problem in
my system, it's obvious to me that the problem must be in Boost.Python or
something related to OpenGL. Since I have no other OpenGL problems, my bets
are on Boost.Python. I modified the simple example in
libs/python/example/quickstart/extending.cpp of the boost 1.34 source to
get this error behavior. Here is the hello class:

  class hello
  {
    public:
      hello(const std::string& country) { this->country = country; }
      std::string greet() const {
      std::cout << "cout" << std::cerr.eof() << " " << std::cerr.fail()
<< " " << std::cerr.bad() << " " << std::cerr.exceptions() << "\n";
      std::cerr << "cerr" << "\n";
      return "Hello from " + country;
      }
    private:
      std::string country;
  };

And this is what I do in python:

import extending; a = extending.hello('aoeu'); a.greet()

The eof, fail, etc are 0 in both, but I get the segfault when -lGL is there. 

I found out that I'm actually not the only one:
http://mail.datenhain.de/pipermail/libavg-users/2007-March/000085.html
http://vdrift.net/Forum/viewtopic.php?p=5128

So, any idea how to track this problem down would be appreciated. I hope
this can result in a meaningful bug report in the end. 

-Gazi 




From srossross at gmail.com  Wed May 30 18:26:09 2007
From: srossross at gmail.com (Sean Ross-Ross)
Date: Wed, 30 May 2007 10:26:09 -0600
Subject: [C++-sig] Extract
Message-ID: <5CEE5A34-AF18-45ED-955F-DFABBDBE8784@gmail.com>

I have a templated class PatameterList that mimics a c++ map or a  
python dict.
it has a method

	template<class T>
	void set( string name,  T item )

how can I add a method in that will extract any c++ class from python  
and put it into the PatameterList

I was thinking along the lines of

adding a method
	
	void my_set( PatameterList const& self,string, name,  
boost::python::object o)
	{
		self.set(name , extract< whatever > ( o ) );
	}
	...
	class_< PatameterList > ( ...
	...
	.def( "set" , &my_set )
	...

where I don't know how to implement the "whatever" arg.

~Sean
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070530/875cd0c2/attachment.htm>

From seefeld at sympatico.ca  Wed May 30 19:24:26 2007
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Wed, 30 May 2007 13:24:26 -0400
Subject: [C++-sig] Extract
In-Reply-To: <5CEE5A34-AF18-45ED-955F-DFABBDBE8784@gmail.com>
References: <5CEE5A34-AF18-45ED-955F-DFABBDBE8784@gmail.com>
Message-ID: <465DB34A.7030401@sympatico.ca>

Sean Ross-Ross wrote:
> I have a templated class PatameterList that mimics a c++ map or a python
> dict.
> it has a method
> 
> template<class T>
> void set( string name,  T item )
> 
> how can I add a method in that will extract any c++ class from python
> and put it into the PatameterList
> 
> I was thinking along the lines of 
> 
> adding a method 
> void my_set( PatameterList const& self,string, name,
> boost::python::object o)
> {
> self.set(name , extract< whatever > ( o ) );
> }
> ...
> class_< PatameterList > ( ...
> ...
> .def( "set" , &my_set )
> ...
> 
> where I don't know how to implement the "whatever" arg.

I'm a little confused, and I'm not sure what exactly you want to do (as you seem
to be mixing templates and types).

If ParameterList is a class template, you have to instantiate a type from it
in order to export ot to python. Once this is done, 'T' (and thus, 'whatever'),
are bound to specific types, so all the rest should 'just work':

template <typename T>
class ParameterList
{
public:
  void set(std::string const &name, T value);
...
};

void my_set(ParameterList<Foo> &list, std::string const &name, object value)
{ list.set(name, extract<Foo>(value));}

class_<ParameterList<Foo> >
  ...
  .def("set", my_set);


However, this my_set indirection shouldn't even be necessary, since the conversion
between Python and C++ of Foo objects is implicit (assuming Foo is properly exported).
The cases where you'd need an explicit conversion like this is when you want to pass
tuples, or lists, and you need to extract the elements one-by-one inside C++.

HTH,
		Stefan


-- 

      ...ich hab' noch einen Koffer in Berlin...


From gazihan at cse.wustl.edu  Wed May 30 20:08:12 2007
From: gazihan at cse.wustl.edu (Gazihan Alankus)
Date: Wed, 30 May 2007 13:08:12 -0500
Subject: [C++-sig] Boost.Python std::cerr segfault when built with -lGL
References: <f37qch$p5q$1@sea.gmane.org> <465771E7.2060104@sympatico.ca>
	<f37u7e$4nq$1@sea.gmane.org> <f3jcjf$vu6$1@sea.gmane.org>
Message-ID: <f3keif$65m$1@sea.gmane.org>

Gazihan Alankus wrote:

> I found out that I'm actually not the only one:
> http://mail.datenhain.de/pipermail/libavg-users/2007-March/000085.html
> http://vdrift.net/Forum/viewtopic.php?p=5128

Quick addition: Both these other people are using Boost.Python and I
couldn't find anyone with the same problem that don't use Boost.Python. I'm
having this on Gentoo Linux with a laptop with intel 855 graphics card.
Here are versions of everything I tried these with:

gcc 4.1.2 
python 2.4 and 2.3
mesa 6.5.2 (this is where the opengl library comes from)
boost 1.33.1 and 1.34.0

I'm trying to find out what the other people have so we can pin this down. I
just filed a bug report:
http://svn.boost.org/trac/boost/ticket/1018

-Gazi 




From seefeld at sympatico.ca  Wed May 30 20:13:38 2007
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Wed, 30 May 2007 14:13:38 -0400
Subject: [C++-sig] Boost.Python std::cerr segfault when built with -lGL
In-Reply-To: <f3keif$65m$1@sea.gmane.org>
References: <f37qch$p5q$1@sea.gmane.org>
	<465771E7.2060104@sympatico.ca>	<f37u7e$4nq$1@sea.gmane.org>
	<f3jcjf$vu6$1@sea.gmane.org> <f3keif$65m$1@sea.gmane.org>
Message-ID: <465DBED2.2040302@sympatico.ca>

Gazihan Alankus wrote:
> Gazihan Alankus wrote:
> 
>> I found out that I'm actually not the only one:
>> http://mail.datenhain.de/pipermail/libavg-users/2007-March/000085.html
>> http://vdrift.net/Forum/viewtopic.php?p=5128
> 
> Quick addition: Both these other people are using Boost.Python and I
> couldn't find anyone with the same problem that don't use Boost.Python. I'm
> having this on Gentoo Linux with a laptop with intel 855 graphics card.
> Here are versions of everything I tried these with:
> 
> gcc 4.1.2 
> python 2.4 and 2.3
> mesa 6.5.2 (this is where the opengl library comes from)
> boost 1.33.1 and 1.34.0
> 
> I'm trying to find out what the other people have so we can pin this down. I
> just filed a bug report:
> http://svn.boost.org/trac/boost/ticket/1018

Can you send a complete (but minimal !) test case exhibiting the problem ?
I may have a look into it.

Thanks,
		Stefan

-- 

      ...ich hab' noch einen Koffer in Berlin...


From srossross at gmail.com  Wed May 30 20:19:57 2007
From: srossross at gmail.com (Sean Ross-Ross)
Date: Wed, 30 May 2007 12:19:57 -0600
Subject: [C++-sig] Extract
In-Reply-To: <465DB34A.7030401@sympatico.ca>
References: <5CEE5A34-AF18-45ED-955F-DFABBDBE8784@gmail.com>
	<465DB34A.7030401@sympatico.ca>
Message-ID: <26FE1678-45C8-4C63-9384-81CDF23B0EB5@gmail.com>

> If ParameterList is a class template, you have to instantiate a  
> type from it
> in order to export ot to python. Once this is done, 'T' (and thus,  
> 'whatever'),
> are bound to specific types, so all the rest should 'just work':

No, Parameter List is not a class template but its method set is  
templated

	class ParameterList {
		
		...
		template<typename T>
   		void set( string const& name, T const& value, std::string const&  
docString = "");
		...

And yes I could create a method to use only a limited number of types  
but I would like to pass a python object in as "value" and have boost  
try and extract a c++ object to pass to set.
I have no control over this class, it is part of a software package.

~Sean
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070530/73a3083c/attachment.htm>

From seefeld at sympatico.ca  Wed May 30 20:39:07 2007
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Wed, 30 May 2007 14:39:07 -0400
Subject: [C++-sig] Extract
In-Reply-To: <26FE1678-45C8-4C63-9384-81CDF23B0EB5@gmail.com>
References: <5CEE5A34-AF18-45ED-955F-DFABBDBE8784@gmail.com>	<465DB34A.7030401@sympatico.ca>
	<26FE1678-45C8-4C63-9384-81CDF23B0EB5@gmail.com>
Message-ID: <465DC4CB.6050805@sympatico.ca>

Sean Ross-Ross wrote:

> And yes I could create a method to use only a limited number of types
> but I would like to pass a python object in as "value" and have boost
> try and extract a c++ object to pass to set.

Python isn't statically typed, but it is strongly typed. Thus, you
have to write a (possibly big) series of if/else statements (or conceptually equivalent)
that lets you dispatch to the appropriate set<whatever>() method. There is
no way to automatize that and let boost handle it.

void my_set(ParameterList &list, std::string const &name, object value)
{
  if (is<foo>(value)) list.set(name, extract<foo>(value));
  else if (is<bar>(value)) list.set(name, extract<bar>(value));
  else if (is<baz>(value)) list.set(name, extract<baz>(value));

etc., you get the idea. (it is up to you to implement the is<T>(object) function...)

Regards,
		Stefan

-- 

      ...ich hab' noch einen Koffer in Berlin...


From gazihan at cse.wustl.edu  Wed May 30 20:43:16 2007
From: gazihan at cse.wustl.edu (Gazihan Alankus)
Date: Wed, 30 May 2007 13:43:16 -0500
Subject: [C++-sig] Boost.Python std::cerr segfault when built with -lGL
References: <f37qch$p5q$1@sea.gmane.org>
	<465771E7.2060104@sympatico.ca>	<f37u7e$4nq$1@sea.gmane.org>
	<f3jcjf$vu6$1@sea.gmane.org> <f3keif$65m$1@sea.gmane.org>
	<465DBED2.2040302@sympatico.ca>
Message-ID: <f3kgk7$cag$1@sea.gmane.org>

Stefan Seefeld wrote:

> Can you send a complete (but minimal !) test case exhibiting the problem ?
> I may have a look into it.
> 
> Thanks,
> Stefan
> 

Ok, I'm attaching a very simple source file. Here is how I compiled it along
the recommendations of bjam -o: (-g or not didn't matter)

"g++"  -ftemplate-depth-128 -O0 -fno-inline -Wall -fPIC    -I"../../../.." -I"/usr/include/python2.4" -c -o "src.o" "src.cpp"

"g++"    -o "mod.so" -Wl,-h -Wl,mod.so -shared -Wl,--start-group "src.o" /usr/lib/libboost_python.so -L/usr/local/lib -lGL  -lutil -lpthread -ldl -Wl,--end-group

Here is how I used it:


$ python
Python 2.4.4 (#1, May 25 2007, 22:22:35)
[GCC 4.1.1 (Gentoo 4.1.1-r3)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import mod
>>> mod.func()
cerr!Segmentation fault


Thank you for your interest. If you don't get the same error, I can run any
test programs, give you traces or if you prefer I can open up a user
account for you in my laptop so that you can ssh in. 

-Gazi 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: src.cpp
Type: text/x-c++src
Size: 174 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20070530/b0396dd2/attachment.cpp>

From seefeld at sympatico.ca  Wed May 30 20:54:50 2007
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Wed, 30 May 2007 14:54:50 -0400
Subject: [C++-sig] Boost.Python std::cerr segfault when built with -lGL
In-Reply-To: <f3kgk7$cag$1@sea.gmane.org>
References: <f37qch$p5q$1@sea.gmane.org>	<465771E7.2060104@sympatico.ca>	<f37u7e$4nq$1@sea.gmane.org>	<f3jcjf$vu6$1@sea.gmane.org>
	<f3keif$65m$1@sea.gmane.org>	<465DBED2.2040302@sympatico.ca>
	<f3kgk7$cag$1@sea.gmane.org>
Message-ID: <465DC87A.6080604@sympatico.ca>

Gazihan Alankus wrote:

> Thank you for your interest. If you don't get the same error, I can run any
> test programs, give you traces or if you prefer I can open up a user
> account for you in my laptop so that you can ssh in. 

I followed your suggested procedure, and everything seems to run fine. No segfault.
I would expect the problem to be somewhere deep down in system libraries and the
compiler (toolchain), not directly related to boost or mesa. (Mesa, as a C library,
doesn't even know about std::cerr...)

(I'm using FC6, with GCC 4.1.1 as system compiler.)

Sorry, but I don't think I can help any further at this time.

Regards,
		Stefan

-- 

      ...ich hab' noch einen Koffer in Berlin...