From adruab at hotmail.com  Mon Mar  1 02:02:53 2004
From: adruab at hotmail.com (Adrian Bentley)
Date: Sun, 29 Feb 2004 17:02:53 -0800
Subject: [C++-sig] Boost... Exception safe (can I turn it off?),
	and controlled object construction?
Message-ID: <BAY7-DAV8eSJIEoPXCF0001f12d@hotmail.com>

Hi,

I'd like to turn off the exception forwarding feature of boost python (trying to adapt for use in high performance where exceptions are not desired).  What I'd really like to do is be able to handle the exception classically, by giving it a function to call or the like.  I've read through some of the documentation and haven't found anything like this. Any pointers would be welcome.

Also, I'm trying to use the language in kind of a weird way.  It's probably a bad idea but I'm running with it anyway :)...  I'm trying to embed python as a scripting language for my object system thing (game engine framework), and I'd like to be able to extend objects using python (extend embed at the same time! :P).

Example of something like what I'm trying to do:

//In pyobject file
class BaseObject
{
    void *data; //where the variables will all go
    //All it's stuff in here, some which requires manual construction and such
};

Pythonclass : BaseObject
{
    //add stuff here
}

//in actual program
class Py_wrapper : public BaseObject
{
    PyObject    *pyo;    //gotten from embedded python

    //forward variable requests to python object etc....
};

// End example

Ok, now that you all know I'm crazy the reason why I want to do this is that the base object is designed to communicate a certain way with things it's attached to, which probably wouldn't cooperate with the way python works, so I need to manually construct the base object (also not sure how to do this in boost python).

Basically what I'd like is to have python override things like a class normally would, and track it in my object system.  I'm kind of doubting there's a way to do that, hence then extra layer of indirection py_wrapper (which should work).

So yeah, if anyone has any ideas I'd really appreciate it.  Feel free to enlighten me to a better method of interoperation (I'm kinda new to multi language stuff).  

Any help is appreciated, thanks,
Adruab > Adrian Bentley
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040229/81c39ca5/attachment.htm>

From ingo at fargonauten.de  Mon Mar  1 10:08:19 2004
From: ingo at fargonauten.de (Ingo Luetkebohle)
Date: Mon, 01 Mar 2004 10:08:19 +0100
Subject: [C++-sig] embedding examples
Message-ID: <1078132099.32100.6.camel@localhost>

Hi,

I have trouble compiling the 'Using the interpreter' examples from the
tutorial.

Code like this:
  handle<> main_module(borrowed( PyImport_AddModule("__main__") ));
  dict main_namespace(handle<>(borrowed(
PyModule_GetDict(main_module.get()) )));
results in 
  error: variable declaration is not allowed here

using gcc 3.2 and gcc 3.3 on Linux.

When I use
  handle<> main_module(borrowed( PyImport_AddModule("__main__") ));
  handle<> main_nsh(borrowed( PyModule_GetDict(main_module.get()) ));
  dict main_namespace(main_nsh);

it compiles and the resulting dict is usable.

So, this is not a showstopper but it doesn't work as documented and
results in longer code.  Is there anything I can do to get the intended
behaviour?

regards

-- 
Ingo

Soll doch jeder bleiben, wie er gerne w?re.





From darylew at hotmail.com  Mon Mar  1 10:26:43 2004
From: darylew at hotmail.com (Daryle Walker)
Date: Mon, 01 Mar 2004 04:26:43 -0500
Subject: [C++-sig] Software development levels
In-Reply-To: <u8yilepd4.fsf@boost-consulting.com>
Message-ID: <BC686C03.7584%darylew@hotmail.com>

On 2/29/04 1:01 PM, "David Abrahams" <dave at boost-consulting.com> wrote:

> Daryle Walker <darylew at hotmail.com> writes:
[SNIP]
>> If it's missing features, I would consider it alpha-level software.  Recall
>> that the alpha/beta/final scale relates to feature completeness, NOT code
>> quality!  
> 
> I don't "recall that", and it seems to contradict what you say
> below...

I'm sorry, I forgot that the alpha/beta/final scale isn't formally taught in
general.

>> The two usually go hand in hand, but not always.  (There are
>> programs that are awesome even in the alpha stage, and craptacular final
>> stage programs.  The latter leads to accusations of "rushing the SW out the
>> door and forcing the beta-testing on [paying] users".)
> 
> If it's all about features and not quality, what do the terms "alpha
> testing" and "beta testing" mean?

Alpha stage--not feature complete
Beta stage--feature complete, but may have some bugs

Alpha testing--checking if the features that are there work
Beta testing--shaking out all the bugs

-- 
Daryle Walker
Mac, Internet, and Video Game Junkie
darylew AT hotmail DOT com




From pierre.barbier at cirad.fr  Mon Mar  1 10:56:48 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Mon, 01 Mar 2004 10:56:48 +0100
Subject: [C++-sig] Pickling problem
Message-ID: <1078135007.742.6.camel@pbarbier>

Hi !

I have a problem with pickling a proxy class. I join the C++ code as an
attached file.

I have three classes :
 - Base : an abstract class
 - Derived : a subclass of Base
 - Proxy : the proxy class that contains an instance of Base

Derived objects can be pickled and the pickling of Proxy just return a
tuple with a Base object in it. Here is a python sample :


>>> import test_abstract
>>> p = test_abstract.Proxy(test_abstract.Derived(10))
>>> import cPickle
>>> cPickle.dumps(p.getBase())
'ctest_abstract\nDerived\np1\n(I10\ntRp2\n.'
>>> cPickle.dumps(p)
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
cPickle.PicklingError: Can't pickle <type 'Boost.Python.instance'>:
import of module Boost.Python failed

I really don't know what to do !

Thanks,

Pierre

-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: test_abstract.cc
Type: text/x-c++
Size: 3972 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040301/915ea7d5/attachment.bin>

From pierre.barbier at cirad.fr  Mon Mar  1 11:42:59 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Mon, 01 Mar 2004 11:42:59 +0100
Subject: [C++-sig] Pickling problem
In-Reply-To: <1078135007.742.6.camel@pbarbier>
References: <1078135007.742.6.camel@pbarbier>
Message-ID: <1078137779.693.13.camel@pbarbier>

I think I've found the problem (or at least part of it).

The method getBase() of the class Proxy returns a Base pointer. When
called from python, boost detect the dynamic type of the object (here
Derived) and returns the real object. But, when I create a
boost::python::object from C++ this resolution is not done and the
object is known as a Base one. So, when pickling, it tries to pickle a
Base object (which cannot be pickled). Is there a way to force the
detection of the dynamic type when creating a boost::python::object ?

I would like to have this behaviour :

C++ code:

boost::python::object get_base(Proxy& p)
{
  return boost::python::object(p.getBase());
}

BOOST_PYTHON_MODULE(test_abstract)
{
  ...
  def("get_base", get_base);
  ...
}

Python:

>>> from test_abstract import *
>>> p = Proxy(Derived(10))
>>> get_base(p)
Derived(10)

But for now I have :

>>> from test_abstract import *
>>> p = Proxy(Derived(10))
>>> get_base(p)
<test_abstract.Base object at 0x4022d5dc>

How can I have the first behaviour ?

Pierre

On Mon, 2004-03-01 at 10:56, Pierre Barbier de Reuille wrote:
> Hi !
> 
> I have a problem with pickling a proxy class. I join the C++ code as an
> attached file.
> 
> I have three classes :
>  - Base : an abstract class
>  - Derived : a subclass of Base
>  - Proxy : the proxy class that contains an instance of Base
> 
> Derived objects can be pickled and the pickling of Proxy just return a
> tuple with a Base object in it. Here is a python sample :
> 
> 
> >>> import test_abstract
> >>> p = test_abstract.Proxy(test_abstract.Derived(10))
> >>> import cPickle
> >>> cPickle.dumps(p.getBase())
> 'ctest_abstract\nDerived\np1\n(I10\ntRp2\n.'
> >>> cPickle.dumps(p)
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
> cPickle.PicklingError: Can't pickle <type 'Boost.Python.instance'>:
> import of module Boost.Python failed
> 
> I really don't know what to do !
> 
> Thanks,
> 
> Pierre
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From steam at nurfuerspam.de  Mon Mar  1 12:30:34 2004
From: steam at nurfuerspam.de (Hanz Meizer)
Date: Mon, 01 Mar 2004 12:30:34 +0100
Subject: [C++-sig] Defining and adding functions in pyste.
Message-ID: <c1v6o3$30c$1@sea.gmane.org>

Hi,

I've got a working boost file and want to integrate my changes into the 
pyste file that created it. The problem: I've defined functions inside 
of the boost file and added those functions as member functions of a 
class like that:

namespace {
	std::string foobar() {
	}
}

// Module 
======================================================================
void Export_Test()
{
     class_< Test >("Test")
         .def("foobar", &foobar)
...

I added the declaration of foobar to the pyste file using

declaration_code("""
namespace {
	std::string foobar() {
	}
}
""")

This is working and showing in the output as expected. Unfortunately, 
using add_method gives me bad results because pyste doesn't know about 
the foobar function. ow do I have to setup my pyste file that it 
generates the same layout as my boost file?

H.




From dave at boost-consulting.com  Mon Mar  1 14:09:32 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 01 Mar 2004 08:09:32 -0500
Subject: [C++-sig] Re: Boost... Exception safe (can I turn it off?),
 and controlled object construction?
References: <BAY7-DAV8eSJIEoPXCF0001f12d@hotmail.com>
Message-ID: <uhdx8bto3.fsf@boost-consulting.com>

"Adrian Bentley" <adruab at hotmail.com> writes:

> Hi,
>
> I'd like to turn off the exception forwarding feature 

Not sure which one you're referring to here, but...

> of boost python (trying to adapt for use in high performance where
> exceptions are not desired).  

Even if you could turn off the feature, I doubt you'd notice any
measurable difference in performance because the Python interpreter is
so much slower than C++.  Any performance loss due to the try/catch
block at the boundary between the two languages will almost surely be
swamped by the cost of executing code in the Python interpreter.

> What I'd really like to do is be able to handle the exception
> classically, by giving it a function to call or the like.  I've read
> through some of the documentation and haven't found anything like
> this. Any pointers would be welcome.

There's nothing of that kind.  Boost.Python uses the C++ standard
library, and the C++ standard library throws exceptions, so trying to
avoid them would be fruitless.

> Also, I'm trying to use the language in kind of a weird way.  It's
> probably a bad idea but I'm running with it anyway :)...  I'm trying
> to embed python as a scripting language for my object system thing
> (game engine framework), and I'd like to be able to extend objects
> using python (extend embed at the same time! :P).

That's pretty typical, actually.


> Example of something like what I'm trying to do:
>
> //In pyobject file class BaseObject { void *data; //where the
> variables will all go //All it's stuff in here, some which requires
> manual construction and such
> };
>
> Pythonclass : BaseObject { //add stuff here
> }
>
> //in actual program class Py_wrapper : public BaseObject { PyObject
> *pyo; //gotten from embedded python
>
>     //forward variable requests to python object etc....
> };
>
> // End example

That doesn't illustrate very much, but I don't know why anyone would
use PyObject* when they could use boost::python::object.

> Ok, now that you all know I'm crazy the reason why I want to do this
> is that the base object is designed to communicate a certain way with
> things it's attached to, which probably wouldn't cooperate with the
> way python works

Maybe you should find out for sure before deciding.

> , so I need to manually construct the base object (also not sure how
> to do this in boost python).

I've no clue what you're asking about, sorry.

> Basically what I'd like is to have python override things like a class
> normally would, and track it in my object system.  I'm kind of
> doubting there's a way to do that, hence then extra layer of
> indirection py_wrapper (which should work).

http://www.boost.org/libs/python/doc/tutorial/doc/class_virtual_functions.html
http://www.boost.org/libs/python/doc/tutorial/doc/deriving_a_python_class.html

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Mon Mar  1 14:15:00 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 01 Mar 2004 08:15:00 -0500
Subject: [C++-sig] Re: embedding examples
References: <1078132099.32100.6.camel@localhost>
Message-ID: <ubrngbtez.fsf@boost-consulting.com>

Ingo Luetkebohle <ingo at fargonauten.de> writes:

> Hi,
>
> I have trouble compiling the 'Using the interpreter' examples from the
> tutorial.
>
> Code like this:
>   handle<> main_module(borrowed( PyImport_AddModule("__main__") ));
>   dict main_namespace(handle<>(borrowed(
> PyModule_GetDict(main_module.get()) )));
> results in 
>   error: variable declaration is not allowed here
>
> using gcc 3.2 and gcc 3.3 on Linux.
>
> When I use
>   handle<> main_module(borrowed( PyImport_AddModule("__main__") ));
>   handle<> main_nsh(borrowed( PyModule_GetDict(main_module.get()) ));
>   dict main_namespace(main_nsh);
>
> it compiles and the resulting dict is usable.
>
> So, this is not a showstopper but it doesn't work as documented and
> results in longer code.  Is there anything I can do to get the intended
> behaviour?

[joel, can you fix this?]

I think this will work.  Please let us know if it doesn't:

    // Retrieve the main module
    object main_module = python::extract<python::object>(
        PyImport_AddModule("__main__")
    );
    
    // Retrieve the main module's namespace
    dict main_namespace(
         extract<dict>(main_module.attr("__dict__"))
    );

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Mon Mar  1 14:16:21 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 01 Mar 2004 08:16:21 -0500
Subject: [C++-sig] Re: Software development levels
References: <u8yilepd4.fsf@boost-consulting.com>
	<BC686C03.7584%darylew@hotmail.com>
Message-ID: <u65dobtcq.fsf@boost-consulting.com>

Daryle Walker <darylew at hotmail.com> writes:

> On 2/29/04 1:01 PM, "David Abrahams" <dave at boost-consulting.com> wrote:
>
>> Daryle Walker <darylew at hotmail.com> writes:
> [SNIP]
>>> If it's missing features, I would consider it alpha-level software.  Recall
>>> that the alpha/beta/final scale relates to feature completeness, NOT code
>>> quality!  
>> 
>> I don't "recall that", and it seems to contradict what you say
>> below...
>
> I'm sorry, I forgot that the alpha/beta/final scale isn't formally taught in
> general.
>
>>> The two usually go hand in hand, but not always.  (There are
>>> programs that are awesome even in the alpha stage, and craptacular final
>>> stage programs.  The latter leads to accusations of "rushing the SW out the
>>> door and forcing the beta-testing on [paying] users".)
>> 
>> If it's all about features and not quality, what do the terms "alpha
>> testing" and "beta testing" mean?
>
> Alpha stage--not feature complete
> Beta stage--feature complete, but may have some bugs
>
> Alpha testing--checking if the features that are there work
> Beta testing--shaking out all the bugs

Just to clarify, the module is "missing a few features" *that people
might want*, but it's not missing any features that it was originally
designed to provide.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Mon Mar  1 14:28:26 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 01 Mar 2004 08:28:26 -0500
Subject: [C++-sig] Re: Pickling problem
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier>
Message-ID: <uznb0ae85.fsf@boost-consulting.com>

Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:

> I think I've found the problem (or at least part of it).
>
> The method getBase() of the class Proxy returns a Base pointer. When
> called from python, boost detect the dynamic type of the object (here
> Derived) and returns the real object. But, when I create a
> boost::python::object from C++ this resolution is not done and the
> object is known as a Base one. So, when pickling, it tries to pickle a
> Base object (which cannot be pickled). Is there a way to force the
> detection of the dynamic type when creating a boost::python::object ?
>
> I would like to have this behaviour :
>
> C++ code:
>
> boost::python::object get_base(Proxy& p)
> {
>   return boost::python::object(p.getBase());
> }
>
> BOOST_PYTHON_MODULE(test_abstract)
> {
>   ...
>   def("get_base", get_base);
>   ...
> }

Perhaps:

BOOST_PYTHON_MODULE(test_abstract)
{
  ...
  def("get_base", &Proxy::getBase, return_internal_reference<>());
  ...
}

??

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From joel at boost-consulting.com  Mon Mar  1 14:34:47 2004
From: joel at boost-consulting.com (Joel de Guzman)
Date: Mon, 01 Mar 2004 21:34:47 +0800
Subject: [C++-sig] Re: embedding examples
In-Reply-To: <ubrngbtez.fsf@boost-consulting.com>
References: <1078132099.32100.6.camel@localhost>
	<ubrngbtez.fsf@boost-consulting.com>
Message-ID: <40433BF7.5030806@boost-consulting.com>

David Abrahams wrote:

> Ingo Luetkebohle <ingo at fargonauten.de> writes:
> 
> 
>>Hi,
>>
>>I have trouble compiling the 'Using the interpreter' examples from the
>>tutorial.
>>
>>Code like this:
>>  handle<> main_module(borrowed( PyImport_AddModule("__main__") ));
>>  dict main_namespace(handle<>(borrowed(
>>PyModule_GetDict(main_module.get()) )));
>>results in 
>>  error: variable declaration is not allowed here
>>
>>using gcc 3.2 and gcc 3.3 on Linux.
>>
>>When I use
>>  handle<> main_module(borrowed( PyImport_AddModule("__main__") ));
>>  handle<> main_nsh(borrowed( PyModule_GetDict(main_module.get()) ));
>>  dict main_namespace(main_nsh);
>>
>>it compiles and the resulting dict is usable.
>>
>>So, this is not a showstopper but it doesn't work as documented and
>>results in longer code.  Is there anything I can do to get the intended
>>behaviour?
> 
> 
> [joel, can you fix this?]
> 
> I think this will work.  Please let us know if it doesn't:
> 
>     // Retrieve the main module
>     object main_module = python::extract<python::object>(
>         PyImport_AddModule("__main__")
>     );
>     
>     // Retrieve the main module's namespace
>     dict main_namespace(
>          extract<dict>(main_module.attr("__dict__"))
>     );

I think Dirk Gerrits is the authority here. He's the author of that
part of the tutorial. Dirk?

-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net



From pierre.barbier at cirad.fr  Mon Mar  1 15:04:31 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Mon, 01 Mar 2004 15:04:31 +0100
Subject: [C++-sig] Pickling problem
In-Reply-To: <1078137779.693.13.camel@pbarbier>
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier>
Message-ID: <1078149112.693.18.camel@pbarbier>

Ok, I've found the solution of my problem. 
If I create a function with this implementation :

PyObject* test_base( Proxy& p )
{
  return reference_existing_object::apply<Base*>::type()( p.getBase() );
}

The PyObject* will have the correct type (ie. not Base* but the dynamic
type of the result).

Pierre

-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From dave at boost-consulting.com  Mon Mar  1 15:28:42 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 01 Mar 2004 09:28:42 -0500
Subject: [C++-sig] Re: Pickling problem
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier>
	<1078149112.693.18.camel@pbarbier>
Message-ID: <uu1188wv9.fsf@boost-consulting.com>

Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:

> Ok, I've found the solution of my problem. 
> If I create a function with this implementation :
>
> PyObject* test_base( Proxy& p )
> {
>   return reference_existing_object::apply<Base*>::type()( p.getBase() );
> }
>
> The PyObject* will have the correct type (ie. not Base* but the dynamic
> type of the result).

The solution I posted is cleaner and won't leave you with dangling
references.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From pierre.barbier at cirad.fr  Mon Mar  1 15:37:28 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Mon, 01 Mar 2004 15:37:28 +0100
Subject: [C++-sig] Re: Pickling problem
In-Reply-To: <uu1188wv9.fsf@boost-consulting.com>
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier> <1078149112.693.18.camel@pbarbier>
	<uu1188wv9.fsf@boost-consulting.com>
Message-ID: <1078151848.742.24.camel@pbarbier>

I didn't receive any other answer ! Did you post it for another question
? If not, can you repost it ?

By the way, I have another problem : 

In my class Base I define a virtual method 'operator=='. In the wrapper,
if I follow the tutorial I write :

class BaseWrap : public Base
{
...
  virtual bool operator==(const Base& other) const
    {
      return call_method<bool>(self, "__eq__", other);
    }
...
}

But doing that, "other" is view as an 'other' class when it will surely
be a Python class. I tried to write the method like this :

class BaseWrap : public Base
{
...
  virtual bool operator==(const Base& other) const
    {
      const BaseWrap* bw = dynamic_cast<const BaseWrap*>( &other );
      PyObject* real_other = converter( &other );
      object obj;
      if( bw != NULL )
        {
          real_other = bw->self;
          obj = object(handle<>(borrowed(real_other)));
        }
      else
        {
          real_other = converter( &other );
          obj = object(handle<>(real_other));
        }
      return call_method<bool>(self, "__eq__", obj);
    }
...
}

But it crashes in python when I try to access the argument. I'm sure I'm
missing something, but I can't see what !

Thanks,

Pierre

On Mon, 2004-03-01 at 15:28, David Abrahams wrote:
> Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:
> 
> > Ok, I've found the solution of my problem. 
> > If I create a function with this implementation :
> >
> > PyObject* test_base( Proxy& p )
> > {
> >   return reference_existing_object::apply<Base*>::type()( p.getBase() );
> > }
> >
> > The PyObject* will have the correct type (ie. not Base* but the dynamic
> > type of the result).
> 
> The solution I posted is cleaner and won't leave you with dangling
> references.
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From ingo at fargonauten.de  Mon Mar  1 16:11:38 2004
From: ingo at fargonauten.de (Ingo Luetkebohle)
Date: Mon, 01 Mar 2004 16:11:38 +0100
Subject: [C++-sig] Re: embedding examples
In-Reply-To: <ubrngbtez.fsf@boost-consulting.com>
References: <1078132099.32100.6.camel@localhost>
	<ubrngbtez.fsf@boost-consulting.com>
Message-ID: <1078153896.13885.15.camel@localhost>

Hello David,

thanks for the hints.  There was a slight run-time problem with the
code, which I solved as explained below.  I hope my "fix" doesn't have
any side-effects.

Am Mo, den 01.03.2004 schrieb David Abrahams um 14:15: 
>     // Retrieve the main module's namespace
>     dict main_namespace(
>          extract<dict>(main_module.attr("__dict__"))
>     );

If I use that verbatim, an exception is thrown and PyErr_Print() prints

	TypeError: No to_python (by-value) converter found for C++ type:
N5boost6python7extractINS0_4dictEEE

I'm assuming that is because 'attr' returns an object_attribute, not a
PyObject* and for extract to work, some behind-the-scenes conversion
creates the above error.  Correct?

Luckily

	dict main_namespace(main_module.attr("__dict__"));

works fine and seems simpler, too.

regards,

-- 
Ingo

Soll doch jeder bleiben, wie er gerne w?re.





From pierre.barbier at cirad.fr  Mon Mar  1 16:09:06 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Mon, 01 Mar 2004 16:09:06 +0100
Subject: [C++-sig] Re: Pickling problem
In-Reply-To: <1078151848.742.24.camel@pbarbier>
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier> <1078149112.693.18.camel@pbarbier>
	<uu1188wv9.fsf@boost-consulting.com> <1078151848.742.24.camel@pbarbier>
Message-ID: <1078153746.693.28.camel@pbarbier>

Just the python part to reproduce my problem :

from test_abstract import *

class TestDerived(Base):
  def __init__(self, a):
    Base.__init__(self)
    self.a = a
  def __eq__(self, other):
    print "TestDerived.__eq__"
    print "with %s and %s" % (self, other)
    if isinstance(other, TestDerived):
      return self.a == other.a
    return False
  def __neq__(self, other):
    print "TestDerived.__neq__"
    print "with %s and %s" % (self, other)
    if isinstance(other, TestDerived):
      return self.a != other.a
    return True
  def getCopy(self):
    return TestDerived(self.a)
  def __hash__(self):
    return hash(self.a)

t1 = TestDerived(10)
p1 = Proxy(t1)
p2 = Proxy(TestDerived(10))
p3 = Proxy(TestDerived(11))
print p1 == p1 # This is ok
print p1 == p2 # Crash here
print p1 == p3 # And there

Thanks,

Pierre


-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From dave at boost-consulting.com  Mon Mar  1 16:38:46 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 01 Mar 2004 10:38:46 -0500
Subject: [C++-sig] Re: Pickling problem
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier>
	<1078149112.693.18.camel@pbarbier> <uu1188wv9.fsf@boost-consulting.com>
	<1078151848.742.24.camel@pbarbier>
Message-ID: <uu1187f21.fsf@boost-consulting.com>

Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:

> I didn't receive any other answer ! Did you post it for another question
> ? If not, can you repost it ?

http://article.gmane.org/gmane.comp.python.c%2B%2B/5973

It's untested, but if it works, it's much better.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From pierre.barbier at cirad.fr  Mon Mar  1 16:38:43 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Mon, 01 Mar 2004 16:38:43 +0100
Subject: [C++-sig] Re: Pickling problem
In-Reply-To: <uu1187f21.fsf@boost-consulting.com>
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier> <1078149112.693.18.camel@pbarbier>
	<uu1188wv9.fsf@boost-consulting.com> <1078151848.742.24.camel@pbarbier>
	<uu1187f21.fsf@boost-consulting.com>
Message-ID: <1078155522.693.30.camel@pbarbier>

This does not work because return_internal_reference<> requires a
pointer or reference. However, I think that :

PyObject* test_base( Proxy& p )
{
  return
return_internal_reference<1>::result_converter::apply<Base*>::type()(
p.getBase() );
}

gives the same result that you suggested.

On Mon, 2004-03-01 at 16:38, David Abrahams wrote:
> Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:
> 
> > I didn't receive any other answer ! Did you post it for another question
> > ? If not, can you repost it ?
> 
> http://article.gmane.org/gmane.comp.python.c%2B%2B/5973
> 
> It's untested, but if it works, it's much better.
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From dave at boost-consulting.com  Mon Mar  1 16:40:42 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 01 Mar 2004 10:40:42 -0500
Subject: [C++-sig] Re: Pickling problem
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier>
	<1078149112.693.18.camel@pbarbier> <uu1188wv9.fsf@boost-consulting.com>
	<1078151848.742.24.camel@pbarbier>
Message-ID: <uoerg7eyt.fsf@boost-consulting.com>

Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:

> By the way, I have another problem : 
>
> In my class Base I define a virtual method 'operator=='. In the wrapper,
> if I follow the tutorial I write :
>
<snip>

You didn't post a definition for "converter".  Please post a complete
reproducible test case.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From pierre.barbier at cirad.fr  Mon Mar  1 16:45:44 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Mon, 01 Mar 2004 16:45:44 +0100
Subject: [C++-sig] Re: Pickling problem
In-Reply-To: <uoerg7eyt.fsf@boost-consulting.com>
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier> <1078149112.693.18.camel@pbarbier>
	<uu1188wv9.fsf@boost-consulting.com> <1078151848.742.24.camel@pbarbier>
	<uoerg7eyt.fsf@boost-consulting.com>
Message-ID: <1078155944.742.35.camel@pbarbier>

oops ... sorry ... I send as attachement the complete file.
For the Python test case, here it is :

from test_abstract import *
class TestDerived(Base):
  def __init__(self, a):
    Base.__init__(self)
    self.a = a
  def __eq__(self, other):
    print "TestDerived.__eq__"
    print "with %s and %s" % (self, other)
    if isinstance(other, TestDerived):
      return self.a == other.a
    return False
  def __neq__(self, other):
    print "TestDerived.__neq__"
    print "with %s and %s" % (self, other)
    if isinstance(other, TestDerived):
      return self.a != other.a
    return True
  def getCopy(self):
    return TestDerived(self.a)
  def __hash__(self):
    return hash(self.a)

t1 = TestDerived(10)
p1 = Proxy(t1)
p2 = Proxy(TestDerived(10))
p3 = Proxy(TestDerived(11))
print p1 == p1 # works
print p1 == p2 # crash
print p1 == p3 # crash



On Mon, 2004-03-01 at 16:40, David Abrahams wrote:
> Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:
> 
> > By the way, I have another problem : 
> >
> > In my class Base I define a virtual method 'operator=='. In the wrapper,
> > if I follow the tutorial I write :
> >
> <snip>
> 
> You didn't post a definition for "converter".  Please post a complete
> reproducible test case.
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: test_abstract.cc
Type: text/x-c++
Size: 5666 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040301/2d9b980a/attachment.bin>

From pierre.barbier at cirad.fr  Mon Mar  1 16:59:49 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Mon, 01 Mar 2004 16:59:49 +0100
Subject: [C++-sig] Re: Pickling problem
In-Reply-To: <1078155944.742.35.camel@pbarbier>
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier> <1078149112.693.18.camel@pbarbier>
	<uu1188wv9.fsf@boost-consulting.com> <1078151848.742.24.camel@pbarbier>
	<uoerg7eyt.fsf@boost-consulting.com> <1078155944.742.35.camel@pbarbier>
Message-ID: <1078156789.693.40.camel@pbarbier>

Ok, the problem seems to be with the ownership of the class derived from
Base. It seems I did not understand well how to transfer the ownership
from Python to C++.

The use of auto_ptr does not seems to be enough as soon as I derive the
class in Python. Is there a way to completly transfer the ownership of
the object (even the python part of it) to the C++ program ?

-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From dave at boost-consulting.com  Mon Mar  1 17:02:48 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 01 Mar 2004 11:02:48 -0500
Subject: [C++-sig] Re: Pickling problem
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier>
	<1078149112.693.18.camel@pbarbier> <uu1188wv9.fsf@boost-consulting.com>
	<1078151848.742.24.camel@pbarbier> <uu1187f21.fsf@boost-consulting.com>
	<1078155522.693.30.camel@pbarbier>
Message-ID: <u4qt87dxz.fsf@boost-consulting.com>

Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:

> This does not work because return_internal_reference<> requires a
> pointer or reference. 

Yeah... but if getBase doesn't return a pointer or reference, then it
returns by value.  And if it returns by value, the static type of the
returned object is the same as its dynamic type, so I don't see how
you expect to end up with a Derived object.

> However, I think that :
>
> PyObject* test_base( Proxy& p )
> {
>   return
> return_internal_reference<1>::result_converter::apply<Base*>::type()(
> p.getBase() );
> }
>
> gives the same result that you suggested.

Not really.  That gives the same result you were getting before with
reference_existing_object.


-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From pierre.barbier at cirad.fr  Mon Mar  1 17:17:02 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Mon, 01 Mar 2004 17:17:02 +0100
Subject: [C++-sig] Re: Pickling problem
In-Reply-To: <1078156789.693.40.camel@pbarbier>
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier> <1078149112.693.18.camel@pbarbier>
	<uu1188wv9.fsf@boost-consulting.com> <1078151848.742.24.camel@pbarbier>
	<uoerg7eyt.fsf@boost-consulting.com> <1078155944.742.35.camel@pbarbier>
	<1078156789.693.40.camel@pbarbier>
Message-ID: <1078157821.693.48.camel@pbarbier>

I tried to modify my class constructors / destructor like that :

struct BaseWrap : public Base
{
  ...
  BaseWrap( PyObject* self_ ) : self( self_ ) { Py_INCREF( self ); }
  BaseWrap( PyObject* self_, const Base& copy ) : Base( copy ), self(
self_ ) { Py_INCREF( self ); }
  virtual ~BaseWrap()
    {
      Py_DECREF( self );
    }
  ...
}

And now it seems to work. Is there a problem with that ?

Pierre

On Mon, 2004-03-01 at 16:59, Pierre Barbier de Reuille wrote:
> Ok, the problem seems to be with the ownership of the class derived from
> Base. It seems I did not understand well how to transfer the ownership
> from Python to C++.
> 
> The use of auto_ptr does not seems to be enough as soon as I derive the
> class in Python. Is there a way to completly transfer the ownership of
> the object (even the python part of it) to the C++ program ?
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From pierre.barbier at cirad.fr  Mon Mar  1 17:07:04 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Mon, 01 Mar 2004 17:07:04 +0100
Subject: [C++-sig] Re: Pickling problem
In-Reply-To: <u4qt87dxz.fsf@boost-consulting.com>
References: <1078135007.742.6.camel@pbarbier>
	<1078137779.693.13.camel@pbarbier> <1078149112.693.18.camel@pbarbier>
	<uu1188wv9.fsf@boost-consulting.com> <1078151848.742.24.camel@pbarbier>
	<uu1187f21.fsf@boost-consulting.com> <1078155522.693.30.camel@pbarbier>
	<u4qt87dxz.fsf@boost-consulting.com>
Message-ID: <1078157224.693.43.camel@pbarbier>

On Mon, 2004-03-01 at 17:02, David Abrahams wrote:
> Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:
> 
> > This does not work because return_internal_reference<> requires a
> > pointer or reference. 
> 
> Yeah... but if getBase doesn't return a pointer or reference, then it
> returns by value.  And if it returns by value, the static type of the
> returned object is the same as its dynamic type, so I don't see how
> you expect to end up with a Derived object.

No, what I meant is that I cannot use return_internal_reference with a
boost::python::object as a result. Nor can I use it with a PyObject* !
I know I can directly send a Base* object, but the goal of this function
is to test the creation of a boost::python::object from C++.

> 
> > However, I think that :
> >
> > PyObject* test_base( Proxy& p )
> > {
> >   return
> > return_internal_reference<1>::result_converter::apply<Base*>::type()(
> > p.getBase() );
> > }
> >
> > gives the same result that you suggested.
> 
> Not really.  That gives the same result you were getting before with
> reference_existing_object.

Arg ... so I don't know what to do :(

-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From dirk at gerrits.homeip.net  Mon Mar  1 23:32:15 2004
From: dirk at gerrits.homeip.net (Dirk Gerrits)
Date: Mon, 01 Mar 2004 23:32:15 +0100
Subject: [C++-sig] Re: embedding examples
In-Reply-To: <40433BF7.5030806@boost-consulting.com>
References: <1078132099.32100.6.camel@localhost>	<ubrngbtez.fsf@boost-consulting.com>
	<40433BF7.5030806@boost-consulting.com>
Message-ID: <c20dle$9i$1@sea.gmane.org>

Joel de Guzman wrote:
>> Ingo Luetkebohle <ingo at fargonauten.de> writes:
>>
>>
>>> Hi,
>>>
>>> I have trouble compiling the 'Using the interpreter' examples from the
>>> tutorial.

[snip]

> I think Dirk Gerrits is the authority here. He's the author of that
> part of the tutorial. Dirk?

Well the (sad?) truth is that I haven't used Python or Boost.Python for
quite a while (several months). Also, my views on embedding versus
extending have changed.

Issues with the embedding tutorial code have come up before, and IIRC
Dave was able to provide fixes. Dave, Joel, other BPL contributors, you
have my permission to change the embedding tutorial. If you can't manage
it, just let me know and I'll see what I can do, but I think it would be
best for Boost.Python if some one else took care of the embedding
tutorial from now on.

Kind regards,

Dirk Gerrits





From joel at boost-consulting.com  Tue Mar  2 14:06:53 2004
From: joel at boost-consulting.com (Joel de Guzman)
Date: Tue, 02 Mar 2004 21:06:53 +0800
Subject: [C++-sig] Re: embedding examples
In-Reply-To: <c20dle$9i$1@sea.gmane.org>
References: <1078132099.32100.6.camel@localhost>	<ubrngbtez.fsf@boost-consulting.com>	<40433BF7.5030806@boost-consulting.com>
	<c20dle$9i$1@sea.gmane.org>
Message-ID: <404486ED.3050205@boost-consulting.com>

Dirk Gerrits wrote:

> Joel de Guzman wrote:
> 
>> I think Dirk Gerrits is the authority here. He's the author of that
>> part of the tutorial. Dirk?
> 
> 
> Well the (sad?) truth is that I haven't used Python or Boost.Python for
> quite a while (several months). Also, my views on embedding versus
> extending have changed.
> 
> Issues with the embedding tutorial code have come up before, and IIRC
> Dave was able to provide fixes. Dave, Joel, other BPL contributors, you
> have my permission to change the embedding tutorial. If you can't manage
> it, just let me know and I'll see what I can do, but I think it would be
> best for Boost.Python if some one else took care of the embedding
> tutorial from now on.

Ok, I'll take care of it then.

-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net



From dave at boost-consulting.com  Tue Mar  2 15:50:09 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Tue, 02 Mar 2004 09:50:09 -0500
Subject: [C++-sig] Re: embedding examples
References: <1078132099.32100.6.camel@localhost>
	<ubrngbtez.fsf@boost-consulting.com>
	<40433BF7.5030806@boost-consulting.com> <c20dle$9i$1@sea.gmane.org>
	<404486ED.3050205@boost-consulting.com>
Message-ID: <ubrnfqp5q.fsf@boost-consulting.com>

Joel de Guzman <joel at boost-consulting.com> writes:

> Dirk Gerrits wrote:
>
>> Joel de Guzman wrote:
>> 
>>> I think Dirk Gerrits is the authority here. He's the author of that
>>> part of the tutorial. Dirk?
>> Well the (sad?) truth is that I haven't used Python or Boost.Python
>> for
>> quite a while (several months). Also, my views on embedding versus
>> extending have changed.
>> Issues with the embedding tutorial code have come up before, and
>> IIRC
>> Dave was able to provide fixes. Dave, Joel, other BPL contributors, you
>> have my permission to change the embedding tutorial. If you can't manage
>> it, just let me know and I'll see what I can do, but I think it would be
>> best for Boost.Python if some one else took care of the embedding
>> tutorial from now on.
>
> Ok, I'll take care of it then.

Thanks, Joel!

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From ingo at fargonauten.de  Tue Mar  2 17:03:00 2004
From: ingo at fargonauten.de (Ingo Luetkebohle)
Date: Tue, 02 Mar 2004 17:03:00 +0100
Subject: [C++-sig] Re: embedding examples
In-Reply-To: <ubrngbtez.fsf@boost-consulting.com>
References: <1078132099.32100.6.camel@localhost>
	<ubrngbtez.fsf@boost-consulting.com>
Message-ID: <1078243379.816.5.camel@localhost>

Am Mo, den 01.03.2004 schrieb David Abrahams um 14:15:
>     // Retrieve the main module
>     object main_module = python::extract<python::object>(
>         PyImport_AddModule("__main__")
>     );

One word of caution: extract does not seem to check for NULL pointers. 
PyImport_AddModule never returns NULL, but other functions do.  E.g., I
was using PyImport_ImportModule and when it can't find the module, it
returns NULL.

To work around that, I use handle<> on the very first call like that,
then initialize an object and stick with Boost.Python calls afterwards.

Thanks again for the help :)

-- 
Ingo

Soll doch jeder bleiben, wie er gerne w?re.





From sirbender at freenet.de  Tue Mar  2 19:30:06 2004
From: sirbender at freenet.de (Billy Gnosis)
Date: Tue, 02 Mar 2004 19:30:06 +0100
Subject: [C++-sig] Could anyone send me an example code for C++/Python
 Interfacing ???
Message-ID: <4044D2AE.5090500@freenet.de>

Hi...


I am writing plugins for an already existing application. An API is 
supplied in C++ to "interact" with the necessary parts of the application.

I have written a Pthon programm that does some text parsing tasks and 
returns a some lists and an Nnumeric array as the desired result.

Now I would like to _call _this Python program from within my C++ plugin 
_with a single argument:_    _a filename_. The Python program now should 
do the Text processing trick and_ return the lists (could convert to 
Numeric arrays before, if it helps) and arrays to the C++ programm_. 
(optimally this should happen without the user knowing that python is 
working partially in the background - no Dos Box opening etc.).

I have underlined the parts of the process in the text above I don't 
know how to do...I hope there is a simpler solution than using a neat 
interfacing support  like SWIG etc...

Now, I want to ask somebody if he perhaps has a little SIMPLE example 
code in C++ and Python where something similar is done. I would 
appreciate some example files very much,  as I could then directly test 
out what's going on.
Surely I am also very happy with any other counsel...

Thanks,

billy




From sirbender at freenet.de  Tue Mar  2 21:17:24 2004
From: sirbender at freenet.de (Billy Gnosis)
Date: Tue, 02 Mar 2004 21:17:24 +0100
Subject: [C++-sig] Could anyone send me an example code for C++/Python
	Interfacing ???
In-Reply-To: <4044D2AE.5090500@freenet.de>
References: <4044D2AE.5090500@freenet.de>
Message-ID: <4044EBD4.4060003@freenet.de>

Billy Gnosis wrote:

> Hi...
>
>
> I am writing plugins for an already existing application. An API is 
> supplied in C++ to "interact" with the necessary parts of the 
> application.
>
> I have written a Pthon programm that does some text parsing tasks and 
> returns a some lists and an Nnumeric array as the desired result.
>
> Now I would like to /*_call _this Python program from within my C++ 
> plugin _with a single argument:_    _a filename_*/. The Python program 
> now should do the Text processing trick and*/_ return the lists (could 
> convert to Numeric arrays_ _before that, if it helps) and arrays to 
> the C++ programm_/*. (optimally this should happen without the user 
> knowing that python is working partially in the background - no Dos 
> Box opening etc.).
>
> I have underlined the parts of the process in the text above I don't 
> know how to do...I hope there is a simpler solution than using a neat 
> interfacing support  like SWIG etc...
>
> Now, I want to ask somebody if he perhaps has a little SIMPLE example 
> code in C++ and Python where something similar is done. I would 
> appreciate some example files very much,  as I could then directly 
> test out what's going on.
> Surely I am also very happy with any other counsel...
>
> Thanks,
>
> billy
>
>
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>
>
>





From nicodemus at esss.com.br  Tue Mar  2 23:11:59 2004
From: nicodemus at esss.com.br (Nicodemus)
Date: Tue, 02 Mar 2004 19:11:59 -0300
Subject: [C++-sig] Defining and adding functions in pyste.
In-Reply-To: <c1v6o3$30c$1@sea.gmane.org>
References: <c1v6o3$30c$1@sea.gmane.org>
Message-ID: <404506AF.3050700@esss.com.br>

Hanz Meizer wrote:

> <snip>
> This is working and showing in the output as expected. Unfortunately, 
> using add_method gives me bad results because pyste doesn't know about 
> the foobar function. ow do I have to setup my pyste file that it 
> generates the same layout as my boost file?


You have to include foobar in another header file, and use that to 
include your classe:

Test = Class("Test", "foobar.h")
add_method(Test, "foobar")

HTH,
Nicodemus.




From adruab at hotmail.com  Wed Mar  3 03:55:08 2004
From: adruab at hotmail.com (Adrian Bentley)
Date: Tue, 2 Mar 2004 18:55:08 -0800
Subject: [C++-sig] Re: Boost... Exception safe (can I turn it off?),
	and controlled object construction?
References: <BAY7-DAV8eSJIEoPXCF0001f12d@hotmail.com>
	<uhdx8bto3.fsf@boost-consulting.com>
Message-ID: <BAY7-DAV25v3zOdnjQ800026898@hotmail.com>

Sorry about being somewhat unclear.

In terms of what I wanted to do with exceptions, I was under the impression
that the boost python code boiled down to using the C Python API (which
allows you to set a python global exception handler... I think).  There are
alternative ways I can wrap the interpreting calls to prevent exceptions
from being thrown.  However, I want to compile the code without exceptions
turned on, which would prevent me from being able to use try catch blocks.

In terms of my object madness:
> > //In pyobject file class BaseObject { void *data; //where the
> > variables will all go //All it's stuff in here, some which requires
> > manual construction and such
> > };
> >
> > Pythonclass : BaseObject { //add stuff here
> > }
> >
> > //in actual program class Py_wrapper : public BaseObject { PyObject
> > *pyo; //gotten from embedded python
> >
> >     //forward variable requests to python object etc....
> > };
> >
> > // End example
>
> That doesn't illustrate very much, but I don't know why anyone would
> use PyObject* when they could use boost::python::object.

Heh, yeah, I'm not sure why I spontaneously forgot to write
boost::python::object instead.

> > Ok, now that you all know I'm crazy the reason why I want to do this
> > is that the base object is designed to communicate a certain way with
> > things it's attached to, which probably wouldn't cooperate with the
> > way python works
>
> Maybe you should find out for sure before deciding.
>
> > , so I need to manually construct the base object (also not sure how
> > to do this in boost python).
>
> I've no clue what you're asking about, sorry.

To elaborate, and clarify (since I don't seem to be very good at that):

> > Basically what I'd like is to have python override things like a class
> > normally would, and track it in my object system.  I'm kind of
> > doubting there's a way to do that, hence then extra layer of
> > indirection py_wrapper (which should work).

Clarification: I don't imagine there is any way to "cast" the python object
to a BaseObject * or the like, is there?  I'd imagine there wouldn't be,
given that python must use a method of inheritance incompatible with C++.
In the case that I would NOT be able to cast as such, if I wanted to
register any python derived BaseObjects with my system, I'd have to use the
wrapping class with the object inside.

Thanks for the pointers (and being patient with me :P),
Adruab > Adrian Bentley



From gilles.orazi at varianinc.com  Wed Mar  3 14:35:17 2004
From: gilles.orazi at varianinc.com (Gilles Orazi)
Date: Wed, 03 Mar 2004 14:35:17 +0100
Subject: [C++-sig] Virtual base class and concrete class
Message-ID: <5.2.0.9.2.20040303135525.00b5ebd0@begonia.varianinc.com>

An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040303/93f3be14/attachment.htm>

From pierre.barbier at cirad.fr  Wed Mar  3 14:50:16 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Wed, 03 Mar 2004 14:50:16 +0100
Subject: [C++-sig] Automation of Python object retrieval
Message-ID: <1078321816.742.164.camel@pbarbier>

I have a problem with the boost::python::object contruction. I'd want to
know if there is a way to automate the retrieval of the existing Python
object when I hold a Pointer on the C++ base class and that the class is
extended in Python. Here's an example of what I do now :

================
In C++
================

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

#include <string>
#include <memory>

using namespace boost::python;

struct Base
{
  virtual ~Base() {}
  virtual std::string toStr() const
    {
      return "base";
    }
  virtual bool operator==( Base& b )
    {
      return toStr() == b.toStr();
    }
};

struct BaseWrap : public Base
{
  BaseWrap( PyObject* self_ ) : self(self_)
    {
      Py_INCREF(self); /* Ensure the Python object exists as least as
long as the C++ object does */
    }
  BaseWrap( PyObject* self_, const Base& copy) : Base(copy), self(self_)
    {
      Py_INCREF(self);
    }
  virtual ~BaseWrap()
    {
      Py_DECREF(self);
    }
  virtual std::string toStr() const
    {
      return call_method<std::string>(self, "toStr");
    }
  std::string default_toStr() const
    {
      return Base::toStr();
    }
  virtual bool operator==( Base& b )
    {
      return call_method<bool>(self, "__eq__", b); // Naive approach
    }
  bool default_eq( Base& b)
    {
      return Base::operator==( b );
    }
  PyObject *self;
};

struct Holder
{
  Holder( Base* b = NULL ) : hold( b ) {}
  bool operator==( Holder& other )
    {
      return *hold == *other.hold;
    }
  std::string toStr()
    {
      return hold->toStr();
    }
  Base& getHold() { return *hold; }
  Base* hold;
};

Holder* makeHolder( std::auto_ptr<Base> b )
{
  Holder* obj = new Holder( b.get() );
  b.release();
  return obj;
};

BOOST_PYTHON_MODULE(mymod)
{
  class_<Base, std::auto_ptr<BaseWrap> >("Base", init<>())
    .def("__eq__", &Base::operator==, &BaseWrap::default_eq)
    .def("toStr", &Base::toStr, &BaseWrap::default_toStr)
    ;
  implicitly_convertible<std::auto_ptr<BaseWrap>, std::auto_ptr<Base>
>();
  class_<Holder, boost::noncopyable>( "Holder" )
    .def( "__init__", make_constructor( makeHolder ) )
    .def( "__eq__", &Holder::operator== )
    .def( "toStr", &Holder::toStr )
    .def( "getHold", &Holder::getHold, return_internal_reference<1>() )
    ;
}


================
In Python:
================

from mymod import *
class Derived(Base):
  def __init__(self, value):
    Base.__init__(self)
    self._value = value
  def toStr(self):
    return "Derived with value %s" % self._value
  def __eq__(self,other):
    if isinstance(other, Derived):
      return self._value == other._value
    return False

h = Holder(Derived(10))
h1 = Holder(Derived(10))
h2 = Holder(Derived("10"))
h3 = Holder(Derived(11))

print h == h1 # => 0
print h == h2 # => 0
print h == h  # => 0

======================

All the equality tests fails because the other in detected as a Base
object and not a Derived object !!!

But, if I replace the BaseWrap::operator== by :

  virtual bool operator==( Base& b )
    {
      static reference_existing_object::apply<Base&>::type convert;
      BaseWrap* bw = dynamic_cast<BaseWrap*>( &b );
      object other;
      if( bw != NULL )
        {
          other = object( handle<>( borrowed( bw->self ) ) );
        }
      else
        {
          other = object( handle<>( convert( b ) ) );
        }
      return call_method<bool>(self, "__eq__", other);
    }

I get the wanted behaviour. But I wonder if there is a cleaner method to
do that ? Or at least a method to automate that. It's always the same :
take the self PyObject* from the wrapper and use it to build the the
boost::python::object.

Thanks,

Pierre

-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From steam at nurfuerspam.de  Wed Mar  3 16:46:51 2004
From: steam at nurfuerspam.de (Hanz Meizer)
Date: Wed, 03 Mar 2004 16:46:51 +0100
Subject: [C++-sig] Re: Conversion of C++ return type.
In-Reply-To: <ufzcx742v.fsf@boost-consulting.com>
References: <c1khs3$4i8$1@sea.gmane.org> <ufzcx742v.fsf@boost-consulting.com>
Message-ID: <c24ugj$u85$1@sea.gmane.org>

> Good guess.
> 
> Unfortunately I can't track down the details for you, but here's the
> gist: you need to register a custom to_python_converter for Any.
> Then, inside its conversion function, you need to get the typeid of
> the type stored by the Any (if you can't get that you're out of luck),
> and invoke the registry's lookup function to get the to-python
> conversion chain for the stored type.  Then you need to "manually"
> invoke that converter. Expect the resulting code to be fairly ugly
> and low-level.  Sorry :(
> 
> 

The Any objects usually contain very basic types, or vectors of very 
basic types. So I think what would be really nice is to have two 
functions, one converting an Any object to a PyObject that contains a 
pythonification of the Any member objects, and another one that gets 
called with a python object and returns the corresponding Any object. 
The latter would be used implicitly when a C++ Method is called from 
python (for example std::string printAny(const Any& car) ).

For example:
I have a function Any* getProperty() that returns an Any that either 
conains a vector<double> or a single double. The conversion function 
then would take the Any, and return a PyObject to python that either 
contins a list of doubles or a simple double.

Unfortunately, diggig through the examples and mailing list archives 
didn't enlighten me enough, my last test was something like:

namespace
{
   struct any_to_python
   {
     static PyObject* convert(Any const& x)
     {
       std:cerr << "convert called, rejoice!" << endl;
       exit(1);
     }
};

...

class_< Any, boost::noncopyable >("Any", no_init)
;

to_python_converter< Any, any_to_python >();

The intention is to have any_to_python::convert check what is inside of 
the Any and return the approprioate, constructed python object. 
Unfortunately, convert is never called. If I uncomment the class_<> ...; 
thing, python says:

No Python class registered for C++ class Any

If I sue the class_<>...; thing, I get some error because python tries 
to call __str__ on the Any object and fails. Help! :)

Regards,
H.




From halbert at bbn.com  Wed Mar  3 17:32:59 2004
From: halbert at bbn.com (Dan Halbert)
Date: Wed, 3 Mar 2004 11:32:59 -0500
Subject: [C++-sig] register_exception_translator problem in 1.31.0
Message-ID: <200403031632.i23GWxs00444@rolex.bbn.com>

I am trying to use the exception translator feature stock BPL version
from Boost 1.31.0. My code is like this:

  void
  translate(My::Own::Exception const &e)
  {
      PyErr_SetString( ... );
  }


  BOOST_PYTHON_MODULE(mod)
  {
       // ...
      register_exception_translator<My::Own::Exception>(translate);
  }

However, I get a compile error with gcc 3.2.2 on Linux:

  ../py_my_own.cpp: In function `void init_module_mod()':
  ../py_my_own.cpp:68: no matching function for call to `
		        register_exception_translator(void (&)(My::Own::Exception&))'

Studying the BPL include files has gotten me nowhere. Can you tell me
what I'm doing wrong?

Thanks,
Dan



From dave at boost-consulting.com  Wed Mar  3 20:34:44 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Wed, 03 Mar 2004 14:34:44 -0500
Subject: [C++-sig] Re: register_exception_translator problem in 1.31.0
References: <200403031632.i23GWxs00444@rolex.bbn.com>
Message-ID: <ur7w9hgh7.fsf@boost-consulting.com>

Dan Halbert <halbert at bbn.com> writes:

> I am trying to use the exception translator feature stock BPL version
> from Boost 1.31.0. My code is like this:
>
>   void
>   translate(My::Own::Exception const &e)
>   {
>       PyErr_SetString( ... );
>   }
>
>
>   BOOST_PYTHON_MODULE(mod)
>   {
>        // ...
>       register_exception_translator<My::Own::Exception>(translate);
Add an ampersand ('&') here-------------------------------^
>   }
>
> However, I get a compile error with gcc 3.2.2 on Linux:
>
>   ../py_my_own.cpp: In function `void init_module_mod()':
>   ../py_my_own.cpp:68: no matching function for call to `
> 		        register_exception_translator(void (&)(My::Own::Exception&))'
>
> Studying the BPL include files has gotten me nowhere. Can you tell me
> what I'm doing wrong?

HTH,

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From jeff.holle at verizon.net  Wed Mar  3 20:55:41 2004
From: jeff.holle at verizon.net (Jeff Holle)
Date: Wed, 03 Mar 2004 11:55:41 -0800
Subject: [C++-sig] A problem with pyste.
Message-ID: <4046383D.8000304@verizon.net>

I'm using Pyste version 0.9.28 and gcc v3.3 on a linux system.

I've a problem doing the most complex thing I've done with pyste so far.
It involves exposing a couple of templated functions, which can each be 
instantiated with 3 types of classes.

I've attached both the pyste configuration file and the referenced *.h file.

The errors produced are the following:
  /tmp/tmp9zPX8o.h:135: error: syntax error before `;' token
  /tmp/tmp9zPX8o.h: In function `void 
__instantiate_putNamespace_Parameter()':
  /tmp/tmp9zPX8o.h:137: error: `putNamespace_Parameter' undeclared 
(first use this function)

Note that putNamespace_Parameter is explicitly specified within 
NamespaceDef.pyste.
I tried to use the default name instead of renaming it and the error 
message changed to can't find "UMLModel_putNamespace_Parameter".

Could somebody with more experience with pyste please inform me what I'm 
doing wrong.

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: NamespaceDef.pyste
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040303/8f2789ab/attachment.txt>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: NamespaceDef.h
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040303/8f2789ab/attachment-0001.txt>

From halbert at bbn.com  Wed Mar  3 21:18:34 2004
From: halbert at bbn.com (Dan Halbert)
Date: Wed, 3 Mar 2004 15:18:34 -0500
Subject: [C++-sig] Re: register_exception_translator problem in 1.31.0
In-Reply-To: <ur7w9hgh7.fsf@boost-consulting.com> (message from David Abrahams
	on Wed, 03 Mar 2004 14:34:44 -0500)
References: <200403031632.i23GWxs00444@rolex.bbn.com>
	<ur7w9hgh7.fsf@boost-consulting.com>
Message-ID: <200403032018.i23KIY400820@rolex.bbn.com>

  >>       register_exception_translator<My::Own::Exception>(translate);
  >Add an ampersand ('&') here-------------------------------^

Ah, so the tutorial example is wrong. Thank you!
(Bad example is at http://www.boost.org/libs/python/doc/tutorial/doc/exception_translation.html)



From dave at boost-consulting.com  Wed Mar  3 22:49:37 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Wed, 03 Mar 2004 16:49:37 -0500
Subject: [C++-sig] Re: register_exception_translator problem in 1.31.0
References: <200403031632.i23GWxs00444@rolex.bbn.com>
	<ur7w9hgh7.fsf@boost-consulting.com>
	<200403032018.i23KIY400820@rolex.bbn.com>
Message-ID: <ullmhfvny.fsf@boost-consulting.com>

Dan Halbert <halbert at bbn.com> writes:

>   >>       register_exception_translator<My::Own::Exception>(translate);
>   >Add an ampersand ('&') here-------------------------------^
>
> Ah, so the tutorial example is wrong. Thank you!
> (Bad example is at http://www.boost.org/libs/python/doc/tutorial/doc/exception_translation.html)

http://www.boost.org/libs/python/doc/v2/exception_translator.html
gets it right.


Hmm.  Do me a favor, please.  In exception_translator.hpp, change:

  template <class ExceptionType, class Translate>
  void register_exception_translator(Translate const& translate, boost::type<ExceptionType>* = 0)
                                               ^^^^^^
to:

  template <class ExceptionType, class Translate>
  void register_exception_translator(Translate translate, boost::type<ExceptionType>* = 0)

and try your example again.  I expect it works, and if so I'm inclined
to change the signature of register_exception_translator.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From halbert at bbn.com  Thu Mar  4 00:32:06 2004
From: halbert at bbn.com (Dan Halbert)
Date: Wed, 3 Mar 2004 18:32:06 -0500
Subject: [C++-sig] Re: register_exception_translator problem in 1.31.0
In-Reply-To: <ullmhfvny.fsf@boost-consulting.com> (message from David Abrahams
	on Wed, 03 Mar 2004 16:49:37 -0500)
References: <200403031632.i23GWxs00444@rolex.bbn.com>
	<ur7w9hgh7.fsf@boost-consulting.com>
	<200403032018.i23KIY400820@rolex.bbn.com>
	<ullmhfvny.fsf@boost-consulting.com>
Message-ID: <200403032332.i23NW6q01280@rolex.bbn.com>

David Abrahams <dave at boost-consulting.com> wrote:
  >>   >>       register_exception_translator<My::Own::Exception>(translate);
  >>   >Add an ampersand ('&') here-------------------------------^
  >>

  >Hmm.  Do me a favor, please.  In exception_translator.hpp, change:
  >
  >  template <class ExceptionType, class Translate>
  >  void register_exception_translator(Translate const& translate, boost::type<ExceptionType>* = 0)
  >                                               ^^^^^^
  >to:
  >
  >  template <class ExceptionType, class Translate>
  >  void register_exception_translator(Translate translate, boost::type<ExceptionType>* = 0)
  >
  >and try your example again.  I expect it works, and if so I'm inclined
  >to change the signature of register_exception_translator.

OK, I did try this, and now it works either with "&translate" or plain
"translate" above. So that does seem better.

Dan




From halbert at bbn.com  Thu Mar  4 00:45:49 2004
From: halbert at bbn.com (Dan Halbert)
Date: Wed, 3 Mar 2004 18:45:49 -0500
Subject: [C++-sig] pyste: excluding a constructor
Message-ID: <200403032345.i23Njnl01298@rolex.bbn.com>

OK, now a pyste question, with the 1.31.0 pyste:

I have code like this:

========================================
class Opaque; // Forward declaration, not otherwise declared in this .h

class Foo {
public:
	Foo(int);
	Foo(Opaque*);
        // ...
}
========================================

I get "use of undefined type" and "forward declaration" g++ errors about
Opaque. I see two possibilities:

1. Can I use a policy, maybe, to avoid this error?

2. I don't necessarily care about pythonizing the second constructor.
I could exclude Opaque, but how do I exclude the second constructor?

Nicodemus mentioned some meta-programming constructs he was adding
sometime last summer which looked like they could be used to iterate
through the constructors and exclude the second one. But I tried
something like his example and couldn't get it to work. I studied
infos.py but didn't glean a solution.

Thanks very much,
Dan

	



From s_sourceforge at nedprod.com  Thu Mar  4 02:51:34 2004
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Thu, 04 Mar 2004 01:51:34 -0000
Subject: [C++-sig] pyste: excluding a constructor
In-Reply-To: <200403032345.i23Njnl01298@rolex.bbn.com>
Message-ID: <40468BA6.19225.1BAAA3D2@localhost>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 3 Mar 2004 at 18:45, Dan Halbert wrote:

> I get "use of undefined type" and "forward declaration" g++ errors
> about Opaque. I see two possibilities:
> 
> 1. Can I use a policy, maybe, to avoid this error?

Just specify the return_opaque_pointer() policy. I may have the name 
slightly wrong - look it up in pyste.py.

Cheers,
Niall





-----BEGIN PGP SIGNATURE-----
Version: idw's PGP-Frontend 4.9.6.1 / 9-2003 + PGP 8.0.2

iQA/AwUBQEaLpsEcvDLFGKbPEQKGpwCeI/aFyteBoeV50cNi779pbWwu/ecAoKZ4
FXeRUAKtnrpspmKMxIcy9nYn
=6doy
-----END PGP SIGNATURE-----



From pierre.barbier at cirad.fr  Thu Mar  4 09:19:04 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Thu, 04 Mar 2004 09:19:04 +0100
Subject: [C++-sig] embedding examples
In-Reply-To: <1078132099.32100.6.camel@localhost>
References: <1078132099.32100.6.camel@localhost>
Message-ID: <1078388344.742.175.camel@pbarbier>

Hi !

I use the interpreter with the CVS version of Boost. In my code I use
these lines to get the __main__ dict (and it works ...):

object main_module =
object(handle<>(borrowed(PyImport_AddModule("__main__"))));
object main_namespace = extract<dict>(
object(handle<>(borrowed(PyModule_GetDict(main_module.ptr())))) );

Pierre

On Mon, 2004-03-01 at 10:08, Ingo Luetkebohle wrote:
> Hi,
> 
> I have trouble compiling the 'Using the interpreter' examples from the
> tutorial.
> 
> Code like this:
>   handle<> main_module(borrowed( PyImport_AddModule("__main__") ));
>   dict main_namespace(handle<>(borrowed(
> PyModule_GetDict(main_module.get()) )));
> results in 
>   error: variable declaration is not allowed here
> 
> using gcc 3.2 and gcc 3.3 on Linux.
> 
> When I use
>   handle<> main_module(borrowed( PyImport_AddModule("__main__") ));
>   handle<> main_nsh(borrowed( PyModule_GetDict(main_module.get()) ));
>   dict main_namespace(main_nsh);
> 
> it compiles and the resulting dict is usable.
> 
> So, this is not a showstopper but it doesn't work as documented and
> results in longer code.  Is there anything I can do to get the intended
> behaviour?
> 
> regards
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From Holger.Joukl at LBBW.de  Thu Mar  4 17:28:25 2004
From: Holger.Joukl at LBBW.de (Holger Joukl)
Date: Thu, 4 Mar 2004 17:28:25 +0100
Subject: [C++-sig] building boost.python problems
Message-ID: <OF9522F46B.A36EDE12-ONC1256E4D.00538714@lbbw.sko.de>


Dear boost.pythoneers,
first of all, sorry for the lengthish post (and more so, for the
confidential blabla my employer attaches to all my outgoing mails).

I?m having a little trouble building even the simplest tutorial example,
running python 2.3.3 on a solaris 2.6 system and boost.python 1.31.0:
This is what I get when following the tutorial build instructions:

 hjoukl at dev-b .../boost_1_31_0 $ cd libs/python/example/tutorial/
0 hjoukl at dev-b .../tutorial $ export PYTHON_ROOT=/opt/tools
0 hjoukl at dev-b .../tutorial $ export PYTHON_VERSION=2.3
0 hjoukl at dev-b .../tutorial $ which gcc
/opt/tools/bin/gcc
0 hjoukl at dev-b .../tutorial $ ./bjam
...found 1972 targets...
...updating 24 targets...
gcc-C++-action
../../../../bin/boost/libs/python/build/libboost_python.so/gcc/debug/shared-linkable-true/numeric.o
In file included from /opt/tools/include/python2.3/Python.h:8,
                 from
/data/projects/hjoukl/new_boost/boost_1_31_0/boost/python/detail/wrap_python.hpp:121,
                 from
/data/projects/hjoukl/new_boost/boost_1_31_0/boost/python/detail/prefix..hpp:13,
                 from
/data/projects/hjoukl/new_boost/boost_1_31_0/boost/python/numeric.hpp:9,
                 from
/data/projects/hjoukl/new_boost/boost_1_31_0/libs/python/build/../src/numeric.cpp:7:
/opt/tools/include/python2.3/pyconfig.h:825: warning: `_FILE_OFFSET_BITS'
redefined
/usr/include/sys/feature_tests.h:80: warning: this is the location of the
previous definition
/opt/tools/bin/objcopy: unrecognized section flag `debug'
/opt/tools/bin/objcopy: supported flags: alloc, load, readonly, code, data,
rom, contents

...

    set -e
    g++   -c -Wall -ftemplate-depth-100  -DBOOST_PYTHON_DYNAMIC_LIB
-DBOOST_PYTHON_SOURCE  -g -O0 -fno-inline -fPIC
-I"../../../../bin/boost/libs/python/build"  -I
"/opt/tools/include/python2.3" -I
"/data/projects/hjoukl/new_boost/boost_1_31_0"  -o
"../../../../bin/boost/libs/python/build/libboost_python.so/gcc/debug/shared-linkable-true/object_operators.o"

"/data/projects/hjoukl/new_boost/boost_1_31_0/libs/python/build/../src/object_operators.cpp"
    /opt/tools/bin/objcopy --set-section-flags .debug_str=contents,debug
"../../../../bin/boost/libs/python/build/libboost_python.so/gcc/debug/shared-linkable-true/object_operators.o"

...failed gcc-C++-action
../../../../bin/boost/libs/python/build/libboost_python.so/gcc/debug/shared-linkable-true/object_operators.o...
...removing
../../../../bin/boost/libs/python/build/libboost_python.so/gcc/debug/shared-linkable-true/object_operators.o
...skipped
<@boost!libs!python!build/libboost_python.so/gcc/debug/shared-linkable-true>libboost_python.so
 for lack of
<@boost!libs!python!build/libboost_python.so/gcc/debug/shared-linkable-true>numeric.o...
...skipped
<@boost!libs!python!example!tutorial/hello.so/gcc/debug/shared-linkable-true>hello.so
 for lack of
<@boost!libs!python!build/libboost_python.so/gcc/debug/shared-linkable-true>libboost_python.so...
...failed updating 22 targets...
...skipped 2 targets...

As I understand this, boost_python shared library is looked for in a debug
version, which is not found as for some other reason,
building the debug objects failed. There is indeed a "normal" boost_python
lib:

1 hjoukl at dev-b .../tutorial $ ls -l
../../../../bin/boost/libs/python/build/libboost_python.so/gcc/release/shared-linkable-true/libboost_python-gcc-1_31.so
-rwxr-xr-x   1 hjoukl   intern     448536 Mar  2 13:00
../../../../bin/boost/libs/python/build/libboost_python.so/gcc/release/shared-linkable-true/libboost_python-gcc-1_31.so*


Also the objcopy step seems not to work, probably for the same reason?
No hello.so is created and the jam build system remains a mystery to me at
the moment.

Any hints are appreciated,
G
  Holger

Der Inhalt dieser E-Mail ist vertraulich. Falls Sie nicht der angegebene
Empf?nger sind oder falls diese E-Mail irrt?mlich an Sie adressiert wurde,
verst?ndigen Sie bitte den Absender sofort und l?schen Sie die E-Mail
sodann. Das unerlaubte Kopieren sowie die unbefugte ?bermittlung sind nicht
gestattet. Die Sicherheit von ?bermittlungen per E-Mail kann nicht
garantiert werden. Falls Sie eine Best?tigung w?nschen, fordern Sie bitte
den Inhalt der E-Mail als Hardcopy an.

The contents of this  e-mail are confidential. If you are not the named
addressee or if this transmission has been addressed to you in error,
please notify the sender immediately and then delete this e-mail.  Any
unauthorized copying and transmission is forbidden. E-Mail transmission
cannot be guaranteed to be secure. If verification is required, please
request a hard copy version.






From rwgk at yahoo.com  Thu Mar  4 20:55:25 2004
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Thu, 4 Mar 2004 11:55:25 -0800 (PST)
Subject: [C++-sig] building boost.python problems
In-Reply-To: <OF9522F46B.A36EDE12-ONC1256E4D.00538714@lbbw.sko.de>
Message-ID: <20040304195525.51371.qmail@web20202.mail.yahoo.com>

FWIW: When I tried the last time a few months ago Boost.Python built OK under
Solaris 5.9 with gcc 3.3.1, using Python 2.3. I also know about a successful
build with gcc 3.2.2 and Python 2.3.2 (don't know the Solaris version). A
SCons-based build system was used in both cases, but your error message doesn't
seem to be related to the build system.

What is the output of g++ --version on your machine?

Ralf


__________________________________
Do you Yahoo!?
Yahoo! Search - Find what you?re looking for faster
http://search.yahoo.com



From grafik.list at redshift-software.com  Thu Mar  4 21:33:24 2004
From: grafik.list at redshift-software.com (Rene Rivera)
Date: Thu, 04 Mar 2004 14:33:24 -0600
Subject: [C++-sig] RE: building boost.python problems
Message-ID: <40479294.8020801@redshift-software.com>

Sorry for the top post, deleted the original email too fast :-(

...

> Dear boost.pythoneers,
> first of all, sorry for the lengthish post (and more so, for the
> confidential blabla my employer attaches to all my outgoing mails).
> 
> I?m having a little trouble building even the simplest tutorial example,
> running python 2.3.3 on a solaris 2.6 system and boost.python 1.31.0:
> This is what I get when following the tutorial build instructions:
> 
>  hjoukl at dev-b .../boost_1_31_0 $ cd libs/python/example/tutorial/
> 0 hjoukl at dev-b .../tutorial $ export PYTHON_ROOT=/opt/tools
> 0 hjoukl at dev-b .../tutorial $ export PYTHON_VERSION=2.3
> 0 hjoukl at dev-b .../tutorial $ which gcc
> /opt/tools/bin/gcc
> 0 hjoukl at dev-b .../tutorial $ ./bjam
> ...found 1972 targets...
> ...updating 24 targets...
> gcc-C++-action
> ../../../../bin/boost/libs/python/build/libboost_python.so/gcc/debug/shared-linkable-true/numeric.o
> In file included from /opt/tools/include/python2.3/Python.h:8,
>                  from
> /data/projects/hjoukl/new_boost/boost_1_31_0/boost/python/detail/wrap_python.hpp:121,
>                  from
> /data/projects/hjoukl/new_boost/boost_1_31_0/boost/python/detail/prefix..hpp:13,
>                  from
> /data/projects/hjoukl/new_boost/boost_1_31_0/boost/python/numeric.hpp:9,
>                  from
> /data/projects/hjoukl/new_boost/boost_1_31_0/libs/python/build/../src/numeric.cpp:7:
> /opt/tools/include/python2.3/pyconfig.h:825: warning: `_FILE_OFFSET_BITS'
> redefined
> /usr/include/sys/feature_tests.h:80: warning: this is the location of the
> previous definition
> /opt/tools/bin/objcopy: unrecognized section flag `debug'
> /opt/tools/bin/objcopy: supported flags: alloc, load, readonly, code, data,
> rom, contents

I think all your problems stems from the objcopy not working. This prevents 
the linking from continuing and basically nothing getting produced. Two 
options to fix this are to: disable objcopy on Solaris, or figure out which 
section flag to use instead of "debug". But some explanation first... The 
objcopy is a workaround for the n^2 linking problems of GCC, it has the effect 
of disabling the problematic symbol reduction algorithm in the linker. For BPL 
this is important because of the large number of _debug_ symbols generated by 
the templates.

OK, so to the first option disabling the objcopy. Try this change...

Index: gcc-tools.jam
===================================================================
RCS file: /cvsroot/boost/boost/tools/build/v1/gcc-tools.jam,v
retrieving revision 1.86
diff -u -2 -r1.86 gcc-tools.jam
--- gcc-tools.jam       26 Jan 2004 20:21:22 -0000      1.86
+++ gcc-tools.jam       4 Mar 2004 20:12:57 -0000
@@ -160,4 +160,5 @@
          {
          NO_GNU_LN = true ; # sun seems not to use the GNU linker with gcc
+        NO_OBJCOPY = true ;
          }
      case Linux :
@@ -350,6 +351,10 @@
  # used to manipulate the object files produced by GCC to
  # prevent the merging of debug symbols (which happens to be n^2 slow)
-flags gcc .OBJCOPY : [ GLOB $(GCC_BIN_DIRECTORY) $(PATH) : objcopy ] ;
-flags gcc OBJCOPY_FLAGS <debug-symbols>on : "--set-section-flags 
.debug_str=contents,debug" ;
+if ! $(NO_OBJCOPY)
+{
+    flags gcc .OBJCOPY : [ GLOB $(GCC_BIN_DIRECTORY) $(PATH) : objcopy ] ;
+    flags gcc OBJCOPY_FLAGS <debug-symbols>on : "--set-section-flags 
.debug_str=contents,debug" ;
+}
+
  if ! $(on-windows)
  {

For the second option, you could try using "--set-section-flags 
.debug_str=contents" (that is removing the ",debug"). Which this patch does...

Index: gcc-tools.jam
===================================================================
RCS file: /cvsroot/boost/boost/tools/build/v1/gcc-tools.jam,v
retrieving revision 1.86
diff -u -1 -r1.86 gcc-tools.jam
--- gcc-tools.jam       26 Jan 2004 20:21:22 -0000      1.86
+++ gcc-tools.jam       4 Mar 2004 20:23:30 -0000
@@ -161,2 +161,3 @@
          NO_GNU_LN = true ; # sun seems not to use the GNU linker with gcc
+        flags gcc OBJCOPY_FLAGS <debug-symbols>on : "--set-section-flags 
.debug_str=contents" ;
          }
@@ -166,2 +167,3 @@
          SONAME = -Wl,-soname, ;
+        flags gcc OBJCOPY_FLAGS <debug-symbols>on : "--set-section-flags 
.debug_str=contents,debug" ;
          }
@@ -352,3 +354,3 @@
  flags gcc .OBJCOPY : [ GLOB $(GCC_BIN_DIRECTORY) $(PATH) : objcopy ] ;
-flags gcc OBJCOPY_FLAGS <debug-symbols>on : "--set-section-flags 
.debug_str=contents,debug" ;
+
  if ! $(on-windows)


And of course I'd like to know which one works so I can apply the change to 
future releases.


...HTH :-)

-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org - grafik/redshift-software.com - 102708583/icq



From pearu at cens.ioc.ee  Thu Mar  4 22:16:58 2004
From: pearu at cens.ioc.ee (Pearu Peterson)
Date: Thu, 4 Mar 2004 23:16:58 +0200 (EET)
Subject: [C++-sig] wrapping class X with private constructor X::X(const X&)
Message-ID: <Pine.LNX.4.21.0403042246240.16684-100000@cens.kybi>


Hi,

I am trying to wrap a library class that is special in the sense
that its constructor is specified as private and this causes
compilation to fail. Is it true that boost.python tries
to use X::X(const X&) internally? Is is possible to avoid it?
(I tried to use no_init and init<>() with no success.)

Thanks,
Pearu

Details
-------

GCC: 3.3.3
Here is a fragment of my boost.python wrapper:

  class_<agg::matrix<agg::int8u> >("matrix");
  //where `int8u` is typedef `unsigned char`

Here's the definition of agg::matrix:

template<class T> class matrix
{
    public:
        matrix();
        matrix(T* buf, unsigned width, unsigned height, int stride);

    <snip>

    private:
        matrix(const matrix&);

    <snip>

}

and here follows the error message:

bpl_src/rendering_buffer.cpp: In function `void
export_rendering_buffer()':
bpl_src/rendering_buffer.cpp:8: warning: unused variable `char 
   rendering_buffer_doc[26]'
agg2/include/agg_matrix.h: In constructor `
   boost::python::objects::value_holder<Value>::value_holder(PyObject*,
A0) 
   [with A0 = boost::reference_wrapper<const agg::matrix<int8u> >, Value = 
   agg::matrix<int8u>]':
boost_1_31_0/boost/python/object/make_instance.hpp:69:   instantiated from
`static Holder* boost::python::objects::make_instance<T,
Holder>::construct(void*, PyObject*, boost::reference_wrapper<const
T>) [with T = agg::matrix<int8u>, Holder =
boost::python::objects::value_holder<agg::matrix<int8u> >]'
boost_1_31_0/boost/python/object/make_instance.hpp:43:   instantiated from
`static PyObject* boost::python::objects::make_instance_impl<T, Holder,
Derived>::execute(Arg&) [with Arg = const boost::reference_wrapper<const
agg::matrix<int8u> >, T = agg::matrix<int8u>, Holder =
boost::python::objects::value_holder<agg::matrix<int8u> >, Derived =
boost::python::objects::make_instance<agg::matrix<int8u>,
boost::python::objects::value_holder<agg::matrix<int8u> > >]'
boost_1_31_0/boost/python/object/class_wrapper.hpp:27:   instantiated from
`static PyObject* boost::python::objects::class_cref_wrapper<Src,
MakeInstance>::convert(const Src&) [with Src = agg::matrix<int8u>,
MakeInstance = boost::python::objects::make_instance<agg::matrix<int8u>,
boost::python::objects::value_holder<agg::matrix<int8u> > >]'
boost_1_31_0/boost/python/converter/as_to_python_function.hpp:28:   instantiated
from `static PyObject* boost::python::converter::as_to_python_function<T,
ToPython>::convert(const void*) [with T = agg::matrix<int8u>, ToPython =
boost::python::objects::class_cref_wrapper<agg::matrix<int8u>,
boost::python::objects::make_instance<agg::matrix<int8u>,
boost::python::objects::value_holder<agg::matrix<int8u> > > >]'
boost_1_31_0/boost/python/to_python_converter.hpp:34:   instantiated from
`boost::python::to_python_converter<T,
Conversion>::to_python_converter() [with T = agg::matrix<int8u>,
Conversion =
boost::python::objects::class_cref_wrapper<agg::matrix<int8u>,
boost::python::objects::make_instance<agg::matrix<int8u>,
boost::python::objects::value_holder<agg::matrix<int8u> > > >]'
boost_1_31_0/boost/python/class.hpp:105:   instantiated from `void
boost::python::detail::register_class_to_python(boost::mpl::bool_<true>,
SelectHolder, T*) [with T = agg::matrix<int8u>, SelectHolder =
boost::python::objects::select_holder<agg::matrix<int8u>,
boost::python::detail::not_specified>]'
boost_1_31_0/boost/python/class.hpp:631:   instantiated from `void
boost::python::class_<T, X1, X2, X3>::register_() const [with T =
agg::matrix<int8u>, X1 = boost::python::detail::not_specified, X2 =
boost::python::detail::not_specified, X3 =
boost::python::detail::not_specified]'
boost_1_31_0/boost/python/class.hpp:660:   instantiated from
`boost::python::class_<T, X1, X2, X3>::class_(const char*,
boost::python::no_init_t) [with T = agg::matrix<int8u>, X1 =
boost::python::detail::not_specified, X2 =
boost::python::detail::not_specified, X3 =
boost::python::detail::not_specified]'
bpl_src/rendering_buffer.cpp:11:   instantiated from here
agg2/include/agg_matrix.h:124: error: `agg::matrix<T>::matrix(const 
   agg::matrix<T>&) [with T = int8u]' is private
boost_1_31_0/boost/python/object/value_holder.hpp:117: error: within this 
   context





From dave at boost-consulting.com  Thu Mar  4 22:33:50 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 04 Mar 2004 16:33:50 -0500
Subject: [C++-sig] Re: wrapping class X with private constructor X::X(const
	X&)
References: <Pine.LNX.4.21.0403042246240.16684-100000@cens.kybi>
Message-ID: <ur7w870w1.fsf@boost-consulting.com>

Pearu Peterson <pearu at cens.ioc.ee> writes:

> I am trying to wrap a library class that is special in the sense
> that its constructor is specified as private and this causes
> compilation to fail. Is it true that boost.python tries
> to use X::X(const X&) internally? 

Yes; it has to figure out how to return Xs from functions, should you
wrap such a function.

> Is is possible to avoid it?

yes (see below).

> (I tried to use no_init and init<>() with no success.)
>
> Thanks,
> Pearu
>
> Details
> -------
>
> GCC: 3.3.3
> Here is a fragment of my boost.python wrapper:
>
>   class_<agg::matrix<agg::int8u> >("matrix");
add ", boost::noncopyable" -------^


-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From mike at nospam.com  Thu Mar  4 22:36:36 2004
From: mike at nospam.com (Mike Rovner)
Date: Thu, 4 Mar 2004 13:36:36 -0800
Subject: [C++-sig] Re: wrapping class X with private constructor X::X(const
	X&)
References: <Pine.LNX.4.21.0403042246240.16684-100000@cens.kybi>
Message-ID: <c287h3$q7k$1@sea.gmane.org>

Pearu Peterson wrote:
> Is it true that boost.python tries
> to use X::X(const X&) internally? Is is possible to avoid it?

Yes. Yes, use boost::noncopyable as in

class_<Klass, boost::noncopyable>("Klass", no_init) ...

HTH,
Mike






From Holger.Joukl at LBBW.de  Fri Mar  5 09:39:20 2004
From: Holger.Joukl at LBBW.de (Holger Joukl)
Date: Fri, 5 Mar 2004 09:39:20 +0100
Subject: [C++-sig] building boost.python problems 
Message-ID: <OF90198FC6.8CD7A47F-ONC1256E4E.002F458C@lbbw.sko.de>


What is the output of g++ --version on your machine?

 g++ -v
Reading specs from /opt/tools/lib/gcc-lib/sparc-sun-solaris2.6/2.95.2/specs
gcc version 2.95.2 19991024 (release)

Oops, forgot that in my original email. I will try out Rene?s proposed
solutions and
post the results.
Holger
_________________________________

Holger Joukl
Landesbank Baden-W?rttemberg
1651 Financial Markets Technologies
fon +49 (711) 124 - 7078
fax +49 (711) 124 - 3759
_________________________________

Der Inhalt dieser E-Mail ist vertraulich. Falls Sie nicht der angegebene
Empf?nger sind oder falls diese E-Mail irrt?mlich an Sie adressiert wurde,
verst?ndigen Sie bitte den Absender sofort und l?schen Sie die E-Mail
sodann. Das unerlaubte Kopieren sowie die unbefugte ?bermittlung sind nicht
gestattet. Die Sicherheit von ?bermittlungen per E-Mail kann nicht
garantiert werden. Falls Sie eine Best?tigung w?nschen, fordern Sie bitte
den Inhalt der E-Mail als Hardcopy an.

The contents of this  e-mail are confidential. If you are not the named
addressee or if this transmission has been addressed to you in error,
please notify the sender immediately and then delete this e-mail.  Any
unauthorized copying and transmission is forbidden. E-Mail transmission
cannot be guaranteed to be secure. If verification is required, please
request a hard copy version.






From Holger.Joukl at LBBW.de  Fri Mar  5 11:07:56 2004
From: Holger.Joukl at LBBW.de (Holger Joukl)
Date: Fri, 5 Mar 2004 11:07:56 +0100
Subject: [C++-sig] RE: building boost.python problems
Message-ID: <OF412B10A9.8E9EC25F-ONC1256E4E.00374E8B@lbbw.sko.de>


Hmm, I can?t seem to apply Rene Rivera?s patches; I always
get this:
2 hjoukl at dev-b .../v1 $ patch -i gcc-tools.patch
  Looks like a unified context diff.
  I can't seem to find a patch in there anywhere.

Will try inserting the changes manually now.
Regards,
 Holger

Der Inhalt dieser E-Mail ist vertraulich. Falls Sie nicht der angegebene
Empf?nger sind oder falls diese E-Mail irrt?mlich an Sie adressiert wurde,
verst?ndigen Sie bitte den Absender sofort und l?schen Sie die E-Mail
sodann. Das unerlaubte Kopieren sowie die unbefugte ?bermittlung sind nicht
gestattet. Die Sicherheit von ?bermittlungen per E-Mail kann nicht
garantiert werden. Falls Sie eine Best?tigung w?nschen, fordern Sie bitte
den Inhalt der E-Mail als Hardcopy an.

The contents of this  e-mail are confidential. If you are not the named
addressee or if this transmission has been addressed to you in error,
please notify the sender immediately and then delete this e-mail.  Any
unauthorized copying and transmission is forbidden. E-Mail transmission
cannot be guaranteed to be secure. If verification is required, please
request a hard copy version.






From Holger.Joukl at LBBW.de  Fri Mar  5 12:34:37 2004
From: Holger.Joukl at LBBW.de (Holger Joukl)
Date: Fri, 5 Mar 2004 12:34:37 +0100
Subject: [C++-sig] building boost.python problems
Message-ID: <OF6ABADB73.4D1EA0C7-ONC1256E4E.003DAE9C@lbbw.sko.de>


My problems to apply Rene?s patches resulted from some mail-related issues
I think.
Some lines were wrapped where they are not in the sources and also a
problem with tabs.
Don?t know why, maybe our unspeakable Lotus Notes. Or lines already wrapped
by the
sending mail client.

Anyway, I applied the changes manually. BOTH solutions worked fine for me.

I repost the diffs here, just for completeness. Rene?s originals are
the authoritative ones (the line numbers are a little different, too), of
course.

P1:
1 hjoukl at dev-b .../v1 $ diff -u -1 gcc-tools.jam.ORIGINAL
gcc-tools.jam.rivera1_patched
--- gcc-tools.jam.ORIGINAL      Fri Mar  5 09:53:35 2004
+++ gcc-tools.jam.rivera1_patched       Fri Mar  5 11:26:02 2004
@@ -162,2 +162,3 @@
         NO_GNU_LN = true ; # sun seems not to use the GNU linker with gcc
+        NO_OBJCOPY = true ;
         }
@@ -352,4 +353,8 @@
 # prevent the merging of debug symbols (which happens to be n^2 slow)
-flags gcc .OBJCOPY : [ GLOB $(GCC_BIN_DIRECTORY) $(PATH) : objcopy ] ;
-flags gcc OBJCOPY_FLAGS <debug-symbols>on : "--set-section-flags
.debug_str=contents,debug" ;
+if ! $(NO_OBJCOPY)
+{
+    flags gcc .OBJCOPY : [ GLOB $(GCC_BIN_DIRECTORY) $(PATH) : objcopy ] ;
+    flags gcc OBJCOPY_FLAGS <debug-symbols>on : "--set-section-flags
.debug_str=contents,debug" ;
+}
+
 if ! $(on-windows)


P2:
1 hjoukl at dev-b .../v1 $ diff -u -1 gcc-tools.jam.ORIGINAL
gcc-tools.jam.rivera2_patched
--- gcc-tools.jam.ORIGINAL      Fri Mar  5 09:53:35 2004
+++ gcc-tools.jam.rivera2_patched       Fri Mar  5 12:16:32 2004
@@ -162,2 +162,3 @@
         NO_GNU_LN = true ; # sun seems not to use the GNU linker with gcc
+        flags gcc OBJCOPY_FLAGS <debug-symbols>on : "--set-section-flags
.debug_str=contents" ;
         }
@@ -167,2 +168,3 @@
         SONAME = -Wl,-soname, ;
+        flags gcc OBJCOPY_FLAGS <debug-symbols>on : "--set-section-flags
.debug_str=contents,debug" ;
         }
@@ -353,3 +355,3 @@
 flags gcc .OBJCOPY : [ GLOB $(GCC_BIN_DIRECTORY) $(PATH) : objcopy ] ;
-flags gcc OBJCOPY_FLAGS <debug-symbols>on : "--set-section-flags
.debug_str=contents,debug" ;
+
 if ! $(on-windows)
1 hjoukl at dev-b .../v1 $

So, a big thank you, Rene!
Now I am gonna try to wrap the TIB/Rv API. Follow-up questions are sure to
come, soon :-)

Holger

Der Inhalt dieser E-Mail ist vertraulich. Falls Sie nicht der angegebene
Empf?nger sind oder falls diese E-Mail irrt?mlich an Sie adressiert wurde,
verst?ndigen Sie bitte den Absender sofort und l?schen Sie die E-Mail
sodann. Das unerlaubte Kopieren sowie die unbefugte ?bermittlung sind nicht
gestattet. Die Sicherheit von ?bermittlungen per E-Mail kann nicht
garantiert werden. Falls Sie eine Best?tigung w?nschen, fordern Sie bitte
den Inhalt der E-Mail als Hardcopy an.

The contents of this  e-mail are confidential. If you are not the named
addressee or if this transmission has been addressed to you in error,
please notify the sender immediately and then delete this e-mail.  Any
unauthorized copying and transmission is forbidden. E-Mail transmission
cannot be guaranteed to be secure. If verification is required, please
request a hard copy version.






From grafik.list at redshift-software.com  Fri Mar  5 16:22:27 2004
From: grafik.list at redshift-software.com (Rene Rivera)
Date: Fri, 05 Mar 2004 09:22:27 -0600
Subject: [C++-sig] building boost.python problems
In-Reply-To: <OF6ABADB73.4D1EA0C7-ONC1256E4E.003DAE9C@lbbw.sko.de>
References: <OF6ABADB73.4D1EA0C7-ONC1256E4E.003DAE9C@lbbw.sko.de>
Message-ID: <40489B33.4010000@redshift-software.com>

Holger Joukl wrote:
> My problems to apply Rene?s patches resulted from some mail-related issues
> I think.
> Some lines were wrapped where they are not in the sources and also a
> problem with tabs.
> Don?t know why, maybe our unspeakable Lotus Notes. Or lines already wrapped
> by the
> sending mail client.

Yes likely some word wraps on both my end and your end :-)

> Anyway, I applied the changes manually. BOTH solutions worked fine for me.

Good.

> So, a big thank you, Rene!

You're welcome. Question for you... Was either solution faster doing the link? 
As I'd rather go with the faster one.


-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org - grafik/redshift-software.com - 102708583/icq



From Holger.Joukl at LBBW.de  Fri Mar  5 16:41:51 2004
From: Holger.Joukl at LBBW.de (Holger Joukl)
Date: Fri, 5 Mar 2004 16:41:51 +0100
Subject: [c++-sig] private copy constructor with callback from c++
Message-ID: <OF6829A3CA.9A2CFBC5-ONC1256E4E.0054C51F@lbbw.sko.de>

Yet another long post...
I am trying to wrap Tibco?s TIB/Rv API, using BPL 1.31.0 and pyste, gcc
2.95.2, solaris 2.6.

There happens to be a call from the C++ side to TibrvMsgCallback::onEvent
(with BPL
it is actually TibrvMsgCallback_Wrapper::onEvent?), giving a TibrvListener*
as an argument.
Now the problem seems to me that the TibrvListener class (inheriting from
TibrvEvent), has a private copy
constructor.
This results in a core dump when performing the call.
Do I have chance of avoiding this problem?

Funny thing is, I first tried to build all the stuff by hand, not using
bjam -- there I managed to not
run into core, but got an error message that no to_python converter was
registered for
TibrvListener/TibrvEvent; so maybe I could write a custom converter?

Thank you very much for any help
  Holger

Btw I do not have access to my box until tuesday, so no hurry/do not wonder
if I will not
respond quickly.

Here is the wrapper code:


// Boost Includes
==============================================================
#include <boost/python.hpp>
#include <boost/cstdint.hpp>

// Includes
====================================================================
#include <tibrv/tibrvcpp.h>

// Using
=======================================================================
using namespace boost::python;

// Declarations
================================================================



// Boost Includes
==============================================================
#include <boost/python.hpp>
#include <boost/cstdint.hpp>

// Includes
====================================================================
#include <tibrv/tibrvcpp.h>

// Using
=======================================================================
using namespace boost::python;

// Declarations
================================================================

struct void_;

BOOST_PYTHON_OPAQUE_SPECIALIZED_TYPE_ID(void_);

void_ *TibrvEvent_getClosure_wrapper(TibrvEvent* event)
{
    return (void_ *) event->getClosure();
}

// Boost Includes
==============================================================
#include <boost/python.hpp>
#include <boost/cstdint.hpp>

// Includes
====================================================================
#include <tibrv/tibrvcpp.h>

// Using
=======================================================================
using namespace boost::python;

// Declarations
================================================================

struct void_;

BOOST_PYTHON_OPAQUE_SPECIALIZED_TYPE_ID(void_);

void_ *TibrvEvent_getClosure_wrapper(TibrvEvent* event)
{
    return (void_ *) event->getClosure();
}

//BOOST_PYTHON_OPAQUE_SPECIALIZED_TYPE_ID(void)



namespace  {

struct TibrvEvent_Wrapper: TibrvEvent
{
    TibrvStatus destroy(TibrvEventOnComplete* p0) {
        return call_method< TibrvStatus >(self, "destroy", p0);
    }

    TibrvStatus default_destroy_0() {
        return TibrvEvent::destroy();
    }

    TibrvStatus default_destroy_1(TibrvEventOnComplete* p0) {
        return TibrvEvent::destroy(p0);
    }

    PyObject* self;
};

BOOST_PYTHON_MEMBER_FUNCTION_OVERLOADS(TibrvListener_create_overloads_4_5,
create, 4, 5)


}// namespace


// Module
======================================================================
void Export_TibrvEvent()
{
    class_< TibrvEvent, boost::noncopyable, TibrvEvent_Wrapper
>("TibrvEvent", no_init)
        .def("destroy", &TibrvEvent::destroy,
&TibrvEvent_Wrapper::default_destroy_1)
        .def("destroy", &TibrvEvent_Wrapper::default_destroy_0)
        .def("isValid", &TibrvEvent::isValid)
        .def("getQueue", &TibrvEvent::getQueue, return_value_policy<
reference_existing_object >())
        .def("getClosure", &TibrvEvent_getClosure_wrapper,
return_value_policy< return_opaque_pointer >())
        .def("isListener", &TibrvEvent::isListener)
        .def("isTimer", &TibrvEvent::isTimer)
        .def("isIOEvent", &TibrvEvent::isIOEvent)
        .def("getType", &TibrvEvent::getType)
        .def("getHandle", &TibrvEvent::getHandle)
    ;

    class_< TibrvListener, bases< TibrvEvent >, boost::noncopyable
>("TibrvListener", init<  >())
        .def("create", &TibrvListener::create,
TibrvListener_create_overloads_4_5())
        .def("getTransport", &TibrvListener::getTransport,
return_value_policy< reference_existing_object >())
        .def("getSubject", &TibrvListener::getSubject)
    ;

}


#include <iostream>

// Boost Includes
==============================================================
#include <boost/python.hpp>
#include <boost/cstdint.hpp>

// Includes
====================================================================
#include <tibrv/tibrvcpp.h>

// Using
=======================================================================
using namespace boost::python;

// Declarations
================================================================
namespace  {

struct TibrvCallback_Wrapper: TibrvCallback
{
    TibrvCallback_Wrapper(PyObject* self_, const TibrvCallback& p0):
        TibrvCallback(p0), self(self_) {}

    TibrvCallback_Wrapper(PyObject* self_):
        TibrvCallback(), self(self_) {}

    void onEvent(TibrvEvent* p0, TibrvMsg& p1) {
        call_method< void >(self, "onEvent", p0, p1);
    }

    PyObject* self;
};

struct TibrvMsgCallback_Wrapper: TibrvMsgCallback
{
    TibrvMsgCallback_Wrapper(PyObject* self_, const TibrvMsgCallback& p0):
        TibrvMsgCallback(p0), self(self_) {}

    TibrvMsgCallback_Wrapper(PyObject* self_):
        TibrvMsgCallback(), self(self_) {}

    void onMsg(TibrvListener* p0, TibrvMsg& p1) {
        cout << "-->TibrvMsgCallback_Wrapper::onMsg\n";
        call_method< void >(self, "onMsg", p0, p1);
        cout << "<--TibrvMsgCallback_Wrapper::onMsg\n";
    }

    void onEvent(TibrvEvent* p0, TibrvMsg& p1) {
        cout << "-->TibrvMsgCallback_Wrapper::onEvent\n";
        call_method< void >(self, "onEvent", p0, p1);
        cout << "<--TibrvMsgCallback_Wrapper::onEvent\n";
    }

    PyObject* self;
};


Der Inhalt dieser E-Mail ist vertraulich. Falls Sie nicht der angegebene
Empf?nger sind oder falls diese E-Mail irrt?mlich an Sie adressiert wurde,
verst?ndigen Sie bitte den Absender sofort und l?schen Sie die E-Mail
sodann. Das unerlaubte Kopieren sowie die unbefugte ?bermittlung sind nicht
gestattet. Die Sicherheit von ?bermittlungen per E-Mail kann nicht
garantiert werden. Falls Sie eine Best?tigung w?nschen, fordern Sie bitte
den Inhalt der E-Mail als Hardcopy an.

The contents of this  e-mail are confidential. If you are not the named
addressee or if this transmission has been addressed to you in error,
please notify the sender immediately and then delete this e-mail.  Any
unauthorized copying and transmission is forbidden. E-Mail transmission
cannot be guaranteed to be secure. If verification is required, please
request a hard copy version.






From dave at boost-consulting.com  Fri Mar  5 17:13:39 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Fri, 05 Mar 2004 11:13:39 -0500
Subject: [C++-sig] Re: private copy constructor with callback from c++
References: <OF6829A3CA.9A2CFBC5-ONC1256E4E.0054C51F@lbbw.sko.de>
Message-ID: <uwu5z46h8.fsf@boost-consulting.com>

"Holger Joukl" <Holger.Joukl at LBBW.de> writes:

> Now the problem seems to me that the TibrvListener class (inheriting
> from TibrvEvent), has a private copy constructor.  This results in a
> core dump when performing the call.

I think you've mis-diagnosed the problem.  There's basically no way
the accessibility of the copy ctor can have an effect on runtime
behavior.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From ron.clarke at hp.com  Fri Mar  5 02:43:54 2004
From: ron.clarke at hp.com (Ron Clarke)
Date: Thu, 4 Mar 2004 17:43:54 -0800
Subject: [C++-sig] Accessing string vector in Python
Message-ID: <c28m1e$i02$1@sea.gmane.org>

I am trying to wrap a class method (using Boost.Python 1.31, MS VS.NET2003,
Python 2.3) that returns a vector of std::string objects. Here is the C++
definition:

typedef std :: vector<std::string> StringVector;
...
StringVector* myClass::getOwners()
{...
}

When wrapping this method for Python use, I thought I could use the
vector_indexing_suite like so:
BOOST_PYTHON_MODULE(MyModule)
{
...
class_<StringVector>("StringVector")
    .def(vector_indexing_suite<StringVector>() )
    ;
 ...
}

This builds OK, and the StringVector is accessible in Python. However,
Python does not understand the individual elements within the vector. When I
try to access StringVector elements, Python complains:
sv = img.pubDict.getOwners()
for i in range(len(sv)):
    print sv[i]
TypeError: No Python class registered for C++ class class
std::basic_string<char,struct std::char_traits<char>,class
std::allocator<char> >

How do I get Python to understand/convert to string objects it can handle? I
tried registering a converter, but that wasn't the answer
(implicitly_convertible<std::string, boost::python::str>(); ).

I cannot make the StringVector contents boost::python::str objects, since
the code being wrapped must also be callable by C++ users. Any suggestions
are greatly appreciated.

Thanks in advance.

-- 
Ron Clarke
ron.clarke at hp.com






From johng at quakes.uq.edu.au  Mon Mar  8 03:00:46 2004
From: johng at quakes.uq.edu.au (John Gerschwitz)
Date: Mon, 08 Mar 2004 12:00:46 +1000
Subject: [C++-sig] problem with embedded python
Message-ID: <404BD3CE.3060409@quakes.uq.edu.au>

Thanks for your reply Pierre,

> The error is when linking ... to be able to help you, I would need the
> linking command.
>
Yes I was wonderng whether I am missing a library or whether it is a 
library ordering problem.

Here is the link step.

icc -o SerialAccessorTest.exe ./obj/SerialAccessorTest.o 
./obj/AbstractAccessor.o ./obj/PointData.o ./obj/ParallelAccessor.o 
./obj/SerialAccessorTestCase.o ./obj/SerialAccessor.o ./obj/Data.o 
-L/raid2/john/newEsys/esys/tools/CppUnitTest/lib 
-L/raid2/john/newEsys/esys/utils/lib 
-L/raid2/john/newEsys/esys/escript/lib 
-L/raid2/john/newEsys/esys/tools/boost/bin/boost/libs/python/build/libboost_python.so/intel-ia64/debug/shared-linkable-true 
  -L/usr/local/lib/python2.2/config 
-L/opt/intel/cmplrs/80.058/intel_cc_80/lib  -lCppUnitTest  -lUtils 
-lescript  -lboost_python-intel-d-1_31  -lpython2.2  -ldl  -lutil  -lguide

> On Thu, 2004-02-26 at 07:41, John Gerschwitz wrote:
>> Can anybody tell me why I am getting the following error. The problem 
>> C++ code
>> looks like this . The problem occurs trying to import numarray.
>> 
>>    Py_Initialize();
>>     handle<> numarrayModule(PyImport_ImportModule("numarray"));
>>     numeric::array pointData(make_tuple(1,2,3,4));
>> 
>> Thanks John
>> 
>> Traceback (most recent call last):
>>   File "/usr/local/lib/python2.2/site-packages/numarray/__init__.py", 
>> line 1, in ?
>>     from numarrayall import *
>>   File "/usr/local/lib/python2.2/site-packages/numarray/numarrayall.py", 
>> line 1, in ?
>>     from numerictypes import *
>>   File 
>> "/usr/local/lib/python2.2/site-packages/numarray/numerictypes.py", line 
>> 33, in ?
>>     from typeconv import typeConverters as _typeConverters
>>   File "/usr/local/lib/python2.2/site-packages/numarray/typeconv.py", 
>> line 6, in ?
>>     import _conv
>> ImportError: /usr/local/lib/python2.2/site-packages/numarray/_conv.so: 
>> undefined symbol: PyCObject_Type
>> 
>> 
>> 
>> _______________________________________________
>> C++-sig mailing list
>> C++-sig at python.org
>> http://mail.python.org/mailman/listinfo/c++-sig
> -- 
> Pierre Barbier de Reuille
> 
> INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
> Botanique et Bio-informatique de l'Architecture des Plantes
> TA40/PSII, Boulevard de la Lironde
> 34398 MONTPELLIER CEDEX 5, France
> 
> tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 





From halbert at bbn.com  Mon Mar  8 18:12:34 2004
From: halbert at bbn.com (Dan Halbert)
Date: Mon, 8 Mar 2004 12:12:34 -0500
Subject: [C++-sig] Pyste 1.31.0 and exceptions
Message-ID: <200403081712.i28HCYP27270@rolex.bbn.com>

Hi, I am trying to use the Pyste from Boost 1.31.0 with code that uses
exceptions. My impression was that this version of Pyste does support
automatic exception handling, but maybe I am wrong.

I was using gccxml-0.4.2, which does not output any info about exceptions.
I switched to gccxml-0.6.0, which does.

I have this test case:
==============================================
foo.pyste:

foo = AllFromHeader("foo.h")
==============================================
foo.h:

void g() throw (int);                       // line #1
class C { public: void f() throw (int); };  // line #2
==============================================

If I comment out line #2, then I get no errors, but the generated C++
code does not mention the exception for g().

If I comment out line #1 and leave #2 (or uncomment both), then Pyste
gets an error, with the Python stack trace shown below.

I am using "int" exceptions for simplicity in this test case, but I
also get a similar failure with class exceptions.

Thanks,
Dan

==============================================
Traceback (most recent call last):
  File "/d4m/tools/i686ling/Python-2.3.3/bin/pyste.py", line 4, in ?
    pyste.main()
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/pyste.py", line 405, in main
    status = Begin()
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/pyste.py", line 244, in Begin
    return GenerateCode(parser, module, out, interfaces, multiple)
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/pyste.py", line 372, in GenerateCode
    export.GenerateCode(codeunit, exported_names)
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/Exporter.py", line 50, in GenerateCode
    self.Export(codeunit, exported_names)
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/HeaderExporter.py", line 44, in Export
    self.HandleDeclaration(decl, codeunit, exported_names)
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/HeaderExporter.py", line 61, in HandleDeclaration
    self.HandleExporter(decl, exporter_class, codeunit, exported_names)
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/HeaderExporter.py", line 77, in HandleExporter
    exporter.GenerateCode(codeunit, exported_names)
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/Exporter.py", line 50, in GenerateCode
    self.Export(codeunit, exported_names)
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/ClassExporter.py", line 92, in Export
    self.InheritMethods(exported_names)
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/ClassExporter.py", line 119, in InheritMethods
    pointers = [x.PointerDeclaration(True) for x in self.class_ if isinstance(x, Method)]
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/declarations.py", line 320, in PointerDeclaration
    return '(%s (%s::*)(%s) %s%s)&%s' %\
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/declarations.py", line 226, in Exceptions
    return " throw(%s)" % ', '.join (self.throws)
TypeError: sequence item 0: expected string, FundamentalType found




From rclemley at booksys.com  Mon Mar  8 19:06:48 2004
From: rclemley at booksys.com (Rob Lemley)
Date: Mon, 8 Mar 2004 12:06:48 -0600
Subject: [C++-sig] problem with embedded python
In-Reply-To: <404BD3CE.3060409@quakes.uq.edu.au>
References: <404BD3CE.3060409@quakes.uq.edu.au>
Message-ID: <200403081206.48915.rclemley@booksys.com>

On Sunday 07 March 2004 20:00, John Gerschwitz wrote:
> > The error is when linking ... to be able to help you, I would need the
> > linking command.
>
> Yes I was wonderng whether I am missing a library or whether it is a
> library ordering problem.

It looks like the error occurs at runtime linking because of missing options
when building the python-embedding executable.

Python knows how to do it and the distutils module will tell you how to do it.

See section '5.6 Linking Requirements' in the 'Extending and Embedding'
document from the main documentation webpage for your version of python.
(http://python.org/doc/2.2.3/ext/link-reqs.html)

-RCL




From s_sourceforge at nedprod.com  Mon Mar  8 20:20:05 2004
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Mon, 08 Mar 2004 19:20:05 -0000
Subject: [C++-sig] Pyste 1.31.0 and exceptions
In-Reply-To: <200403081712.i28HCYP27270@rolex.bbn.com>
Message-ID: <404CC765.22345.340405DB@localhost>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 8 Mar 2004 at 12:12, Dan Halbert wrote:

> foo = AllFromHeader("foo.h")

Last time I heard AllFromHeader() was still not working.

Cheers,
Niall





-----BEGIN PGP SIGNATURE-----
Version: idw's PGP-Frontend 4.9.6.1 / 9-2003 + PGP 8.0.2

iQA/AwUBQEzHZ8EcvDLFGKbPEQIDSgCfe/v6wRYl5hevexc2gzOk9Ksk+6oAoIHT
21VDl94Kvf00oJCziXIHcicJ
=kJSd
-----END PGP SIGNATURE-----



From halbert at bbn.com  Mon Mar  8 21:01:58 2004
From: halbert at bbn.com (Dan Halbert)
Date: Mon, 8 Mar 2004 15:01:58 -0500
Subject: [C++-sig] Pyste 1.31.0 and exceptions
In-Reply-To: <404CC765.22345.340405DB@localhost> (s_sourceforge@nedprod.com)
References: <404CC765.22345.340405DB@localhost>
Message-ID: <200403082001.i28K1wj27655@rolex.bbn.com>

  >From: "Niall Douglas" <s_sourceforge at nedprod.com>
  >On 8 Mar 2004 at 12:12, Dan Halbert wrote:
  >
  >> foo = AllFromHeader("foo.h")
  >
  >Last time I heard AllFromHeader() was still not working.

Actually, it seems to work pretty well for me, though I promise to be
careful. But even if I change my example to what is below, and don't
use AllFromHeader(), I get the same error as I listed previously. --Dan

==============================================
foo.pyste:

##foo = AllFromHeader("foo.h")
Function("g", "foo.h")
Class("C", "foo.h")
==============================================
foo.h:

void g() throw (int);                       // line #1
class C { public: void f() throw (int); };  // line #2
==============================================

  ...
  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/declarations.py", line 226, in Exceptions
    return " throw(%s)" % ', '.join (self.throws)
TypeError: sequence item 0: expected string, FundamentalType found




From ostiguy at fnal.gov  Mon Mar  8 21:40:27 2004
From: ostiguy at fnal.gov (Francois Ostiguy)
Date: Mon, 8 Mar 2004 14:40:27 -0600 (CST)
Subject: [C++-sig] c++ to python automatic conversion
Message-ID: <Pine.GSO.4.58.0403081429350.2777@wally.fnal.gov>


Hi -

I am having difficulty with c++ to python conversion. Consider the
following snippet:

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

using namespace boost::python;

struct B {
    int n;
    int m;
    B(): n(0), m(0) {
    }
};

struct A {

    A() {}
    B* makeB() {
      return new B;
    }
};


BOOST_PYTHON_MODULE(demo)
{

class_<A>("A")
    .def("makeB",  &A::makeB, return_value_policy<manage_new_object>() );

class_<B>("B");

}
=====================================================================================

Then, the following works fine (no error)

Python 2.3.3 (#3, Mar  8 2004, 11:58:32)
[GCC 3.3.3] on sunos5
Type "help", "copyright", "credits" or "license" for more information.
>>> from demo import *
>>> a = A()
>>> b = a.makeB()
>>> type(b)
<class 'demo.B'>

However, if I put class A and class B in two different files (modules) say
demo1 and demo2 repectively, I get

Python 2.3.3 (#3, Mar  8 2004, 11:58:32)
[GCC 3.3.3] on sunos5
Type "help", "copyright", "credits" or "license" for more information.
>>> from demo1 import A
>>> from demo2 import B
>>> a = A()
>>> b = a.makeB()
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
TypeError: No Python class registered for C++ class B

Do I have to provide an explicit converter ?
Can you explain how this can be done ?

As usual, any help is greatly appreciated.

-Francois

----------------------------------------------------------------------------
Dr. Jean-Francois OSTIGUY                              voice: (630) 840-2231
Beam Physics Dept MS220                                  FAX: (630) 840-6039
Fermi National Accelerator Laboratory                email: ostiguy at fnal.gov
Batavia IL 60510-0500                           WWW:www-ap.fnal.gov/~ostiguy




From dave at boost-consulting.com  Mon Mar  8 22:24:27 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 08 Mar 2004 16:24:27 -0500
Subject: [C++-sig] Re: c++ to python automatic conversion
References: <Pine.GSO.4.58.0403081429350.2777@wally.fnal.gov>
Message-ID: <uu10z58xg.fsf@boost-consulting.com>

Francois Ostiguy <ostiguy at fnal.gov> writes:

> Then, the following works fine (no error)
>
> Python 2.3.3 (#3, Mar  8 2004, 11:58:32)
> [GCC 3.3.3] on sunos5
> Type "help", "copyright", "credits" or "license" for more information.
>>>> from demo import *
>>>> a = A()
>>>> b = a.makeB()
>>>> type(b)
> <class 'demo.B'>
>
> However, if I put class A and class B in two different files (modules) say
> demo1 and demo2 repectively, I get
>
> Python 2.3.3 (#3, Mar  8 2004, 11:58:32)
> [GCC 3.3.3] on sunos5
> Type "help", "copyright", "credits" or "license" for more information.
>>>> from demo1 import A
>>>> from demo2 import B
>>>> a = A()
>>>> b = a.makeB()
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
> TypeError: No Python class registered for C++ class B
>
> Do I have to provide an explicit converter ?
> Can you explain how this can be done ?
>
> As usual, any help is greatly appreciated.

What compiler did you use to build Boost.Python and your modules?

How did you build your modules?  Are they both linked to the same
libboost_python.so?

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From rwgk at yahoo.com  Mon Mar  8 23:03:46 2004
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Mon, 8 Mar 2004 14:03:46 -0800 (PST)
Subject: [C++-sig] Re: c++ to python automatic conversion
In-Reply-To: <uu10z58xg.fsf@boost-consulting.com>
Message-ID: <20040308220346.93101.qmail@web20212.mail.yahoo.com>

This reminds me of the serious trouble I had with FSF gcc (i.e. *not* Apple's
port) under Mac OS X. See here:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14232

To check if you have a similar same problem edit the file

boost/libs/python/src/converter/registry.cpp

Somewhere near the top insert:

#define BOOST_PYTHON_TRACE_REGISTRY

Recompile and relink libboost_python.so. Run your simple test and post the
output.

I hope this is not it but trying this out will only take a few minutes.

Ralf


--- David Abrahams <dave at boost-consulting.com> wrote:
> Francois Ostiguy <ostiguy at fnal.gov> writes:
> 
> > Then, the following works fine (no error)
> >
> > Python 2.3.3 (#3, Mar  8 2004, 11:58:32)
> > [GCC 3.3.3] on sunos5
> > Type "help", "copyright", "credits" or "license" for more information.
> >>>> from demo import *
> >>>> a = A()
> >>>> b = a.makeB()
> >>>> type(b)
> > <class 'demo.B'>
> >
> > However, if I put class A and class B in two different files (modules) say
> > demo1 and demo2 repectively, I get
> >
> > Python 2.3.3 (#3, Mar  8 2004, 11:58:32)
> > [GCC 3.3.3] on sunos5
> > Type "help", "copyright", "credits" or "license" for more information.
> >>>> from demo1 import A
> >>>> from demo2 import B
> >>>> a = A()
> >>>> b = a.makeB()
> > Traceback (most recent call last):
> >   File "<stdin>", line 1, in ?
> > TypeError: No Python class registered for C++ class B
> >
> > Do I have to provide an explicit converter ?
> > Can you explain how this can be done ?
> >
> > As usual, any help is greatly appreciated.
> 
> What compiler did you use to build Boost.Python and your modules?
> 
> How did you build your modules?  Are they both linked to the same
> libboost_python.so?


__________________________________
Do you Yahoo!?
Protect your identity with Yahoo! Mail AddressGuard
http://antispam.yahoo.com/whatsnewfree



From ostiguy at fnal.gov  Tue Mar  9 00:16:16 2004
From: ostiguy at fnal.gov (Francois Ostiguy)
Date: Mon, 8 Mar 2004 17:16:16 -0600 (CST)
Subject: [C++-sig] Re: c++ to python automatic conversion
In-Reply-To: <20040308220346.93101.qmail@web20212.mail.yahoo.com>
References: <20040308220346.93101.qmail@web20212.mail.yahoo.com>
Message-ID: <Pine.GSO.4.58.0403081705130.2777@wally.fnal.gov>



>What compiler did you use to build Boost.Python and your modules?

>How did you build your modules?  Are they both linked to the same
>libboost_python.so?



The modules were built using gcc 3.3. They were also both linked to the
same libboost_python ... but inadvertently to the static version (.a) of
that library. When I link to libboost_python.so,  everything works fine.
I suppose the library contains a list of registered classes. Linking with
the static version of the lib resulted produced two instances of that list
instead of the desired single shared instance.

Many Thanks !


-Francois

----------------------------------------------------------------------------
Dr. Jean-Francois OSTIGUY                              voice: (630) 840-2231
Beam Physics Dept MS220                                  FAX: (630) 840-6039
Fermi National Accelerator Laboratory                email: ostiguy at fnal.gov
Batavia IL 60510-0500                           WWW:www-ap.fnal.gov/~ostiguy




From dave at boost-consulting.com  Tue Mar  9 01:02:21 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 08 Mar 2004 19:02:21 -0500
Subject: [C++-sig] Re: c++ to python automatic conversion
References: <20040308220346.93101.qmail@web20212.mail.yahoo.com>
	<Pine.GSO.4.58.0403081705130.2777@wally.fnal.gov>
Message-ID: <uhdwy51ma.fsf@boost-consulting.com>

Francois Ostiguy <ostiguy at fnal.gov> writes:

>>What compiler did you use to build Boost.Python and your modules?
>
>>How did you build your modules?  Are they both linked to the same
>>libboost_python.so?
>
>
>
> The modules were built using gcc 3.3. They were also both linked to the
> same libboost_python ... but inadvertently to the static version (.a) of
> that library. When I link to libboost_python.so,  everything works fine.
> I suppose the library contains a list of registered classes. Linking with
> the static version of the lib resulted produced two instances of that list
> instead of the desired single shared instance.
>
> Many Thanks !

Hint: follow the formula that the examples in libs/python/example
use.  Run bjam with the -n or -d+2 option to see the build/run
commands.  If you want to use your own build system, just make sure
it replicates the same actions.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From nicodemus at esss.com.br  Tue Mar  9 02:08:23 2004
From: nicodemus at esss.com.br (Nicodemus)
Date: Mon, 08 Mar 2004 22:08:23 -0300
Subject: [C++-sig] Pyste 1.31.0 and exceptions
In-Reply-To: <200403082001.i28K1wj27655@rolex.bbn.com>
References: <404CC765.22345.340405DB@localhost>
	<200403082001.i28K1wj27655@rolex.bbn.com>
Message-ID: <404D1907.7060906@esss.com.br>

Dan Halbert wrote:

>  >From: "Niall Douglas" <s_sourceforge at nedprod.com>
>  >On 8 Mar 2004 at 12:12, Dan Halbert wrote:
>  >
>  >> foo = AllFromHeader("foo.h")
>  >
>  >Last time I heard AllFromHeader() was still not working.
>
>Actually, it seems to work pretty well for me, though I promise to be
>careful. But even if I change my example to what is below, and don't
>use AllFromHeader(), I get the same error as I listed previously. --Dan
>
>==============================================
>foo.pyste:
>
>##foo = AllFromHeader("foo.h")
>Function("g", "foo.h")
>Class("C", "foo.h")
>==============================================
>foo.h:
>
>void g() throw (int);                       // line #1
>class C { public: void f() throw (int); };  // line #2
>==============================================
>
>  ...
>  File "/d4m/tools/i686ling/Python-2.3.3/lib/python2.3/site-packages/Pyste/declarations.py", line 226, in Exceptions
>    return " throw(%s)" % ', '.join (self.throws)
>TypeError: sequence item 0: expected string, FundamentalType found
>
>

Thanks for reporting this. A patch that fixes this is applied to CVS.
AllFromHeader doesn't work for some classes hierarquies, for simple uses 
it works fine. I still can't find time to make it work... way too busy. 8(

Regards,
Nicodemus.


Index: declarations.py
===================================================================
RCS file: /cvsroot/boost/boost/libs/python/pyste/src/Pyste/declarations.py,v
retrieving revision 1.6
diff -r1.6 declarations.py
226c226
<             return " throw(%s)" % ', '.join (self.throws)
---
 >             return " throw(%s)" % ', '.join ([x.FullName() for x in 
self.throws])
239c239
<             return '(%s (*)(%s))&%s' % (result, params, self.FullName())
---
 >             return '(%s (*)(%s)%s)&%s' % (result, params, 
self.Exceptions(), self.FullName())
349c349
<         is_public = self.visibility = Scope.public
---
 >         is_public = self.visibility == Scope.public





From paul at paulgrenyer.co.uk  Tue Mar  9 12:57:52 2004
From: paul at paulgrenyer.co.uk (Paul Grenyer)
Date: Tue, 9 Mar 2004 11:57:52 GMT
Subject: [C++-sig] invalid literal for int()
Message-ID: <200403091157.i29Bvqh27322@ns.cricketthosting.co.uk>

Hi All

I've got a bit of a weird one. I'm passing a string value my embedded 
python environment like this:

m_main_namespace[fieldDetails.name.c_str()] = str();

Where fieldDetails.name == "TITLE".

I am trying to convert the value of title to an integer inside the 
python code that is run in the embedded environment:

intValue = int( TITLE[0:2] )

but the error I get is: "invalid literal for int()".

Anyone know how I can solve this?

Regards
Paul

Paul Grenyer
Email: paul at paulgrenyer.co.uk
Web: http://www.paulgrenyer.co.uk

Have you met Aeryn: http://www.paulgrenyer.co.uk/aeryn/?
Version 0.3.0 beta now available for download.





From paul at paulgrenyer.co.uk  Tue Mar  9 13:24:42 2004
From: paul at paulgrenyer.co.uk (Paul Grenyer)
Date: Tue, 9 Mar 2004 12:24:42 GMT
Subject: [C++-sig] invalid literal for int()
Message-ID: <200403091224.i29COgf28918@ns.cricketthosting.co.uk>

Hi

The plot has just thickened. The following doesn't produce any errors:

import string

houseno = ADD1[0:2]

if TITLE == "MR":
    if int(houseno) > 0 and int(houseno) < 16:
                outputFile = "Mr.out"
else:
    outputFile = "rest.out"

However, if I remove the first if statement:

import string

houseno = ADD1[0:2]

if int(houseno) > 0 and int(houseno) < 16:
    outputFile = "Mr.out"
else:
    outputFile = "rest.out"

I am back to the "invalid literal for int()" error.

I must be doing something very wrong somewhere.

Regards
Paul

Paul Grenyer
Email: paul at paulgrenyer.co.uk
Web: http://www.paulgrenyer.co.uk

Have you met Aeryn: http://www.paulgrenyer.co.uk/aeryn/?
Version 0.3.0 beta now available for download.





From dave at boost-consulting.com  Tue Mar  9 14:10:26 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Tue, 09 Mar 2004 08:10:26 -0500
Subject: [C++-sig] Re: invalid literal for int()
References: <200403091224.i29COgf28918@ns.cricketthosting.co.uk>
Message-ID: <uptbm17zx.fsf@boost-consulting.com>

"Paul Grenyer" <paul at paulgrenyer.co.uk> writes:

> Hi
>
> The plot has just thickened. The following doesn't produce any errors:
>
> import string
>
> houseno = ADD1[0:2]
>
> if TITLE == "MR":
>     if int(houseno) > 0 and int(houseno) < 16:
>                 outputFile = "Mr.out"
> else:
>     outputFile = "rest.out"
>
> However, if I remove the first if statement:
>
> import string
>
> houseno = ADD1[0:2]
>
> if int(houseno) > 0 and int(houseno) < 16:
>     outputFile = "Mr.out"
> else:
>     outputFile = "rest.out"
>
> I am back to the "invalid literal for int()" error.
>
> I must be doing something very wrong somewhere.

These are pure Python errors, having nothing to do with C++:

  >>> int('foo')
  Traceback (most recent call last):
    File "<stdin>", line 1, in ?
  ValueError: invalid literal for int(): foo


-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From paul at paulgrenyer.co.uk  Tue Mar  9 14:24:26 2004
From: paul at paulgrenyer.co.uk (Paul Grenyer)
Date: Tue, 9 Mar 2004 13:24:26 GMT
Subject: [C++-sig] Re: invalid literal for int()
Message-ID: <200403091324.i29DOQE32523@ns.cricketthosting.co.uk>

Hi Dave

> These are pure Python errors, having nothing to do with C++:
> 
>   >>> int('foo')
>   Traceback (most recent call last):
>     File "<stdin>", line 1, in ?
>   ValueError: invalid literal for int(): foo

Right. I've rechecked our code. I thought we were just checking syntax; 
we're not. That explains everything.

Thanks!

Regards
Paul

Paul Grenyer
Email: paul at paulgrenyer.co.uk
Web: http://www.paulgrenyer.co.uk

Have you met Aeryn: http://www.paulgrenyer.co.uk/aeryn/?
Version 0.3.0 beta now available for download.





From ostiguy at fnal.gov  Tue Mar  9 22:17:08 2004
From: ostiguy at fnal.gov (Francois Ostiguy)
Date: Tue, 9 Mar 2004 15:17:08 -0600 (CST)
Subject: [C++-sig] iterators and abstact base class
Message-ID: <Pine.GSO.4.58.0403091503550.385@wally.fnal.gov>


Hi -

This has to be an FAQ. I searched the list archives and I was not able
locate a satisfactory solution to the following problem:

Assume an abstract base class A and a derived class B. I have an iterator
I for a container C holding B-type objects.  The iterator returns a
pointer to the base class A.
At runtime, I.next() returns a type A which cannot be instantiated.
I would like to iterate through my container and call a (virtual) function
on each element.  Can you point me to an idiom/technique that would
allow me to accomplish this ?

As always, any help is greatly appreciated.

-Francois

----------------------------------------------------------------------------
Dr. Jean-Francois OSTIGUY                              voice: (630) 840-2231
Beam Physics Dept MS220                                  FAX: (630) 840-6039
Fermi National Accelerator Laboratory                email: ostiguy at fnal.gov
Batavia IL 60510-0500                           WWW:www-ap.fnal.gov/~ostiguy




From RaoulGough at yahoo.co.uk  Tue Mar  9 23:05:05 2004
From: RaoulGough at yahoo.co.uk (Raoul Gough)
Date: Tue, 09 Mar 2004 22:05:05 +0000
Subject: [C++-sig] Re: Accessing string vector in Python
References: <c28m1e$i02$1@sea.gmane.org>
Message-ID: <usmghvfqm.fsf@yahoo.co.uk>

"Ron Clarke" <ron.clarke at hp.com> writes:

> I am trying to wrap a class method (using Boost.Python 1.31, MS VS.NET2003,
> Python 2.3) that returns a vector of std::string objects. Here is the C++
> definition:
>
> typedef std :: vector<std::string> StringVector;
> ...
> StringVector* myClass::getOwners()
> {...
> }

Returning the vector via a plain pointer? I guess you've omitted the
return policy that this would require.

>
> When wrapping this method for Python use, I thought I could use the
> vector_indexing_suite like so:
> BOOST_PYTHON_MODULE(MyModule)
> {
> ...
> class_<StringVector>("StringVector")
>     .def(vector_indexing_suite<StringVector>() )
>     ;
>  ...
> }
>
> This builds OK, and the StringVector is accessible in Python. However,
> Python does not understand the individual elements within the vector. When I
> try to access StringVector elements, Python complains:
> sv = img.pubDict.getOwners()
> for i in range(len(sv)):
>     print sv[i]
> TypeError: No Python class registered for C++ class class
> std::basic_string<char,struct std::char_traits<char>,class
> std::allocator<char> >
>
> How do I get Python to understand/convert to string objects it can handle? I
> tried registering a converter, but that wasn't the answer
> (implicitly_convertible<std::string, boost::python::str>(); ).
>
> I cannot make the StringVector contents boost::python::str objects, since
> the code being wrapped must also be callable by C++ users. Any suggestions
> are greatly appreciated.
>
> Thanks in advance.

Sounds weird to me. Have you tried a very simple function that returns
a std::string? The vector indexing suite provides a __getitem__ method
that (in this case) will return a string, and that should all just
work automatically. Actually, maybe it returns a proxy object for the
string - can't exactly remember how the original indexing suite does
this...

Does something like the following work?

std::string get_string () { return std::string ("hello world"); }

  //...
  boost::python::def ("get_string", get_string);

# in python
s = get_string
print s

This should work, of course.

-- 
Raoul Gough.
export LESS='-X'




From dave at boost-consulting.com  Tue Mar  9 23:34:53 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Tue, 09 Mar 2004 17:34:53 -0500
Subject: [C++-sig] Re: iterators and abstact base class
References: <Pine.GSO.4.58.0403091503550.385@wally.fnal.gov>
Message-ID: <ullm9y7hu.fsf@boost-consulting.com>

Francois Ostiguy <ostiguy at fnal.gov> writes:

> Hi -
>
> This has to be an FAQ. I searched the list archives and I was not able
> locate a satisfactory solution to the following problem:
>
> Assume an abstract base class A and a derived class B. I have an iterator
> I for a container C holding B-type objects.  The iterator returns a
> pointer to the base class A.
> At runtime, I.next() 

?? This is a standard C++ iterator, or something else??

> returns a type A which cannot be instantiated.

Clearly it can't return A; it can return a reference to A, though.

> I would like to iterate through my container and call a (virtual) function
> on each element.  Can you point me to an idiom/technique that would
> allow me to accomplish this ?

Is this a Python/C++ question, a straight C++ question, or something
else?  Too many details are missing.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From halbert at bbn.com  Wed Mar 10 00:43:38 2004
From: halbert at bbn.com (Dan Halbert)
Date: Tue, 9 Mar 2004 18:43:38 -0500
Subject: [C++-sig] passing buffer of bytes from Python to C++
Message-ID: <200403092343.i29NhcE30830@rolex.bbn.com>

One of the C++ methods I've wrapped with BPL has a signature something like

    process_data(unsigned char const* buffer, unsigned long buffer_length)

In other words, it's just taking a buffer of bytes (it's 8-bit audio
data), which I'll be reading from a file or a network connection in
Python.

I can make the Python data object whatever I want (string, list,
tuple), but I see that BPL doesn't convert any of these to a char
pointer automatically. Is there a way to generate a simple char*
buffer with Python and BPL?

Later I'll also want to do the same thing for buffers of 16-bit
values, so is that also possible?


Thanks, and continuing thanks for putting up with all my questions,
Dan



From s_sourceforge at nedprod.com  Wed Mar 10 01:47:12 2004
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Wed, 10 Mar 2004 00:47:12 -0000
Subject: [C++-sig] passing buffer of bytes from Python to C++
In-Reply-To: <200403092343.i29NhcE30830@rolex.bbn.com>
Message-ID: <404E6590.3623.5EA7163@localhost>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 9 Mar 2004 at 18:43, Dan Halbert wrote:

> I can make the Python data object whatever I want (string, list,
> tuple), but I see that BPL doesn't convert any of these to a char
> pointer automatically. Is there a way to generate a simple char*
> buffer with Python and BPL?

Isn't there a FAQ entry about this?

Cheers,
Niall





-----BEGIN PGP SIGNATURE-----
Version: idw's PGP-Frontend 4.9.6.1 / 9-2003 + PGP 8.0.2

iQA/AwUBQE5lkcEcvDLFGKbPEQJ+UQCfWU+CAKeRGOOjm9xjB3lrCnX9+hYAnR2c
IZEhvwK+1aF4cyvnD7UdCp7g
=v1ex
-----END PGP SIGNATURE-----



From dave at boost-consulting.com  Wed Mar 10 13:45:34 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Wed, 10 Mar 2004 07:45:34 -0500
Subject: [C++-sig] Re: passing buffer of bytes from Python to C++
References: <200403092343.i29NhcE30830@rolex.bbn.com>
Message-ID: <u65dcx441.fsf@boost-consulting.com>

Dan Halbert <halbert at bbn.com> writes:

> One of the C++ methods I've wrapped with BPL has a signature something like
>
>     process_data(unsigned char const* buffer, unsigned long buffer_length)
>
> In other words, it's just taking a buffer of bytes (it's 8-bit audio
> data), which I'll be reading from a file or a network connection in
> Python.
>
> I can make the Python data object whatever I want (string, list,
> tuple), but I see that BPL doesn't convert any of these to a char
> pointer automatically. Is there a way to generate a simple char*
> buffer with Python and BPL?

Python strings will be automatically converted to char const*
                                                       ^^^^^

> Later I'll also want to do the same thing for buffers of 16-bit
> values, so is that also possible?

I'm afraid there's only a conversion from Python unicode strings to
std::wstring, but not to wchar_t const*.


-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From ak at ixion.net  Wed Mar 10 13:52:53 2004
From: ak at ixion.net (Andreas Kloeckner)
Date: Wed, 10 Mar 2004 13:52:53 +0100
Subject: [C++-sig] Tutorial example for derivation in python doesn't work
Message-ID: <20040310125253.GA2548@aramis.ath.cx>

Hi all,

I've tried to use python-implemented C++ interfaces as described in the
tutorial, with no luck. 

More specifically, I tried to just redo the example described in the
tutorial pages "Class Virtual Functions" and "Deriving a Python Class".
I copied-and-pasted  the examples there into two files (which you can
find attached to this email, along with my build script), built them,
and voila, I get the following error message:

Traceback (most recent call last):
  File "virtual_functions.py", line 7, in ?
      print call_f(Derived())
RuntimeError: This class cannot be instantiated from Python

Upon giving the derived class an __init__ function as commented out
below, the error changes to:

Traceback (most recent call last):
  File "virtual_functions.py", line 9, in ?
    print call_f(Derived())
Boost.Python.ArgumentError: Python argument types in
  virtual_functions_ext.call_f(Derived)
did not match C++ signature:
  call_f(Base {lvalue})

My configuration is
  Boost 1.31.0
  Python 2.3.3
  gcc 3.3.3

Andreas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: virtual_functions.cpp
Type: text/x-c++src
Size: 458 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040310/6de7b98f/attachment.cpp>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: virtual_functions.py
Type: application/x-python
Size: 145 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040310/6de7b98f/attachment.bin>
-------------- next part --------------
#!/bin/bash
g++-3.3 -shared -I$HOME/pool/include/python2.3 -I$HOME/work/boost $1.cpp -o $1_ext.so -lboost_python


From ostiguy at fnal.gov  Wed Mar 10 17:10:08 2004
From: ostiguy at fnal.gov (Francois Ostiguy)
Date: Wed, 10 Mar 2004 10:10:08 -0600 (CST)
Subject: [C++-sig] Re: iterators and abstact base class
In-Reply-To: <ullm9y7hu.fsf@boost-consulting.com>
References: <Pine.GSO.4.58.0403091503550.385@wally.fnal.gov>
	<ullm9y7hu.fsf@boost-consulting.com>
Message-ID: <Pine.GSO.4.58.0403100951420.385@wally.fnal.gov>

On Tue, 9 Mar 2004, David Abrahams wrote:

> > Hi -
> >
> > This has to be an FAQ. I searched the list archives and I was not able
> > locate a satisfactory solution to the following problem:
> >
> > Assume an abstract base class A and a derived class B. I have an iterator
> > I for a container C holding B-type objects.  The iterator returns a
> > pointer to the base class A.
> > At runtime, I.next()
>
> ?? This is a standard C++ iterator, or something else??
>
> > returns a type A which cannot be instantiated.
>
> Clearly it can't return A; it can return a reference to A, though.
>
> > I would like to iterate through my container and call a (virtual) function
> > on each element.  Can you point me to an idiom/technique that would
> > allow me to accomplish this ?
>
> Is this a Python/C++ question, a straight C++ question, or something
> else?  Too many details are missing.
>


Sorry for not being specific enough. This is really a boost.python
question. The iterator is a custom iterator (the code was written
in the early 1990s) which returns a pointer to the base class.
Of course, the object returned is always an instance of a derived class.
So if class C is my container class and next is the "iterator"

class A;              // abstract class
class B: public A {
...
};

C c;                 // container class, contains objects instances of
                     // classes derived from A
...

Iterator i(c);

A* a = i. next();     //  generic way to retrieve the next element in the
                          container
a.function();         //  calls a virtual function on the element


I want to achieve in python

b1 = B(..)    # appropriate contructor
b2 = B(..)    #
c = C(b1, b2) # constructs a C which contains b1 and b2
i = I(c)      # i is an iterator for c

a = i.next()  <<<<==== this is the problem. i.next() must return something
                       that can be held into python object a. In C++
                       i.next() returns a type A*.

a.function()

...

How do I define a python type that can be used to hold what i.next() returns ?

As always, your help is greatly appreciated.

-Francois

----------------------------------------------------------------------------
Dr. Jean-Francois OSTIGUY                              voice: (630) 840-2231
Beam Physics Dept MS220                                  FAX: (630) 840-6039
Fermi National Accelerator Laboratory                email: ostiguy at fnal.gov
Batavia IL 60510-0500                           WWW:www-ap.fnal.gov/~ostiguy




From dave at boost-consulting.com  Wed Mar 10 20:29:17 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Wed, 10 Mar 2004 14:29:17 -0500
Subject: [C++-sig] Re: iterators and abstact base class
References: <Pine.GSO.4.58.0403091503550.385@wally.fnal.gov>
	<ullm9y7hu.fsf@boost-consulting.com>
	<Pine.GSO.4.58.0403100951420.385@wally.fnal.gov>
Message-ID: <uptbkqz5e.fsf@boost-consulting.com>

Francois Ostiguy <ostiguy at fnal.gov> writes:

> Sorry for not being specific enough. This is really a boost.python
> question. The iterator is a custom iterator (the code was written
> in the early 1990s) which returns a pointer to the base class.
> Of course, the object returned is always an instance of a derived class.
> So if class C is my container class and next is the "iterator"
>
> class A;              // abstract class
> class B: public A {
> ...
> };
>
> C c;                 // container class, contains objects instances of
>                      // classes derived from A
> ...
>
> Iterator i(c);
>
> A* a = i. next();     //  generic way to retrieve the next element in the
>                           container
> a.function();         //  calls a virtual function on the element

Surely it's a->function();
??

> I want to achieve in python
>
> b1 = B(..)    # appropriate contructor
> b2 = B(..)    #
> c = C(b1, b2) # constructs a C which contains b1 and b2
> i = I(c)      # i is an iterator for c
>
> a = i.next()  <<<<==== this is the problem. i.next() must return something
>                        that can be held into python object a. In C++
>                        i.next() returns a type A*.
>
> a.function()
>
> ...
>
> How do I define a python type that can be used to hold what i.next() returns ?

I don't understand the problem.  Why not wrap "A" using class_; then
wrap next() with reference_existing_object or another appropriate call
policy?

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From mikeowen at llnl.gov  Thu Mar 11 02:47:13 2004
From: mikeowen at llnl.gov (J. Michael Owen)
Date: Wed, 10 Mar 2004 17:47:13 -0800
Subject: [C++-sig] Cross module inheritance failing with the Intel compiler
	on Linux
Message-ID: <auto-000035317057@popcorn.llnl.gov>

I'm trying to use the Intel C++ complier (7.0) on Linux with Boost.Python
(latest release, 1.31.0), but running into problems with cross module
inheritance.  At the end of this message I've appended a stripped down example
of the problem in 4 files: A.hh, B.hh, A.cc, and B.cc.  A.hh defines a base
class "Base", while B.hh defines a class "Derived" which inherits publicly from
"Base".  I compile these into two separate modules (A.cc -> A.so, B.cc -> B.so),
and then fire up python and execute the following:

>>> import A
>>> import B

If I build everything with g++ all is fine and I can use the "Derived" class in
python.  However, if I build with Intel's icc (both Boost.Python and this
example code of course) everything compiles, but I get the following run time
error:

>>> import A
>>> import B
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
RuntimeError: extension class wrapper for base class 4Base has not been created yet
>>>

Are other folks successfully using Intel icc with Boost.Python?  Does anyone
have any suggestions how I could correct this problem?

Thanks!

Mike.

************************************ A.hh **************************************
#include <iostream>
class Base {
public:
  Base() { std::cerr << "Base()" << std::endl; }
  virtual ~Base() { std::cerr << "~Base()" << std::endl;}
  virtual void printMessage() const { std::cerr << "Base instance @ " << this << std::endl; }
};

************************************ B.hh **************************************
#include <iostream>
#include "A.hh"
class Derived: public Base {
public:
  Derived(): Base() { std::cerr << "Derived()" << std::endl; }
  virtual ~Derived() { std::cerr << "~Derived()" << std::endl;}
  virtual void printMessage() const { std::cerr << "Derived instance @ " << this << std::endl; }
};

************************************ A.cc **************************************
#include "A.hh"
#include <boost/python.hpp>
using namespace boost::python;
BOOST_PYTHON_MODULE(A) {
  class_<Base>("Base", init<>())
    .def("printMessage", &Base::printMessage)
    ;
}

************************************ B.cc **************************************
#include "B.hh"
#include <boost/python.hpp>
using namespace boost::python;
BOOST_PYTHON_MODULE(B) {
  class_<Derived, bases<Base> >("Derived", init<>());
}

********************************************************************************
-- 
 "Hey...where are the sunflower seeds?" |       J. Michael Owen         
        o_o /                           |          
        (")                             |       Email:  mikeowen at llnl.gov
       \/'\/                            |
____(__(,_,)_______________________________________________________________



From rwgk at yahoo.com  Thu Mar 11 05:40:17 2004
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Wed, 10 Mar 2004 20:40:17 -0800 (PST)
Subject: [C++-sig] Cross module inheritance failing with the Intel
	compiler on Linux
In-Reply-To: <auto-000035317057@popcorn.llnl.gov>
Message-ID: <20040311044017.91244.qmail@web20203.mail.yahoo.com>

--- "J. Michael Owen" <mikeowen at llnl.gov> wrote:
> >>> import A
> >>> import B
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
> RuntimeError: extension class wrapper for base class 4Base has not been
> created yet
> >>>

I can reproduce the error with this compiler:

Intel(R) C++ Compiler for 32-bit applications, Version 7.1   Build 20031225Z

But it works with this compiler:

Intel(R) C++ Compiler for 32-bit applications, Version 8.0   Build 20040122Z
Package ID: l_cc_pc_8.0.058_pe059

>>> import A
>>> import B
>>> B.Derived()
Base()
Derived()
<B.Derived object at 0x4035b504>
>>> <ctrl-d>
~Derived()
~Base()

I know cross-module inheritance works with Intel 7.1 for non-virtual classes. I
don't have any idea what to do about the failure for virtual classes. Can you
switch to the 8.0 compiler?

Ralf


__________________________________
Do you Yahoo!?
Yahoo! Search - Find what you?re looking for faster
http://search.yahoo.com



From rwgk at yahoo.com  Thu Mar 11 09:03:51 2004
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Thu, 11 Mar 2004 00:03:51 -0800 (PST)
Subject: [C++-sig] Cross module inheritance failing with the Intel
	compiler on Linux
In-Reply-To: <20040311044017.91244.qmail@web20203.mail.yahoo.com>
Message-ID: <20040311080351.18704.qmail@web20203.mail.yahoo.com>

--- "Ralf W. Grosse-Kunstleve" <rwgk at yahoo.com> wrote:
> I know cross-module inheritance works with Intel 7.1 for non-virtual classes.

I guess it doesn't make a difference to the OP, but to set the record straight
I have to correct myself. All my non-virtual inheritance relationships don't
cross module boundaries. The original problem persists even for non-virtual
classes.

Anyway, I think I have the solution: compile everything with

-DBOOST_PYTHON_TYPE_ID_NAME

Alternatively, change

boost/boost/python/type_id.hpp

so the code starting at line 32 looks like this:

# if (defined(__GNUC__) && __GNUC__ >= 3) \
 || defined(_AIX) \
 || (   defined(__sgi) && defined(__host_mips)) \
 || (defined(linux) && defined(__INTEL_COMPILER) && defined(__ICC))
#  define BOOST_PYTHON_TYPE_ID_NAME
# endif

David, what do you think about the defines that I chose?

Ralf


__________________________________
Do you Yahoo!?
Yahoo! Search - Find what you?re looking for faster
http://search.yahoo.com



From Holger.Joukl at LBBW.de  Thu Mar 11 10:28:15 2004
From: Holger.Joukl at LBBW.de (Holger Joukl)
Date: Thu, 11 Mar 2004 10:28:15 +0100
Subject: [C++-sig] boost.build: How to set runtime library paths in a Jamfile
	(wrapping 3rd party libs)
Message-ID: <OF37C1883F.6B21D240-ONC1256E54.00330732-C1256E54.0033FA43@lbbw.sko.de>


Hi,
Jamfile seems to work like this:
Jamfile entry ---translates to---> build/link command line

<include>/include/path ---> -I/include/path
<library-path>/path/name---> -L/path/name
<find-library> mylib  ---> -lmylib

How can I set a runtime library dir? I did not find the information in the
docs, nor
did the boost *.jam files help me much.
Any hint on how to do that (or where to find the proper documentation)
is appreciated!
Thanks,
  Holger


Der Inhalt dieser E-Mail ist vertraulich. Falls Sie nicht der angegebene
Empf?nger sind oder falls diese E-Mail irrt?mlich an Sie adressiert wurde,
verst?ndigen Sie bitte den Absender sofort und l?schen Sie die E-Mail
sodann. Das unerlaubte Kopieren sowie die unbefugte ?bermittlung sind nicht
gestattet. Die Sicherheit von ?bermittlungen per E-Mail kann nicht
garantiert werden. Falls Sie eine Best?tigung w?nschen, fordern Sie bitte
den Inhalt der E-Mail als Hardcopy an.

The contents of this  e-mail are confidential. If you are not the named
addressee or if this transmission has been addressed to you in error,
please notify the sender immediately and then delete this e-mail.  Any
unauthorized copying and transmission is forbidden. E-Mail transmission
cannot be guaranteed to be secure. If verification is required, please
request a hard copy version.






From dave at boost-consulting.com  Thu Mar 11 12:45:37 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 11 Mar 2004 06:45:37 -0500
Subject: [C++-sig] Re: boost.build: How to set runtime library paths in a
 Jamfile (wrapping 3rd party libs)
References: <OF37C1883F.6B21D240-ONC1256E54.00330732-C1256E54.0033FA43@lbbw.sko.de>
Message-ID: <u7jxrppy6.fsf@boost-consulting.com>

"Holger Joukl" <Holger.Joukl at LBBW.de> writes:

> Hi,
> Jamfile seems to work like this:
> Jamfile entry ---translates to---> build/link command line
>
> <include>/include/path ---> -I/include/path
> <library-path>/path/name---> -L/path/name
> <find-library> mylib  ---> -lmylib
>
> How can I set a runtime library dir? I did not find the information
> in the docs, nor did the boost *.jam files help me much.  Any hint
> on how to do that (or where to find the proper documentation) is
> appreciated!  Thanks, Holger


What do you mean by "a runtime library dir"?

Please illustrate with specific command-line elements.

Thanks,
-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From Holger.Joukl at LBBW.de  Thu Mar 11 13:36:34 2004
From: Holger.Joukl at LBBW.de (Holger Joukl)
Date: Thu, 11 Mar 2004 13:36:34 +0100
Subject: [C++-sig] Re: Re: boost.build: How to set runtime library paths in a
	Jamfile (wrapping 3rd party libs)
Message-ID: <OF04DCDABF.80E45B88-ONC1256E54.00448776-C1256E54.004537CE@lbbw.sko.de>


Hi,
sorry if the terminology is wrong/unclear.
What I meant is the library search path for the runtime linker,
given to gcc with the -R flag:
E.g.
  g++ -o myProg -L /my/strange/dir -R /my/strange/dir -lmyStrangeLib main.o
funcs.o

Just to get rid of LD_LIBRARY_PATH, I am looking for the Jamfile equivalent
of
"-R /my/strange/dir" on the commandline.
I guess it is possible to make such a search path known in a Jamfile, but I
didn?t see
the solution.

Thanks,
  Holger
_________________________________

Holger Joukl
Landesbank Baden-W?rttemberg
1651 Financial Markets Technologies
fon +49 (711) 124 - 7078
fax +49 (711) 124 - 3759
_________________________________

Der Inhalt dieser E-Mail ist vertraulich. Falls Sie nicht der angegebene
Empf?nger sind oder falls diese E-Mail irrt?mlich an Sie adressiert wurde,
verst?ndigen Sie bitte den Absender sofort und l?schen Sie die E-Mail
sodann. Das unerlaubte Kopieren sowie die unbefugte ?bermittlung sind nicht
gestattet. Die Sicherheit von ?bermittlungen per E-Mail kann nicht
garantiert werden. Falls Sie eine Best?tigung w?nschen, fordern Sie bitte
den Inhalt der E-Mail als Hardcopy an.

The contents of this  e-mail are confidential. If you are not the named
addressee or if this transmission has been addressed to you in error,
please notify the sender immediately and then delete this e-mail.  Any
unauthorized copying and transmission is forbidden. E-Mail transmission
cannot be guaranteed to be secure. If verification is required, please
request a hard copy version.






From dave at boost-consulting.com  Thu Mar 11 15:09:50 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 11 Mar 2004 09:09:50 -0500
Subject: [C++-sig] Re: Cross module inheritance failing with the Intel
 compiler on Linux
References: <auto-000035317057@popcorn.llnl.gov>
Message-ID: <uekrzo4pd.fsf@boost-consulting.com>

"J. Michael Owen" <mikeowen at llnl.gov> writes:

> I'm trying to use the Intel C++ complier (7.0) on Linux with Boost.Python
> (latest release, 1.31.0), but running into problems with cross module
> inheritance.  At the end of this message I've appended a stripped down example
> of the problem in 4 files: A.hh, B.hh, A.cc, and B.cc.  A.hh defines a base
> class "Base", while B.hh defines a class "Derived" which inherits publicly from
> "Base".  I compile these into two separate modules (A.cc -> A.so, B.cc -> B.so),
> and then fire up python and execute the following:
>
>>>> import A
>>>> import B
>
> If I build everything with g++ all is fine and I can use the "Derived" class in
> python.  However, if I build with Intel's icc (both Boost.Python and this
> example code of course) everything compiles, but I get the following run time
> error:
>
>>>> import A
>>>> import B
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
> RuntimeError: extension class wrapper for base class 4Base has not been created yet
>>>>
>
> Are other folks successfully using Intel icc with Boost.Python?  Does anyone
> have any suggestions how I could correct this problem?


Here's the story:

I don't know about Intel 7.0.

In the latest release, we added the <stdlib> feature to Boost.Build in
order to allow us to use Intel 8.x with the GCC standard library
(-cxxlib-gcc ICC command-line option).  This was unfortunately needed
in order to get extensions built with Intel 8.x to work with a stock
(GCC-built) Python on Linux.  Unfortunately, the -cxxlib-gcc feature
is broken in Intel 7.1, so the strategem fails.

If you want to use a Python built with the Intel compiler, you should
use a Boost.Python library built with <stdlib>default instead of
<stdlib>gcc in your bjam BUILD variable.

I *think* that'll work with Intel 7.x, but I can't promise.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Thu Mar 11 15:11:26 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 11 Mar 2004 09:11:26 -0500
Subject: [C++-sig] Re: Cross module inheritance failing with the Intel
 compiler on Linux
References: <20040311044017.91244.qmail@web20203.mail.yahoo.com>
	<20040311080351.18704.qmail@web20203.mail.yahoo.com>
Message-ID: <u65dbo4mp.fsf@boost-consulting.com>

"Ralf W. Grosse-Kunstleve" <rwgk at yahoo.com> writes:

> Anyway, I think I have the solution: compile everything with
>
> -DBOOST_PYTHON_TYPE_ID_NAME
>
> Alternatively, change
>
> boost/boost/python/type_id.hpp
>
> so the code starting at line 32 looks like this:
>
> # if (defined(__GNUC__) && __GNUC__ >= 3) \
>  || defined(_AIX) \
>  || (   defined(__sgi) && defined(__host_mips)) \
>  || (defined(linux) && defined(__INTEL_COMPILER) && defined(__ICC))
> #  define BOOST_PYTHON_TYPE_ID_NAME
> # endif
>
> David, what do you think about the defines that I chose?

Did you test this with Intel 7.1 by running the Boost.Python
regressions?  I'd be surprised if it worked; I could've sworn it was
the first thing I checked when trying to solve this problem.  IIRC we
still have problems with exception handling.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Thu Mar 11 15:23:05 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 11 Mar 2004 09:23:05 -0500
Subject: [C++-sig] Re: boost.build: How to set runtime library paths in a
 Jamfile (wrapping 3rd party libs)
References: <OF04DCDABF.80E45B88-ONC1256E54.00448776-C1256E54.004537CE@lbbw.sko.de>
Message-ID: <uu10vmpiu.fsf@boost-consulting.com>

"Holger Joukl" <Holger.Joukl at LBBW.de> writes:

> Hi,
> sorry if the terminology is wrong/unclear.
> What I meant is the library search path for the runtime linker,
> given to gcc with the -R flag:
> E.g.
>   g++ -o myProg -L /my/strange/dir -R /my/strange/dir -lmyStrangeLib main.o
> funcs.o
>
> Just to get rid of LD_LIBRARY_PATH, I am looking for the Jamfile equivalent
> of
> "-R /my/strange/dir" on the commandline.
> I guess it is possible to make such a search path known in a Jamfile, but I
> didn?t see
> the solution.

I'm afraid I don't know.  I've never heard of the "-R" option before;
Boost.Build uses LD_LIBRARY_PATH to prepare the search path for the
runtime linker.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From halbert at bbn.com  Thu Mar 11 15:58:29 2004
From: halbert at bbn.com (Dan Halbert)
Date: Thu, 11 Mar 2004 09:58:29 -0500
Subject: [C++-sig] Re: passing buffer of bytes from Python to C++
In-Reply-To: <u65dcx441.fsf@boost-consulting.com> (message from David Abrahams
	on Wed, 10 Mar 2004 07:45:34 -0500)
References: <200403092343.i29NhcE30830@rolex.bbn.com>
	<u65dcx441.fsf@boost-consulting.com>
Message-ID: <200403111458.i2BEwTO02178@rolex.bbn.com>

  >From: David Abrahams <dave at boost-consulting.com>
  >Dan Halbert <halbert at bbn.com> writes:
  >
  >> One of the C++ methods I've wrapped with BPL has a signature something like
  >>
  >>     process_data(unsigned char const* buffer, unsigned long buffer_length)
  >
  >Python strings will be automatically converted to char const*

Thanks, David. Just to provide closure here, the problem I had was the
"unsigned char", which is not converted automatically, though "char" is.
I have wrapped the method to avoid this.

Dan



From Holger.Joukl at LBBW.de  Thu Mar 11 16:09:05 2004
From: Holger.Joukl at LBBW.de (Holger Joukl)
Date: Thu, 11 Mar 2004 16:09:05 +0100
Subject: [C++-sig] Re: Re: boost.build: How to set runtime library paths in a
	Jamfile (wrapping 3rd party libs)
Message-ID: <OFF35ADA7F.46080C55-ONC1256E54.00511C75-C1256E54.00532EB6@lbbw.sko.de>


>Dave wrote:
>I'm afraid I don't know.  I've never heard of the "-R" option before;
>Boost.Build uses LD_LIBRARY_PATH to prepare the search path for the
>runtime linker.

Wouldn?t that mean you always have to set the LD_LIBRARY_PATH when
importing
an extension module that e.g. wraps a 3rd party library that is not
installed
in /usr/lib?
Same thing for libboost_python.so, must be in a standard place or it?s
LD_LIBRARY_PATH again.

I created such an extension module with bjam:
When I elfdump or ldd the resulting shared object, there is no information
about
where to find the libraries it depends on (including libboost_python.so).

Can I maybe pass the "-R ..." information as <cxxflags>?

Thanks,
  Holger

Der Inhalt dieser E-Mail ist vertraulich. Falls Sie nicht der angegebene
Empf?nger sind oder falls diese E-Mail irrt?mlich an Sie adressiert wurde,
verst?ndigen Sie bitte den Absender sofort und l?schen Sie die E-Mail
sodann. Das unerlaubte Kopieren sowie die unbefugte ?bermittlung sind nicht
gestattet. Die Sicherheit von ?bermittlungen per E-Mail kann nicht
garantiert werden. Falls Sie eine Best?tigung w?nschen, fordern Sie bitte
den Inhalt der E-Mail als Hardcopy an.

The contents of this  e-mail are confidential. If you are not the named
addressee or if this transmission has been addressed to you in error,
please notify the sender immediately and then delete this e-mail.  Any
unauthorized copying and transmission is forbidden. E-Mail transmission
cannot be guaranteed to be secure. If verification is required, please
request a hard copy version.






From grafik.list at redshift-software.com  Thu Mar 11 16:32:23 2004
From: grafik.list at redshift-software.com (Rene Rivera)
Date: Thu, 11 Mar 2004 09:32:23 -0600
Subject: [C++-sig] Re: Re: boost.build: How to set runtime library paths
	in a	Jamfile (wrapping 3rd party libs)
In-Reply-To: <OFF35ADA7F.46080C55-ONC1256E54.00511C75-C1256E54.00532EB6@lbbw.sko.de>
References: <OFF35ADA7F.46080C55-ONC1256E54.00511C75-C1256E54.00532EB6@lbbw.sko.de>
Message-ID: <40508687.8040401@redshift-software.com>

Holger Joukl wrote:
>>Dave wrote:
>>I'm afraid I don't know.  I've never heard of the "-R" option before;
>>Boost.Build uses LD_LIBRARY_PATH to prepare the search path for the
>>runtime linker.
> 
> 
> Wouldn?t that mean you always have to set the LD_LIBRARY_PATH when
> importing
> an extension module that e.g. wraps a 3rd party library that is not
> installed
> in /usr/lib?

Don't think so, as Python uses different search path for dynamically loading 
modules.

> Same thing for libboost_python.so, must be in a standard place or it?s
> LD_LIBRARY_PATH again.

Yes to that. But this is something that installers (like RPM) do for you by 
running ldconfig, or equivalent.

> I created such an extension module with bjam:
> When I elfdump or ldd the resulting shared object, there is no information
> about
> where to find the libraries it depends on (including libboost_python.so).

Correct, and that's likely the optimal design as it makes the modules 
transportable. Which means you can install them to your preferred location 
instead of whatever the location you decided upon when they got built.

> Can I maybe pass the "-R ..." information as <cxxflags>?

You can use "<linkflags>", but be warned this is platform and linker specific. 
For example you call the -R flag but for the GNU linker it really should be 
the -rpath flag, and you must pass it to the linker with "-Wl,-rpath,<path>".

-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org - grafik/redshift-software.com - 102708583/icq



From dave at boost-consulting.com  Thu Mar 11 16:41:09 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 11 Mar 2004 10:41:09 -0500
Subject: [C++-sig] Re: boost.build: How to set runtime library paths in
 a	Jamfile (wrapping 3rd party libs)
References: <OFF35ADA7F.46080C55-ONC1256E54.00511C75-C1256E54.00532EB6@lbbw.sko.de>
	<40508687.8040401@redshift-software.com>
Message-ID: <uishbmlwq.fsf@boost-consulting.com>

Rene Rivera <grafik.list at redshift-software.com> writes:

> Holger Joukl wrote:
>>>Dave wrote:
>>>I'm afraid I don't know.  I've never heard of the "-R" option before;
>>>Boost.Build uses LD_LIBRARY_PATH to prepare the search path for the
>>>runtime linker.
>> Wouldn?t that mean you always have to set the LD_LIBRARY_PATH when
>> importing
>> an extension module that e.g. wraps a 3rd party library that is not
>> installed
>> in /usr/lib?
>
> Don't think so, as Python uses different search path for dynamically
> loading modules.

I don't think you understood.  If the module wraps a 3rd party .so,
that must be in LD_LIBRARY_PATH.

>> Same thing for libboost_python.so, must be in a standard place or it?s
>> LD_LIBRARY_PATH again.
>
> Yes to that. But this is something that installers (like RPM) do for
> you by running ldconfig, or equivalent.
>
>> I created such an extension module with bjam: When I elfdump or ldd
>> the resulting shared object, there is no information about where to
>> find the libraries it depends on (including libboost_python.so).
>
> Correct, and that's likely the optimal design as it makes the modules
> transportable. Which means you can install them to your preferred
> location instead of whatever the location you decided upon when they
> got built.

Right.  Most people don't want absolute paths to libraries encoded in
their objects.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From ingo at fargonauten.de  Thu Mar 11 16:55:08 2004
From: ingo at fargonauten.de (Ingo Luetkebohle)
Date: Thu, 11 Mar 2004 16:55:08 +0100
Subject: [C++-sig] Re: boost.build: How to set runtime library paths in	a
	Jamfile (wrapping 3rd party libs)
In-Reply-To: <uishbmlwq.fsf@boost-consulting.com>
References: <OFF35ADA7F.46080C55-ONC1256E54.00511C75-C1256E54.00532EB6@lbbw.sko.de>
	<40508687.8040401@redshift-software.com>
	<uishbmlwq.fsf@boost-consulting.com>
Message-ID: <1079020508.9245.11.camel@localhost>

Am Do, den 11.03.2004 schrieb David Abrahams um 16:41:
> Right.  Most people don't want absolute paths to libraries encoded in
> their objects.

While that might be so, it says nothing about ld's '-R' (or '-rpath' on
Linux) option.  That option prepends a directory to the run-time linker
search path.  There is no absolute path to a library, the library is
searched as usual, but in additional locations.

Most people I know prefer that to LD_LIBRARY_PATH very much ;-)

One reason is that it can be applied to a single shared object or
executable, instead of applying unconditionally to everything.  Like,
"for *this* library, look in '/usr/local/boost/lib' for additional
dependencies".  

Another pro is that it works with setuid binaries while LD_LIBRARY_PATH
usually does not, for obvious reasons.

While moving libraries is an issue, one can always re-link with a
different RPATH.  libtool does that, for example.

regards

-- 
Ingo

Soll doch jeder bleiben, wie er gerne w?re.





From grafik.list at redshift-software.com  Thu Mar 11 17:25:46 2004
From: grafik.list at redshift-software.com (Rene Rivera)
Date: Thu, 11 Mar 2004 10:25:46 -0600
Subject: [C++-sig] Re: boost.build: How to set runtime library paths in
	a	Jamfile (wrapping 3rd party libs)
In-Reply-To: <1079020508.9245.11.camel@localhost>
References: <OFF35ADA7F.46080C55-ONC1256E54.00511C75-C1256E54.00532EB6@lbbw.sko.de>	<40508687.8040401@redshift-software.com>	<uishbmlwq.fsf@boost-consulting.com>
	<1079020508.9245.11.camel@localhost>
Message-ID: <4050930A.3010903@redshift-software.com>

Ingo Luetkebohle wrote:

> Am Do, den 11.03.2004 schrieb David Abrahams um 16:41:
> 
>>Right.  Most people don't want absolute paths to libraries encoded in
>>their objects.
> 
> 
> While that might be so, it says nothing about ld's '-R' (or '-rpath' on
> Linux) option.  That option prepends a directory to the run-time linker
> search path.  There is no absolute path to a library, the library is
> searched as usual, but in additional locations.

I'm sorry but that makes no sense to me :-( Yes you can add search locations, 
but you can't predict all the search locations that a user might want to move 
a library to. For example I prefer to install all libraries of a program with 
the program, and there is no way to predict where that is.

> Most people I know prefer that to LD_LIBRARY_PATH very much ;-)

Most people don't know about the alternatives available with ELF 
linker-loaders ;-)

> One reason is that it can be applied to a single shared object or
> executable, instead of applying unconditionally to everything.  Like,
> "for *this* library, look in '/usr/local/boost/lib' for additional
> dependencies".  

But it would not know to search.. "/home/user/boost_1_31_0/lib"

> Another pro is that it works with setuid binaries while LD_LIBRARY_PATH
> usually does not, for obvious reasons.

And for good reason... ldconfig, and other system level configuration of the 
load search path and library cache is the preferred way to manage that.

> While moving libraries is an issue, one can always re-link with a
> different RPATH.  libtool does that, for example.

You can only relink if you have a linker. So if you only distribute binaries 
you are out of luck.

I went over all these issues and more a long time ago in trying to figure out 
how to manage distribution for my current project and there are many things I 
can say about the Unix/Linux link loading model, but the nicest I can say is 
that it's broken. In the end I resorted to using the highly under documented 
feature of ELF for adding searchs paths relative to the loading module. So now 
I can get the effect I have on Windows of adding the location of my executable 
(or any other relative subdir) to the library search path...

	<gcc><*><linkflags>-Wl,-rpath,'${ORIGIN}'


-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org - grafik/redshift-software.com - 102708583/icq



From rwgk at yahoo.com  Thu Mar 11 18:19:10 2004
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Thu, 11 Mar 2004 09:19:10 -0800 (PST)
Subject: [C++-sig] Re: Cross module inheritance failing with the Intel
	compiler on Linux
In-Reply-To: <u65dbo4mp.fsf@boost-consulting.com>
Message-ID: <20040311171910.46706.qmail@web20207.mail.yahoo.com>

--- David Abrahams <dave at boost-consulting.com> wrote:
> Did you test this with Intel 7.1 by running the Boost.Python
> regressions?  I'd be surprised if it worked; I could've sworn it was
> the first thing I checked when trying to solve this problem.  IIRC we
> still have problems with exception handling.

I just re-did the tests with Intel 7.1 (using my type_id.hpp patch) and they
all work except for crossmod_exception.py, as you expect.
At the moment I am thinking:

  - Mike's example tells us that we should be using BOOST_PYTHON_TYPE_ID_NAME

  - BOOST_PYTHON_TYPE_ID_NAME does not affect EH

Correct?

As an aside, I am also thinking that Intel 8.0 works for Mike's example because
it defines __GNUC__ which leads to BOOST_PYTHON_TYPE_ID_NAME being defined.
Does this make sense?

Ralf


__________________________________
Do you Yahoo!?
Yahoo! Search - Find what you?re looking for faster
http://search.yahoo.com



From ostiguy at fnal.gov  Thu Mar 11 18:43:16 2004
From: ostiguy at fnal.gov (Francois Ostiguy)
Date: Thu, 11 Mar 2004 11:43:16 -0600 (CST)
Subject: [C++-sig] Re: iterators and abstact base class
In-Reply-To: <uptbkqz5e.fsf@boost-consulting.com>
References: <Pine.GSO.4.58.0403091503550.385@wally.fnal.gov>
	<ullm9y7hu.fsf@boost-consulting.com>
	<Pine.GSO.4.58.0403100951420.385@wally.fnal.gov>
	<uptbkqz5e.fsf@boost-consulting.com>
Message-ID: <Pine.GSO.4.58.0403111133110.12402@wally.fnal.gov>

On Wed, 10 Mar 2004, David Abrahams wrote:

> > I want to achieve in python
> >
> > b1 = B(..)    # appropriate contructor
> > b2 = B(..)    #
> > c = C(b1, b2) # constructs a C which contains b1 and b2
> > i = I(c)      # i is an iterator for c
> >
> > a = i.next()  <<<<==== this is the problem. i.next() must return something
> >                        that can be held into python object a. In C++
> >                        i.next() returns a type A*.
> >
> > a.function()
> >
> > ...
> >
> > How do I define a python type that can be used to hold what i.next() returns ?
>
> I don't understand the problem.  Why not wrap "A" using class_; then
> wrap next() with reference_existing_object or another appropriate call
> policy?
>

I am affraid the problem is sitting in front of my keyboard.
I guess I did not fully appreciate the fact that introspection allows
one to determine the dynamic type of an object and to return a reference
to the full type intead of a reference to a base class object.
Thanks again!



-Francois

----------------------------------------------------------------------------
Dr. Jean-Francois OSTIGUY                              voice: (630) 840-2231
Beam Physics Dept MS220                                  FAX: (630) 840-6039
Fermi National Accelerator Laboratory                email: ostiguy at fnal.gov
Batavia IL 60510-0500                           WWW:www-ap.fnal.gov/~ostiguy




From dholth at fastmail.fm  Thu Mar 11 19:40:24 2004
From: dholth at fastmail.fm (Daniel Holth)
Date: Thu, 11 Mar 2004 13:40:24 -0500
Subject: [C++-sig] Pyste add_method()
Message-ID: <1079030423.10839.10.camel@bluefish>

I think I've found the solution to my old problem:

sync = Class("ogg::sync", "oggcc.h") # part 1

syncwrite = Function("ogg::sync_write", "oggpy_wrappers.h") # part 2
add_method(sync, "ogg::sync_write")
rename(sync.sync_write, "write")
exclude(syncwrite)

Doesn't work for me: it will say something like:
RuntimeError: no ogg::sync_write declaration found!

Instead, if I say the class is in the file that declares the wrapper,
(oggpy_wrappers.h also includes oggcc.h):

sync = Class("ogg::sync", "oggpy_wrappers.h")

(insert part 2 here)

it works nicely.

- Daniel Holth




From dave at boost-consulting.com  Thu Mar 11 21:09:26 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 11 Mar 2004 15:09:26 -0500
Subject: [C++-sig] Re: Cross module inheritance failing with the Intel
 compiler on Linux
References: <u65dbo4mp.fsf@boost-consulting.com>
	<20040311171910.46706.qmail@web20207.mail.yahoo.com>
Message-ID: <uoer3kux5.fsf@boost-consulting.com>

"Ralf W. Grosse-Kunstleve" <rwgk at yahoo.com> writes:

> --- David Abrahams <dave at boost-consulting.com> wrote:
>> Did you test this with Intel 7.1 by running the Boost.Python
>> regressions?  I'd be surprised if it worked; I could've sworn it was
>> the first thing I checked when trying to solve this problem.  IIRC we
>> still have problems with exception handling.
>
> I just re-did the tests with Intel 7.1 (using my type_id.hpp patch) and they
> all work except for crossmod_exception.py, as you expect.
> At the moment I am thinking:
>
>   - Mike's example tells us that we should be using BOOST_PYTHON_TYPE_ID_NAME
>
>   - BOOST_PYTHON_TYPE_ID_NAME does not affect EH
>
> Correct?
>
> As an aside, I am also thinking that Intel 8.0 works for Mike's example because
> it defines __GNUC__ which leads to BOOST_PYTHON_TYPE_ID_NAME being defined.
> Does this make sense?

Yep, I buy your theories.  Please feel free to make patches at will.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From nicodemus at esss.com.br  Fri Mar 12 01:53:37 2004
From: nicodemus at esss.com.br (Nicodemus)
Date: Thu, 11 Mar 2004 21:53:37 -0300
Subject: [C++-sig] Pyste add_method()
In-Reply-To: <1079030423.10839.10.camel@bluefish>
References: <1079030423.10839.10.camel@bluefish>
Message-ID: <40510A11.2060803@esss.com.br>

Daniel Holth wrote:

>I think I've found the solution to my old problem:
>
>sync = Class("ogg::sync", "oggcc.h") # part 1
>
>syncwrite = Function("ogg::sync_write", "oggpy_wrappers.h") # part 2
>add_method(sync, "ogg::sync_write")
>rename(sync.sync_write, "write")
>exclude(syncwrite)
>
>Doesn't work for me: it will say something like:
>RuntimeError: no ogg::sync_write declaration found!
>
>Instead, if I say the class is in the file that declares the wrapper,
>(oggpy_wrappers.h also includes oggcc.h):
>
>sync = Class("ogg::sync", "oggpy_wrappers.h")
>
>(insert part 2 here)
>
>it works nicely.
>  
>

That's correct. I posted the same solution in response to a post from 
Hanz Meizer a few weeks ago. 8)

Regards,
Nicodemus.




From colin_irwin at hotmail.com  Fri Mar 12 05:44:22 2004
From: colin_irwin at hotmail.com (Colin Irwin)
Date: Fri, 12 Mar 2004 15:44:22 +1100
Subject: [C++-sig] boost.python problem: 'virtual functions and default
	implementations' tutorial doesn't work
Message-ID: <c2rf70$p0s$1@sea.gmane.org>

I'm having a problem following the tutorial that comes with the
Boost::Python library, in particular defining virtual functions with default
implementations.  I've been using this as a test, as I have a set of C++
classes that I want to be available in Python.  These classes use a
top-level interface that has callback functions in the style of inheritable
functions.  I would like to be able to define a Python class, deriving this
off the C++ top-level interface class and in particular inheriting these
callback functions to deal with certain events.  I have a test setup as
follows (which pretty much follows the tutorial example) and I've tried to
compile it using MSVC .NET 2003 (v7.1) with Boost 1.31.0:

class Base
{
    public:
        virtual void function();
};

class BaseWrapper : public Base
{
    public:
        BaseWrapper(PyObject* self_) :
            self(self_) { }
        virtual void function() { call_method<void>(this->self,
"function"); };
        void default_function() { Base::function(); };

    private:
        PyObject* self;
};

BOOST_PYTHON_MODULE(test_module)
{

    class_<Base, BaseWrapper>("Base")
        .def("function", &Base::function, &BaseWrapper::default_function)
    ;
}

When I compile it, I get the following error:
..\..\42_External_Units\421_Boost\boost\python\object\value_holder.hpp(137)
: error C2661: 'BaseWrapper::BaseWrapper' : no overloaded function takes 2
arguments

... (a whole set of 'while compiling...') ...

test.cpp(35) : see reference to class template instantiation
'boost::python::class_<T,X1>' being compiled

with

[

T=Base,

X1=BaseWrapper

]



I've been able to resolve the error by using the following python module
definition instead:

BOOST_PYTHON_MODULE(test_module)
{

    class_<Base, BaseWrapper, boost::noncopyable>("Base")
        .def("function", &Base::function, &BaseWrapper::default_function)
    ;
}

Now this noncopyable is not mentioned in the tutorial file and indeed, I
don't think it does exactly what I want in this case.  By this I mean it
prevents copying any of the derived objects in Python doesn't it?
Where have I gone wrong?  The tutorial file doesn't seem to work as
expected.  My fix doesn't do what I want it do (I think this is right, if
I've inadvertently prevented copying).  How should I structure my module
definition/code?

TIA

Colin








From colin_irwin at hotmail.com  Tue Mar  9 06:58:38 2004
From: colin_irwin at hotmail.com (Colin Irwin)
Date: Tue, 9 Mar 2004 16:58:38 +1100
Subject: [C++-sig] Virtual functions and default implementations
	(Boost::Python tutorial example doesn't work)
Message-ID: <c2jme9$klk$1@sea.gmane.org>

I'm having a problem following the tutorial that comes with the
Boost::Python library, in particular defining virtual functions with default
implementations.  I've been using this as a test, as I have a set of C++
classes that I want to be available in Python.  These classes use a
top-level interface that has callback functions in the style of inheritable
functions.  I would like to be able to define a Python class, deriving this
off the C++ top-level interface class and in particular inheriting these
callback functions to deal with certain events.  I have a test setup as
follows (which pretty much follows the tutorial example) and I've tried to
compile it using MSVC .NET 2003 (v7.1) with Boost 1.31.0:

class Base
{
    public:
        virtual void function();
};

class BaseWrapper : public Base
{
    public:
        BaseWrapper(PyObject* self_) :
            self(self_) { }
        virtual void function() { call_method<void>(this->self,
"function"); };
        void default_function() { Base::function(); };

    private:
        PyObject* self;
};

BOOST_PYTHON_MODULE(test_module)
{

    class_<Base, BaseWrapper>("Base")
        .def("function", &Base::function, &BaseWrapper::default_function)
    ;
}

When I compile it, I get the following error:
..\..\42_External_Units\421_Boost\boost\python\object\value_holder.hpp(137)
: error C2661: 'BaseWrapper::BaseWrapper' : no overloaded function takes 2
arguments

... (a whole set of 'while compiling...') ...

test.cpp(35) : see reference to class template instantiation
'boost::python::class_<T,X1>' being compiled

with

[

T=Base,

X1=BaseWrapper

]



I've been able to resolve the error by using the following python module
definition instead:

BOOST_PYTHON_MODULE(test_module)
{

    class_<Base, BaseWrapper, boost::noncopyable>("Base")
        .def("function", &Base::function, &BaseWrapper::default_function)
    ;
}

Now this noncopyable is not mentioned in the tutorial file and indeed, I
don't think it does exactly what I want in this case.  By this I mean it
prevents copying any of the derived objects in Python doesn't it?
Where have I gone wrong?  The tutorial file doesn't seem to work as
expected.  My fix doesn't do what I want it do (I think this is right, if
I've inadvertently prevented copying).  How should I structure my module
definition/code?

TIA

Colin






From Holger.Joukl at LBBW.de  Fri Mar 12 09:18:27 2004
From: Holger.Joukl at LBBW.de (Holger Joukl)
Date: Fri, 12 Mar 2004 09:18:27 +0100
Subject: [C++-sig] Re: boost.build: How to set runtime library paths
	in	a Jamfile (wrapping 3rd party libs)
Message-ID: <OF7432B9B3.EA2AC13C-ONC1256E55.002A761D-C1256E55.002D9665@lbbw.sko.de>


Rene wrote:
>feature of ELF for adding searchs paths relative to the loading module.. So
now
>I can get the effect I have on Windows of adding the location of my
executable
>(or any other relative subdir) to the library search path...

> <gcc><*><linkflags>-Wl,-rpath,'${ORIGIN}'

If I understand correctly, this translate to an equivalent to "gcc
-R<$ORIGIN>" with $ORIGIN being a
relative path -- that is, simply using -R option with relative paths.
While being fine for executables, I suspect that this is not really
appropriate for boost.python wrapped
libraries, where
- libboost_python.so would probably reside in
$PYTHON_ROOT/lib/python2.x/site-packages
- the wrapper library shared object would probably reside in
$PYTHON_ROOT/lib/python2.x/site-packages
- the wrapped library resides anywhere on your system

Running python and using this extension is now dependant from where you
invoke the interpreter;
it will only work if the wrapped lib can be found in the relative path
$ORIGIN.
So it is either LD_LIBRARY_PATH again (not good) or adding some search
paths with -R in
the link step, the search paths being s.th. like site-dependant standard
library installation directories.
E.g. a shared filesystem named /apps/prod where I would install central
components (like python :-)

Anyway, thanks a lot for all the hints!

Der Inhalt dieser E-Mail ist vertraulich. Falls Sie nicht der angegebene
Empf?nger sind oder falls diese E-Mail irrt?mlich an Sie adressiert wurde,
verst?ndigen Sie bitte den Absender sofort und l?schen Sie die E-Mail
sodann. Das unerlaubte Kopieren sowie die unbefugte ?bermittlung sind nicht
gestattet. Die Sicherheit von ?bermittlungen per E-Mail kann nicht
garantiert werden. Falls Sie eine Best?tigung w?nschen, fordern Sie bitte
den Inhalt der E-Mail als Hardcopy an.

The contents of this  e-mail are confidential. If you are not the named
addressee or if this transmission has been addressed to you in error,
please notify the sender immediately and then delete this e-mail.  Any
unauthorized copying and transmission is forbidden. E-Mail transmission
cannot be guaranteed to be secure. If verification is required, please
request a hard copy version.






From pierre.barbier at cirad.fr  Fri Mar 12 09:37:18 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Fri, 12 Mar 2004 09:37:18 +0100
Subject: [C++-sig] boost.python problem: 'virtual functions and default
	implementations' tutorial doesn't work
In-Reply-To: <c2rf70$p0s$1@sea.gmane.org>
References: <c2rf70$p0s$1@sea.gmane.org>
Message-ID: <1079080638.1024.1.camel@pbarbier>

The problem comes from a change in the handling of the copy constructor
in Boost.Python. Now, if you want to have a copy contructor for your
class, you have to write a contructor in your wrapper as :

BaseWrapper( PyObject* self_, const Base& copy);

In fact, for every constructor you have, you need to add a first
argument being the PyObject* of the current object.

Pierre

Le ven 12/03/2004 ? 05:44, Colin Irwin a ?crit :
> I'm having a problem following the tutorial that comes with the
> Boost::Python library, in particular defining virtual functions with default
> implementations.  I've been using this as a test, as I have a set of C++
> classes that I want to be available in Python.  These classes use a
> top-level interface that has callback functions in the style of inheritable
> functions.  I would like to be able to define a Python class, deriving this
> off the C++ top-level interface class and in particular inheriting these
> callback functions to deal with certain events.  I have a test setup as
> follows (which pretty much follows the tutorial example) and I've tried to
> compile it using MSVC .NET 2003 (v7.1) with Boost 1.31.0:
> 
> class Base
> {
>     public:
>         virtual void function();
> };
> 
> class BaseWrapper : public Base
> {
>     public:
>         BaseWrapper(PyObject* self_) :
>             self(self_) { }
>         virtual void function() { call_method<void>(this->self,
> "function"); };
>         void default_function() { Base::function(); };
> 
>     private:
>         PyObject* self;
> };
> 
> BOOST_PYTHON_MODULE(test_module)
> {
> 
>     class_<Base, BaseWrapper>("Base")
>         .def("function", &Base::function, &BaseWrapper::default_function)
>     ;
> }
> 
> When I compile it, I get the following error:
> ..\..\42_External_Units\421_Boost\boost\python\object\value_holder.hpp(137)
> : error C2661: 'BaseWrapper::BaseWrapper' : no overloaded function takes 2
> arguments
> 
> ... (a whole set of 'while compiling...') ...
> 
> test.cpp(35) : see reference to class template instantiation
> 'boost::python::class_<T,X1>' being compiled
> 
> with
> 
> [
> 
> T=Base,
> 
> X1=BaseWrapper
> 
> ]
> 
> 
> 
> I've been able to resolve the error by using the following python module
> definition instead:
> 
> BOOST_PYTHON_MODULE(test_module)
> {
> 
>     class_<Base, BaseWrapper, boost::noncopyable>("Base")
>         .def("function", &Base::function, &BaseWrapper::default_function)
>     ;
> }
> 
> Now this noncopyable is not mentioned in the tutorial file and indeed, I
> don't think it does exactly what I want in this case.  By this I mean it
> prevents copying any of the derived objects in Python doesn't it?
> Where have I gone wrong?  The tutorial file doesn't seem to work as
> expected.  My fix doesn't do what I want it do (I think this is right, if
> I've inadvertently prevented copying).  How should I structure my module
> definition/code?
> 
> TIA
> 
> Colin
> 
> 
> 
> 
> 
> 
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From grafik.list at redshift-software.com  Fri Mar 12 10:01:20 2004
From: grafik.list at redshift-software.com (Rene Rivera)
Date: Fri, 12 Mar 2004 03:01:20 -0600
Subject: [C++-sig] Re: boost.build: How to set runtime library paths	in
	a Jamfile (wrapping 3rd party libs)
In-Reply-To: <OF7432B9B3.EA2AC13C-ONC1256E55.002A761D-C1256E55.002D9665@lbbw.sko.de>
References: <OF7432B9B3.EA2AC13C-ONC1256E55.002A761D-C1256E55.002D9665@lbbw.sko.de>
Message-ID: <40517C60.7010606@redshift-software.com>

Holger Joukl wrote:

> Rene wrote:
> 
>>feature of ELF for adding searchs paths relative to the loading module.. So
> 
> now
> 
>>I can get the effect I have on Windows of adding the location of my
> 
> executable
> 
>>(or any other relative subdir) to the library search path...
> 
> 
>><gcc><*><linkflags>-Wl,-rpath,'${ORIGIN}'
> 
> 
> If I understand correctly, this translate to an equivalent to "gcc
> -R<$ORIGIN>" with $ORIGIN being a
> relative path -- that is, simply using -R option with relative paths.

You understand incorrectly... There is no equivalent using regular searh paths 
and relative directories. The ELF linker-loader (ld.so in Linux) replaces the 
ORIGIN link variable with the path of the object doing the loading, and adds 
it to the list of search paths.

And by the way GCC will translate your -R to -rpath when it feeds it to the 
linker.

> While being fine for executables, I suspect that this is not really
> appropriate for boost.python wrapped
> libraries, where
> - libboost_python.so would probably reside in
> $PYTHON_ROOT/lib/python2.x/site-packages

You are making an assumption about locality. Why would you place it there?

> - the wrapper library shared object would probably reside in
> $PYTHON_ROOT/lib/python2.x/site-packages

Again why?

> - the wrapped library resides anywhere on your system

Like my home directory for example ;-)

> Running python and using this extension is now dependant from where you
> invoke the interpreter;

Not from where you invoke, but from where the *.so (or exec) doing the loading 
resides.

> it will only work if the wrapped lib can be found in the relative path
> $ORIGIN.

Or if it can find it in the regular search path of the system.

> So it is either LD_LIBRARY_PATH again (not good) or adding some search
> paths with -R in
> the link step, the search paths being s.th. like site-dependant standard
> library installation directories.
> E.g. a shared filesystem named /apps/prod where I would install central
> components (like python :-)

But if you use the ORIGIN link-var you can install to 
/apps/prod/something/myapp and put plugins and libboost_python*.so in a 
plugins subdir. All you need is to add the '${ORIGIN}/plugins' -rpath.

> Anyway, thanks a lot for all the hints!

You're Welcome :-)

-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org - grafik/redshift-software.com - 102708583/icq



From stefan.seefeld at orthosoft.ca  Fri Mar 12 15:38:35 2004
From: stefan.seefeld at orthosoft.ca (Stefan Seefeld)
Date: Fri, 12 Mar 2004 09:38:35 -0500
Subject: [C++-sig] instantiating python::object wrapping C++ objects
Message-ID: <252bdc2ce7eb198071924fe7ceadc30d4051cc2c@Orthosoft.ca>

hi there,

I want to write a function that wraps C++ object references
into python::object instances. The corresponding python::class_
is defined in a different module, but I don't know what the inverse
of python::extract<> is, i.e. how I can write a function

python::object do_wrap(MyType *v);

All the needed functionality seems to be there, so it appears
I'm missing something obvious here, but what ?

Thanks for your help,

		Stefan




From pierre.barbier at cirad.fr  Fri Mar 12 15:51:08 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Fri, 12 Mar 2004 15:51:08 +0100
Subject: [C++-sig] instantiating python::object wrapping C++ objects
In-Reply-To: <252bdc2ce7eb198071924fe7ceadc30d4051cc2c@Orthosoft.ca>
References: <252bdc2ce7eb198071924fe7ceadc30d4051cc2c@Orthosoft.ca>
Message-ID: <1079103058.1062.8.camel@pbarbier>

If you take extra care about the life of the C++ object, you can use
this:

object do_wrap(MyType *v)
{
  static reference_existing_object::apply<MyType*>::type converter;
  PyObject* obj = converter(v);
  return object(handle<>(obj));
}

I suppose the "converter" return a new reference so you don't need to
use the "borrowed" object. Also, there are converters in other return
policies, but I don't know if they are different !

Pierre

Le ven 12/03/2004 ? 15:38, Stefan Seefeld a ?crit :
> hi there,
> 
> I want to write a function that wraps C++ object references
> into python::object instances. The corresponding python::class_
> is defined in a different module, but I don't know what the inverse
> of python::extract<> is, i.e. how I can write a function
> 
> python::object do_wrap(MyType *v);
> 
> All the needed functionality seems to be there, so it appears
> I'm missing something obvious here, but what ?
> 
> Thanks for your help,
> 
> 		Stefan
> 
> 
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From dave at boost-consulting.com  Fri Mar 12 16:38:59 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Fri, 12 Mar 2004 10:38:59 -0500
Subject: [C++-sig] Re: boost.python problem: 'virtual functions and default
 implementations' tutorial doesn't work
References: <c2rf70$p0s$1@sea.gmane.org>
Message-ID: <uk71qcbxo.fsf@boost-consulting.com>

"Colin Irwin" <colin_irwin at hotmail.com> writes:

> I've been able to resolve the error by using the following python module
> definition instead:
>
> BOOST_PYTHON_MODULE(test_module)
> {
>
>     class_<Base, BaseWrapper, boost::noncopyable>("Base")
>         .def("function", &Base::function, &BaseWrapper::default_function)
>     ;
> }
>
> Now this noncopyable is not mentioned in the tutorial file and indeed, I
> don't think it does exactly what I want in this case.  By this I mean it
> prevents copying any of the derived objects in Python doesn't it?

No, it simply tells Boost.Python not to try create a by-value
converter from C++ Base objects to Python, such as would be used to
wrap a C++ function returing a Base object by value.

> Where have I gone wrong?  The tutorial file doesn't seem to work as
> expected.  

Then the tutorial is broken and needs to be fixed.

> My fix doesn't do what I want it do (I think this is right, if
> I've inadvertently prevented copying).  How should I structure my module
> definition/code?

As suggested by Pierre, the additional constructor in BaseWrapper
should work:

     BaseWrapper( PyObject* self_, const Base& copy);


-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From halbert at jubilee.bbn.com  Sun Mar 14 18:03:34 2004
From: halbert at jubilee.bbn.com (Dan Halbert)
Date: Sun, 14 Mar 2004 12:03:34 -0500
Subject: [C++-sig] Pyste: set_wrapper on overloaded function or method
Message-ID: <200403141703.i2EH3YA29952@hbo.bbn.com>

I was trying to create a wrapper for an overloaded method using the
Pyste "set_wrapper" capability (pyste from Boost 1.31.0). However, I ran
into a problem which can be illustrated by the following example:

========================================
testing.pyste:
--------------

testing = AllFromHeader("testing.h")
set_wrapper(testing.f, "f_wrapper")
set_wrapper(testing.C.g, "g_wrapper")
========================================
testing.h:
----------

int f(int i=1) { return i;}

struct C {
    int g(int i=2) { return i;}
};

int f_wrapper(float fl=3.0) { return (int) fl; }
int g_wrapper(C* c, float fl=4.0) { return (int) fl; }
========================================

Pyste produces a .cpp which includes this code:

//...
BOOST_PYTHON_MEMBER_FUNCTION_OVERLOADS(C_g_overloads_0_1, g, 0, 1)
BOOST_PYTHON_FUNCTION_OVERLOADS(f_overloads_0_1, f, 0, 1)
BOOST_PYTHON_FUNCTION_OVERLOADS(f_wrapper_overloads_0_1, f_wrapper, 0, 1)
BOOST_PYTHON_FUNCTION_OVERLOADS(g_wrapper_overloads_1_2, g_wrapper, 1, 2)
// ...
BOOST_PYTHON_MODULE(testing)
{
    class_< C >("C", init<  >())
        .def(init< const C& >())
        .def("g", &g_wrapper, C_g_overloads_0_1())    // ** Line g **
    ;

    def("f", &f_wrapper, f_overloads_0_1());          // ** Line f **
    // ...
}

========================================

The problem is that in Lines g and f above, the third args to ".def"
are the wrong overloads to use. They are the overloads for the
original method or function, not the wrapper overloads. The code
should be:

        .def("g", &g_wrapper, g_wrapper_overloads_1_2())    // corrected Line g

    def("f", &f_wrapper, f_wrapper_overloads_0_1());        // corrected Line f

If f_wrapper and g_wrapper had no overloads, there should be no third
arg at all.

I tried to fix this by patching ClassExporter.py around line 376,
where the third arg to ".def" is constructed. I checked for a wrapper
and tried to generate the overloads arg for the wrapper, not the
method. But I was stymied because the wrapper object is not a
full-fledged declaration object, and I can't get its min and max args.
I'm still looking at the code, but perhaps someone more familar with
it could suggest a fix.

The same kind of fix would need to go into FunctionExporter.py.

(In the example above, I would normally have done an
Include("testing_wrapper.h"). I left it out for simplicity and also to
force the BOOST_PYTHON_FUNCTION_OVERLOADS to occur.)

Regards,
Dan



From steam at nurfuerspam.de  Mon Mar 15 11:34:07 2004
From: steam at nurfuerspam.de (Hanz Meizer)
Date: Mon, 15 Mar 2004 11:34:07 +0100
Subject: [C++-sig] Re: Conversion of C++ return type.
In-Reply-To: <c24ugj$u85$1@sea.gmane.org>
References: <c1khs3$4i8$1@sea.gmane.org> <ufzcx742v.fsf@boost-consulting.com>
	<c24ugj$u85$1@sea.gmane.org>
Message-ID: <c340m3$eu5$1@sea.gmane.org>

Hi,

Just wanted to notify the list that I found the solution. Besides some 
other errors I made one of the wrapped functions 
return_internal_reference. This didn't invoke the to_python_converter I 
registered, but tried to create a wrapped Any object, where no wrapper 
was existant of course. I made all functions copy_(non)const_reference 
and it seems that in this case the to_python_converter get's called all 
the time. That is reasonable, but I think it should be mentioned in the 
documentation so that people can check if all their functions use 
copy_const_reference, bacause if you're new to boost.python there are so 
many things that can be the cause that you just don't kow where to search

H.




From halbert at bbn.com  Mon Mar 15 16:35:00 2004
From: halbert at bbn.com (Dan Halbert)
Date: Mon, 15 Mar 2004 10:35:00 -0500
Subject: [C++-sig] Pyste add_method()
In-Reply-To: <40510A11.2060803@esss.com.br> (message from Nicodemus on Thu, 11
	Mar 2004 21:53:37 -0300)
References: <1079030423.10839.10.camel@bluefish> <40510A11.2060803@esss.com.br>
Message-ID: <200403151535.i2FFZ0X19529@rolex.bbn.com>

  >From: Nicodemus <nicodemus at esss.com.br>
  >Daniel Holth wrote:
  >>I think I've found the solution to my old problem:
  >>
  >>sync = Class("ogg::sync", "oggcc.h") # part 1
  >>
  >>syncwrite = Function("ogg::sync_write", "oggpy_wrappers.h") # part 2
  >>add_method(sync, "ogg::sync_write")
  >>rename(sync.sync_write, "write")
  >>exclude(syncwrite)
  >>
  >>Doesn't work for me: it will say something like:
  >>RuntimeError: no ogg::sync_write declaration found!
  >>
  >>Instead, if I say the class is in the file that declares the wrapper,
  >>(oggpy_wrappers.h also includes oggcc.h):
  >>
  >>sync = Class("ogg::sync", "oggpy_wrappers.h")
  >>
  >>(insert part 2 here)
  >>
  >>it works nicely.

This reminds me of the issue I reported about set_wrapper yesterday,
and how it wasn't clear to me how to fix the problem.

Right now, both set_wrapper() and add_method() take a string as a
second argument to name the function, instead of a full-fledged
"declaration info" object, or whatever it's called. We say:

      Include("wrappers.h")
      set_wrapper(my_function, "my_function_wrapper")

Similarly we say:

      add_method(MyClass, "my_new_method")


But suppose the second argument was not a string, but a definition object.
Maybe I'm wrong, but it seems to me clearer to say something like:

      my_function_wrapper = Function("my_function_wrapper", "wrappers.h")
      set_wrapper(my_function, my_function_wrapper)
and
      my_new_method = Function("my_new_method", "new_methods.h")
      add_method(MyClass, my_new_method)

Internally to Pyste, this would give set_wrapper() and add_method() more
information to work with.

Dan
      

	



From ndbecker2 at verizon.net  Mon Mar 15 21:24:29 2004
From: ndbecker2 at verizon.net (Neal D. Becker)
Date: Mon, 15 Mar 2004 15:24:29 -0500
Subject: [C++-sig] std::vector
Message-ID: <c353dt$iue$1@sea.gmane.org>

I'm back to looking at integration of python with C++ code, which uses
std::vector, particulary vector<double> and vector<complex<double>>.

What would be the best/easiest/recommended way to get python to pickle an
std::vector?





From tether at mitlns.mit.edu  Mon Mar 15 22:23:30 2004
From: tether at mitlns.mit.edu (Stephen Tether)
Date: Mon, 15 Mar 2004 15:23:30 -0600
Subject: [C++-sig] Pyste in boost 1.31.0 has trouble with exception specs
Message-ID: <40561ED2.10703@mitlns.mit.edu>

I tried running Pyste on the following files test.hh and test.pyste and 
it threw an TypeError exception.

// Begin test.hh
#include <stdexcept>

class Foo {
public:
   virtual void bar() throw(std::exception);
};
// End test.hh


# Begin test.pyste
Class("Foo", "test.hh")
# End test.pyste

The exception traceback:
Traceback (most recent call last):
   File "/evbsoft/python/v2_3_3c1/bin/pyste.py", line 8, in ?
     pyste.main()
   File 
"/evbsoft/python/v2_3_3c1/lib/python2.3/site-packages/Pyste/pyste.py", 
line 405, in main
     status = Begin()
   File 
"/evbsoft/python/v2_3_3c1/lib/python2.3/site-packages/Pyste/pyste.py", 
line 244, in Begin
     return GenerateCode(parser, module, out, interfaces, multiple)
   File 
"/evbsoft/python/v2_3_3c1/lib/python2.3/site-packages/Pyste/pyste.py", 
line 372, in GenerateCode
     export.GenerateCode(codeunit, exported_names)
   File 
"/evbsoft/python/v2_3_3c1/lib/python2.3/site-packages/Pyste/Exporter.py", 
line 50, in GenerateCode
     self.Export(codeunit, exported_names)
   File 
"/evbsoft/python/v2_3_3c1/lib/python2.3/site-packages/Pyste/ClassExporter.py", 
line 92, in Export
     self.InheritMethods(exported_names)
   File 
"/evbsoft/python/v2_3_3c1/lib/python2.3/site-packages/Pyste/ClassExporter.py", 
line 119, in InheritMethods
     pointers = [x.PointerDeclaration(True) for x in self.class_ if 
isinstance(x, Method)]
   File 
"/evbsoft/python/v2_3_3c1/lib/python2.3/site-packages/Pyste/declarations.py", 
line 320, in PointerDeclaration
     return '(%s (%s::*)(%s) %s%s)&%s' %\
   File 
"/evbsoft/python/v2_3_3c1/lib/python2.3/site-packages/Pyste/declarations.py", 
line 226, in Exceptions
     return " throw(%s)" % ', '.join (self.throws)
TypeError: sequence item 0: expected string, Type found


It looks like a list comprehension or map() call at line 226 of 
declarations.py:
     return " throw(%s)" % ', '.join ([x.name for x in self.throws])

If I make that change I get the following output:

// Boost Includes 
==============================================================
#include <boost/python.hpp>
#include <boost/cstdint.hpp>

// Includes 
====================================================================
#include <test.hh>

// Using 
=======================================================================
using namespace boost::python;

// Declarations 
================================================================
namespace  {

struct Foo_Wrapper: Foo
{
     Foo_Wrapper(PyObject* self_, const Foo& p0):
         Foo(p0), self(self_) {}

     Foo_Wrapper(PyObject* self_):
         Foo(), self(self_) {}

     void bar() throw(std::exception) {
         call_method< void >(self, "bar");
     }

     void default_bar() {
         Foo::bar();
     }

     PyObject* self;
};


}// namespace


// Module 
======================================================================
BOOST_PYTHON_MODULE(test)
{
     class_< Foo, Foo_Wrapper >("Foo", init<  >())
         .def(init< const Foo& >())
         .def("bar", &Foo::bar, &Foo_Wrapper::default_bar)
     ;

}
// End of test.cpp

- Steve Tether



From rwgk at yahoo.com  Mon Mar 15 23:54:23 2004
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Mon, 15 Mar 2004 14:54:23 -0800 (PST)
Subject: [C++-sig] std::vector
In-Reply-To: <c353dt$iue$1@sea.gmane.org>
Message-ID: <20040315225423.44660.qmail@web20208.mail.yahoo.com>

--- "Neal D. Becker" <ndbecker2 at verizon.net> wrote:
> I'm back to looking at integration of python with C++ code, which uses
> std::vector, particulary vector<double> and vector<complex<double>>.
> 
> What would be the best/easiest/recommended way to get python to pickle an
> std::vector?

Easiest (this requires boost from CVS, will not work with 1.31.0):

class_<std::vector<double> >("std_vector_double")
  .def(vector_indexing_suite<std::vector<double> >())
  .enable_pickling()
;

In Python:

  # import the wrapped world class
  from your_ext import std_vector_double

  # definition of __getinitargs__
  def std_vector_double_getinitargs(self):
    return tuple(self)

  # now inject __getinitargs__ (Python is a dynamic language!)
  std_vector_double.__getinitargs__ = std_vector_double_getinitargs

This requires a std_vector_double constructor that takes a tuple as the
argument. I am not sure if the indexing suite provides this; probably not. Make
one yourself or define __getstate__ and __setstate__. See:

http://cci.lbl.gov/~rwgk/shortcuts/boost/libs/python/doc/v2/pickle.html

I highly recommend implementing the constructor and leaving __getstate__ and
__setstate__ alone.

This easy solution is very inefficient because each element is converted to a
Python object. The scitbx solves this problem with a small, custom
serialization library. The entire array is converted to one Python object, a
Python string. It is fast and the pickled arrays are quite compact and
portable. I am afraid I never found the time do write the documentation; hope
one day someone will take the scitbx code and join it with the indexing suite.
Here it is the relevant code if someone wants to have a look:

http://cvs.sourceforge.net/viewcvs.py/cctbx/scitbx/include/scitbx/serialization/base_256.h?view=markup
http://cvs.sourceforge.net/viewcvs.py/cctbx/scitbx/include/scitbx/boost_python/pickle_single_buffered.h?view=markup
http://cvs.sourceforge.net/viewcvs.py/cctbx/scitbx/include/scitbx/boost_python/pickle_double_buffered.h?view=markup

Ralf


__________________________________
Do you Yahoo!?
Yahoo! Mail - More reliable, more storage, less spam
http://mail.yahoo.com



From greglandrum at mindspring.com  Tue Mar 16 17:25:12 2004
From: greglandrum at mindspring.com (greg Landrum)
Date: Tue, 16 Mar 2004 08:25:12 -0800
Subject: [C++-sig] vector_indexing_suite and std::string
Message-ID: <5.1.0.14.2.20040316080922.040df988@mail.earthlink.net>


The attached files create and test an extension module that uses the indexing_suite to return std::vectors containing ints and std::strings to Python.  The int vectors work just  fine, but the std::string vectors generate:
TypeError: No Python class registered for C++ class std::string
errors when the individual elements are accessed.

It's clearly not a problem with returning std::strings, because the bareString() function does that without any difficulties.

Ron Clarke posted about this a couple of weeks ago: 
http://mail.python.org/pipermail/c++-sig/2004-March/006966.html
but no solution or work-around was provided, so I figured I'd try again.

System details:
Boost: 1.31
Python: 2.2.3
OS: both Redhat8.0 (g++ v3.2)and Win2K (VC++ 7.1)

Thanks,
-greg
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: indexing_with_strings.cpp
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040316/16714db6/attachment.txt>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: indexing_with_strings.py
Type: application/octet-stream
Size: 564 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040316/16714db6/attachment.obj>

From jbrandmeyer at earthlink.net  Tue Mar 16 21:19:50 2004
From: jbrandmeyer at earthlink.net (Jonathan Brandmeyer)
Date: Tue, 16 Mar 2004 15:19:50 -0500
Subject: [C++-sig] vector_indexing_suite and slicing
Message-ID: <1079468390.2134.55.camel@illuvatar>

A few months ago there was a discussion on this list related to the
semantics of setting a (possibly extended) slice when the setting range
is empty[1].  As part of that discussion I submitted a patch to change
the behavior of a Python list [2].  The resolution on that was to
clarify that list.__setitem__(slice, obj)  has the following behavior
(no changes):

When called with a simple slice that defines an empty range, perform an
insertion at the first index.

When called with an extended slice regardless of the length of the
range, perform a replacement of the objects indexed by the slice with
objects from obj.  obj must have exactly the same number of elements as
indexed by the slice.

To close the loop, the attached patch will fix the bugs I noted in [3]
while following the guidance provided by the Python folks.  Unless there
are any objections, I'll commit this to HEAD and the RC_1_31_0 branch
tonight.

-Jonathan Brandmeyer


[1]: summary
http://mail.python.org/pipermail/c++-sig/2004-January/006589.html
[2]: bug tracker for python
https://sourceforge.net/tracker/?func=detail&atid=305470&aid=873305&group_id=5470
[3]: http://mail.python.org/pipermail/c++-sig/2004-January/006552.html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: indexing_suite.patch
Type: text/x-patch
Size: 4288 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040316/93cad8ac/attachment.bin>

From nicodemus at esss.com.br  Tue Mar 16 23:14:31 2004
From: nicodemus at esss.com.br (Nicodemus)
Date: Tue, 16 Mar 2004 19:14:31 -0300
Subject: [C++-sig] Pyste: set_wrapper on overloaded function or method
In-Reply-To: <200403141703.i2EH3YA29952@hbo.bbn.com>
References: <200403141703.i2EH3YA29952@hbo.bbn.com>
Message-ID: <40577C47.9010703@esss.com.br>

Dan Halbert wrote:

>I was trying to create a wrapper for an overloaded method using the
>Pyste "set_wrapper" capability (pyste from Boost 1.31.0). However, I ran
>into a problem which can be illustrated by the following example:
>
><snip example>
>The problem is that in Lines g and f above, the third args to ".def"
>are the wrong overloads to use. They are the overloads for the
>original method or function, not the wrapper overloads. The code
>should be:
>
>        .def("g", &g_wrapper, g_wrapper_overloads_1_2())    // corrected Line g
>
>    def("f", &f_wrapper, f_wrapper_overloads_0_1());        // corrected Line f
>
>If f_wrapper and g_wrapper had no overloads, there should be no third
>arg at all.
>
>I tried to fix this by patching ClassExporter.py around line 376,
>where the third arg to ".def" is constructed. I checked for a wrapper
>and tried to generate the overloads arg for the wrapper, not the
>method. But I was stymied because the wrapper object is not a
>full-fledged declaration object, and I can't get its min and max args.
>I'm still looking at the code, but perhaps someone more familar with
>it could suggest a fix.
>
>The same kind of fix would need to go into FunctionExporter.py.
>
>(In the example above, I would normally have done an
>Include("testing_wrapper.h"). I left it out for simplicity and also to
>force the BOOST_PYTHON_FUNCTION_OVERLOADS to occur.)
>
>  
>

Hi Dan,

I'm not ignoring this message, it's just that I haven't got time to look 
it through... I've been pretty busy. But expect a reply in the next few 
days.

Regards,
Nicodemus.




From jbrandmeyer at earthlink.net  Wed Mar 17 00:16:13 2004
From: jbrandmeyer at earthlink.net (Jonathan Brandmeyer)
Date: Tue, 16 Mar 2004 18:16:13 -0500
Subject: [C++-sig] vector_indexing_suite and std::string
In-Reply-To: <5.1.0.14.2.20040316080922.040df988@mail.earthlink.net>
References: <5.1.0.14.2.20040316080922.040df988@mail.earthlink.net>
Message-ID: <1079478972.2134.60.camel@illuvatar>

On Tue, 2004-03-16 at 11:25, greg Landrum wrote:
> The attached files create and test an extension module that uses the indexing_suite to return std::vectors containing ints and std::strings to Python.  The int vectors work just  fine, but the std::string vectors generate:
> TypeError: No Python class registered for C++ class std::string
> errors when the individual elements are accessed.
> 
> It's clearly not a problem with returning std::strings, because the bareString() function does that without any difficulties.
> 
> Ron Clarke posted about this a couple of weeks ago: 
> http://mail.python.org/pipermail/c++-sig/2004-March/006966.html
> but no solution or work-around was provided, so I figured I'd try again.
> 
> System details:
> Boost: 1.31
> Python: 2.2.3
> OS: both Redhat8.0 (g++ v3.2)and Win2K (VC++ 7.1)
> 
> Thanks,
> -greg

Its not in the indexing suite, per se, its a problem with the
implementation of register_ptr_to_python().  Here is a test that
demonstrates the problem.  Kindly ignore the memory leak, its just for
demonstration purposes.

HTH,
Jonathan Brandmeyer
-------------- next part --------------
A non-text attachment was scrubbed...
Name: wrap_test.cpp
Type: text/x-c++
Size: 869 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040316/d3d79ba9/attachment.bin>

From jbrandmeyer at earthlink.net  Wed Mar 17 02:48:52 2004
From: jbrandmeyer at earthlink.net (Jonathan Brandmeyer)
Date: Tue, 16 Mar 2004 20:48:52 -0500
Subject: [C++-sig] vector_indexing_suite and std::string
In-Reply-To: <1079478972.2134.60.camel@illuvatar>
References: <5.1.0.14.2.20040316080922.040df988@mail.earthlink.net>
	<1079478972.2134.60.camel@illuvatar>
Message-ID: <1079488132.2134.71.camel@illuvatar>

On Tue, 2004-03-16 at 18:16, Jonathan Brandmeyer wrote:

> Its not in the indexing suite, per se, its a problem with the
> implementation of register_ptr_to_python().  Here is a test that
> demonstrates the problem.  Kindly ignore the memory leak, its just for
> demonstration purposes.
> 
> HTH,
> Jonathan Brandmeyer
> 
> ______________________________________________________________________

Never mind, that isn't a bug with register_ptr_to_python(), it's a
feature, as Dave tried to explain to me before ;).  For vectors of
std::string and std::complex, use the following workaround to make the
indexing suite work:

.def( vector_indexing_suite< std::vector<std::string>, true>())
                                                       ^^^^

A permanent fix would involve making sure that this type expression
evaluates true when Data is either std::string or std::complex<T> (from
indexing_suite.hpp line 121):

        typedef mpl::or_<
            mpl::bool_<NoProxy>
          , mpl::not_<is_class<Data> > > 
        no_proxy;

-Jonathan Brandmeyer




From greglandrum at mindspring.com  Wed Mar 17 06:15:05 2004
From: greglandrum at mindspring.com (greg Landrum)
Date: Tue, 16 Mar 2004 21:15:05 -0800
Subject: [C++-sig] vector_indexing_suite and std::string
In-Reply-To: <1079488132.2134.71.camel@illuvatar>
References: <1079478972.2134.60.camel@illuvatar>
	<5.1.0.14.2.20040316080922.040df988@mail.earthlink.net>
	<1079478972.2134.60.camel@illuvatar>
Message-ID: <5.1.0.14.2.20040316211423.03fc70a8@mail.mindspring.com>

At 05:48 PM 3/16/2004, Jonathan Brandmeyer wrote:

>Never mind, that isn't a bug with register_ptr_to_python(), it's a
>feature, as Dave tried to explain to me before ;).  For vectors of
>std::string and std::complex, use the following workaround to make the
>indexing suite work:
>
>.def( vector_indexing_suite< std::vector<std::string>, true>())
>                                                       ^^^^

That cleared up the problem.  Thanks Jonathan!

-greg




From A.Srivastav at zensar.com  Wed Mar 17 09:16:42 2004
From: A.Srivastav at zensar.com (Ajitkumar Srivastav)
Date: Wed, 17 Mar 2004 13:46:42 +0530
Subject: [C++-sig] crash problem
Message-ID: <4171EC88BB77E34EAE1012DBEB6C7EC70A2512@ZENMAILHQ1.ind.zensar.com>

Hello,

I have used python 2.3 in my C++ application.
I generally make calls to python library to perform my intended task. It happens properly.
But after that immediatly the memory gets corrupted and application crashes.
Pls help.

thnx in advance
Ajit Shrivastav

Zensar Technologies Ltd.
Mile Post no. 4
Pune- 411014
Tel no. - 9520-26633711 x121
Mobile no. - 98900 37203

www.zensar.com
Your Transformation Partner

It's amazing that the amount of news that happens in the world
everyday always  just exactly fits the newspaper.




From joel at boost-consulting.com  Thu Mar 18 14:06:32 2004
From: joel at boost-consulting.com (Joel de Guzman)
Date: Thu, 18 Mar 2004 21:06:32 +0800
Subject: [C++-sig] vector_indexing_suite and std::string
In-Reply-To: <5.1.0.14.2.20040316211423.03fc70a8@mail.mindspring.com>
References: <1079478972.2134.60.camel@illuvatar>	<5.1.0.14.2.20040316080922.040df988@mail.earthlink.net>	<1079478972.2134.60.camel@illuvatar>
	<5.1.0.14.2.20040316211423.03fc70a8@mail.mindspring.com>
Message-ID: <40599ED8.6030503@boost-consulting.com>

greg Landrum wrote:
> At 05:48 PM 3/16/2004, Jonathan Brandmeyer wrote:
> 
> 
>>Never mind, that isn't a bug with register_ptr_to_python(), it's a
>>feature, as Dave tried to explain to me before ;).  For vectors of
>>std::string and std::complex, use the following workaround to make the
>>indexing suite work:
>>
>>.def( vector_indexing_suite< std::vector<std::string>, true>())
>>                                                      ^^^^
> 
> 
> That cleared up the problem.  Thanks Jonathan!

Thanks for answering Greg, Jonathan. Hopefully, with the
new indexing suite (by Raoul) all these will be non issues.

-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net



From joel at boost-consulting.com  Thu Mar 18 14:09:42 2004
From: joel at boost-consulting.com (Joel de Guzman)
Date: Thu, 18 Mar 2004 21:09:42 +0800
Subject: [C++-sig] vector_indexing_suite and slicing
In-Reply-To: <1079468390.2134.55.camel@illuvatar>
References: <1079468390.2134.55.camel@illuvatar>
Message-ID: <40599F96.6020602@boost-consulting.com>

Jonathan Brandmeyer wrote:

> A few months ago there was a discussion on this list related to the
> semantics of setting a (possibly extended) slice when the setting range
> is empty[1].  As part of that discussion I submitted a patch to change
> the behavior of a Python list [2].  The resolution on that was to
> clarify that list.__setitem__(slice, obj)  has the following behavior
> (no changes):
> 
> When called with a simple slice that defines an empty range, perform an
> insertion at the first index.
> 
> When called with an extended slice regardless of the length of the
> range, perform a replacement of the objects indexed by the slice with
> objects from obj.  obj must have exactly the same number of elements as
> indexed by the slice.
> 
> To close the loop, the attached patch will fix the bugs I noted in [3]
> while following the guidance provided by the Python folks.  Unless there
> are any objections, I'll commit this to HEAD and the RC_1_31_0 branch
> tonight.
> 
> -Jonathan Brandmeyer
> 
> 
> [1]: summary
> http://mail.python.org/pipermail/c++-sig/2004-January/006589.html
> [2]: bug tracker for python
> https://sourceforge.net/tracker/?func=detail&atid=305470&aid=873305&group_id=5470
> [3]: http://mail.python.org/pipermail/c++-sig/2004-January/006552.html

No objections here. I wonder though about the status of Raoul's
new indexing suite. I was under the impression that his work will
supersede the current indexing suite. Raoul?

Cheers,
-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net



From gwl at u.washington.edu  Fri Mar 19 00:48:33 2004
From: gwl at u.washington.edu (Graeme Lufkin)
Date: Thu, 18 Mar 2004 15:48:33 -0800
Subject: [C++-sig] Re: embedding examples
References: <1078132099.32100.6.camel@localhost>
Message-ID: <pan.2004.03.18.23.48.32.224288@u.washington.edu>

On Mon, 01 Mar 2004 10:08:19 +0100, Ingo Luetkebohle wrote:
> I have trouble compiling the 'Using the interpreter' examples from the
> tutorial.
> 
> Code like this:
>   handle<> main_module(borrowed( PyImport_AddModule("__main__") )); dict
>   main_namespace(handle<>(borrowed(
> PyModule_GetDict(main_module.get()) ))); results in
>   error: variable declaration is not allowed here
> 
> using gcc 3.2 and gcc 3.3 on Linux.
> 
> When I use
>   handle<> main_module(borrowed( PyImport_AddModule("__main__") ));
>   handle<> main_nsh(borrowed( PyModule_GetDict(main_module.get()) ));
>   dict main_namespace(main_nsh);
> 
> it compiles and the resulting dict is usable.
> 
> So, this is not a showstopper but it doesn't work as documented and
> results in longer code.  Is there anything I can do to get the intended
> behaviour?
> 
> regards

	I've got this problem as well, and have investigated its roots.  The
problem is that the programmer thinks he's creating an anonymous temporary
handle<> object, while the compiler thinks you're declaring a variable. 
For example:
PyObject* pobj = 0;
handle<>(pobj);
	I think I'm creating an anonymous temporary of type handle<>,
constructing it with pobj (yes, this would fail, because handle<> throws
on null, but that's not the point).  The compiler discards the parentheses
and thus tries to redeclare pobj as a variable of type handle<>, which is
an error.
	Let's use a more familiar class template for some more examples.
//declare n as int
int n = 3;
//declare v to be a vector, construct it with n elements 
vector<int> v(3);
//construct an anonymous temporary of n elements, setting them to 2
vector<int>(n, 2);
//construct an anonymous temporary of 3 elements 
vector<int>(3);
//attempt to redeclare n as a vector<int> 
vector<int>(n);
	Before I tried it, I would have guessed the last
item would do the same thing as the second-to-last item. gcc and Intel
know better than I, apparently.
	Enough of this fun, what do I do about handles?  The reason I
need the handle<> in the first place is so that the PyObject pointer gets
Py_DECREF'd when I'm done with it.  So all calls to the Python API that
return a PyObject* should get wrapped in a handle<>, possibly borrowed. 
But, as the examples in the tutorial show, you often don't need the
PyObject*, you just want to make sure it gets reference counted correctly.

For example, in the code from the tutorial listed below, instead of
creating and immediately destroying an anonymous temporary handle<>, the
compiler tries to declare a variable called PyRun_String as a handle<> and
construct it with a constructor that doesn't exist.  If I give the
handle<> a name, it works as I expected.

handle<>( PyRun_String("hello = file('hello.txt','w')\n"
                           "hello.write('Hello world!')\n"
                           "hello.close()", Py_file_input,
                           main_namespace.get(), main_namespace.get()) );

	So finally to my question: Since these anonymous temporaries are useful,
what should I do?  Is there a fundamental C++ rule that I'm missing or
fighting against?  Also could the embedding part of the tutorial be
updated to no longer contain code that doesn't compile?

	- Graeme Lufkin
	gwl at u.washington.edu
"This sentence contains exactly threee erors."




From JCrystal at lucascareers.com  Fri Mar 19 20:04:00 2004
From: JCrystal at lucascareers.com (Crystal, Jennifer)
Date: Fri, 19 Mar 2004 14:04:00 -0500
Subject: [C++-sig] Sr. Software Developer - New York, NY
Message-ID: <5E78C9CE14A9C94899DEC75CF243B89904F354AA@ex01.lucas.internal>

Hi, my name is Jennifer Crystal and I am an Executive Recruiter with the
LucasGroup. The Lucas Group has been in business for 33 years and is
ranked in the top 20 search firms nationwide. Headquartered in Atlanta,
we have 12 locations across the US with 20+ recruiting niches. While
national in scope, we do not franchise offices, and thereby maintain
core values and rigorous professional standards throughout our
organization. Our 250+ recruiters are experts in their specialties, with
extensive industry contacts and average 10-year tenure with our company.
The LucasGroup was also chosen by the Wall Street Journal as the
Executive Search Firm of choice and has established a beneficial
partnership. 

I am searching for Sr. Software Engineer in New York, NY:

SR. SOFTWARE DEVELOPER 

JOB PURPOSE/OBJECTIVE

Sr. Software development / programming for ecommerce portal and customer
applications.

 

ESSENTIAL DUTIES AND RESPONSIBILITIES:

 

Will work in small to mid-size teams to develop Web-based
mission-critical applications using Predominately Python, Zope, Linux,
Oracle.

 

QUALIFICATION REQUIREMENTS:

 

To perform this job successfully, and individual must be able to perform
each essential duty satisfactorily. The requirements listed below are
representative of the knowledge, skill, and abilities required:

 

*	Superior knowledge of Python (3+ years) 

*	Oracle 8i or other DBMS specifically integration (3+ years) 

*	Formal project life cycle training 

*	Linux knowledge (3+ years) 

*	Objective Oriented Design (5+ years) and analysis 

*	Client Server background (5+ years in VB 6, Java or similar) 

*	Experience with large and small development teams

 

EDUCATION AND ECPERIENCE:

 

*	Preferable Education Requirement: Bachelor's degree or higher in
computer science or engineering field. 

*	Minimum Experience Requirement: 5+ years previous successful
development/programming experience on the web.. 

*	Preferable Experience Requirement: Prefer experience in
ecommerce portal environment. Exposure to relational databases, data
warehousing, internet-based application development or distributed
object architectures. At least 2 years experience with internet
technologies and significant additional application development
experience in client-sever environments. 

*	Other: Experience with SQL, VB, java, Linux/Unix, Oracle and
integration of databases with Cold Fusion onto an ecommerce web
platform, Apache.

If you or someone you know might be interested in any of the positions,
please send me a MS Word resume to jcrystal at lucascareers.com or contact
me at 800-466-4489 x171. 
 
Thanks!

Jennifer Crystal 
Senior Recruiter, Technology 
LucasGroup 
Recruiting Excellence Since 1970 

3384 Peachtree Road, Suite 700 
Atlanta, GA 30326 
404-239-5630 x171 
800-466-4489 x171 
jcrystal at lucascareers.com 
www.lucascareers.com <http://www.lucascareers.com/>  


 <outbind://40/www.LucasCareers.com> 

LucasGroup, In Partnership with THE WALL STREET JOURNAL. & CareerJournal
Online.

***The information contained in this email is proprietary and
confidential and is intended solely for the use of the named addressee.
Do not disclose, copy, distribute, or disseminate it to any other party
without the expressed consent of the sending party. If you have received
this message in error please return the message to the sender by
replying to it and then delete the message from your computer.

 

Jennifer Crystal 
Senior Recruiter, Technology 
LucasGroup 
Recruiting Excellence Since 1970 

3384 Peachtree Road, Suite 700 
Atlanta, GA 30326 
404-239-5630 x171 
800-466-4489 x171 
jcrystal at lucascareers.com 
www.lucascareers.com <http://www.lucascareers.com/>  


 <outbind://40/www.LucasCareers.com> 

LucasGroup, In Partnership with THE WALL STREET JOURNAL. & CareerJournal
Online.

***The information contained in this email is proprietary and
confidential and is intended solely for the use of the named addressee.
Do not disclose, copy, distribute, or disseminate it to any other party
without the expressed consent of the sending party. If you have received
this message in error please return the message to the sender by
replying to it and then delete the message from your computer.

 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040319/97bf7d3c/attachment.htm>

From RaoulGough at yahoo.co.uk  Sun Mar 21 20:52:54 2004
From: RaoulGough at yahoo.co.uk (Raoul Gough)
Date: Sun, 21 Mar 2004 19:52:54 +0000
Subject: [C++-sig] Re: vector_indexing_suite and slicing
References: <1079468390.2134.55.camel@illuvatar>
	<40599F96.6020602@boost-consulting.com>
Message-ID: <uptb6j7u1.fsf@yahoo.co.uk>

Joel de Guzman <joel at boost-consulting.com> writes:

> Jonathan Brandmeyer wrote:
>
>> A few months ago there was a discussion on this list related to the
>> semantics of setting a (possibly extended) slice when the setting range
>> is empty[1].  As part of that discussion I submitted a patch to change
>> the behavior of a Python list [2].  The resolution on that was to
>> clarify that list.__setitem__(slice, obj)  has the following behavior
>> (no changes):
>> When called with a simple slice that defines an empty range, perform
>> an
>> insertion at the first index.
>> When called with an extended slice regardless of the length of the
>> range, perform a replacement of the objects indexed by the slice with
>> objects from obj.  obj must have exactly the same number of elements as
>> indexed by the slice.
>> To close the loop, the attached patch will fix the bugs I noted in
>> [3]
>> while following the guidance provided by the Python folks.  Unless there
>> are any objections, I'll commit this to HEAD and the RC_1_31_0 branch
>> tonight.
>> -Jonathan Brandmeyer
>> [1]: summary
>> http://mail.python.org/pipermail/c++-sig/2004-January/006589.html
>> [2]: bug tracker for python
>> https://sourceforge.net/tracker/?func=detail&atid=305470&aid=873305&group_id=5470
>> [3]: http://mail.python.org/pipermail/c++-sig/2004-January/006552.html
>
> No objections here. I wonder though about the status of Raoul's
> new indexing suite. I was under the impression that his work will
> supersede the current indexing suite. Raoul?

I'm pretty sure I mentioned my work in the same thread back in
January, and that the slice behaviour of the new indexing suite was
already Python compliant.

I haven't had time recently to get the documentation in order, but I
intend to get that done before the ACCU conference in mid-April. Once
the docco is ready, we could probably consider merging indexing_v2
onto the mainline, but I guess there's no harm in patching the
original indexing suite in the meantime.

-- 
Raoul Gough.
export LESS='-X'




From joel at boost-consulting.com  Mon Mar 22 01:01:21 2004
From: joel at boost-consulting.com (Joel de Guzman)
Date: Mon, 22 Mar 2004 08:01:21 +0800
Subject: [C++-sig] Re: vector_indexing_suite and slicing
In-Reply-To: <uptb6j7u1.fsf@yahoo.co.uk>
References: <1079468390.2134.55.camel@illuvatar>	<40599F96.6020602@boost-consulting.com>
	<uptb6j7u1.fsf@yahoo.co.uk>
Message-ID: <405E2CD1.2040703@boost-consulting.com>

Raoul Gough wrote:

> Joel de Guzman <joel at boost-consulting.com> writes:
> 
>>>To close the loop, the attached patch will fix the bugs I noted in
>>>[3]
>>>while following the guidance provided by the Python folks.  Unless there
>>>are any objections, I'll commit this to HEAD and the RC_1_31_0 branch
>>>tonight.

[snip]

>>No objections here. I wonder though about the status of Raoul's
>>new indexing suite. I was under the impression that his work will
>>supersede the current indexing suite. Raoul?
> 
> 
> I'm pretty sure I mentioned my work in the same thread back in
> January, and that the slice behaviour of the new indexing suite was
> already Python compliant.
> 
> I haven't had time recently to get the documentation in order, but I
> intend to get that done before the ACCU conference in mid-April. Once
> the docco is ready, we could probably consider merging indexing_v2
> onto the mainline, but I guess there's no harm in patching the
> original indexing suite in the meantime.

I agree. I look forward to the new indexing suite. I still owe
you a thorough review ;-)

Cheers,
-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net



From dave at boost-consulting.com  Mon Mar 22 01:17:49 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Sun, 21 Mar 2004 19:17:49 -0500
Subject: [C++-sig] Re: crash problem
References: <4171EC88BB77E34EAE1012DBEB6C7EC70A2512@ZENMAILHQ1.ind.zensar.com>
Message-ID: <u7jxdivki.fsf@boost-consulting.com>

"Ajitkumar Srivastav" <A.Srivastav at zensar.com> writes:

> Hello,
>
> I have used python 2.3 in my C++ application.
>
> I generally make calls to python library to perform my intended
> task. It happens properly.  But after that immediatly the memory
> gets corrupted and application crashes.

Usually you need to describe what you're doing a bit before anyone can
help you.  Most often, it requires reducing your problem to a small,
reproducible case.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Mon Mar 22 04:09:47 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Sun, 21 Mar 2004 22:09:47 -0500
Subject: [C++-sig] Re: Conversion of C++ return type.
References: <c1khs3$4i8$1@sea.gmane.org> <ufzcx742v.fsf@boost-consulting.com>
	<c24ugj$u85$1@sea.gmane.org> <c340m3$eu5$1@sea.gmane.org>
Message-ID: <uoeqph91g.fsf@boost-consulting.com>

Hanz Meizer <steam at nurfuerspam.de> writes:

> Hi,
>
> Just wanted to notify the list that I found the solution. Besides some
> other errors I made one of the wrapped functions
> return_internal_reference. This didn't invoke the to_python_converter
> I registered, but tried to create a wrapped Any object, where no
> wrapper was existant of course. I made all functions
> copy_(non)const_reference and it seems that in this case the
> to_python_converter get's called all the time. That is reasonable, but
> I think it should be mentioned in the documentation 

What, specifically?

> so that people can
> check if all their functions use copy_const_reference, bacause if
> you're new to boost.python there are so many things that can be the
> cause that you just don't kow where to search

Please propose a patch to the documentation, so I know exactly what
you'd like to see explained, and what language you think would've
prevented your confusion.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From bhartsho at yahoo.com  Mon Mar 22 08:48:45 2004
From: bhartsho at yahoo.com (brett hartshorn)
Date: Sun, 21 Mar 2004 23:48:45 -0800 (PST)
Subject: [C++-sig] Boost newbie help
Message-ID: <20040322074845.47550.qmail@web13421.mail.yahoo.com>

I'm having a problem wrapping a class that looks like this:

class A {
    A _myVar;
    A (const A &);   //don't ask why this is not defined

  public:
    A() { _myVar = myFunction(); }

}

It will generate errors when building that say:
error: `A::A(const A&)' is private
/usr/include/boost/python/object/value_holder.hpp:111: error: within this context

If i remove the strange constructor that does nothing, then it builds without any problems.
Anybody have any ideas how to fix this, other than removing that constructor?

thanks for the help,
-brett

__________________________________
Do you Yahoo!?
Yahoo! Finance Tax Center - File online. File on time.
http://taxes.yahoo.com/filing.html



From bhartsho at yahoo.com  Mon Mar 22 09:00:53 2004
From: bhartsho at yahoo.com (brett hartshorn)
Date: Mon, 22 Mar 2004 00:00:53 -0800 (PST)
Subject: [C++-sig] How to specify a double return type?
Message-ID: <20040322080053.31492.qmail@web13426.mail.yahoo.com>

I have a type called dReal, that can either be a float or double number type, depending on the
build options.  A member function in my class returns a dReal type, this is giving me problems in
Boost.  I am still new to boost, so i am not sure what to do.  

Here is the error i get:
/usr/include/boost/python/detail/invoke.hpp:93: error: no match for call to `(
   boost::python::detail::specify_a_return_value_policy_to_wrap_functions_returning<const
   dReal*>) (const dReal*)'

I've tried to use:
   return_internal_reference<>()
   return_value_policy<manage_new_object>())
   return_value_policy<reference_existing_object>()
but to no avail.

Also i have member functions that take in a dReal number.  What is the best way to pass that
function a dReal type?  Or can i just pass floats and doubles in python and boost will treat those
as a dReal?

thanks,
-brett

__________________________________
Do you Yahoo!?
Yahoo! Finance Tax Center - File online. File on time.
http://taxes.yahoo.com/filing.html



From rwgk at yahoo.com  Mon Mar 22 15:48:05 2004
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Mon, 22 Mar 2004 06:48:05 -0800 (PST)
Subject: [C++-sig] Boost newbie help
In-Reply-To: <20040322074845.47550.qmail@web13421.mail.yahoo.com>
Message-ID: <20040322144805.85314.qmail@web20205.mail.yahoo.com>

I believe you just have to use boost::noncopyable.

http://www.boost.org/libs/python/doc/v2/class.html

To see many examples:

cd boost/libs/python/test
grep noncopyable *

Ralf

--- brett hartshorn <bhartsho at yahoo.com> wrote:
> I'm having a problem wrapping a class that looks like this:
> 
> class A {
>     A _myVar;
>     A (const A &);   //don't ask why this is not defined
> 
>   public:
>     A() { _myVar = myFunction(); }
> 
> }
> 
> It will generate errors when building that say:
> error: `A::A(const A&)' is private
> /usr/include/boost/python/object/value_holder.hpp:111: error: within this
> context
> 
> If i remove the strange constructor that does nothing, then it builds without
> any problems.
> Anybody have any ideas how to fix this, other than removing that constructor?


__________________________________
Do you Yahoo!?
Yahoo! Finance Tax Center - File online. File on time.
http://taxes.yahoo.com/filing.html



From rwgk at yahoo.com  Mon Mar 22 15:54:09 2004
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Mon, 22 Mar 2004 06:54:09 -0800 (PST)
Subject: [C++-sig] How to specify a double return type?
In-Reply-To: <20040322080053.31492.qmail@web13426.mail.yahoo.com>
Message-ID: <20040322145409.13312.qmail@web20211.mail.yahoo.com>

Is dReal a typedef for double or float? If so: Python float are immutable. You
cannot return pointers to immutable types. Somehow you have to change your
interface, e.g. through the use of "thin wrappers."
Please explain in more detail what exactly you want to do; hopefully this will
lead us to the best solution.
Ralf

--- brett hartshorn <bhartsho at yahoo.com> wrote:
> I have a type called dReal, that can either be a float or double number type,
> depending on the
> build options.  A member function in my class returns a dReal type, this is
> giving me problems in
> Boost.  I am still new to boost, so i am not sure what to do.  
> 
> Here is the error i get:
> /usr/include/boost/python/detail/invoke.hpp:93: error: no match for call to
> `(
>   
>
boost::python::detail::specify_a_return_value_policy_to_wrap_functions_returning<const
>    dReal*>) (const dReal*)'
> 
> I've tried to use:
>    return_internal_reference<>()
>    return_value_policy<manage_new_object>())
>    return_value_policy<reference_existing_object>()
> but to no avail.
> 
> Also i have member functions that take in a dReal number.  What is the best
> way to pass that
> function a dReal type?  Or can i just pass floats and doubles in python and
> boost will treat those
> as a dReal?


__________________________________
Do you Yahoo!?
Protect your identity with Yahoo! Mail AddressGuard
http://antispam.yahoo.com/whatsnewfree



From nico_ml at mgdesign.org  Mon Mar 22 16:02:45 2004
From: nico_ml at mgdesign.org (Nicolas Lelong)
Date: Mon, 22 Mar 2004 16:02:45 +0100
Subject: [C++-sig] Boost newbie help
References: <20040322074845.47550.qmail@web13421.mail.yahoo.com>
Message-ID: <00ca01c4101e$bc92ff30$de00a8c0@nico>

> If i remove the strange constructor that does nothing, then it builds
without any problems.
> Anybody have any ideas how to fix this, other than removing that
constructor?

use a declaration like this one :

class_<A, boost::noncopyable>("A");

HTH,

Nicolas.




From dave at boost-consulting.com  Tue Mar 23 05:17:46 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 22 Mar 2004 23:17:46 -0500
Subject: [C++-sig] Re: vector_indexing_suite and slicing
References: <1079468390.2134.55.camel@illuvatar>
	<40599F96.6020602@boost-consulting.com>
	<uptb6j7u1.fsf@yahoo.co.uk> <405E2CD1.2040703@boost-consulting.com>
Message-ID: <ullls9oyd.fsf@boost-consulting.com>

Joel de Guzman <joel at boost-consulting.com> writes:

> Raoul Gough wrote:
>
>> Joel de Guzman <joel at boost-consulting.com> writes:
>> 
>>>>To close the loop, the attached patch will fix the bugs I noted in
>>>>[3]
>>>>while following the guidance provided by the Python folks.  Unless there
>>>>are any objections, I'll commit this to HEAD and the RC_1_31_0 branch
>>>>tonight.
>
> [snip]
>
>>>No objections here. I wonder though about the status of Raoul's
>>>new indexing suite. I was under the impression that his work will
>>>supersede the current indexing suite. Raoul?
>> I'm pretty sure I mentioned my work in the same thread back in
>> January, and that the slice behaviour of the new indexing suite was
>> already Python compliant.
>> I haven't had time recently to get the documentation in order, but I
>> intend to get that done before the ACCU conference in mid-April. Once
>> the docco is ready, we could probably consider merging indexing_v2
>> onto the mainline, but I guess there's no harm in patching the
>> original indexing suite in the meantime.
>
> I agree. I look forward to the new indexing suite. I still owe
> you a thorough review ;-)

Since we're not near a release at this point, I'm happy to have it
checked into the main trunk once you have docs.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From bhartsho at yahoo.com  Tue Mar 23 06:58:09 2004
From: bhartsho at yahoo.com (brett hartshorn)
Date: Mon, 22 Mar 2004 21:58:09 -0800 (PST)
Subject: [C++-sig] Re: How to specify a double return type?
Message-ID: <20040323055809.17691.qmail@web13425.mail.yahoo.com>

Thanks Ralf and Nicolas for answering my first question.  boost::noncopyable fixed my problem!

Sadly, dReal is a typedef.  I have tried to make 'thin wrappers' but i am not sure if i'm doing it
right.  Here's an example, Foo is the class i am wrapping, Bar is my thin wrapper.

typedef double dReal;

class Foo {
  public:
    const dReal * getSomething() const {
      return otherFunction();
    }

};
class Bar : public Foo {
  public:
    const double * getSomething() const {
       return Foo::getSomething();
    }
};

// the wrapper

class_<Bar> bar("Bar");
 bar
   .def("getSomething", &Bar::getSomething, return_internal_reference<>())
   ;

The tail end of my errors when compiling:

instantiated from here
/usr/include/boost/python/object/make_instance.hpp:25: error: invalid
   application of `sizeof' to an incomplete type

-brett

>Is dReal a typedef for double or float? If so: Python float are immutable. You
>cannot return pointers to immutable types. Somehow you have to change your
>interface, e.g. through the use of "thin wrappers."
>Please explain in more detail what exactly you want to do; hopefully this will
>lead us to the best solution.
>Ralf

__________________________________
Do you Yahoo!?
Yahoo! Finance Tax Center - File online. File on time.
http://taxes.yahoo.com/filing.html



From rwgk at yahoo.com  Tue Mar 23 16:10:09 2004
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Tue, 23 Mar 2004 07:10:09 -0800 (PST)
Subject: [C++-sig] Re: How to specify a double return type?
In-Reply-To: <20040323055809.17691.qmail@web13425.mail.yahoo.com>
Message-ID: <20040323151009.93603.qmail@web20203.mail.yahoo.com>

--- brett hartshorn <bhartsho at yahoo.com> wrote:
> Thanks Ralf and Nicolas for answering my first question.  boost::noncopyable
> fixed my problem!
> 
> Sadly, dReal is a typedef.  I have tried to make 'thin wrappers' but i am not
> sure if i'm doing it
> right.  Here's an example, Foo is the class i am wrapping, Bar is my thin
> wrapper.
> 
> typedef double dReal;
> 
> class Foo {
>   public:
>     const dReal * getSomething() const {
>       return otherFunction();
>     }
> 
> };
> class Bar : public Foo {
>   public:
>     const double * getSomething() const {
>        return Foo::getSomething();
>     }
> };

You are making a "thick wrapper". And as far as Boost.Python is concerned there
is no difference between wrapping Foo::getSomething and Bar::getsomething.

I believe you have to do something like this (untested):

// a thin wrapper
double
Foo_getSomething(Foo const& foo_instance)
{
  return *foo_instance.getSomething(); // dereference the pointer
}

class_<Foo>("Foo")
  .def("getSomething", Foo_getSomething)
;

You can .def an unbound C++ function as a Python method. Cool?

In Python foo_instance.getSomething() returns the *current* value of your
dReal. If the value inside your Foo class changes you have to call
foo_instance.getSomething() again to see the change in Python. E.g. instead of
holding on to the const double* (as you might in C++) you have to hold on to
the foo_instance instead.

HTH,
        Ralf


__________________________________
Do you Yahoo!?
Yahoo! Finance Tax Center - File online. File on time.
http://taxes.yahoo.com/filing.html



From pierre.barbier at cirad.fr  Tue Mar 23 16:11:25 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Tue, 23 Mar 2004 16:11:25 +0100
Subject: [C++-sig] Exception not caught
Message-ID: <1080054685.917.52.camel@pbarbier>

Hello,

I've registred an exception :

struct ObjectCreationException
{ };

with the following function :


void obj_create_trans( const ObjectCreationException& err )
{
  PyErr_SetString( PyExc_UserWarning, "Error when creating object, see
standard error for more information" );
}

and in the module definition :

boost::python::register_exception_translator<ObjectCreationException>(&obj_create_trans);

But, when my C++ program throw this exception, the program just abort.
What can be the problem ?

Thanks,

Pierre

-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From dave at boost-consulting.com  Wed Mar 24 01:56:42 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Tue, 23 Mar 2004 19:56:42 -0500
Subject: [C++-sig] Re: Exception not caught
References: <1080054685.917.52.camel@pbarbier>
Message-ID: <u8yhr6p11.fsf@boost-consulting.com>

Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:

> But, when my C++ program throw this exception, the program just abort.
> What can be the problem ?

I suggest using a debugger to find out.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From tlovexyj at 21cn.com  Wed Mar 24 10:03:28 2004
From: tlovexyj at 21cn.com (=?gb2312?B?x6fA78LtuM4=?=)
Date: Wed, 24 Mar 2004 17:03:28 +0800
Subject: [C++-sig] Fw: PyErr_Print error track in c++
Message-ID: <VM959675829694.23546@send2.inner-21cn.com>

python-chinese at lists.python.cn, ???

	?????C++???????py????PyErr_Print???????
	???python???????PyObject_CallObject????module?????????
	???????module?????????????????????????????

	???py???

import sys, time
from cd2 import *

# ??????
class errCatcher:
    def __init__(self, filename):
        self.info = ''
        self.name = filename
        tmp = open(filename, 'w')
        tmp.close()
        
    def write(self, stuff):
        self.info += stuff

    def showmsg(self):
        f = open(self.name, 'a')
        f.write(time.ctime())
        f.write('\n--------------------------------------------------\n')
        f.write(self.info)
        f.write('\n\n')
        f.close()
        MsgBox(self.info)
        self.info = ''


# ??????
class outCatcher:
    def __init__(self, filename):
        self.name = filename
        tmp = open(filename, 'w')
        tmp.close()
        
    def write(self, stuff):
        f = open(self.name, 'a')
        f.write(stuff)
        f.close()

sys.stderr = errCatcher('python_err.log')
sys.stdout = outCatcher('python_out.log')




???C++??python??????
	
void RunFuncInModule(const Char *szFuncName, const Char *szModuleName)
{
	try
	{
		python::handle<> main_module( python::borrowed( PyImport_AddModule( "__main__" ) ) );
		python::handle<> main_namespace( python::borrowed( PyModule_GetDict( main_module.get() ) ) );

		python::handle<> local_module( python::borrowed( PyImport_ImportModule( const_cast<Char *>( szModuleName ) ) ) );

#ifdef _DEBUG
		python::handle<> ( python::borrowed( PyImport_ReloadModule( local_module.get() ) ) );
#endif

		python::handle<> local_namespace( python::borrowed( PyModule_GetDict( local_module.get() ) ) );
		
		PyObject *func = python::expect_non_null( PyDict_GetItemString( local_namespace.get(), const_cast<Char *>( szFuncName ) ) );
		if ( PyCallable_Check( func ) != 0 )
		{
			python::handle<>( PyObject_CallObject( func, NULL ) );
		}
	}
	catch (python::error_already_set)
	{
		python::handle_exception();
		PyErr_Print();
		PyRun_SimpleString("sys.stderr.showmsg()");
	}
}

  
?????????
??

  
????????????
????????tlovexyj at 21cn.com
??????????2004-03-24 

From pierre.barbier at cirad.fr  Wed Mar 24 10:10:07 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Wed, 24 Mar 2004 10:10:07 +0100
Subject: [C++-sig] Re: Exception not caught
In-Reply-To: <u8yhr6p11.fsf@boost-consulting.com>
References: <1080054685.917.52.camel@pbarbier>
	<u8yhr6p11.fsf@boost-consulting.com>
Message-ID: <1080119393.727.23.camel@pbarbier>

I found the problem : it's a threading pb :( I raise the exception in a
thread not being the python one. I'll try to find a way to raise the
python exception event if launched from the other thread (during the
execution of the code raising the exception, the python thread is
blocked on a lock waiting the the end of the processing).

Thanks,

Pierre

Le mer 24/03/2004 ? 01:56, David Abrahams a ?crit :
> Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:
> 
> > But, when my C++ program throw this exception, the program just abort.
> > What can be the problem ?
> 
> I suggest using a debugger to find out.
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From ndbecker2 at verizon.net  Thu Mar 25 01:19:38 2004
From: ndbecker2 at verizon.net (Neal Becker)
Date: Wed, 24 Mar 2004 19:19:38 -0500
Subject: [C++-sig] simple numarray question
Message-ID: <c3t8ir$a0r$1@sea.gmane.org>

I'm trying out numarray with boost::python.  numarray-0.9, python-2.3.3.
Any clues what's wrong with this simple test (I'm trying to print out
#elements in the array)

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

using namespace boost;
using namespace boost::python;

// sets the first element in a 1d numeric array
void set_first_element(numeric::array& y, double value)
{
  //  y[python::make_tuple(0,0)] = value;
  y[0] = value;
  std::clog << y.nelements() << std::endl;
}

BOOST_PYTHON_MODULE(Test3)
{
  def("f", set_first_element);
}

from numarray import *
>>> from Test3 import *
>>> x=array([1.2, 3.4])
>>> f(x, 2.8)
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
TypeError: No registered converter was able to produce a C++ rvalue of type
long from this Python object of type builtin_function_or_method




From tlovexyj at 21cn.com  Thu Mar 25 02:30:58 2004
From: tlovexyj at 21cn.com (=?utf-8?B?5Y2D6YeM6ams6IKd?=)
Date: Thu, 25 Mar 2004 09:30:58 +0800
Subject: [C++-sig] PyErr_Print error track in c++
Message-ID: <HH959674907369.26463@send1.inner-21cn.com>

hello, c++-sig at python.org

	I write own class to track PyErr_Print message for my c++ application.
	And I use PyObject_CallObject to call some function in some module.
	But, when some error in the module or the function of the module, who i use PyImport_ImportModule to import or call, it cannot report error message for me. :(

	What can be the problem ?

	below is track code:

import sys, time
from cd2 import *

# ??????
class errCatcher:
    def __init__(self, filename):
        self.info = ''
        self.name = filename
        tmp = open(filename, 'w')
        tmp.close()
        
    def write(self, stuff):
        self.info += stuff

    def showmsg(self):
        f = open(self.name, 'a')
        f.write(time.ctime())
        f.write('\n--------------------------------------------------\n')
        f.write(self.info)
        f.write('\n\n')
        f.close()
        MsgBox(self.info)
        self.info = ''


# ??????
class outCatcher:
    def __init__(self, filename):
        self.name = filename
        tmp = open(filename, 'w')
        tmp.close()
        
    def write(self, stuff):
        f = open(self.name, 'a')
        f.write(stuff)
        f.close()

sys.stderr = errCatcher('python_err.log')
sys.stdout = outCatcher('python_out.log')




c++ code embed python:
	
void RunFuncInModule(const Char *szFuncName, const Char *szModuleName)
{
	try
	{
		python::handle<> main_module( python::borrowed( PyImport_AddModule( "__main__" ) ) );
		python::handle<> main_namespace( python::borrowed( PyModule_GetDict( main_module.get() ) ) );

		python::handle<> local_module( python::borrowed( PyImport_ImportModule( const_cast<Char *>( szModuleName ) ) ) );

#ifdef _DEBUG
		python::handle<> ( python::borrowed( PyImport_ReloadModule( local_module.get() ) ) );
#endif

		python::handle<> local_namespace( python::borrowed( PyModule_GetDict( local_module.get() ) ) );
		
		PyObject *func = python::expect_non_null( PyDict_GetItemString( local_namespace.get(), const_cast<Char *>( szFuncName ) ) );
		if ( PyCallable_Check( func ) != 0 )
		{
			python::handle<>( PyObject_CallObject( func, NULL ) );
		}
	}
	catch (python::error_already_set)
	{
		python::handle_exception();
		PyErr_Print();
		PyRun_SimpleString("sys.stderr.showmsg()");
	}
}

??????????2004-03-24

From paul at paulgrenyer.co.uk  Thu Mar 25 14:53:05 2004
From: paul at paulgrenyer.co.uk (Paul Grenyer)
Date: Thu, 25 Mar 2004 13:53:05 GMT
Subject: [C++-sig] MSVC 7.1 Multithreaded Dll Memory Leak
Message-ID: <200403251353.i2PDr5d31317@ns.cricketthosting.co.uk>

Hi All

I've had a look at the FAQ and used google to look through the archives 
but I can't find anything about this problem or anything similar to it.

My basic problem is that when I use the Multithreaded Dll (both Debug 
and Release) runtime with boost python I get a memory leak. Even with 
the simplest of code. For example:

#include <boost/python.hpp>
#include <iostream>
#include <cstdlib>
#include <FindMemoryLeaks.h>

int main()
{
    Py_Initialize();
	
    {
        using namespace boost::python;
        handle<> main_module(borrowed( PyImport_AddModule
("__main__") ));
    }

    Py_Finalize();

    return EXIT_SUCCESS;
}

If I use the single threaded runtime or the multithreaded static 
runtime there is no memory leak.

I've checked and rebuilt boost python to ensure that it is built with 
the multithreaded dll runtime, but that didn't solve the problem. 
Therefore I suspect that the problem is with the python23.dll and the 
corresponding library?

Has anyone else come across this?

Regards
Paul

Paul Grenyer
Email: paul at paulgrenyer.co.uk
Web: http://www.paulgrenyer.co.uk

Have you met Aeryn: http://www.paulgrenyer.co.uk/aeryn/?
Version 0.3.0 beta now available for download.





From e at cyan.com  Thu Mar 25 18:35:25 2004
From: e at cyan.com (Eric Anderson)
Date: Thu, 25 Mar 2004 09:35:25 -0800
Subject: [C++-sig] MSVC 7.1 Multithreaded Dll Memory Leak
References: <200403251353.i2PDr5d31317@ns.cricketthosting.co.uk>
Message-ID: <0ddf01c4128f$90116960$bf01a8c0@ea1xp>


> My basic problem is that when I use the Multithreaded Dll (both Debug
> and Release) runtime with boost python I get a memory leak. Even with
> the simplest of code.

I also use MSVC 7.1 w/ Multithreaded DLL.  I seem to have the same problem.
The leak I'm seeing is at:

boost_1_30_0\boost\function\function_base.hpp(262)

I have not taken the time to track down the cause yet.

Eric





From gbiddison at velocity11.com  Thu Mar 25 19:33:30 2004
From: gbiddison at velocity11.com (Giles Biddison)
Date: Thu, 25 Mar 2004 10:33:30 -0800
Subject: [C++-sig] MSVC 7.1 Multithreaded Dll Memory Leak
Message-ID: <C0D87860D14664429FCBB0B2C349AD76742281@ilpostino.velocity11.com>


(apologies -- I'm not sure if top or bottom post is the norm for this
list)
I can verify this problem.  I've been reluctant to call it a memory
leak, because it only seems to occur after calling Py_Finalize().  I did
not go so far as to attempt to use anything other than the multithreaded
runtime dll.

Giles Biddison
gbiddison at velocity11.com

-----Original Message-----
Date: Thu, 25 Mar 2004 13:53:05 GMT
From: "Paul Grenyer" <paul at paulgrenyer.co.uk>
Subject: [C++-sig] MSVC 7.1 Multithreaded Dll Memory Leak
To: c++-sig at python.org
Message-ID: <200403251353.i2PDr5d31317 at ns.cricketthosting.co.uk>
Content-Type: text/plain; charset=iso-8859-1

Hi All

I've had a look at the FAQ and used google to look through the archives 
but I can't find anything about this problem or anything similar to it.

My basic problem is that when I use the Multithreaded Dll (both Debug 
and Release) runtime with boost python I get a memory leak. Even with 
the simplest of code. For example:

#include <boost/python.hpp>
#include <iostream>
#include <cstdlib>
#include <FindMemoryLeaks.h>

int main()
{
    Py_Initialize();
	
    {
        using namespace boost::python;
        handle<> main_module(borrowed( PyImport_AddModule
("__main__") ));
    }

    Py_Finalize();

    return EXIT_SUCCESS;
}

If I use the single threaded runtime or the multithreaded static 
runtime there is no memory leak.

I've checked and rebuilt boost python to ensure that it is built with 
the multithreaded dll runtime, but that didn't solve the problem. 
Therefore I suspect that the problem is with the python23.dll and the 
corresponding library?

Has anyone else come across this?

Regards
Paul

Paul Grenyer
Email: paul at paulgrenyer.co.uk
Web: http://www.paulgrenyer.co.uk

Have you met Aeryn: http://www.paulgrenyer.co.uk/aeryn/?
Version 0.3.0 beta now available for download.





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

_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig


End of C++-sig Digest, Vol 8, Issue 26
**************************************



From paul at paulgrenyer.co.uk  Thu Mar 25 19:37:23 2004
From: paul at paulgrenyer.co.uk (Paul Grenyer)
Date: Thu, 25 Mar 2004 18:37:23 -0000
Subject: [C++-sig] MSVC 7.1 Multithreaded Dll Memory Leak
References: <200403251353.i2PDr5d31317@ns.cricketthosting.co.uk>
	<0ddf01c4128f$90116960$bf01a8c0@ea1xp>
Message-ID: <003101c41298$36f79260$0302a8c0@paul>

Hi

> > My basic problem is that when I use the Multithreaded Dll (both Debug
> > and Release) runtime with boost python I get a memory leak. Even with
> > the simplest of code.
>
> I also use MSVC 7.1 w/ Multithreaded DLL.  I seem to have the same
problem.
> The leak I'm seeing is at:
>
> boost_1_30_0\boost\function\function_base.hpp(262)

Interesting. I'll take a look at that once I get back in the office
tomorrow.

Thanks!

Regards
Paul

Paul Grenyer
Email: paul at paulgrenyer.co.uk
Web: http://www.paulgrenyer.co.uk

Have you met Aeryn: http://www.paulgrenyer.co.uk/aeryn/?
Version 0.3.0 beta now ready for download




From paul at paulgrenyer.co.uk  Thu Mar 25 19:42:23 2004
From: paul at paulgrenyer.co.uk (Paul Grenyer)
Date: Thu, 25 Mar 2004 18:42:23 -0000
Subject: [C++-sig] MSVC 7.1 Multithreaded Dll Memory Leak
References: <C0D87860D14664429FCBB0B2C349AD76742281@ilpostino.velocity11.com>
Message-ID: <003a01c41298$e9b74e90$0302a8c0@paul>

Hi

> I can verify this problem.  I've been reluctant to call it a memory
> leak, because it only seems to occur after calling Py_Finalize().  I did
> not go so far as to attempt to use anything other than the multithreaded
> runtime dll.

That's interesting as well.  I don't think it does only occur after
Py_Finalize.

I first notice the problem during a test application that executes over 1.6
million python scripts one after the other. Py_Initialize and Py_Finalize
are only called once, at the beginning and end of the application,
respectively. The memory allocation is clearly rising all the time. As soon
as I switch to another runtime the leak goes away.

I am currently running all our unit tests with the static multithreaded
runtime. If they pass then we may well switch to that for the time being,
but I would prefer to fix the problem and keep using the dll multithreaded
runtime.

Regards
Paul

Paul Grenyer
Email: paul at paulgrenyer.co.uk
Web: http://www.paulgrenyer.co.uk

Have you met Aeryn: http://www.paulgrenyer.co.uk/aeryn/?
Version 0.3.0 beta now ready for download




From dave at boost-consulting.com  Thu Mar 25 23:08:41 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 25 Mar 2004 17:08:41 -0500
Subject: [C++-sig] Re: Fw: PyErr_Print error track in c++
References: <VM959675829694.23546@send2.inner-21cn.com>
Message-ID: <u65cs1swm.fsf@boost-consulting.com>

"????" <tlovexyj at 21cn.com> writes:

> python-chinese at lists.python.cn, ???
>
> 	?????C++???????py????PyErr_Print???????

I can't read the message; most of the characters show up for me as
empty rectangles.  I suspect that applies to most other participants
on the list.  Please re-post in plain-ASCII encoded English if you
want us to help.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Thu Mar 25 23:13:05 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 25 Mar 2004 17:13:05 -0500
Subject: [C++-sig] Re: simple numarray question
References: <c3t8ir$a0r$1@sea.gmane.org>
Message-ID: <uzna4ziby.fsf@boost-consulting.com>

Neal Becker <ndbecker2 at verizon.net> writes:

> I'm trying out numarray with boost::python.  numarray-0.9,
> python-2.3.3.

Which compiler?

> Any clues what's wrong with this simple test (I'm trying to print out
> #elements in the array)
>
> #include <boost/python.hpp>
> #include <boost/python/numeric.hpp>
> #include <boost/python/tuple.hpp>
> #include <iostream>
>
> using namespace boost;
> using namespace boost::python;
>
> // sets the first element in a 1d numeric array
> void set_first_element(numeric::array& y, double value)
> {
>   //  y[python::make_tuple(0,0)] = value;
>   y[0] = value;
>   std::clog << y.nelements() << std::endl;

I suggest pruning out parts of this function until you are able to
pinpoint the error.  For example:

         y.nelements();
then:
         y[0] = value;

then:
        y[0];

> }
>
> BOOST_PYTHON_MODULE(Test3)
> {
>   def("f", set_first_element);
> }
>
> from numarray import *
>>>> from Test3 import *
>>>> x=array([1.2, 3.4])
>>>> f(x, 2.8)
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
> TypeError: No registered converter was able to produce a C++ rvalue of type
> long from this Python object of type builtin_function_or_method

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Thu Mar 25 23:21:01 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 25 Mar 2004 17:21:01 -0500
Subject: [C++-sig] Re: PyErr_Print error track in c++
References: <HH959674907369.26463@send1.inner-21cn.com>
Message-ID: <uu10czhyq.fsf@boost-consulting.com>

".??.?.??.?.?.??" <tlovexyj at 21cn.com> writes:

> hello, c++-sig at python.org
>
> 	I write own class to track PyErr_Print message for my c++ application.
> 	And I use PyObject_CallObject to call some function in some
> 	module.

Usually it's better to use the function call operator of
python::object.

> 	But, when some error in the module or the function of the module,
> 	who i use PyImport_ImportModule to import or call, it cannot
> 	report error message for me. :(

What specific symptom are you seeing, and which specific line of your
sample code causes that symptom.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From ndbecker2 at verizon.net  Fri Mar 26 04:07:09 2004
From: ndbecker2 at verizon.net (Neal Becker)
Date: Thu, 25 Mar 2004 22:07:09 -0500
Subject: [C++-sig] Re: simple numarray question
References: <c3t8ir$a0r$1@sea.gmane.org> <uzna4ziby.fsf@boost-consulting.com>
Message-ID: <c406ou$q7o$2@sea.gmane.org>

David Abrahams wrote:

> Neal Becker <ndbecker2 at verizon.net> writes:
> 
>> I'm trying out numarray with boost::python.  numarray-0.9,
>> python-2.3.3.
> 
> Which compiler?
> 

gcc version 3.3.3 20040311 (Red Hat Linux 3.3.3-3)
boost_1_31_0

>> Any clues what's wrong with this simple test (I'm trying to print out
>> #elements in the array)
>>
>> #include <boost/python.hpp>
>> #include <boost/python/numeric.hpp>
>> #include <boost/python/tuple.hpp>
>> #include <iostream>
>>
>> using namespace boost;
>> using namespace boost::python;
>>
>> // sets the first element in a 1d numeric array
>> void set_first_element(numeric::array& y, double value)
>> {
>>   //  y[python::make_tuple(0,0)] = value;
>>   y[0] = value;
>>   std::clog << y.nelements() << std::endl;
> 
> I suggest pruning out parts of this function until you are able to
> pinpoint the error.  For example:
> 
>          y.nelements();
> then:
>          y[0] = value;
> 
> then:
>         y[0];
> 
>> }
>>
>> BOOST_PYTHON_MODULE(Test3)
>> {
>>   def("f", set_first_element);
>> }
>>
>> from numarray import *
>>>>> from Test3 import *
>>>>> x=array([1.2, 3.4])
>>>>> f(x, 2.8)
>> Traceback (most recent call last):
>>   File "<stdin>", line 1, in ?
>> TypeError: No registered converter was able to produce a C++ rvalue of
>> type long from this Python object of type builtin_function_or_method
> 

So I guess you're implying that as far as you can see, this is OK?  I have
approximately zero experience with boost-python so far, so I need to know
if it's my code or my environment that is the problem.





From bhartsho at yahoo.com  Fri Mar 26 07:24:34 2004
From: bhartsho at yahoo.com (brett hartshorn)
Date: Thu, 25 Mar 2004 22:24:34 -0800 (PST)
Subject: [C++-sig] boost python wrapper parser bug?
Message-ID: <20040326062434.63236.qmail@web13425.mail.yahoo.com>

I think i found a bug in boost python.  My bug sounds very much like this one:
http://mail.python.org/pipermail/c++-sig/2002-December/003125.html

I am wrapping a series of classes, they are all very much the same, they all inherit from the same
base class, they all have protected deconstrutors.  I am wrapping like:
class_<myclass, boost::noncopyable>foo("myclass", no_init)
and all is working great, except for this one class called Extensions.  It appears there is
nothing very different about this class, but for some reason boost does not like - hates in fact -
its protected deconstructor.  I've even talked to the author of the class and here is what he had
to say:

"""What I find strange is that only osgGL2::Extensions::~Extensions() 
experiences this problem.  I wonder what makes it unique?  Can you 
confirm that e.g. osg::BlendColor::Extensions::~Extensions() (or any of 
the others mentioned below) does wrap successfully (since it is also a 
protected dtor)?  Perhaps it has something to do with the levels of 
nesting of namespaces?

However, the code is valid c++ and successfully compiles and runs on 
multiple platforms.  So I still believe it's a bug in the boost wrapper 
parser, tho I'm not in a position to create a definitive testcase.  You 
might want to mention to the boost folks.

Hmm, looking closer at the original error message, something else strange:
 >>>osgGL2/Extensions.cpp:37:   instantiated from here
 >>>/usr/local/include/osgGL2/Extensions:178: error: `virtual
 >>>   osgGL2::Extensions::~Extensions()' is protected

That dtor is not virtual, I don't know why the error message claims it is?

Cheers
-- mew"""

if anyone wants to see the code in question you have to download it the PyOSG wrappers from here:
http://opart.org/pyosg/
the wrapper file you want to look at is in pyosg_devel/osgGL2/Extensions.cpp
the class header it wraps is in the OSG package under include/osgGL2/Extensions

I don't think the error is on my side in the wrapper code, but if it is found to be, i applogize
in advance for this post.

-brett


__________________________________
Do you Yahoo!?
Yahoo! Finance Tax Center - File online. File on time.
http://taxes.yahoo.com/filing.html



From paul at paulgrenyer.co.uk  Fri Mar 26 09:55:15 2004
From: paul at paulgrenyer.co.uk (Paul Grenyer)
Date: Fri, 26 Mar 2004 08:55:15 GMT
Subject: [C++-sig] MSVC 7.1 Multithreaded Dll Memory Leak [long post]
Message-ID: <200403260855.i2Q8tFJ20451@ns.cricketthosting.co.uk>

Hi

> I am currently running all our unit tests with the static 
> multithreaded runtime. If they pass then we may well switch to that 
> for the time being, but I would prefer to fix the problem and keep 
> using the dll multithreaded runtime.

I'm not sure how relevant to most people this post will be, but for 
those of you that are interested....

All of our C++ unit tests, including those that use the embedded Python 
interpretter ran successfully when linked against the static runtime, 
with one exception (pun intended) where an exception was thrown due to 
an invalid string position. That should be an easy fix and this 
exercise has hopefully thrown up a bug I can fix.

However, all of our boost python extensions (I didn't write the stuff 
and I'm not sure of the right terminology. I mean C++ code wrapped in 
boost python to expose it to python) failed with unknown C++ errors or 
something similar. I suspect this is due to the fact that they are Dlls 
and therefore presumably need the Dll runtime.

Ah! I thought, build everything with the static runtime except the 
boost python extensions. No! The boost pythong extensions use other 
libraries in our system that have already been built with the static 
runtime, so they fail to link.

So, I'm stuck with either having to do two different builds to build 
our whole system (not nice. espcially when you have one large project 
capable of building and running all tests) or try to get the memory 
leak fixed.

Dave, is this memory leak a recognised problem? Is anything being done 
to fix it? As the problem is in function.hpp is it worth posting about 
it to the main boost list?

Thanks!



Regards
Paul

Paul Grenyer
Email: paul at paulgrenyer.co.uk
Web: http://www.paulgrenyer.co.uk

Have you met Aeryn: http://www.paulgrenyer.co.uk/aeryn/?
Version 0.3.0 beta now available for download.





From benny at kramekweb.com  Fri Mar 26 13:26:33 2004
From: benny at kramekweb.com (Benny Kramek)
Date: Fri, 26 Mar 2004 12:26:33 +0000
Subject: [C++-sig] Boost::Signals - Python functions as slots
Message-ID: <20040326122633.19442.qmail@hosting335.com>

I'm using Boost::Signals in C++. I want to be able to attach python
functions as slots to the C++ signals. Here is a very simple test that
doesn't work. Everything compiles OK, but I get a Python error.

### C++ class

class Button 
{
    public:
        void doOnClick(const boost::signal<void ()>::slot_type& slot);

    private:
        boost::signal<void ()> onClick;
};

void Button::doOnClick(const boost::signal<void ()>::slot_type& slot)
{
    onClick.connect(slot);
}

### Boost::Python Module

class_<Button>("Button")
    .def("doOnClick", &Button::doOnClick)
    ;

### Python test

def f():
    print 'f'

button = Button()
button.doOnClick(f)

### Python error

    button.doOnClick(f)
Boost.Python.ArgumentError: Python argument types in
    Button.doOnClick(Button, function)
did not match C++ signature:
    doOnClick(Button {lvalue}, N5boost4slotINS_8functionIFvvESaIvEEEEE)

The demangled name
N5boost4slotINS_8functionIFvvESaIvEEEEE
is
boost::slot<boost::function<void ()(), std::allocator<void> > >

I'm using gcc 3.2.3 and boost 1.31.0

Thanks for any help,
Benny

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20040326/5b0f3671/attachment.htm>

From raoulgough at yahoo.co.uk  Fri Mar 26 13:39:45 2004
From: raoulgough at yahoo.co.uk (=?iso-8859-1?q?Raoul=20Gough?=)
Date: Fri, 26 Mar 2004 12:39:45 +0000 (GMT)
Subject: [C++-sig] Re: MSVC 7.1 Multithreaded Dll Memory Leak [long post]
Message-ID: <20040326123945.63427.qmail@web25208.mail.ukl.yahoo.com>

Paul Grenyer paul at paulgrenyer.co.uk 
Fri Mar 26 03:55:15 EST 2004 :

[snippage]
>However, all of our boost python extensions (I didn't write the stuff 
>and I'm not sure of the right terminology. I mean C++ code wrapped in 
>boost python to expose it to python) failed with unknown C++ errors or

>something similar. I suspect this is due to the fact that they are
>Dlls and therefore presumably need the Dll runtime.

I've never heard of an "unknown C++ error" - how about posting some
more information (compiler, build options, actual error messages :-)

>Dave, is this memory leak a recognised problem? Is anything being done

>to fix it? As the problem is in function.hpp is it worth posting about

>it to the main boost list?

Not sure how this relates to multi-threaded builds, but there is a
known leak related to Py_Finalize and the boost Python library:

http://mail.python.org/pipermail/c++-sig/2003-December/006511.html

"The problem is that Boost.Python hangs on to
some Python references even after Py_Finalize is called.  When the
application exits, those reference counts are decremented, even
though there's no interpreter running."

Regards,
Raoul Gough.


	
	
		
___________________________________________________________
Yahoo! Messenger - Communicate instantly..."Ping" 
your friends today! Download Messenger Now 
http://uk.messenger.yahoo.com/download/index.html



From paul at paulgrenyer.co.uk  Fri Mar 26 13:54:20 2004
From: paul at paulgrenyer.co.uk (Paul Grenyer)
Date: Fri, 26 Mar 2004 12:54:20 GMT
Subject: [C++-sig] Re: MSVC 7.1 Multithreaded Dll Memory Leak [long post]
Message-ID: <200403261254.i2QCsKS01873@ns.cricketthosting.co.uk>

Hi

> >something similar. I suspect this is due to the fact that they are
> >Dlls and therefore presumably need the Dll runtime.
> 
> I've never heard of an "unknown C++ error" - how about posting some
> more information (compiler, build options, actual error messages :-)


It was just a passing comment, I wasn't asking for help. If I had been 
I would have posted more info. :-)

However, now that you've taken an interest, I'll get some more details. 
I had just assumed that Dlls probably needed the dll runtime. Is that 
right, do you know?


> Not sure how this relates to multi-threaded builds, but there is a
> known leak related to Py_Finalize and the boost Python library:
> 
> http://mail.python.org/pipermail/c++-sig/2003-December/006511.html
> 
> "The problem is that Boost.Python hangs on to
> some Python references even after Py_Finalize is called.  When the
> application exits, those reference counts are decremented, even
> though there's no interpreter running."

The problem I am seeing (as I said in a previous post :-) ) is while 
the app is running. Changing the runtime (as I have also said) fixed 
the problem. 

That said, using the static multithreaded runtime appeared to solve the 
problem described in the email above.

Regards
Paul

Paul Grenyer
Email: paul at paulgrenyer.co.uk
Web: http://www.paulgrenyer.co.uk

Have you met Aeryn: http://www.paulgrenyer.co.uk/aeryn/?
Version 0.3.0 beta now available for download.





From raoulgough at yahoo.co.uk  Fri Mar 26 17:41:29 2004
From: raoulgough at yahoo.co.uk (=?iso-8859-1?q?Raoul=20Gough?=)
Date: Fri, 26 Mar 2004 16:41:29 +0000 (GMT)
Subject: [C++-sig] Re: Boost::Signals - Python functions as slots
Message-ID: <20040326164129.96818.qmail@web25205.mail.ukl.yahoo.com>

Benny Kramek benny at kramekweb.com 
Fri Mar 26 07:26:33 EST 2004 

>I'm using Boost::Signals in C++. I want to be able to attach python
>functions as slots to the C++ signals.
> Here is a very simple test that
>doesn't work. Everything compiles OK, but I get a Python error.

[snip]
>void Button::doOnClick(const boost::signal<void ()>::slot_type& slot)
>{
>    onClick.connect(slot);
>}
>
>### Python test
>
>def f():
>    print 'f'
>
>button = Button()
>button.doOnClick(f)
>
>### Python error
>
>    button.doOnClick(f)
>Boost.Python.ArgumentError: Python argument types in
>    Button.doOnClick(Button, function)
>did not match C++ signature:
>    doOnClick(Button {lvalue},
N5boost4slotINS_8functionIFvvESaIvEEEEE)
>
>The demangled name
>N5boost4slotINS_8functionIFvvESaIvEEEEE
>is
>boost::slot<boost::function<void ()(), std::allocator<void> > >
>
>I'm using gcc 3.2.3 and boost 1.31.0

When you pass the Python function f into a C++ routine I guess it is
represented as a boost::python::function object. You can't pass a
boost::python:object where a boost::slot<...> is expected, which is why
you are getting the (run-time) error.

I think you will have to modify your doOnClick member function to
accept a boost::python::function (or maybe boost::python::object)
instead, and use this object to handle the signal. I assume that the
slots/signals code can handle functors (i.e. objects with an
operator()). I'm not sure if you will able to use the Python object
directly for this, you might have to write a thin wrapper like this:

struct python_function {
  boost::python::object my_callable_object;

  python_function (boost::python::object const &obj)
    : my_callable_object (obj)
  {
  }

  void operator()() {
    my_callable_object();
  }
};

Untested code! I'd be more definite about this, but I don't have access
to the Boost.Python code at the moment.

Regards,
Raoul Gough.


	
	
		
___________________________________________________________
Yahoo! Messenger - Communicate instantly..."Ping" 
your friends today! Download Messenger Now 
http://uk.messenger.yahoo.com/download/index.html



From rwgk at yahoo.com  Fri Mar 26 19:16:42 2004
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Fri, 26 Mar 2004 10:16:42 -0800 (PST)
Subject: [C++-sig] boost python wrapper parser bug?
In-Reply-To: <20040326062434.63236.qmail@web13425.mail.yahoo.com>
Message-ID: <20040326181642.41190.qmail@web20205.mail.yahoo.com>

--- brett hartshorn <bhartsho at yahoo.com> wrote:
> However, the code is valid c++ and successfully compiles and runs on 
> multiple platforms.  So I still believe it's a bug in the boost wrapper 
> parser, tho I'm not in a position to create a definitive testcase.  You 
> might want to mention to the boost folks.

"boost wrapper parser" == your very C++ compiler. There is nothing else.

> Hmm, looking closer at the original error message, something else strange:
>  >>>osgGL2/Extensions.cpp:37:   instantiated from here
>  >>>/usr/local/include/osgGL2/Extensions:178: error: `virtual
>  >>>   osgGL2::Extensions::~Extensions()' is protected
> 
> That dtor is not virtual, I don't know why the error message claims it is?

osgGL2::Extensions inherits from osg::Referenced which has a virtual
destructor.

> if anyone wants to see the code in question you have to download it the PyOSG
> wrappers from here:
> http://opart.org/pyosg/
> the wrapper file you want to look at is in pyosg_devel/osgGL2/Extensions.cpp
> the class header it wraps is in the OSG package under
> include/osgGL2/Extensions

Here is what I would do:

Make a full copy of the complete source code tree.

Minimize the Boost.Python bindings (without the error going away). It seems you
have done this already because you say that class_<...> alone triggers the
error.

Start stripping down the code in osgGL2/Extensions until you are left with a
minimal example that still triggers the error.

Ralf


__________________________________
Do you Yahoo!?
Yahoo! Finance Tax Center - File online. File on time.
http://taxes.yahoo.com/filing.html



From dave at boost-consulting.com  Sun Mar 28 00:04:12 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Sat, 27 Mar 2004 18:04:12 -0500
Subject: [C++-sig] Re: simple numarray question
References: <c3t8ir$a0r$1@sea.gmane.org> <uzna4ziby.fsf@boost-consulting.com>
	<c406ou$q7o$2@sea.gmane.org>
Message-ID: <ud66xx577.fsf@boost-consulting.com>

Neal Becker <ndbecker2 at verizon.net> writes:

>>> from numarray import *
>>>>>> from Test3 import *
>>>>>> x=array([1.2, 3.4])
>>>>>> f(x, 2.8)
>>> Traceback (most recent call last):
>>>   File "<stdin>", line 1, in ?
>>> TypeError: No registered converter was able to produce a C++ rvalue of
>>> type long from this Python object of type builtin_function_or_method
>> 
>
> So I guess you're implying that as far as you can see, this is OK?  

Yeah, but that doesn't mean much.

> I have approximately zero experience with boost-python so far, so I
> need to know if it's my code or my environment that is the problem.

I don't have a clue yet.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Sun Mar 28 00:15:33 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Sat, 27 Mar 2004 18:15:33 -0500
Subject: [C++-sig] Re: Boost::Signals - Python functions as slots
References: <20040326164129.96818.qmail@web25205.mail.ukl.yahoo.com>
Message-ID: <u7jx5x4oa.fsf@boost-consulting.com>

Raoul Gough <raoulgough at yahoo.co.uk> writes:

> I think you will have to modify your doOnClick member function to
> accept a boost::python::function (or maybe boost::python::object)
> instead, and use this object to handle the signal. I assume that the
> slots/signals code can handle functors (i.e. objects with an
> operator()). I'm not sure if you will able to use the Python object
> directly for this, you might have to write a thin wrapper like this:
>
> struct python_function {
>   boost::python::object my_callable_object;
>
>   python_function (boost::python::object const &obj)
>     : my_callable_object (obj)
>   {
>   }
>
>   void operator()() {
>     my_callable_object();
>   }
> };
>
> Untested code! I'd be more definite about this, but I don't have access
> to the Boost.Python code at the moment.

     boost::python::object instances are callable directly as
     functions, and therefore there's an implicit conversion from
     boost::python::object instances to the boost::function instances
     used by Boost.Signals.  There's probably an easy way to take
     advantage of that fact to avoid writing so much extra
     code... but I'm speculating.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Sun Mar 28 07:23:44 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Sun, 28 Mar 2004 00:23:44 -0500
Subject: [C++-sig] Re: MSVC 7.1 Multithreaded Dll Memory Leak [long post]
References: <200403260855.i2Q8tFJ20451@ns.cricketthosting.co.uk>
Message-ID: <usmfttuhr.fsf@boost-consulting.com>

"Paul Grenyer" <paul at paulgrenyer.co.uk> writes:

> Hi
>
>> I am currently running all our unit tests with the static 
>> multithreaded runtime. If they pass then we may well switch to that 
>> for the time being, but I would prefer to fix the problem and keep 
>> using the dll multithreaded runtime.
>
> I'm not sure how relevant to most people this post will be, but for 
> those of you that are interested....
>
> All of our C++ unit tests, including those that use the embedded Python 
> interpretter ran successfully when linked against the static runtime, 
> with one exception (pun intended) where an exception was thrown due to 
> an invalid string position. That should be an easy fix and this 
> exercise has hopefully thrown up a bug I can fix.
>
> However, all of our boost python extensions (I didn't write the stuff 
> and I'm not sure of the right terminology. I mean C++ code wrapped in 
> boost python to expose it to python) failed with unknown C++ errors or 
> something similar. 

"Unknown C++ Error" is a pretty vague term.  Almost anything could be
considered to be similar to that.

> I suspect this is due to the fact that they are Dlls 
> and therefore presumably need the Dll runtime.

DLLs don't require the dynamic runtime library, but extensions linked
to Python might require it.

> Ah! I thought, build everything with the static runtime except the 
> boost python extensions. No! The boost pythong extensions use other

 "static" here
 
> libraries in our system that have already been built with the static 
> runtime, so they fail to link.

??

> So, I'm stuck with either having to do two different builds to build 
> our whole system (not nice. espcially when you have one large project 
> capable of building and running all tests) or try to get the memory 
> leak fixed.
>
> Dave, is this memory leak a recognised problem? 

I recognize that some people are reporting something, but I don't know
anything more about it.

> Is anything being done to fix it? 

Not by me.  I don't have a clue where it is.

> As the problem is in function.hpp

I think it's probably too early to draw that conclusion, but...

> is it worth posting about it to the main boost list?

Sure; first I'd try to reproduce the problem with a minimal example.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Sun Mar 28 07:33:16 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Sun, 28 Mar 2004 00:33:16 -0500
Subject: [C++-sig] Re: Automation of Python object retrieval
References: <1078321816.742.164.camel@pbarbier>
Message-ID: <un061tu1v.fsf@boost-consulting.com>

Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:

> I have a problem with the boost::python::object contruction. I'd want to
> know if there is a way to automate the retrieval of the existing Python
> object when I hold a Pointer on the C++ base class and that the class is
> extended in Python. Here's an example of what I do now :
>

<snip big example with no mention of boost::python::object>


> ======================
>
> All the equality tests fails because the other in detected as a Base
> object and not a Derived object !!!

detected where?

> But, if I replace the BaseWrap::operator== by :

<snip some workaround>

> I get the wanted behaviour. But I wonder if there is a cleaner method to
> do that ? 

Sorry, I stared at your example for 5 minutes and couldn't figure out
what you were after.  Is there a reason that Holder can't hold a
shared_ptr<Base> or a boost::python::object?

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From pierre.barbier at cirad.fr  Sun Mar 28 11:22:42 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Sun, 28 Mar 2004 11:22:42 +0200
Subject: [C++-sig] Re: Automation of Python object retrieval
In-Reply-To: <un061tu1v.fsf@boost-consulting.com>
References: <1078321816.742.164.camel@pbarbier>
	<un061tu1v.fsf@boost-consulting.com>
Message-ID: <1080465761.1017.12.camel@pbarbier>

Sorry if I wasn't clear, but the problem is quite complex from my point
of view and I have problems to express it clearly.

Le dim 28/03/2004 ? 07:33, David Abrahams a ?crit :
> Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:
> 
> > I have a problem with the boost::python::object contruction. I'd want to
> > know if there is a way to automate the retrieval of the existing Python
> > object when I hold a Pointer on the C++ base class and that the class is
> > extended in Python. Here's an example of what I do now :
> >
> 
> <snip big example with no mention of boost::python::object>
> 
> 
> > ======================
> >
> > All the equality tests fails because the other in detected as a Base
> > object and not a Derived object !!!
> 
> detected where?

When a function returns a base class pointer, there is a mechanism to
detect the dynamic type of the pointer (whether it's the base class or
any Derived class exported in Python). But if you just construct a
boost::python::object from a pointer, the Python type of the
boost::python::object will be the type of the pointer (whatever the
dynamic type of the object could be).

> 
> > But, if I replace the BaseWrap::operator== by :
> 
> <snip some workaround>
> 
> > I get the wanted behaviour. But I wonder if there is a cleaner method to
> > do that ? 
> 
> Sorry, I stared at your example for 5 minutes and couldn't figure out
> what you were after.  Is there a reason that Holder can't hold a
> shared_ptr<Base> or a boost::python::object?

No, there's no reasons. In the code I use, I export the classes with
smart pointers ... but it does not resolve anything. My goal is the
following:

I have an object pointed by a Base pointer. But the object's dynamic
type could be (and in practive it WILL be) a class derived from Base and
most probably the derivation occured in Python. What I want is :

 - if the dynamic type of the object is a Python class derived from Base
: construct the boost::python::object that holds the PyObject* given to
the class wrapper when constructed
 - if the dynamic type of the object is a C++ class derived from Base
and exported in Python using Boost : construct the boost::python::object
referencing the dynamic type of the object

The method I wrote do that, but it's not very clean. I hope I'm clear
enough here too :-/

Thanks,
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From paul at paulgrenyer.co.uk  Sun Mar 28 12:32:18 2004
From: paul at paulgrenyer.co.uk (Paul Grenyer)
Date: Sun, 28 Mar 2004 11:32:18 +0100
Subject: [C++-sig] Re: MSVC 7.1 Multithreaded Dll Memory Leak [long post]
References: <200403260855.i2Q8tFJ20451@ns.cricketthosting.co.uk>
	<usmfttuhr.fsf@boost-consulting.com>
Message-ID: <00e001c414af$f277b420$0302a8c0@paul>

Hi

> "Unknown C++ Error" is a pretty vague term.  Almost anything could be
> considered to be similar to that.

Absolutely. As I've already mentioned, I was not asking for help on this
point. I left some test running on Friday when I left work so that I can
post the exact errors tomorrow.

>  > I suspect this is due to the fact that they are Dlls
> > and therefore presumably need the Dll runtime.
>
> DLLs don't require the dynamic runtime library, but extensions linked
> to Python might require it.

Right.

>  > Ah! I thought, build everything with the static runtime except the
> > boost python extensions. No! The boost pythong extensions use other
>
>  "static" here

Yes.

>   > libraries in our system that have already been built with the static
> > runtime, so they fail to link.
>
> ??

Ok, without going into our system in too much depth. Our main "application"
is an ATL COM server which hosts a number of COM objects. The COM wrappers
are completely seperated from the "back-end processing" that they provide.
Each peice of backend processing is contained in a static library (this also
allows us to test each piece in isolation).

We're using boost.python to make the back-end processing available to anyone
wanting to write a python script and take advantage of the pieces. Therefore
the static libs must be linked into both the ATL COM server and the
boost.python extension DLL. This can't be done in a single project if the
server and the dll are using different runtimes.

> > Dave, is this memory leak a recognised problem?
>
> I recognize that some people are reporting something, but I don't know
> anything more about it.

Ok.

>  > Is anything being done to fix it?
>
> Not by me.  I don't have a clue where it is.

Fair enough. :-)

> > is it worth posting about it to the main boost list?
>
> Sure; first I'd try to reproduce the problem with a minimal example.

Well, as I think I put in my original post, the memory leak occurs even when
a program contains only a handle declaration. However, I don't know if this
is the post-Py_Finalize memory leak or the growing memory leak I have been
experiencing, but it is present in the dll runtime and not in the static
runtime.

Regards
Paul

Paul Grenyer
Email: paul at paulgrenyer.co.uk
Web: http://www.paulgrenyer.co.uk

Have you met Aeryn: http://www.paulgrenyer.co.uk/aeryn/?
Version 0.3.0 beta now ready for download




From RaoulGough at yahoo.co.uk  Sun Mar 28 18:52:48 2004
From: RaoulGough at yahoo.co.uk (Raoul Gough)
Date: Sun, 28 Mar 2004 17:52:48 +0100
Subject: [C++-sig] Injecting python code to a class
Message-ID: <uisgosylb.fsf@yahoo.co.uk>

I'd like to add the __repr__ and __str__ methods to container classes
(in the indexing suite) and the easiest way to implement them is in
Python, e.g.

def list_repr(container):
    return repr([x for x in container])

It is possible to inject this into an existing Python class
(e.g. "container") as follows:

container.__repr__ = new.instancemethod (list_repr, None, container)

I've been trying to do something like this from C++ code, but I don't
have much experience with embedding Python code in C++, and can't see
the easy way to do it. I've tried the following:

  class_< std::vector<int> > vector_int ("vector_int");

  vector_int.def (indexing::container_suite< std::vector<int> >());

  handle<> lambda_handle
    (PyRun_String ("lambda v: repr([x for x in v])"
                   , Py_single_input, object().ptr(), object().ptr()));

  objects::add_to_namespace (vector_int, "__repr__", object(lambda_handle));

i.e. passing None for the global and local dicts to PyRun_String, but
this produces the following error during module initialization
(wrapped for legibility):

>>> from example import vector_int
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
SystemError: C:\Code\22\Objects\frameobject.c:235:
   bad argument to internal function


Is there an easy way to get at the right global and local dicts for
the PyRun_String call? I'm not sure that this would work even if I
could pass the right parameters there, so is there a better way
altogether to do this?

-- 
Raoul Gough.
export LESS='-X'




From rwgk at yahoo.com  Sun Mar 28 19:27:09 2004
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Sun, 28 Mar 2004 09:27:09 -0800 (PST)
Subject: [C++-sig] Injecting python code to a class
In-Reply-To: <uisgosylb.fsf@yahoo.co.uk>
Message-ID: <20040328172709.19754.qmail@web20210.mail.yahoo.com>

Hi from Cambridge, UK :)

This might not be what you are after, but anyway:

I inject methods from Python all the time using a "boost.python.injector"
similar to the one mentioned in the Boost.Python tutorial. For example:

http://cvs.sourceforge.net/viewcvs.py/cctbx/scitbx/scitbx/math/gaussian.py?rev=1.7&view=auto

The injector is defined here:

http://cvs.sourceforge.net/viewcvs.py/cctbx/boost_adaptbx/boost/python.py?rev=1.1&view=auto

All my extension module imports go through a Python layer to work around
exception handling problems with certain versions of gcc, and because I add all
kinds of stuff from Python anyway. Therefore the injector doesn't add any extra
complications. If you don't want the Python layer I'd write the extra functions
in C++ using the boost::python::object facilities. Just loop over your array in
C++ and append str(element) to a boost::python::list.

BTW: I believe an array __repr__ that spits out the entire array is a very bad
idea. If the user bothers to make a C++ array instead of simply using a Python
list the C++ array is probably big, and the __repr__ output will be a nuisance
because of its size. Seeing the default __repr__ output is more helpful. If the
user really wants the content, print list(array_instance) will do the trick.
If the wrapped arrays are small it is IMO much more efficient (compile-time and
object code size) to use the "tuple_mappings". See FAQ, mention of scitbx in
connection with array wrappers.

Ralf

--- Raoul Gough <RaoulGough at yahoo.co.uk> wrote:
> I'd like to add the __repr__ and __str__ methods to container classes
> (in the indexing suite) and the easiest way to implement them is in
> Python, e.g.
> 
> def list_repr(container):
>     return repr([x for x in container])
> 
> It is possible to inject this into an existing Python class
> (e.g. "container") as follows:
> 
> container.__repr__ = new.instancemethod (list_repr, None, container)
> 
> I've been trying to do something like this from C++ code, but I don't
> have much experience with embedding Python code in C++, and can't see
> the easy way to do it. I've tried the following:
> 
>   class_< std::vector<int> > vector_int ("vector_int");
> 
>   vector_int.def (indexing::container_suite< std::vector<int> >());
> 
>   handle<> lambda_handle
>     (PyRun_String ("lambda v: repr([x for x in v])"
>                    , Py_single_input, object().ptr(), object().ptr()));
> 
>   objects::add_to_namespace (vector_int, "__repr__", object(lambda_handle));
> 
> i.e. passing None for the global and local dicts to PyRun_String, but
> this produces the following error during module initialization
> (wrapped for legibility):
> 
> >>> from example import vector_int
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
> SystemError: C:\Code\22\Objects\frameobject.c:235:
>    bad argument to internal function
> 
> 
> Is there an easy way to get at the right global and local dicts for
> the PyRun_String call? I'm not sure that this would work even if I
> could pass the right parameters there, so is there a better way
> altogether to do this?


__________________________________
Do you Yahoo!?
Yahoo! Finance Tax Center - File online. File on time.
http://taxes.yahoo.com/filing.html



From RaoulGough at yahoo.co.uk  Mon Mar 29 00:36:47 2004
From: RaoulGough at yahoo.co.uk (Raoul Gough)
Date: Sun, 28 Mar 2004 23:36:47 +0100
Subject: [C++-sig] Re: Injecting python code to a class
References: <uisgosylb.fsf@yahoo.co.uk>
	<20040328172709.19754.qmail@web20210.mail.yahoo.com>
Message-ID: <uekrcsio0.fsf@yahoo.co.uk>

"Ralf W. Grosse-Kunstleve" <rwgk at yahoo.com> writes:

> Hi from Cambridge, UK :)

Well, hi from South-East London!

>
> This might not be what you are after, but anyway:
>
> I inject methods from Python all the time using a
> "boost.python.injector" similar to the one mentioned in the
> Boost.Python tutorial. For example:
>
> http://cvs.sourceforge.net/viewcvs.py/cctbx/scitbx/scitbx/math/gaussian.py?rev=1.7&view=auto
>
> The injector is defined here:
>
> http://cvs.sourceforge.net/viewcvs.py/cctbx/boost_adaptbx/boost/python.py?rev=1.1&view=auto
>
> All my extension module imports go through a Python layer to work
> around exception handling problems with certain versions of gcc, and
> because I add all kinds of stuff from Python anyway. Therefore the
> injector doesn't add any extra complications. If you don't want the
> Python layer I'd write the extra functions in C++ using the
> boost::python::object facilities. Just loop over your array in C++
> and append str(element) to a boost::python::list.

Yes, doing it with a loop in C++ is certainly an option, but I was
hoping to avoid this becuase of the simplicity of the Python
solution. The C++ code would have to rely heavily on Python facilities
like list and str, as you suggest, since client code might have
provided Python __str__ or __repr__ methods for the container
elements.

I had a quick look at the injector code you mentioned, but that seems
to rely on Python code as well. Not that that's a bad thing, but I'm
looking for a solution to embed within the C++ extension module code,
which was my original problem - how to insert some Python code as a
method from within the C++ code.

> BTW: I believe an array __repr__ that spits out the entire array is
> a very bad idea. If the user bothers to make a C++ array instead of
> simply using a Python list the C++ array is probably big, and the
> __repr__ output will be a nuisance because of its size. Seeing the
> default __repr__ output is more helpful. If the user really wants
> the content, print list(array_instance) will do the trick.  If the
> wrapped arrays are small it is IMO much more efficient (compile-time
> and object code size) to use the "tuple_mappings". See FAQ, mention
> of scitbx in connection with array wrappers.

Well, I guess it's debatable whether the C++ containers should provide
__str__ and __repr__, but I was thinking along the lines of providing
a near-complete emulation of Python sequences, and "least surprise" on
the Python side.

-- 
Raoul Gough.
export LESS='-X'




From colin_irwin at hotmail.com  Mon Mar 29 02:52:46 2004
From: colin_irwin at hotmail.com (Colin Irwin)
Date: Mon, 29 Mar 2004 10:52:46 +1000
Subject: [C++-sig] boost::variant in Python interface
Message-ID: <c47s0j$a47$1@sea.gmane.org>

I have a C++ interface that I would like to make available in a
corresponding Python module.  The interface contains a member function that
returns a vector of boost::variants, with the variant being one of three
possible types.  I've used the boost::python::vector_indexing_suite class to
wrap the vector up, so far so good.  However, the core of the problem is how
to wrap the variant type up.

To provide a more concrete example, I have provided a sample implementation
below:

class BaseData { ... };
class DataType1 : public BaseData { ... };
class DataType2 : public BaseData { ... };
class DataType3 : public BaseData { ... };

class DataProducer
{
    public:
        typedef boost::variant<
        void ReceiveData(std::vec






From colin_irwin at hotmail.com  Mon Mar 29 03:02:13 2004
From: colin_irwin at hotmail.com (Colin Irwin)
Date: Mon, 29 Mar 2004 11:02:13 +1000
Subject: [C++-sig] Re: boost::variant in Python interface
References: <c47s0j$a47$1@sea.gmane.org>
Message-ID: <c4805t$g9i$1@sea.gmane.org>

Sorry for the incomplete send - let me try this again:


I have a C++ interface that I would like to make available in a
corresponding Python module.  The interface contains a member function that
returns a vector of boost::variants, with the variant being one of  three
possible types.  I've used the boost::python::vector_indexing_suite class to
wrap the vector up, so far so good.  However, the core of the problem is how
to wrap the variant type up.

To provide a more concrete example, I have provided a sample implementation
below:

class BaseData { ... };
class DataType1 : public BaseData { ... };
class DataType2 : public BaseData { ... };
class DataType3 : public BaseData { ... };

class DataProducer
{
    public:
        typedef boost::variant<DataType1, DataType2, DataType3> DataVariant;
        typedef std::vector<DataVariant> DataVector;

        void ReceiveData(DataVector& data_array)
        {
            // Return data in the user specified data_array parameter.
        }

        ...
}

I'm thinking it would be nice (and possible?) to somehow have the fact
hidden that there is a variant type involved.  By this I mean that the
various member data and member functions would simply either be available,
if the variant actually contained a type where these were available, or not
available, if the variant contained a type where these were not available.

What's the best approach to deal with this type of problem?  Has anyone
tried to use a boost::variant type through a Python interface?  Would it be
possible to wrap up the variant type is some way as to hide the
implementation, but still make functionality available?

Thanks in advance.

Colin






From dave at boost-consulting.com  Mon Mar 29 05:02:51 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Sun, 28 Mar 2004 22:02:51 -0500
Subject: [C++-sig] Re: MSVC 7.1 Multithreaded Dll Memory Leak [long post]
References: <200403260855.i2Q8tFJ20451@ns.cricketthosting.co.uk>
	<usmfttuhr.fsf@boost-consulting.com>
	<00e001c414af$f277b420$0302a8c0@paul>
Message-ID: <un060s6ck.fsf@boost-consulting.com>

"Paul Grenyer" <paul at paulgrenyer.co.uk> writes:

>> > is it worth posting about it to the main boost list?
>>
>> Sure; first I'd try to reproduce the problem with a minimal example.
>
> Well, as I think I put in my original post, the memory leak occurs even when
> a program contains only a handle declaration. 

Did you post a minimal example?

> However, I don't know if this is the post-Py_Finalize memory

You can't call Py_Finalize when using Boost.Python; if you do, all
bets are off.  I'd have guessed you already know that, though(?)

> leak or the growing memory leak I have been experiencing, but it is
> present in the dll runtime and not in the static runtime.

Well, it seems like you have a small enough test case that you ought
to be able to track it down yourself (?)

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Mon Mar 29 05:10:10 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Sun, 28 Mar 2004 22:10:10 -0500
Subject: [C++-sig] Re: Automation of Python object retrieval
References: <1078321816.742.164.camel@pbarbier>
	<un061tu1v.fsf@boost-consulting.com>
	<1080465761.1017.12.camel@pbarbier>
Message-ID: <uhdw8s60d.fsf@boost-consulting.com>

Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:

> Sorry if I wasn't clear, but the problem is quite complex from my point
> of view and I have problems to express it clearly.
>
> Le dim 28/03/2004 ? 07:33, David Abrahams a ?crit :
>> Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:
>> 
>> > I have a problem with the boost::python::object contruction. I'd want to
>> > know if there is a way to automate the retrieval of the existing Python
>> > object when I hold a Pointer on the C++ base class and that the class is
>> > extended in Python. Here's an example of what I do now :
>> >
>> 
>> <snip big example with no mention of boost::python::object>
>> 
>> 
>> > ======================
>> >
>> > All the equality tests fails because the other in detected as a Base
>> > object and not a Derived object !!!
>> 
>> detected where?
>
> When a function returns a base class pointer, there is a mechanism to
> detect the dynamic type of the pointer (whether it's the base class or
> any Derived class exported in Python). But if you just construct a
> boost::python::object from a pointer, the Python type of the
> boost::python::object will be the type of the pointer (whatever the
> dynamic type of the object could be).

Oh!  That's a very clear explanation!

It should be fixed in the library, but you can get around it by
wrapping a little function in an object:

         template <class T>
         T identity_(T x) { return x; }

         object identity(&identity_<Base*>);
         
         ...
                
            void f(Base* b)
            {
                object b = identity(b);
                ...

            }

> No, there's no reasons. In the code I use, I export the classes with
> smart pointers ... but it does not resolve anything. My goal is the
> following:
>
> I have an object pointed by a Base pointer. 

This is a C++ object, then.

> But the object's dynamic type could be (and in practive it WILL be)
> a class derived from Base

OK so far.

>  and most probably the derivation occured
> in Python.

Urrh, wait: Python derivation isn't like C++ derivation in any way.
You'll just have a pointer to a BaseWrap if the Python class wrapper
was subclassed.

> What I want is :
>
>  - if the dynamic type of the object is a Python class

Impossible.  I guess you mean: "if it's BaseWrap"

> derived from Base: construct the boost::python::object that holds
> the PyObject* given to the class wrapper when constructed - 

If you're holding a shared_ptr<Base> that was passed as an argument
to a warpped function, that part is automatic.

> if the
> dynamic type of the object is a C++ class derived from Base and
> exported in Python using Boost : construct the boost::python::object
> referencing the dynamic type of the object

That is also automatic if you're using shared_ptr.

> The method I wrote do that, but it's not very clean. I hope I'm clear
> enough here too :-/

Seems to me you want shared_ptr; everything else is just making your
life harder.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Mon Mar 29 05:11:58 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Sun, 28 Mar 2004 22:11:58 -0500
Subject: [C++-sig] Re: Injecting python code to a class
References: <uisgosylb.fsf@yahoo.co.uk>
	<20040328172709.19754.qmail@web20210.mail.yahoo.com>
	<uekrcsio0.fsf@yahoo.co.uk>
Message-ID: <ubrmgs5xd.fsf@boost-consulting.com>

Raoul Gough <RaoulGough at yahoo.co.uk> writes:

>> http://cvs.sourceforge.net/viewcvs.py/cctbx/boost_adaptbx/boost/python.py?rev=1.1&view=auto
>>
>> All my extension module imports go through a Python layer to work
>> around exception handling problems with certain versions of gcc, and
>> because I add all kinds of stuff from Python anyway. Therefore the
>> injector doesn't add any extra complications. If you don't want the
>> Python layer I'd write the extra functions in C++ using the
>> boost::python::object facilities. Just loop over your array in C++
>> and append str(element) to a boost::python::list.
>
> Yes, doing it with a loop in C++ is certainly an option, but I was
> hoping to avoid this becuase of the simplicity of the Python
> solution. The C++ code would have to rely heavily on Python facilities
> like list and str, as you suggest, since client code might have
> provided Python __str__ or __repr__ methods for the container
> elements.
>
> I had a quick look at the injector code you mentioned, but that seems
> to rely on Python code as well. Not that that's a bad thing, but I'm
> looking for a solution to embed within the C++ extension module code,
> which was my original problem - how to insert some Python code as a
> method from within the C++ code.

Can't you use PyRun_String or similar to evaluate some Python code
from C++?

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From paul at paulgrenyer.co.uk  Mon Mar 29 10:09:38 2004
From: paul at paulgrenyer.co.uk (Paul Grenyer)
Date: Mon, 29 Mar 2004 09:09:38 +0100
Subject: [C++-sig] Re: MSVC 7.1 Multithreaded Dll Memory Leak [long post]
Message-ID: <200403290809.i2T89cQ29308@ns.cricketthosting.co.uk>

Hi

> > Well, as I think I put in my original post, the memory leak occurs 
> even when a program contains only a handle declaration. 
> 
> Did you post a minimal example?

Yep. :-)

> > However, I don't know if this is the post-Py_Finalize memory
> 
> You can't call Py_Finalize when using Boost.Python; if you do, all
> bets are off.  I'd have guessed you already know that, though(?)

We'd pretty much come to that conclusion yes. 

> 
> > leak or the growing memory leak I have been experiencing, but it is
> > present in the dll runtime and not in the static runtime.
> 
> Well, it seems like you have a small enough test case that you ought
> to be able to track it down yourself (?)

More than likely. I will at least try.

Regards
Paul

Paul Grenyer
Email: paul at paulgrenyer.co.uk
Web: http://www.paulgrenyer.co.uk

Have you met Aeryn: http://www.paulgrenyer.co.uk/aeryn/?
Version 0.3.0 beta now available for download.





From raoulgough at yahoo.co.uk  Mon Mar 29 14:51:44 2004
From: raoulgough at yahoo.co.uk (=?iso-8859-1?q?Raoul=20Gough?=)
Date: Mon, 29 Mar 2004 13:51:44 +0100 (BST)
Subject: [C++-sig] Re: Injecting python code to a class
Message-ID: <20040329125144.688.qmail@web25205.mail.ukl.yahoo.com>

David Abrahams dave at boost-consulting.com 
Sun Mar 28 22:11:58 EST 2004 wrote:

>Raoul Gough <RaoulGough at yahoo.co.uk> writes:
>[snip]
>> I had a quick look at the injector code you mentioned, but that
seems
>> to rely on Python code as well. Not that that's a bad thing, but I'm
>> looking for a solution to embed within the C++ extension module
code,
>> which was my original problem - how to insert some Python code as a
>> method from within the C++ code.
>
>Can't you use PyRun_String or similar to evaluate some Python code
>from C++?

Actually, that was the idea I had originally, but I wasn't sure where
to get the necessary function arguments from. PyRun_String needs the
global and local dictionaries to operate with - is there any easy way
to get these? It would be running within a def_visitor::visit
function...

Regards,
Raoul Gough.


	
	
		
___________________________________________________________
WIN FREE WORLDWIDE FLIGHTS - nominate a cafe in the Yahoo! Mail Internet Cafe Awards  www.yahoo.co.uk/internetcafes 



From nicodemus at esss.com.br  Mon Mar 29 16:10:37 2004
From: nicodemus at esss.com.br (Nicodemus)
Date: Mon, 29 Mar 2004 11:10:37 -0300
Subject: [C++-sig] Re: Injecting python code to a class
In-Reply-To: <20040329125144.688.qmail@web25205.mail.ukl.yahoo.com>
References: <20040329125144.688.qmail@web25205.mail.ukl.yahoo.com>
Message-ID: <40682E5D.10409@esss.com.br>

Hi Raoul,

Raoul Gough wrote:

>Actually, that was the idea I had originally, but I wasn't sure where
>to get the necessary function arguments from. PyRun_String needs the
>global and local dictionaries to operate with - is there any easy way
>to get these? It would be running within a def_visitor::visit
>function...
>  
>

IIRC, you have to pass NULL as the dictionaries in PyRunString.

HTH,
Nicodemus.



From faheem at email.unc.edu  Mon Mar 29 21:46:45 2004
From: faheem at email.unc.edu (Faheem Mitha)
Date: Mon, 29 Mar 2004 19:46:45 +0000 (UTC)
Subject: [C++-sig] blitz++ and numarray
Message-ID: <slrnc6gv95.3v3.faheem@Chrestomanci.home.earth>

Hi,

I'm considering the use of the Blitz++ C++ library
(http://www.oonumerics.org/blitz/) along with numarray/Numeric. I want
to pass an array down to C++ and then manipulate it using Blitz++.

My current thinking is to use the Numeric interface to pass down the
array, and then copy it to a blitz++ array object. This seems
straightforward except I am a little concerned about those reference
counting issues, which I don't properly understand (I'm a Python
beginner). I do get the impression that it is mainly an issue if I am
working with Python objects in C/C++ code, though, and I am mostly
going to try to avoid doing so by declaring C++ objects within the C++
code and working with those as much as possible.

Other possibilities include somehow using the Boost.Python package,
though I am not sure how. One advantage of this is that it apparently
takes care of the reference counting issue. However, I don't want to
learn yet another API.

Also, weave.blitz() from SciPy looks relevant/useful, but this
automatically generates its own C++ code, and I'd prefer to write my
own.

Can anyone suggested a good basic tutorial which discusses the issues
of reference counting in a C/C++ API context? I have tried reading the
official documentation and will continue to do so, but something
gentler would be useful.

Comments, suggestions gratefully appreciated. Please cc me on any
reply to faheem at email.unc.edu. Thanks in advance.

                                                    Faheem.




From RaoulGough at yahoo.co.uk  Mon Mar 29 22:58:41 2004
From: RaoulGough at yahoo.co.uk (Raoul Gough)
Date: Mon, 29 Mar 2004 21:58:41 +0100
Subject: [C++-sig] Re: Injecting python code to a class
References: <20040329125144.688.qmail@web25205.mail.ukl.yahoo.com>
	<40682E5D.10409@esss.com.br>
Message-ID: <uptavqsji.fsf@yahoo.co.uk>

Nicodemus <nicodemus at esss.com.br> writes:

> Hi Raoul,
>
> Raoul Gough wrote:
>
>>Actually, that was the idea I had originally, but I wasn't sure where
>>to get the necessary function arguments from. PyRun_String needs the
>>global and local dictionaries to operate with - is there any easy way
>>to get these? It would be running within a def_visitor::visit
>>function...
>>
>
> IIRC, you have to pass NULL as the dictionaries in PyRunString.

That was the first thing I tried, but I get the following error:

Traceback (most recent call last):
  File "runexample.py", line 1, in ?
    from example import vector_int
SystemError: PyEval_EvalCodeEx: NULL globals

After some further research, it seems that I should be passing
__builtins__.__dict__ as the global dictionary to PyRun_String, and
the locals builtin can indeed be NULL. This might depend on what your
string actually does, in my case it's just a lambda expression.
Anyway, finding this global __dict__ object is relatively easy, as
mentioned here:

http://mail.python.org/pipermail/python-list/1999-April/000125.html

The code I've ended up with is shown below, slightly simplified. I'm
still not sure that this is the "right" way to do it, but it seems to
work.

// Get the global dictionary
boost::python::object global_dict ()
{
  using namespace boost::python;

  handle<> builtin_handle (PyImport_ImportModule("__builtin__"));
  return object (borrowed (PyModule_GetDict (builtin_handle.get())));
}

// Return the result of evaluating a python expression
boost::python::object python_eval (char const *str)
{
  using namespace boost::python;

  return object (handle<> (PyRun_String (
      const_cast<char *>(str), Py_eval_input, global_dict().ptr(), 0)));
}

// Add a __repr__ method for a container
template<class PythonClass> void add_repr (PythonClass &pyClass)
{
  boost::python::objects::add_to_namespace (
      pyClass, "__repr__", python_eval ("lambda v: repr([x for x in v])"));
}

-- 
Raoul Gough.
export LESS='-X'




From RaoulGough at yahoo.co.uk  Mon Mar 29 23:02:21 2004
From: RaoulGough at yahoo.co.uk (Raoul Gough)
Date: Mon, 29 Mar 2004 22:02:21 +0100
Subject: [C++-sig] Re: Injecting python code to a class
References: <20040329125144.688.qmail@web25205.mail.ukl.yahoo.com>
	<40682E5D.10409@esss.com.br> <uptavqsji.fsf@yahoo.co.uk>
Message-ID: <ullljqsde.fsf@yahoo.co.uk>

Raoul Gough <RaoulGough at yahoo.co.uk> writes:

> [...] it seems that I should be passing __builtins__.__dict__ as the
> global dictionary to PyRun_String, and the locals builtin can indeed
                                             ^^^^^^^^^^^^^^
> be NULL.

Oops, sorry. I meant "local dictionary" rather than "locals builtin"

-- 
Raoul Gough.
export LESS='-X'




From xavier.warin at der.edfgdf.fr  Tue Mar 30 07:25:18 2004
From: xavier.warin at der.edfgdf.fr (Xavier WARIN(Compte LOCAL) - I23)
Date: Tue, 30 Mar 2004 07:25:18 +0200
Subject: [C++-sig] blitz++ and numarray
In-Reply-To: <slrnc6gv95.3v3.faheem@Chrestomanci.home.earth>
References: <slrnc6gv95.3v3.faheem@Chrestomanci.home.earth>
Message-ID: <406904BE.6030808@der.edfgdf.fr>

Hi

Don't do it by yourself.
You'd better use SWIG  to do the wrapping, or better Boost 
(www.boost.org). You can find a wrapping for Numeric at this adresse :
http://www.eos.ubc.ca/research/clouds/num_util.html.
(It is done with Boost)
You will be able to export all C++ classes  to Python,
and use tuple, dict inside your C++ code etc...

Xavier

Faheem Mitha wrote:
> Hi,
> 
> I'm considering the use of the Blitz++ C++ library
> (http://www.oonumerics.org/blitz/) along with numarray/Numeric. I want
> to pass an array down to C++ and then manipulate it using Blitz++.
> 
> My current thinking is to use the Numeric interface to pass down the
> array, and then copy it to a blitz++ array object. This seems
> straightforward except I am a little concerned about those reference
> counting issues, which I don't properly understand (I'm a Python
> beginner). I do get the impression that it is mainly an issue if I am
> working with Python objects in C/C++ code, though, and I am mostly
> going to try to avoid doing so by declaring C++ objects within the C++
> code and working with those as much as possible.
> 
> Other possibilities include somehow using the Boost.Python package,
> though I am not sure how. One advantage of this is that it apparently
> takes care of the reference counting issue. However, I don't want to
> learn yet another API.
> 
> Also, weave.blitz() from SciPy looks relevant/useful, but this
> automatically generates its own C++ code, and I'd prefer to write my
> own.
> 
> Can anyone suggested a good basic tutorial which discusses the issues
> of reference counting in a C/C++ API context? I have tried reading the
> official documentation and will continue to do so, but something
> gentler would be useful.
> 
> Comments, suggestions gratefully appreciated. Please cc me on any
> reply to faheem at email.unc.edu. Thanks in advance.
> 
>                                                     Faheem.
> 
> 
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
> 
> 




From sirbender at freenet.de  Tue Mar 30 13:10:04 2004
From: sirbender at freenet.de (Billy Gnosis)
Date: Tue, 30 Mar 2004 13:10:04 +0200
Subject: [C++-sig] Newbie needs advice for decision.
Message-ID: <4069558C.4060703@freenet.de>

Hi I want to extend and more important embed Python in my Programs.

1. What is the best (easiest way) to extend Python with C++ ? Swig or 
Boost.Python, or is there something easier ?

2. What is the best (easiest way) to embed Python in a C++ Program ? 
Boost.Python, the Python/C Api, or is there something easier out there ?


3.  General Question: I played a little bit wit the high level embedding 
of the python interpreter in my C++ Program (#include "Python.h", 
Py_Initialize()...etc.etc...). Can I somehow compile and link everything 
statically ? Meaning, I have one binary in the end, that already 
includes all necessary parts of the python interpreter - when I give 
this file somebody else, this should be enough to run it  no 
installation of the python interpreter necessary, etc...

Is this possible ?

Cheers,

sb




From pierre.barbier at cirad.fr  Tue Mar 30 14:11:07 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Tue, 30 Mar 2004 14:11:07 +0200
Subject: [C++-sig] Re: Automation of Python object retrieval
In-Reply-To: <uhdw8s60d.fsf@boost-consulting.com>
References: <1078321816.742.164.camel@pbarbier>
	<un061tu1v.fsf@boost-consulting.com>
	<1080465761.1017.12.camel@pbarbier>
	<uhdw8s60d.fsf@boost-consulting.com>
Message-ID: <1080648667.6967.10.camel@pbarbier>

On Mon, 2004-03-29 at 05:10, David Abrahams wrote:

> It should be fixed in the library, but you can get around it by
> wrapping a little function in an object:
> 
>          template <class T>
>          T identity_(T x) { return x; }
> 
>          object identity(&identity_<Base*>);
>          
>          ...
>                 
>             void f(Base* b)
>             {
>                 object b = identity(b);
>                 ...
> 
>             }
> 

mmhh ... I have a problem with this. I'm using g++-3.2 and I've got the
error :

boost/boost/python/detail/invoke.hpp|76| error: no match for 
||    call to `(const 
||   
boost::python::detail::specify_a_return_value_policy_to_wrap_functions_returning<Base*>
||    ) (Base*)'


But I don't know how to specify a return value policy with this syntax.
Then, I had to write :

Base* ( *identity_Base )( Base* ) = &identity_<Base*>;
object identity( identity_Base );

If not, g++ was unable to find the type of identity_<Base*>, any clue on
this ?

Thanks,

Pierre

-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From pierre.barbier at cirad.fr  Tue Mar 30 14:34:28 2004
From: pierre.barbier at cirad.fr (Pierre Barbier de Reuille)
Date: Tue, 30 Mar 2004 14:34:28 +0200
Subject: [C++-sig] Re: Automation of Python object retrieval
In-Reply-To: <1080648667.6967.10.camel@pbarbier>
References: <1078321816.742.164.camel@pbarbier>
	<un061tu1v.fsf@boost-consulting.com>
	<1080465761.1017.12.camel@pbarbier>
	<uhdw8s60d.fsf@boost-consulting.com>
	<1080648667.6967.10.camel@pbarbier>
Message-ID: <1080650068.6967.25.camel@pbarbier>

Ok, I solved the compilation problem by creating the identity object
with :

object identity = make_function(&identity_<Base*>,
return_internal_reference<1>());

But it does not solve any of the problems I have. And if I give a
Derived* pointer to the identity function, the created
boost::python::object will be of class Base !

On Tue, 2004-03-30 at 14:11, Pierre Barbier de Reuille wrote:
> On Mon, 2004-03-29 at 05:10, David Abrahams wrote:
> 
> > It should be fixed in the library, but you can get around it by
> > wrapping a little function in an object:
> > 
> >          template <class T>
> >          T identity_(T x) { return x; }
> > 
> >          object identity(&identity_<Base*>);
> >          
> >          ...
> >                 
> >             void f(Base* b)
> >             {
> >                 object b = identity(b);
> >                 ...
> > 
> >             }
> > 
> 
> mmhh ... I have a problem with this. I'm using g++-3.2 and I've got the
> error :
> 
> boost/boost/python/detail/invoke.hpp|76| error: no match for 
> ||    call to `(const 
> ||   
> boost::python::detail::specify_a_return_value_policy_to_wrap_functions_returning<Base*>
> ||    ) (Base*)'
> 
> 
> But I don't know how to specify a return value policy with this syntax.
> Then, I had to write :
> 
> Base* ( *identity_Base )( Base* ) = &identity_<Base*>;
> object identity( identity_Base );
> 
> If not, g++ was unable to find the type of identity_<Base*>, any clue on
> this ?
> 
> Thanks,
> 
> Pierre
-- 
Pierre Barbier de Reuille

INRA - UMR Cirad/Inra/Cnrs/Univ.MontpellierII AMAP
Botanique et Bio-informatique de l'Architecture des Plantes
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

tel   : (33) 4 67 61 65 77    fax   : (33) 4 67 61 56 68 




From dave at boost-consulting.com  Wed Mar 31 00:05:16 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Tue, 30 Mar 2004 17:05:16 -0500
Subject: [C++-sig] Re: Automation of Python object retrieval
References: <1078321816.742.164.camel@pbarbier>
	<un061tu1v.fsf@boost-consulting.com>
	<1080465761.1017.12.camel@pbarbier>
	<uhdw8s60d.fsf@boost-consulting.com>
	<1080648667.6967.10.camel@pbarbier> <1080650068.6967.25.camel@pbarbier>
Message-ID: <ubrme7zz7.fsf@boost-consulting.com>

Pierre Barbier de Reuille <pierre.barbier at cirad.fr> writes:

> Ok, I solved the compilation problem by creating the identity object
> with :
>
> object identity = make_function(&identity_<Base*>,
> return_internal_reference<1>());
>
> But it does not solve any of the problems I have. And if I give a
> Derived* pointer to the identity function, the created
> boost::python::object will be of class Base !

Provided that Derived has been registered as a Boost.Python extension
class, and Base has a virtual function, it should work.  The crucial
code that does the lookup is in
boost/python/object/make_ptr_instance.hpp

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From dave at boost-consulting.com  Wed Mar 31 02:30:15 2004
From: dave at boost-consulting.com (David Abrahams)
Date: Tue, 30 Mar 2004 19:30:15 -0500
Subject: [C++-sig] Re: Boost Python on Mac OS X
References: <D6091EB1-78A7-11D8-98B0-00306599C846@rochester.rr.com>
Message-ID: <uad1x6ep4.fsf@boost-consulting.com>

Hi James,

It's best to post Boost.Python questions to the c++-sig:

  http://www.boost.org/more/mailing_lists.htm#cplussig

James Mastro <jmastro at rochester.rr.com> writes:

> Hello all,
>
> I'm trying to use Boost Python on Mac OS X 10.3. I'm following along
> with what is at  http://www.boost.org/libs/python/doc/tutorial/doc/
> building_hello_world.html. After using bjam like this:
>
> bjam -sTOOLS=darwin
>
> the following are built:
>
> libboost_python.dylib
> hello.so
>
> This seems to be correct. I'm not totally sure where they are supposed
> to go, so I put them in next to Python in the Python framework so
> Python could find them to import. It found them, but an "import hello"
> in Python resulted in:
>
> ImportError: Failure linking new module:
> ../../../../bin/boost/libs/python/build/libboost_python.dylib/darwin/
> debug/shared-linkable-true/libboost_python.dylib: dyld:
> /Applications/MacPython-2.3/PythonIDE.app/Contents/MacOS/PythonIDE
> can't open library:
> ../../../../bin/boost/libs/python/build/libboost_python.dylib/darwin/
> debug/shared-linkable-true/libboost_python.dylib  (No such file or
> directory, errno = 2)
>
> Before putting the 2 libs in, "import hello" found nothing. So I
> assume  Python is finding the hello module now, but not liking it.
> Can anyone  show me where I've gone wrong? Many thanks.

I don't know much about MacOS X, but I'm guessing there's an
LD_LIBRARY_PATH or some such that tells it where to look for
dynamically linked libraries?  If so, you need to modify it to point
at libboost_python.dylib, or move the library to a location already in
that path.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com




From jbrandmeyer at earthlink.net  Wed Mar 31 03:52:25 2004
From: jbrandmeyer at earthlink.net (Jonathan Brandmeyer)
Date: Tue, 30 Mar 2004 20:52:25 -0500
Subject: [C++-sig] Re: Boost Python on Mac OS X
In-Reply-To: <uad1x6ep4.fsf@boost-consulting.com>
References: <D6091EB1-78A7-11D8-98B0-00306599C846@rochester.rr.com>
	<uad1x6ep4.fsf@boost-consulting.com>
Message-ID: <1080697945.5098.18.camel@illuvatar>

On Tue, 2004-03-30 at 19:30, David Abrahams wrote:
> James Mastro <jmastro at rochester.rr.com> writes:
> 
> > Hello all,
> >
> > I'm trying to use Boost Python on Mac OS X 10.3. I'm following along
> > with what is at  http://www.boost.org/libs/python/doc/tutorial/doc/
> > building_hello_world.html. After using bjam like this:
> >
> > bjam -sTOOLS=darwin
> >
> > the following are built:
> >
> > libboost_python.dylib
> > hello.so
> >
> > This seems to be correct. I'm not totally sure where they are supposed
> > to go, so I put them in next to Python in the Python framework so
> > Python could find them to import. It found them, but an "import hello"
> > in Python resulted in:
> >
> > ImportError: Failure linking new module:
> > ../../../../bin/boost/libs/python/build/libboost_python.dylib/darwin/
> > debug/shared-linkable-true/libboost_python.dylib: dyld:
> > /Applications/MacPython-2.3/PythonIDE.app/Contents/MacOS/PythonIDE
> > can't open library:
> > ../../../../bin/boost/libs/python/build/libboost_python.dylib/darwin/
> > debug/shared-linkable-true/libboost_python.dylib  (No such file or
> > directory, errno = 2)

What are the output of running `otool -L /path/to/hello.so` and `otool
-D /path/to/libboost_python.dylib`?  

The first command lists all of the libraries that hello.so depends on
(similar to ldd on Linux) and the second reports the install name of the
dylib (similar to the 'soname' concept for shared libs on Linux).

> > Before putting the 2 libs in, "import hello" found nothing. So I
> > assume  Python is finding the hello module now, but not liking it.
> > Can anyone  show me where I've gone wrong? Many thanks.
> 
> I don't know much about MacOS X, but I'm guessing there's an
> LD_LIBRARY_PATH or some such that tells it where to look for
> dynamically linked libraries?  If so, you need to modify it to point
> at libboost_python.dylib, or move the library to a location already in
> that path.

-Jonathan Brandmeyer