From pwinston at gmail.com  Fri Aug  1 18:13:23 2008
From: pwinston at gmail.com (Philip Winston)
Date: Fri, 1 Aug 2008 12:13:23 -0400
Subject: [C++-sig] application exit and shared_ptr
In-Reply-To: <87provnrxe.fsf@mcbain.luannocracy.com>
References: <33677e5d0807292014s5cae77cdm7fa02b0fc21e6645@mail.gmail.com>
	<87provnrxe.fsf@mcbain.luannocracy.com>
Message-ID: <33677e5d0808010913t779bcd86h2e48b9e432a13e8c@mail.gmail.com>

Thanks, wow so maybe we are wrong to use shared_ptr at all in our
situation.  Does weak_ptr work as nicely with Boost.Python, because
shared_ptr works great and that's why we were using it.   But weak_ptr
semantics would be ideal.  The main C++ application needs to be in
charge, no interpreter variable should be able to prevent something
from being deleted.  Yet at the same time using a "stale" variable in
Python can't crash the, it can throw but not crash everything.  Does
weak_ptr with Boost.Python behave like this?  I can give it a try but
any advice is helpful.  We have a lot of Boost.Python classes exposed
now, but have been ignoring this fundamental issue.

Philip Winston
pwinston at gmail.com

On Wed, Jul 30, 2008 at 2:33 PM, David Abrahams <dave at boostpro.com> wrote:
>
> on Tue Jul 29 2008, "Philip Winston" <pwinston-AT-gmail.com> wrote:
>
>> I understand it is not possible to call Py_Finalize with Boost.Python.
>>  But is it possible to manually delete all variables from an embedded
>> interactive interpreter?  Would this be a way to to cause Python to
>> release any shared_ptrs it is holding on to?
>
> That's essentially the same problem, because any Python object could be
> holding a shared_ptr (indirectly).
>
>
> --
> Dave Abrahams
> BoostPro Computing
> http://www.boostpro.com
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>


From markus.woschank at gmail.com  Fri Aug  1 23:55:51 2008
From: markus.woschank at gmail.com (Markus Woschank)
Date: Fri, 01 Aug 2008 23:55:51 +0200
Subject: [C++-sig] boost::python and boost::shared_ptr strict weak ordering
	problem
Message-ID: <48938667.1080906@gmail.com>

Hi list,

I'm getting a strange behaviour with shared_ptr objects as arguments of 
wrapped C++ functions.
The problem occurs when i try to use the shared_ptr objects in a 
std::set which (AFAIK) uses
strict weak ordering equivalence [ !(a<b) && !(b<a) ]. The problem also 
only seems to occur if
there are different calls to C++ from python, while the C++ code holds a 
shared_ptr somewhere.

I'm using boost 1.35.0.

Here is a sample program that should demonstrate my problem

C++ code:
---
class Toast : boost::noncopyable
{
  public:
    typedef boost::shared_ptr<Toast> ptr;

    static ptr create()
    { return ptr( new Toast() ); }

  private:
    Toast() {}
};

bool ptr_equals_last(Toast::ptr p)
{
  static Toast::ptr last;
  bool equal = !(p<last) && !(last<p);
  last = p;
  return equal;
}

BOOST_PYTHON_MODULE(ptr_test)
{
  using namespace boost::python;

  class_< Toast, Toast::ptr, boost::noncopyable >("Toast",no_init)
    .def("create",&Toast::create)
    .staticmethod("create");
  ;

  def("ptr_equals_last",ptr_equals_last);
}
---
[ also available (for some time) on http://tmp.netst.at/python_module.cpp ]

python test and output:
---
Python 2.5.2 (r252:60911, Aug  1 2008, 22:31:18)
[GCC 4.1.2 (Gentoo 4.1.2)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
 >>> from ptr_test import Toast,ptr_equals_last
 >>>
 >>> t = Toast.create() # instance for testing
 >>> t
<ptr_test.Toast object at 0x7f10e95b6578>
 >>> ptr_equals_last(None) # should return True ( empty == empty => 
equivalent )
True
 >>> ptr_equals_last(None) # should return True ( empty == empty => e. )
True
 >>>
 >>> ptr_equals_last(t) # should return False ( empty != t => not e. )
False
 >>> ptr_equals_last(t) # should return True ( t == t => e. )
False
 >>> ptr_equals_last(t) # should return True ( t == t => e. )
False
 >>> ptr_equals_last(None) # should return False ( t != empty => not e. )
False
 >>> ptr_equals_last(None) # should return True ( empty == empty => e. )
True
---
[ also available (for some time) on http://tmp.netst.at/test.py ]

PS: After debugging a little bit my wild guess is, that boost.python 
uses the shared_ptr objects with
custom deleter function that decrements the python refcount. And such 
custom shared_ptrs are created
every time a C++ function with an shared_ptr argument is called (while 
incrementing the python refcount).
Which seems to work fine but (IMO) breaks the less than operator of 
boost::shared_ptr documented at
[http://www.boost.org/doc/libs/1_35_0/libs/smart_ptr/shared_ptr.htm#functions] 
- but as already said, it's
only a wild guess.

Cheers,
Markus



From rwgk at yahoo.com  Sat Aug  2 00:57:29 2008
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Fri, 1 Aug 2008 15:57:29 -0700 (PDT)
Subject: [C++-sig] boost::python and boost::shared_ptr strict weak
	ordering problem
Message-ID: <133366.52902.qm@web31401.mail.mud.yahoo.com>

I think your guess is very good. I stumbled over this, too, a while ago:

http://mail.python.org/pipermail/c++-sig/2006-March/010407.html

This lead me to redesign my parent-child classes in order to completely avoid
smart pointers in the interfaces (they are still used "under the hood").

Ralf

> PS: After debugging a little bit my wild guess is, that boost.python
> uses the shared_ptr objects with
> custom deleter function that decrements the python refcount. And such
> custom shared_ptrs are created
> every time a C++ function with an shared_ptr argument is called (while
> incrementing the python refcount).
> Which seems to work fine but (IMO) breaks the less than operator of
> boost::shared_ptr documented at
> [http://www.boost.org/doc/libs/1_35_0/libs/smart_ptr/shared_ptr.htm#functions]
> - but as already said, it's
> only a wild guess.


From william.marie at gmail.com  Sat Aug  2 11:37:05 2008
From: william.marie at gmail.com (=?UTF-8?Q?William_Mari=C3=A9?=)
Date: Sat, 2 Aug 2008 02:37:05 -0700 (PDT)
Subject: [C++-sig]  Generic Function Calls
Message-ID: <18787671.post@talk.nabble.com>


Hi,

I'm using Boost Python and i would like to know if there is an easy way to
call functions generically with N numbers of arguments a bit like this :

void call( boost::python::object bpFunction, ... )
{
     // Here i want to find a way to call the bpFunction with the variable
argument list
}

Thank you
William Mari?
-- 
View this message in context: http://www.nabble.com/Generic-Function-Calls-tp18787671p18787671.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From william.marie at gmail.com  Sat Aug  2 14:17:16 2008
From: william.marie at gmail.com (=?UTF-8?Q?William_Mari=C3=A9?=)
Date: Sat, 2 Aug 2008 05:17:16 -0700 (PDT)
Subject: [C++-sig]  Problem with python-debugging
Message-ID: <18788695.post@talk.nabble.com>


Hi,
I have a problem trying using python-debugging, the generated library for
boost-python is called : 
boost_python-vc80-mt-gyd-1_35.dll

but when i try to build an exe using it i have this error :
LINK : fatal error LNK1104 : cannot open file
'libboost-python-vc80-mt-gd-1_35.lib'

So i don't really understand why bjam doesn't try to link with the gyd
library ?

Could someone tell me why ?

Thanks 
William Mari?
-- 
View this message in context: http://www.nabble.com/Problem-with-python-debugging-tp18788695p18788695.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From seefeld at sympatico.ca  Sat Aug  2 16:59:05 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Sat, 02 Aug 2008 10:59:05 -0400
Subject: [C++-sig] Generic Function Calls
In-Reply-To: <18787671.post@talk.nabble.com>
References: <18787671.post@talk.nabble.com>
Message-ID: <48947639.9010506@sympatico.ca>

William Mari? wrote:
> Hi,
>
> I'm using Boost Python and i would like to know if there is an easy way to
> call functions generically with N numbers of arguments a bit like this :
>
> void call( boost::python::object bpFunction, ... )
> {
>      // Here i want to find a way to call the bpFunction with the variable
> argument list
> }
>   

Just a couple of days ago I have checked in support for this into the 
trunk (it's too late for the upcoming 36 release, but should be part of 
the next one). With that, you will be able to do

void call(boost::python::object callable)
{
  boost::python::list args;
  // ... now fill the list

  callable(*args);
}

HTH,
       Stefan

-- 

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



From m.rogers at hef.ru.nl  Sat Aug  2 23:04:38 2008
From: m.rogers at hef.ru.nl (Michael Rogers)
Date: Sat, 2 Aug 2008 23:04:38 +0200
Subject: [C++-sig] boost python on os x 10.5, intel
Message-ID: <18EC77C7-65C5-4C9A-A31E-72E12AB74E10@hef.ru.nl>

I have figured out how to build and use python extensions under linux.  
When I try to run any of the examples on os x 10.5 (intel), the  
boost_1_35_0/libs/python/example/quickstart for example, it compiles a  
lot of stuff, then gets stuck at this line:

darwin.link.dll bin/darwin/debug/macosx-version-10.4/extending.so


it just sits there, bjam and the python and sh processes that it  
spawns show 0% cpu usage. If I Ctrl-c it, and do it again, then I get:

...using <pbin/darwin/debug/macosx-version-10.4>extending.so...

instead.


This link pretty much describes what is happening to me.

http://article.gmane.org/gmane.comp.lib.boost.build/18302

Does anyone else experience this? Anyone know how to fix it?






From merlin66b at gmail.com  Sun Aug  3 07:28:00 2008
From: merlin66b at gmail.com (Thomas Berg)
Date: Sun, 3 Aug 2008 05:28:00 +0000 (UTC)
Subject: [C++-sig] Problem with python-debugging
References: <18788695.post@talk.nabble.com>
Message-ID: <loom.20080803T051758-782@post.gmane.org>

The problem is not bjam, but boost/config/auto_link.hpp. This header has hard-
coded the naming convention of the boost debug libraries to "-gd", it just 
doesn't know about the "-gyd" convention for python debugging.

I worked around the problem by disabling the autolink-feature, by defining 
"BOOST_ALL_NO_LIB" when building. Bjam already links to the correct "-gyd" 
version (I think), so this just prevents Visual Studio from pulling in the "-
gd" in addition...

Thomas



From happybrowndog at hotmail.com  Sun Aug  3 18:46:13 2008
From: happybrowndog at hotmail.com (happybrowndog)
Date: Sun, 03 Aug 2008 09:46:13 -0700
Subject: [C++-sig] python-debugging=on can't open boost library during Link
	step
Message-ID: <g74ndj$e4q$1@ger.gmane.org>

When I bjam as follows: bjam python-debugging=on 
--preserve-test-targets --debug-configuration

I get a link error that it can't open boost_python-vc90-mt-gd-1_34.lib . 
  Without python-debugging=on, there are no errors.
I'm using the default setup from the Boost distribution for windows.  My 
toolset is msvc=9.0 and I'm building for python 2.5.  I'm using the 
default Jamroot provided, and I'm trying to build the 
getting_started*.cpp files in the example.

Is there a fix for this?

Also, the reason I used python-debugging=on is because I would like to 
debug the python extension later using VC++ debugger.  Is this what the 
option does?



From hohehohe2 at gmail.com  Mon Aug  4 09:18:45 2008
From: hohehohe2 at gmail.com (Koichi Tamura)
Date: Mon, 4 Aug 2008 16:18:45 +0900
Subject: [C++-sig] static/dynamic boost::python conflicts?
In-Reply-To: <b8878c1b0807230304k865e5c2q92026717c83e0522@mail.gmail.com>
References: <b8878c1b0807230304k865e5c2q92026717c83e0522@mail.gmail.com>
Message-ID: <b8878c1b0808040018o3ee1b7d7i851aa082cfb7d970@mail.gmail.com>

Hello,
I've experienced a trouble about boost.python linking (or Python itself) and
I've separated what is causing trouble from my source.
The condition is

- An application (m) loads two libraries dynamically.
- One of  them  (a)  creates a Python module, calls its init function, and
uses it internally. boost.python is statically linked with (a).
- The other (b) linkes boost.python dynamically
- The application (m) loads (b) first, then loads (a)

In this condition, (a)'s module is broken.

If anybody experienced it, or knows what's going on and how to make the
module working, please let me know.
Thanks in advance.

-Koichi


Here's the test code. (a) is a.cpp and (b) is b.cpp, (m) is main.cpp here.

$ cat a.cpp
#include <boost/python.hpp>

using namespace boost::python;

struct World
{
  std::string greet(){return "hello";}
};

BOOST_PYTHON_MODULE(hello)
{
  class_<World>("World")
    .def("greet", &World::greet);
}

extern "C"
{

void pycall()
{
	Py_Initialize();
	PyRun_SimpleString("print 'pycall'");
	inithello();
	object o_main(handle<>(borrowed(PyImport_AddModule("__main__"))));
	object o_hello(handle<>(PyImport_ImportModule("hello")));
	o_main.attr("hel") = o_hello;
	PyRun_SimpleString("print hel.World().greet()");
}

}
$ cat b.cpp
$ cat main.cpp
#include <dlfcn.h>

int main()
{
  void* handle0 = dlopen("b.bundle", RTLD_NOW);
  void* handle = dlopen("a.bundle", RTLD_NOW);

  typedef void (*func)();
  func myfunc;
  myfunc = (func)dlsym(handle, "pycall");
  myfunc();
  dlclose(handle);

  return 0;
}
$ gcc -o a.bundle -bundle
-I/System/Library/Frameworks/Python.framework/Versions/2.5/include/python2.5
-L/System/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/config
-lpython2.5 -lstdc++ a.cpp -lboost_python-mtstatic-1_35
$ gcc -o b.bundle -bundle
-I/System/Library/Frameworks/Python.framework/Versions/2.5/include/python2.5
-L/System/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/config
-lpython2.5 -lstdc++ b.cpp -lboost_python-mt-1_35
$ gcc main.cpp -lstdc++
$ ./a.out
pycall
Traceback (most recent call last):
  File "<string>", line 1, in <module>

AttributeError: 'module' object has no attribute 'World'


I've tested some more.

- It works if I comment out void* handle0 = dlopen("b.bundle", RTLD_NOW);
- It works if I call dlclose(handle0) just after void* handle0 =
dlopen("b.bundle", RTLD_NOW);
- except inithello(); it is working fine.
- It works if I separate the hello module, create hello.so, and import it
later.
- If I do one of the following things:
    1) dynamically link a.bundle and boost::python
    2) statically link b.bundle and boost::python
    3) dynamically link a.bundle and boost::python, and statically link
b.bundle and boost::python

I cannot use PyImport_AppendInittab() nor PyImport_ExtendInittab(), since
Python interpreter may be already running when
my code is first executed. I can separate my extension but I'd rather not.




2008/7/23 Koichi Tamura <hohehohe2 at gmail.com>

> Hello,
> I'm using boost::python and I've got a problem now.
>
>  I'm writing a shared library(bundle, I'm using MacOSX10.5, and it's a
> plug-in for some
> application) where I have a code like this.
>
>     //....on init....
>     if ( ! Py_IsInitialized())
>     {
>         int a = PyImport_AppendInittab("MyModule", &initMyModule);
>         Py_Initialize();
>         PyEval_InitThreads();
>         PyEval_ReleaseLock();
>     }
>
>     //....somewhere else...
>     PyGILState_STATE state = PyGILState_Ensure();
>     module = object(handle<>(PyModule_New(moduleName)));
>     myMod = object(handle<>(PyImport_ImportModule("MyModule")));
>     module.attr("mymod") = myMod;
>     PyGILState_Release(state);
>
> I statically linked boost::python, and it works fine. When imported,
> MyModule is
> filled with classes that boost::python has wrapped.
>
> But when I use another shared library(#) which dynamically links
> boost::python
> together with mine, PyImport_ImportModule("MyModule") returns a module
> which dict
> is empty but __doc__ and __name__.
>
> Instead of (#) I made a test shared library that does nothing but linking
> boost::python dynamically, and it still fails.
>
> Why does all statically linked boost::python and extension is affected by a
> dynamically
> linked boost::python? I'm really puzzled.
>
>
> I would really appreciate if anybody could help.
>
> -Koichi
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080804/bcb51c0a/attachment.htm>

From dave at boostpro.com  Wed Aug  6 17:03:13 2008
From: dave at boostpro.com (David Abrahams)
Date: Wed, 06 Aug 2008 11:03:13 -0400
Subject: [C++-sig] application exit and shared_ptr
In-Reply-To: <33677e5d0808010913t779bcd86h2e48b9e432a13e8c@mail.gmail.com>
	(Philip Winston's message of "Fri, 1 Aug 2008 12:13:23 -0400")
References: <33677e5d0807292014s5cae77cdm7fa02b0fc21e6645@mail.gmail.com>
	<87provnrxe.fsf@mcbain.luannocracy.com>
	<33677e5d0808010913t779bcd86h2e48b9e432a13e8c@mail.gmail.com>
Message-ID: <87ej52chji.fsf@mcbain.luannocracy.com>


on Fri Aug 01 2008, "Philip Winston" <pwinston-AT-gmail.com> wrote:

> Thanks, wow so maybe we are wrong to use shared_ptr at all in our
> situation.  Does weak_ptr work as nicely with Boost.Python, 

No, there's no built-in support for weak_ptr.  It wouldn't be (too) hard
to add, though.

> because
> shared_ptr works great and that's why we were using it.   But weak_ptr
> semantics would be ideal.  The main C++ application needs to be in
> charge, no interpreter variable should be able to prevent something
> from being deleted.  Yet at the same time using a "stale" variable in
> Python can't crash the, it can throw but not crash everything.  Does
> weak_ptr with Boost.Python behave like this?  

No, weak_ptr isn't actually a ptr.  So here's the question: if we
implemented a holder-by-weak_ptr, what would the semantics be when the
object has been released and it is passed to a function accepting a
shared_ptr?  Specifically, should we throw or should we generate an
empty shared_ptr?  Note, if it's the latter, and your C++ code assumes a
valid pointee, it will crash.

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From dave at boostpro.com  Wed Aug  6 17:37:28 2008
From: dave at boostpro.com (David Abrahams)
Date: Wed, 06 Aug 2008 11:37:28 -0400
Subject: [C++-sig] boost python on os x 10.5, intel
In-Reply-To: <18EC77C7-65C5-4C9A-A31E-72E12AB74E10@hef.ru.nl> (Michael
	Rogers's message of "Sat, 2 Aug 2008 23:04:38 +0200")
References: <18EC77C7-65C5-4C9A-A31E-72E12AB74E10@hef.ru.nl>
Message-ID: <87wsiub1dz.fsf@mcbain.luannocracy.com>


on Sat Aug 02 2008, Michael Rogers <m.rogers-AT-hef.ru.nl> wrote:

> This link pretty much describes what is happening to me.
>
> http://article.gmane.org/gmane.comp.lib.boost.build/18302
>
> Does anyone else experience this? Anyone know how to fix it?

Looks like a bug in either Darwin's linker (hangs) or in its
implementation of execwait.  bjam is just waiting for the linker process
to complete.

My suggestion, to find out which: run bjam again with -n -a (build
nothing) or -d+2 (rebuild what's out of date) and look at the linker
command that bjam is issuing.  Then try it yourself from a shell
prompt.  If it hangs, you know it's the linker.  If not, it's either a
bug in execwait or a bug in the way bjam is using it.

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From pwinston at gmail.com  Thu Aug  7 05:30:39 2008
From: pwinston at gmail.com (Philip Winston)
Date: Wed, 6 Aug 2008 23:30:39 -0400
Subject: [C++-sig] application exit and shared_ptr
In-Reply-To: <87ej52chji.fsf@mcbain.luannocracy.com>
References: <33677e5d0807292014s5cae77cdm7fa02b0fc21e6645@mail.gmail.com>
	<87provnrxe.fsf@mcbain.luannocracy.com>
	<33677e5d0808010913t779bcd86h2e48b9e432a13e8c@mail.gmail.com>
	<87ej52chji.fsf@mcbain.luannocracy.com>
Message-ID: <33677e5d0808062030j76ba612ucce66daa8475033f@mail.gmail.com>

> No, weak_ptr isn't actually a ptr.  So here's the question: if we
> implemented a holder-by-weak_ptr, what would the semantics be when the
> object has been released and it is passed to a function accepting a
> shared_ptr?  Specifically, should we throw or should we generate an
> empty shared_ptr?  Note, if it's the latter, and your C++ code assumes a
> valid pointee, it will crash.

In our case throwing would work great.  In our case if someone is
trying to use these variables they are doing something wrong, they are
trying to refer to objects which are long gone.  So it's okay if their
interactive python commands or scripts fail immediately.  We just
don't want the whole application to crash on them.

We're new to embedding Python and this whole concern is mostly
hypothetical at this point.  We realized the issue because at
application exit we were asserting some shared_ptrs are unique.  But
if you've created certain variables in Python, those ptrs are no
longer unique.

Maybe we should just deal with that uncertainty, just have no asserts
and know the interpreter might be holding references.  And then as for
memory usage, yes it's possible for the user to hold stuff in memory,
but how common will that be and how much memory are we talking about?
It might be something we live with an no one really cares or notices.

But to be strictly correct, the weak_ptr support sounds ideal.  Then
someone has a variable "foo" pointing to something, it gets unloaded,
they try to use foo and get an exception.  Very clean.  I think long
term this would be a valuable capability.


Philip Winston
pwinston at gmail.com


From dave at boostpro.com  Thu Aug  7 07:02:28 2008
From: dave at boostpro.com (David Abrahams)
Date: Thu, 07 Aug 2008 01:02:28 -0400
Subject: [C++-sig] application exit and shared_ptr
In-Reply-To: <33677e5d0808062030j76ba612ucce66daa8475033f@mail.gmail.com>
	(Philip Winston's message of "Wed, 6 Aug 2008 23:30:39 -0400")
References: <33677e5d0807292014s5cae77cdm7fa02b0fc21e6645@mail.gmail.com>
	<87provnrxe.fsf@mcbain.luannocracy.com>
	<33677e5d0808010913t779bcd86h2e48b9e432a13e8c@mail.gmail.com>
	<87ej52chji.fsf@mcbain.luannocracy.com>
	<33677e5d0808062030j76ba612ucce66daa8475033f@mail.gmail.com>
Message-ID: <87bq058ljv.fsf@mcbain.luannocracy.com>


on Wed Aug 06 2008, "Philip Winston" <pwinston-AT-gmail.com> wrote:

> But to be strictly correct, the weak_ptr support sounds ideal.  Then
> someone has a variable "foo" pointing to something, it gets unloaded,
> they try to use foo and get an exception.  Very clean.  I think long
> term this would be a valuable capability.

Please enter a ticket in http://svn.boost.org, then.  I don't know when
I'll be able to get to this (paying work is keeping me very busy) but if
we don't have a record it will surely get lost.

Thanks,

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From happybrowndog at hotmail.com  Thu Aug  7 09:32:00 2008
From: happybrowndog at hotmail.com (happybrowndog)
Date: Thu, 07 Aug 2008 00:32:00 -0700
Subject: [C++-sig] import getting_started1 - "no module named
 getting_started1" when using python_d
Message-ID: <g7e8e2$5k8$1@ger.gmane.org>

I used bjam to create the pyd files in the example, all created fine. 
When I use python to import getting_started1, there's no problem.  But 
when I use python_d to do the same thing, I get a "ImportError: no 
module named getting_started1 [29800 refs]" error.

I am using VC++ 9.0, and have built python.exe and python_d.exe from 
source, as well as python25.dll and python25_d.dll.  It is also my bjam 
tool set.

The reason I am interested in running it with python_d.exe is because I 
would like to be able to debug the extension with VC++ .... but I may be 
under the wrong assumption that using python_d.exe is a necessity for 
debugging.  Is this true?

Here is my jamfile:
******************
use-project boost
   : ../../.. ;

project
   : requirements <library>/boost/python//boost_python ;

python-extension getting_started1 : getting_started1.cpp : 
<threading>multi : <variant>debug : <python-debugging>on ;
python-extension getting_started2 : getting_started2.cpp : 
<threading>multi : <variant>debug : <python-debugging>on ;
python-extension std_pair_ext : std_pair.cpp : <threading>multi : 
<variant>debug : <python-debugging>on ;


Here is my bjam command:
********************
c:\boost-jam-3.1.16\bin.ntx86\bjam --preserve-test-targets 
--debug-configuration


Here is my user-config.jam file:
********************
using msvc : : "C:/Program Files/Microsoft Visual Studio 9.0/VC/bin/cl";
using python : 2.5 : C:/Python25 ;


The following is the output of bjam with -o log.txt:
***********************

     call "C:\Program Files\Microsoft Visual Studio 
9.0\VC\vcvarsall.bat" x86 >nul
cl /Zm800 -nologo 
@"bin\msvc-9.0\debug\threading-multi\getting_started1.obj.rsp"

         call "C:\Program Files\Microsoft Visual Studio 
9.0\VC\vcvarsall.bat" x86 >nul
link /NOLOGO /INCREMENTAL:NO /DLL /DEBUG /subsystem:console 
/out:"bin\msvc-9.0\debug\threading-multi\getting_started1.pyd" 
/IMPLIB:"bin\msvc-9.0\debug\threading-multi\getting_started1.lib" 
/LIBPATH:"C:\Python25\libs" 
@"bin\msvc-9.0\debug\threading-multi\getting_started1.pyd.rsp"
         if %ERRORLEVEL% NEQ 0 EXIT %ERRORLEVEL%
         if exist 
"bin\msvc-9.0\debug\threading-multi\getting_started1.pyd.manifest" (
             mt -nologo -manifest 
"bin\msvc-9.0\debug\threading-multi\getting_started1.pyd.manifest" 
"-outputresource:bin\msvc-9.0\debug\threading-multi\getting_started1.pyd;2"
         )

     call "C:\Program Files\Microsoft Visual Studio 
9.0\VC\vcvarsall.bat" x86 >nul
cl /Zm800 -nologo 
@"bin\msvc-9.0\debug\threading-multi\getting_started2.obj.rsp"

         call "C:\Program Files\Microsoft Visual Studio 
9.0\VC\vcvarsall.bat" x86 >nul
link /NOLOGO /INCREMENTAL:NO /DLL /DEBUG /subsystem:console 
/out:"bin\msvc-9.0\debug\threading-multi\getting_started2.pyd" 
/IMPLIB:"bin\msvc-9.0\debug\threading-multi\getting_started2.lib" 
/LIBPATH:"C:\Python25\libs" 
@"bin\msvc-9.0\debug\threading-multi\getting_started2.pyd.rsp"
         if %ERRORLEVEL% NEQ 0 EXIT %ERRORLEVEL%
         if exist 
"bin\msvc-9.0\debug\threading-multi\getting_started2.pyd.manifest" (
             mt -nologo -manifest 
"bin\msvc-9.0\debug\threading-multi\getting_started2.pyd.manifest" 
"-outputresource:bin\msvc-9.0\debug\threading-multi\getting_started2.pyd;2"
         )

     call "C:\Program Files\Microsoft Visual Studio 
9.0\VC\vcvarsall.bat" x86 >nul
cl /Zm800 -nologo @"bin\msvc-9.0\debug\threading-multi\std_pair.obj.rsp"

         call "C:\Program Files\Microsoft Visual Studio 
9.0\VC\vcvarsall.bat" x86 >nul
link /NOLOGO /INCREMENTAL:NO /DLL /DEBUG /subsystem:console 
/out:"bin\msvc-9.0\debug\threading-multi\std_pair_ext.pyd" 
/IMPLIB:"bin\msvc-9.0\debug\threading-multi\std_pair_ext.lib" 
/LIBPATH:"C:\Python25\libs" 
@"bin\msvc-9.0\debug\threading-multi\std_pair_ext.pyd.rsp"
         if %ERRORLEVEL% NEQ 0 EXIT %ERRORLEVEL%
         if exist 
"bin\msvc-9.0\debug\threading-multi\std_pair_ext.pyd.manifest" (
             mt -nologo -manifest 
"bin\msvc-9.0\debug\threading-multi\std_pair_ext.pyd.manifest" 
"-outputresource:bin\msvc-9.0\debug\threading-multi\std_pair_ext.pyd;2"
         )



From kuroutadori at gmail.com  Thu Aug  7 16:48:42 2008
From: kuroutadori at gmail.com (Geoff)
Date: Thu, 07 Aug 2008 23:48:42 +0900
Subject: [C++-sig] Wrapping functions that take void* parameters
Message-ID: <489B0B4A.60304@gmail.com>

I'm trying to wrap a C++ library that is used for serial communications. 
It has functions such as:

ssize_t Read (void * const buffer, size_t count);

and

ssize_t Write (const void * const buffer, size_t count);

There are two issues here that I haven't been able to solve:

1) How to wrap the void pointers. Ideally, I'd like to be able to do 
things like pass in a buffer of raw data or a Python string and have it 
written. Half a day of searching hasn't turned up anything even vaguely 
related to this more recent than 2003, so I was hoping there may have 
been some progress since then.

2) The Read() method expects buffer to point to some pre-allocated space 
into which it can put the data that is read from the serial port. I'm 
not sure how I could wrap this, because just passing in a Python string 
or something doesn't guarantee that that string has enough space to 
store the data. My first instinct would be to write a wrapper function 
that allocates space based on count, then calls Read(), passing in a 
pointer to that space, but I don't know how I could return that to the 
Python side.

Thanks,
Geoff



From gjcarneiro at gmail.com  Thu Aug  7 17:17:31 2008
From: gjcarneiro at gmail.com (Gustavo Carneiro)
Date: Thu, 7 Aug 2008 16:17:31 +0100
Subject: [C++-sig] Wrapping functions that take void* parameters
In-Reply-To: <489B0B4A.60304@gmail.com>
References: <489B0B4A.60304@gmail.com>
Message-ID: <a467ca4f0808070817y1051ae9ey85f462d6d5bf4cbb@mail.gmail.com>

2008/8/7 Geoff <kuroutadori at gmail.com>

> I'm trying to wrap a C++ library that is used for serial communications. It
> has functions such as:
>
> ssize_t Read (void * const buffer, size_t count);
>
> and
>
> ssize_t Write (const void * const buffer, size_t count);
>
> There are two issues here that I haven't been able to solve:
>
> 1) How to wrap the void pointers. Ideally, I'd like to be able to do things
> like pass in a buffer of raw data or a Python string and have it written.
> Half a day of searching hasn't turned up anything even vaguely related to
> this more recent than 2003, so I was hoping there may have been some
> progress since then.


>
> 2) The Read() method expects buffer to point to some pre-allocated space
> into which it can put the data that is read from the serial port. I'm not
> sure how I could wrap this, because just passing in a Python string or
> something doesn't guarantee that that string has enough space to store the
> data. My first instinct would be to write a wrapper function that allocates
> space based on count, then calls Read(), passing in a pointer to that space,
> but I don't know how I could return that to the Python side.


Normally this type of memory buffers are mapped to Python strings, since
Python strings can hold NULL bytes.  Use PyString_FromStringAndSize to
convert from C to Python (return), and "s#" format chars in
PyArg_ParseTuple* to convert from Python to C.

-- 
Gustavo J. A. M. Carneiro
INESC Porto, Telecommunications and Multimedia Unit
"The universe is always one step beyond logic." -- Frank Herbert
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080807/352f5fb0/attachment.htm>

From dave at boostpro.com  Thu Aug  7 21:17:26 2008
From: dave at boostpro.com (David Abrahams)
Date: Thu, 07 Aug 2008 15:17:26 -0400
Subject: [C++-sig] static/dynamic boost::python conflicts?
In-Reply-To: <b8878c1b0808040018o3ee1b7d7i851aa082cfb7d970@mail.gmail.com>
	(Koichi Tamura's message of "Mon, 4 Aug 2008 16:18:45 +0900")
References: <b8878c1b0807230304k865e5c2q92026717c83e0522@mail.gmail.com>
	<b8878c1b0808040018o3ee1b7d7i851aa082cfb7d970@mail.gmail.com>
Message-ID: <873algprcp.fsf@mcbain.luannocracy.com>


on Mon Aug 04 2008, "Koichi Tamura" <hohehohe2-AT-gmail.com> wrote:

> Hello,
> I've experienced a trouble about boost.python linking (or Python itself) and I've separated what
> is causing trouble from my source.
> The condition is
>
> - An application (m) loads two libraries dynamically.
> - One of  them  (a)  creates a Python module, calls its init function, and uses it internally.
> boost.python is statically linked with (a).
> - The other (b) linkes boost.python dynamically
> - The application (m) loads (b) first, then loads (a)
>
> In this condition, (a)'s module is broken.

This must be Linux?

> If anybody experienced it, or knows what's going on and how to make the module working, please
> let me know.

The problem is that Linux has a single global symbol space.
If you're interested, you might look at the threads I posted in at
http://gcc.gnu.org/ml/gcc/2002-05/threads.html#02965

If you want to avoid that effect, you might think about using the GCC
-fvisibility option: http://www.nedprod.com/programs/gccvisibility.html

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From ope-devel at gmx.de  Thu Aug  7 21:18:16 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Thu, 07 Aug 2008 21:18:16 +0200
Subject: [C++-sig] embedded python and redirect
Message-ID: <489B4A78.9040204@gmx.de>

Hi

at this time I try to embed python and redirect stderr/stdout using the
test attached. That's are my first steps in boost.python ;-) Well, the
exception handling does not compile as expected, I've got a lot of
errors - why?

Anyway, this file is mixed from different tutorials from net. Unsolved
is the problem on executing the real python command (py::object vs.
PyObject* strategy)? Any hints here?

How can I check that redirection was successfully?

Thanks,
Olaf
-------------- next part --------------
A non-text attachment was scrubbed...
Name: boost_python.cpp
Type: text/x-c++src
Size: 2832 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080807/91b6dabd/attachment.cpp>

From NurEinDummy at gmx.de  Sun Aug 10 16:38:01 2008
From: NurEinDummy at gmx.de (Dummy)
Date: Sun, 10 Aug 2008 16:38:01 +0200
Subject: [C++-sig] Boost::Python -- Huuge warning when using 'extract'.
Message-ID: <3152E7C87ECF40558A3AE312CCEC5BA2@cellardoor>


Hi, list! :D

A boost::python warning drives me crazy... as it's huuuge...

I'm using VS 2005 and the warning rises when doing an extract like here:

		list huhu;
		huhu.append(4);
		
		int mfour = extract<int>(huhu[0]);


... I pasted the whole warning here:
http://rafb.net/p/YGyPyg88.html



What can I do to get rid of it?



From konrad.hinsen at laposte.net  Mon Aug 11 15:01:14 2008
From: konrad.hinsen at laposte.net (Konrad Hinsen)
Date: Mon, 11 Aug 2008 15:01:14 +0200
Subject: [C++-sig] boost python on os x 10.5, intel
Message-ID: <F635E479-BB4A-4811-BBD2-50F2EF4DF11E@laposte.net>

On 06.08.2008, at 11:57, David Abrahams <dave at boostpro.com> wrote:

> on Sat Aug 02 2008, Michael Rogers <m.rogers-AT-hef.ru.nl> wrote:
>
> > This link pretty much describes what is happening to me.
> >
> > http://article.gmane.org/gmane.comp.lib.boost.build/18302
> >
> > Does anyone else experience this? Anyone know how to fix it?
>
> Looks like a bug in either Darwin's linker (hangs) or in its
> implementation of execwait.  bjam is just waiting for the linker  
> process
> to complete.
>
> My suggestion, to find out which: run bjam again with -n -a (build
> nothing) or -d+2 (rebuild what's out of date) and look at the linker
> command that bjam is issuing.  Then try it yourself from a shell
> prompt.  If it hangs, you know it's the linker.  If not, it's either a
> bug in execwait or a bug in the way bjam is using it.

I am running into exactly the same problem with Boost 1.35.0 under  
MacOS 10.4.11, both Intel and PowerPC platformst.

Following your suggestion, I ran bjam with -n -a and found the  
following linker command:

g++ -dynamiclib -install_name "extending.so" -L"/Library/Frameworks/ 
Python.framework/Versions/2.5/lib" -L"/Library/Frameworks/ 
Python.framework/Versions/2.5/lib/python2.5/config" -o "bin/darwin/ 
debug/macosx-version-10.4/extending.so" "bin/darwin/debug/macosx- 
version-10.4/extending.o" "/usr/local/boost_1_35_0/bin.v2/libs/python/ 
build/darwin/debug/macosx-version-10.4/libboost_python-d-1_35.dylib"   
-lpython2.5    -g -isysroot /Developer/SDKs/MacOSX10.4u.sdk -mmacosx- 
version-min=10.4

When I run this command line directly from the shell, it terminates  
immediately without any error message, producing the library file ./ 
bin/darwin/debug/macosx-version-10.4/extending.so

I started the build procedure from scratch and interrupted bjam when  
it got stuck. I checked that the library had already been produced,  
and that it is identical to the one that I got when running g++  
directly.

Restarting bjam after interrupting it makes it hang again  
immediately, trying to execute the command

DYLD_LIBRARY_PATH=/Library/Frameworks/Python.framework/Versions/2.5/ 
lib:/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/ 
config:/usr/local/boost_1_35_0/bin.v2/libs/python/build/darwin/debug/ 
macosx-version-10.4:/usr/local/boost_1_35_0/libs/python/example/ 
quickstart/bin/darwin/debug/macosx-version-10.4:$DYLD_LIBRARY_PATH
export DYLD_LIBRARY_PATH

So it looks like the problem is in bjam, not in the linker. It still  
in the current development trunk of boost, so I filed a bug report:

http://svn.boost.org/trac/boost/ticket/2185

Konrad.



From sipickles at hotmail.com  Mon Aug 11 15:19:46 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Mon, 11 Aug 2008 14:19:46 +0100
Subject: [C++-sig] Extending Stackless Python
Message-ID: <BAY114-DAV7C73F5C65E4F539A387D3BE710@phx.gbl>

Hi,

I am quite new to Boost::python. I have got the examples in the boost 
directory to work with a standard python installation, however I want to 
use boost::python with stackless python.

When I install Stackless, I then get errors when running bjam:

extending.cpp
C:\Python25\Include\Python.h(79) : fatal error C1083: Cannot open 
include file:
'stackless.h': No such file or directory

Stackless.h is in C:/Python25/stackless. My MSVC IDE can see the file 
okay, it is in the include path.

What am I missing here?

Many thanks

Simon Pickles

-- 
Linux Counter: User# 424693 





From dave at boostpro.com  Mon Aug 11 19:02:58 2008
From: dave at boostpro.com (David Abrahams)
Date: Mon, 11 Aug 2008 09:02:58 -0800
Subject: [C++-sig] boost python on os x 10.5, intel
In-Reply-To: <F635E479-BB4A-4811-BBD2-50F2EF4DF11E@laposte.net> (Konrad
	Hinsen's message of "Mon, 11 Aug 2008 15:01:14 +0200")
References: <F635E479-BB4A-4811-BBD2-50F2EF4DF11E@laposte.net>
Message-ID: <87abfjmqm5.fsf@mcbain.luannocracy.com>


on Mon Aug 11 2008, Konrad Hinsen <konrad.hinsen-AT-laposte.net> wrote:

> So it looks like the problem is in bjam, not in the linker. It still
> in the current development trunk of boost, so I filed a bug report:
>
> http://svn.boost.org/trac/boost/ticket/2185

Thanks for your work on this.

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From william.marie at gmail.com  Tue Aug 12 11:58:20 2008
From: william.marie at gmail.com (=?UTF-8?Q?William_Mari=C3=A9?=)
Date: Tue, 12 Aug 2008 02:58:20 -0700 (PDT)
Subject: [C++-sig]  Python Static link
Message-ID: <18941050.post@talk.nabble.com>


Hi,

I would like to link statically python25.lib to boost-python.lib and then
link statically boost-python.lib to my script.lib. i can't find a way to do
that.

Could someone tell me if it's possible and if yes how to do it ?
Thanks

William
-- 
View this message in context: http://www.nabble.com/Python-Static-link-tp18941050p18941050.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From sipickles at hotmail.com  Tue Aug 12 12:18:53 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Tue, 12 Aug 2008 11:18:53 +0100
Subject: [C++-sig] Building Boost::Stackless_Python - undefined symbol:
 PyUnicodeUCS4_FromEncodedObject
Message-ID: <BAY114-W340E00BB79460FE00F82B9BE700@phx.gbl>


Hi,

Sorry for the cross post - I am lost over whether my problem resides 
with my Boost or Stackless Python build.

I am attempting to extend Stackless Python on Ubuntu 8.04.

I have built the examples in boost_1_35_0/libs/python/example/quickstart 
with bjam. This passes the tests if I run it with the default Ubuntu 
python installation (/usr/lib/python25). However it fails if I run it 
with the stackless python installation (/usr/local/lib/python25).

I get the error:

Trying:
    from extending import *
Expecting nothing
**********************************************************************
File "/home/simon/Dev/testSpace/boostPython/test_extending.py", line 6, 
in test_extending
Failed example:
    from extending import *
Exception raised:
    Traceback (most recent call last):
      File "/usr/local/lib/python2.5/doctest.py", line 1228, in __run
        compileflags, 1) in test.globs
      File "", line 1, in 
        from extending import *
    ImportError: /usr/local/lib/libboost_python-gcc42-d-1_35.so.1.35.0: 
undefined symbol: PyUnicodeUCS4_FromEncodedObject


I built libboost_python-gcc42-d-1_35.so.1.35.0 by pointing bjam at my 
stackless python installation:

sudo bjam -sPYTHON_ROOT=/usr/local/lib/python25 -sPYTHON_VERSION=2.5 stage

It is obviously some sort of Unicode error. I know ubuntu comes with 
unicode support, so is it a problem with my stackless build? How do I 
enable unicode when building stackless or test for its support?

Thanks

Si

-- 
Linux Counter: User# 424693 



_________________________________________________________________
Get Hotmail on your mobile from Vodafone 
http://clk.atdmt.com/UKM/go/107571435/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080812/3e07d9ab/attachment.htm>

From niki at vintech.bg  Tue Aug 12 12:38:21 2008
From: niki at vintech.bg (niki)
Date: Tue, 12 Aug 2008 13:38:21 +0300
Subject: [C++-sig] Building Boost::Stackless_Python - undefined symbol:
 PyUnicodeUCS4_FromEncodedObject
In-Reply-To: <BAY114-W340E00BB79460FE00F82B9BE700@phx.gbl>
References: <BAY114-W340E00BB79460FE00F82B9BE700@phx.gbl>
Message-ID: <48A1681D.3000504@vintech.bg>

Simon Pickles wrote:
> ImportError: /usr/local/lib/libboost_python-gcc42-d-1_35.so.1.35.0:
> undefined symbol: PyUnicodeUCS4_FromEncodedObject

There are 2 kind of Unicode python builds - UCS2 and UCS4. Check that 
python and stackless and boost use same kind.

HTH

Niki Spahiev


From ope-devel at gmx.de  Tue Aug 12 21:14:25 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Tue, 12 Aug 2008 21:14:25 +0200
Subject: [C++-sig] [boost.python] general questions
Message-ID: <48A1E111.8000400@gmx.de>

Hi,

the basic of redirection works in the example attached. Anyway, there
are at least two points:

$ LANG=en g++ -I /usr/include/python2.5 boost_python.cpp -o boost_python
-l python2.5 -l boost_python && ./boost_python
CTor py_redirector("out: ")
CTor py_redirector("err: ")
Testing embedded python:
out: Hello
out: World
err: Traceback (most recent call last):
err:   File "<string>", line 1, in <module>
err: NameError
err: :
err: name 'Hello' is not defined
DTor py_redirector("out: ")

The line break after NameError comes probably from internal symbol
lookup and is '\0' terminated string, isn't it? How can I beautify this
even more resp. correct it? Any hints to this?

Where is my destructor of the 2nd py_redirector("err: ") ??? It occours
only on enabled Py_Finalize(). Seems not to be a good style to hold all
created objects during lifetime (memory leaks ....) until application quits.

How can I set/register some callbacks in C++ (at best at construction
time on C++ call side) inside py_redirector to get the write output to a
GUI (e.g. QTextBrowser)?


Thanks,
Olaf
-------------- next part --------------
A non-text attachment was scrubbed...
Name: boost_python.cpp
Type: text/x-c++src
Size: 4916 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080812/bc7c5c72/attachment.cpp>

From sipickles at hotmail.com  Wed Aug 13 10:07:28 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Wed, 13 Aug 2008 09:07:28 +0100
Subject: [C++-sig] bjam output
Message-ID: <BAY114-W473D8A16A909D9EBDF3F89BE730@phx.gbl>


Hi,

I have finally got my head round boost python and I'm loving it. I am on 
Ubuntu Hardy, stackless python 2.5, gcc 4.2.3.

Two questions:

1) how can I use bjam to output the built libraries into a different 
folder to /bin/gcc-4.2.3/debug?

2) is this actually a debug library? if so, how do I create release 
versions?

Many thanks

Simon

-- 
Linux Counter: User# 424693 



_________________________________________________________________
Win New York holidays with Kellogg?s & Live Search
http://clk.atdmt.com/UKM/go/107571440/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080813/45f4742c/attachment.htm>

From paul.bilokon at lehman.com  Thu Aug 14 10:24:50 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Thu, 14 Aug 2008 09:24:50 +0100
Subject: [C++-sig] Boost.Python: wrapping a method returning const
	std::string &
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B15@lopets1exms003.leh.lbcorp.lehman.com>

Hi c++-sig,

Sorry if this is supposed to be glaringly obvious, but I thought
std::string types were pretty much fully supported in Boost.Python
(looking at Boost 1.35.0).

I'm trying to wrap a method

    const std::string& getName( ) const;

This should be straightforward, right?

I'm doing it like this:

    class_<Ccy>("Ccy", init<>())
        .def("getName", &Ccy::getName)
    ;

But in return I'm getting

Error	6	error C2027: use of undefined type
'boost::python::detail::specify_a_return_value_policy_to_wrap_functions_
returning<T>'
E:\dev\external\boost\boost_1_35_0\boost\python\detail\caller.hpp
223	

Is this because the string is returned by reference? Do I have to wrap
the method? If so, I would be very much obliged if you could direct me
to an example.

Many thanks,
Paul
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From ope-devel at gmx.de  Thu Aug 14 10:45:46 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Thu, 14 Aug 2008 10:45:46 +0200
Subject: [C++-sig] Boost.Python: wrapping a method returning
 const	std::string &
In-Reply-To: <BBABA8164420CB43ABEA9DC81FE7A884065D1B15@lopets1exms003.leh.lbcorp.lehman.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B15@lopets1exms003.leh.lbcorp.lehman.com>
Message-ID: <48A3F0BA.7050208@gmx.de>

Bilokon, Paul schrieb:
> Hi c++-sig,
> 
> Sorry if this is supposed to be glaringly obvious, but I thought
> std::string types were pretty much fully supported in Boost.Python
> (looking at Boost 1.35.0).
> 
> I'm trying to wrap a method
> 
>     const std::string& getName( ) const;
> 
> This should be straightforward, right?
> 
> I'm doing it like this:
> 
>     class_<Ccy>("Ccy", init<>())
>         .def("getName", &Ccy::getName)
>     ;
> 
> But in return I'm getting
> 
> Error	6	error C2027: use of undefined type
> 'boost::python::detail::specify_a_return_value_policy_to_wrap_functions_
> returning<T>'
> E:\dev\external\boost\boost_1_35_0\boost\python\detail\caller.hpp
> 223	
> 
> Is this because the string is returned by reference? Do I have to wrap
> the method? If so, I would be very much obliged if you could direct me
> to an example.

Cool compiler error, it says what he expected: see 
http://www.boost.org/doc/libs/1_35_0/libs/python/doc/v2/copy_const_reference.html. 
Please google, why this is you friend.


Regards,
Olaf



From sipickles at hotmail.com  Thu Aug 14 17:55:19 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Thu, 14 Aug 2008 16:55:19 +0100
Subject: [C++-sig] ImportError: dynamic module does not define init function
Message-ID: <BAY114-DAV529CE17E363832A3301A1BE720@phx.gbl>

Hi,

Sometimes I come across this error when I try to import a library:

ImportError: dynamic module does not define init function 
(initHybridBulletObject)

What exactly does it mean, and how can I fix this?

My code looks like this, and compiles with bjam:


class CHybridBulletObject
{ .... }

BOOST_PYTHON_MODULE(pyboost)
{
    using namespace boost::python;
    class_<CHybridBulletObject>("HybridBulletObject", init<std::string>())
       .def("GetName", &CHybridBulletObject::GetName)
        ;
}


Thanks for any advice

Simon

-- 
Linux User# 424693 






From nat at lindenlab.com  Thu Aug 14 17:56:18 2008
From: nat at lindenlab.com (Nat Goodspeed)
Date: Thu, 14 Aug 2008 11:56:18 -0400
Subject: [C++-sig] Wrapping functions that take void* parameters
In-Reply-To: <489B0B4A.60304@gmail.com>
References: <489B0B4A.60304@gmail.com>
Message-ID: <48A455A2.2030002@lindenlab.com>

Geoff wrote:
> I'm trying to wrap a C++ library that is used for serial communications. 
> It has functions such as:
> 
> ssize_t Read (void * const buffer, size_t count);
> 
> and
> 
> ssize_t Write (const void * const buffer, size_t count);
> 
> There are two issues here that I haven't been able to solve:
> 
> 1) How to wrap the void pointers. Ideally, I'd like to be able to do 
> things like pass in a buffer of raw data or a Python string and have it 
> written. Half a day of searching hasn't turned up anything even vaguely 
> related to this more recent than 2003, so I was hoping there may have 
> been some progress since then.
> 
> 2) The Read() method expects buffer to point to some pre-allocated space 
> into which it can put the data that is read from the serial port. I'm 
> not sure how I could wrap this, because just passing in a Python string 
> or something doesn't guarantee that that string has enough space to 
> store the data. My first instinct would be to write a wrapper function 
> that allocates space based on count, then calls Read(), passing in a 
> pointer to that space, but I don't know how I could return that to the 
> Python side.

Isn't this what the Python buffer object is for?
http://www.python.org/doc/2.5.2/lib/typesseq.html
http://docs.python.org/api/bufferObjects.html


From seefeld at sympatico.ca  Thu Aug 14 18:00:57 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Thu, 14 Aug 2008 12:00:57 -0400
Subject: [C++-sig] ImportError: dynamic module does not define init
	function
In-Reply-To: <BAY114-DAV529CE17E363832A3301A1BE720@phx.gbl>
References: <BAY114-DAV529CE17E363832A3301A1BE720@phx.gbl>
Message-ID: <48A456B9.8010501@sympatico.ca>

Simon Pickles wrote:
> Hi,
>
> Sometimes I come across this error when I try to import a library:
>
> ImportError: dynamic module does not define init function 
> (initHybridBulletObject)
>
> What exactly does it mean, and how can I fix this?

Python requires a module 'Foo' to provide an 'initFoo' function as entry 
point.
This is generated by BOOST_PYTHON_MODULE, so make sure the name you pass 
to that matches the name of your module.

HTH,
       Stefan

-- 

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



From paul.bilokon at lehman.com  Thu Aug 14 17:51:58 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Thu, 14 Aug 2008 16:51:58 +0100
Subject: [C++-sig] Boost.Python: wrapping a method returning
	const	std::string &
In-Reply-To: <48A3F0BA.7050208@gmx.de>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B15@lopets1exms003.leh.lbcorp.lehman.com>
	<48A3F0BA.7050208@gmx.de>
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B1F@lopets1exms003.leh.lbcorp.lehman.com>

Hi Olaf,

Thanks a lot. This was certainly an informative error message.

As per your reply, the error was resolved by replacing

        .def("getName", &Ccy::getName)

with

        .def("getName", &Ccy::getName,
boost::python::return_value_policy<boost::python::copy_const_reference>(
)) 

Regards,
Paul

-----Original Message-----
From: c++-sig-bounces at python.org [mailto:c++-sig-bounces at python.org] On
Behalf Of Olaf Peter
Sent: Thursday, August 14, 2008 9:46 AM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Boost.Python: wrapping a method returning const
std::string &

Bilokon, Paul schrieb:
> Hi c++-sig,
> 
> Sorry if this is supposed to be glaringly obvious, but I thought 
> std::string types were pretty much fully supported in Boost.Python 
> (looking at Boost 1.35.0).
> 
> I'm trying to wrap a method
> 
>     const std::string& getName( ) const;
> 
> This should be straightforward, right?
> 
> I'm doing it like this:
> 
>     class_<Ccy>("Ccy", init<>())
>         .def("getName", &Ccy::getName)
>     ;
> 
> But in return I'm getting
> 
> Error	6	error C2027: use of undefined type
> 'boost::python::detail::specify_a_return_value_policy_to_wrap_function
> s_
> returning<T>'
> E:\dev\external\boost\boost_1_35_0\boost\python\detail\caller.hpp
> 223	
> 
> Is this because the string is returned by reference? Do I have to wrap

> the method? If so, I would be very much obliged if you could direct me

> to an example.

Cool compiler error, it says what he expected: see
http://www.boost.org/doc/libs/1_35_0/libs/python/doc/v2/copy_const_refer
ence.html. 
Please google, why this is you friend.


Regards,
Olaf

_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From sipickles at hotmail.com  Fri Aug 15 12:17:20 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Fri, 15 Aug 2008 11:17:20 +0100
Subject: [C++-sig] ImportError: dynamic module does not define
	init	function
Message-ID: <BAY114-W687727899C1077DFAE2B1BE6D0@phx.gbl>


Ah I see it.

Thanks, getting my head around this now...

Stefan Seefeld wrote:
> Simon Pickles wrote:
>> Hi,
>>
>> Sometimes I come across this error when I try to import a library:
>>
>> ImportError: dynamic module does not define init function 
>> (initHybridBulletObject)
>>
>> What exactly does it mean, and how can I fix this?
>
> Python requires a module 'Foo' to provide an 'initFoo' function as 
> entry point.
> This is generated by BOOST_PYTHON_MODULE, so make sure the name you 
> pass to that matches the name of your module.
>
> HTH,
>       Stefan
>

-- 
Linux Counter: User# 424693 



_________________________________________________________________
Win New York holidays with Kellogg?s & Live Search 
http://clk.atdmt.com/UKM/go/107571440/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080815/28244f96/attachment.htm>

From sipickles at hotmail.com  Fri Aug 15 12:18:21 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Fri, 15 Aug 2008 11:18:21 +0100
Subject: [C++-sig] [SOLVED]Re:  bjam output
Message-ID: <BAY114-W4452C5787C1795E420CA56BE6D0@phx.gbl>


Ah,

Its as simple as adding this to the project section of the jamroot:

.

Simon Pickles wrote:
> Hi,
>
> I have finally got my head round boost python and I'm loving it. I am on
> Ubuntu Hardy, stackless python 2.5, gcc 4.2.3.
>
> Two questions:
>
> 1) how can I use bjam to output the built libraries into a different
> folder to /bin/gcc-4.2.3/debug?
>
> 2) is this actually a debug library? if so, how do I create release
> versions?
>
> Many thanks
>
> Simon
>
> -- 
> Linux Counter: User# 424693
>
>
>
> ------------------------------------------------------------------------
> Get fish-slapping on Messenger! Play Now 
> 
> ------------------------------------------------------------------------
>
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>   

-- 
Linux Counter: User# 424693 



_________________________________________________________________
Win New York holidays with Kellogg?s & Live Search
http://clk.atdmt.com/UKM/go/107571440/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080815/05226976/attachment.htm>

From sipickles at hotmail.com  Fri Aug 15 12:23:32 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Fri, 15 Aug 2008 11:23:32 +0100
Subject: [C++-sig] Overloading member function problem
Message-ID: <BAY114-W5303BE513247C4274EF60BE6D0@phx.gbl>


Hi,

Given a class:

///////////////////
class CLog
{
public:
void Log( const boost::format& );
void Log( const std::string& );
};

void (CLog::*LogS)(const std::string&) = &CLog::Log;
void (CLog::*LogF)(const boost::format&) = &CLog::Log;
////////////////////

And this module Macro:

///////////////////

BOOST_PYTHON_MODULE(HybridLog)
{
using namespace boost::python;

class_("logInterface")
.def("Log", &LogS)
.def("Log", &LogF)
;
}
///////////////////

How come I am getting these errors:

/boost/python/class.hpp:536: error: no matching function for call to 
?get_signature(void (CLog::**&)(const std::string&), CLog*)?
/boost/python/class.hpp:536: error: no matching function for call to 
?get_signature(void (CLog::**&)(const std::string&), CLog*)?

I've tried all sorts of variations!

Thanks

Simon

-- 
Linux Counter: User# 424693 



_________________________________________________________________
Get Hotmail on your mobile from Vodafone 
http://clk.atdmt.com/UKM/go/107571435/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080815/807bba1d/attachment.htm>

From sipickles at hotmail.com  Fri Aug 15 12:42:00 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Fri, 15 Aug 2008 11:42:00 +0100
Subject: [C++-sig] [SOLVED]Re:  bjam output
Message-ID: <BAY114-W1285F1D5E381E9DB6C076FBE6D0@phx.gbl>


hmm

mailman or thunderbird doesn't like text which looks like HTML tags

should have read:

< location>.

or

(less-than)location(greater-than)period

Simon Pickles wrote:
> Ah,
>
> Its as simple as adding this to the project section of the jamroot:
>
> .
>
> Simon Pickles wrote:
>> Hi,
>>
>> I have finally got my head round boost python and I'm loving it. I am on
>> Ubuntu Hardy, stackless python 2.5, gcc 4.2.3.
>>
>> Two questions:
>>
>> 1) how can I use bjam to output the built libraries into a different
>> folder to /bin/gcc-4.2.3/debug?
>>
>> 2) is this actually a debug library? if so, how do I create release
>> versions?
>>
>> Many thanks
>>
>> Simon
>>
>> --
>> Linux Counter: User# 424693
>>
>>
>>
>> ------------------------------------------------------------------------
>> Get fish-slapping on Messenger! Play Now
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> C++-sig mailing list
>> C++-sig at python.org
>> http://mail.python.org/mailman/listinfo/c++-sig
>>
>
> -- 
> Linux Counter: User# 424693
>
>
>
> ------------------------------------------------------------------------
> Get fish-slapping on Messenger! Play Now 
> 
> ------------------------------------------------------------------------
>
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>   

-- 
Linux Counter: User# 424693 



_________________________________________________________________
Win a voice over part with Kung Fu Panda & Live Search?? and?? 100?s of Kung Fu Panda prizes to win with Live Search
http://clk.atdmt.com/UKM/go/107571439/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080815/031df43d/attachment.htm>

From seefeld at sympatico.ca  Fri Aug 15 13:29:25 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Fri, 15 Aug 2008 07:29:25 -0400
Subject: [C++-sig] Overloading member function problem
In-Reply-To: <BAY114-W5303BE513247C4274EF60BE6D0@phx.gbl>
References: <BAY114-W5303BE513247C4274EF60BE6D0@phx.gbl>
Message-ID: <48A56895.30401@sympatico.ca>

Simon Pickles wrote:
>
> How come I am getting these errors:
>
> /boost/python/class.hpp:536: error: no matching function for call to
> ?get_signature(void (CLog::**&)(const std::string&), CLog*)?
> /boost/python/class.hpp:536: error: no matching function for call to
> ?get_signature(void (CLog::**&)(const std::string&), CLog*)?

While boost.python knows (defines) how to convert std::string, it 
doesn't seem to know about boost::format. Try wrapping boost::format 
first, then your overload should be resolvable.

HTH,
       Stefan

-- 

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



From sipickles at hotmail.com  Fri Aug 15 14:09:37 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Fri, 15 Aug 2008 13:09:37 +0100
Subject: [C++-sig] Overloading member function problem
Message-ID: <BAY114-W33432F902E16CEFF716FD6BE6D0@phx.gbl>




Stefan Seefeld wrote:
> Simon Pickles wrote:
>>
>> How come I am getting these errors:
>>
>> /boost/python/class.hpp:536: error: no matching function for call to
>> ?get_signature(void (CLog::**&)(const std::string&), CLog*)?
>> /boost/python/class.hpp:536: error: no matching function for call to
>> ?get_signature(void (CLog::**&)(const std::string&), CLog*)?
>
> While boost.python knows (defines) how to convert std::string, it 
> doesn't seem to know about boost::format. Try wrapping boost::format 
> first, 

What does 'wrapping' mean? :)

As in BOOST_PYTHON_MODULE wrapping?
> then your overload should be resolvable.
>
> HTH,
>       Stefan
>

-- 
Linux Counter: User# 424693 



_________________________________________________________________
Make a mini you on Windows Live Messenger!
http://clk.atdmt.com/UKM/go/107571437/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080815/a3cc07bc/attachment.htm>

From seefeld at sympatico.ca  Fri Aug 15 15:12:38 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Fri, 15 Aug 2008 09:12:38 -0400
Subject: [C++-sig] Overloading member function problem
In-Reply-To: <BAY114-W33432F902E16CEFF716FD6BE6D0@phx.gbl>
References: <BAY114-W33432F902E16CEFF716FD6BE6D0@phx.gbl>
Message-ID: <48A580C6.4000908@sympatico.ca>

Simon Pickles wrote:
>
>
> Stefan Seefeld wrote:
> > Simon Pickles wrote:
> >>
> >> How come I am getting these errors:
> >>
> >> /boost/python/class.hpp:536: error: no matching function for call to
> >> ?get_signature(void (CLog::**&)(const std::string&), CLog*)?
> >> /boost/python/class.hpp:536: error: no matching function for call to
> >> ?get_signature(void (CLog::**&)(const std::string&), CLog*)?
> >
> > While boost.python knows (defines) how to convert std::string, it
> > doesn't seem to know about boost::format. Try wrapping boost::format
> > first,
>
> What does 'wrapping' mean? :)
class_<boost::format> format("format");

HTH,
       Stefan


-- 

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



From troy at resophonic.com  Fri Aug 15 16:04:33 2008
From: troy at resophonic.com (troy d. straszheim)
Date: Fri, 15 Aug 2008 10:04:33 -0400
Subject: [C++-sig] Overloading member function problem
In-Reply-To: <48A56895.30401@sympatico.ca>
References: <BAY114-W5303BE513247C4274EF60BE6D0@phx.gbl>
	<48A56895.30401@sympatico.ca>
Message-ID: <48A58CF1.402@resophonic.com>

Stefan Seefeld wrote:
> Simon Pickles wrote:
>>
>> How come I am getting these errors:
>>
>> /boost/python/class.hpp:536: error: no matching function for call to
>> ?get_signature(void (CLog::**&)(const std::string&), CLog*)?
>> /boost/python/class.hpp:536: error: no matching function for call to
>> ?get_signature(void (CLog::**&)(const std::string&), CLog*)?
> 
> While boost.python knows (defines) how to convert std::string, it 
> doesn't seem to know about boost::format. Try wrapping boost::format 
> first, then your overload should be resolvable.

boost.python won't complain about the inability to convert things
until *runtime*.  The needed converters could easily be someplace
that this compilation unit cannot see (like in a completely different
python wrapper library).

Looks to me like OP just forgot to pass a type argument to class_<>

> class_("logInterface")
> .def("Log", &LogS)
> .def("Log", &LogF)

-t


From sipickles at hotmail.com  Fri Aug 15 17:13:19 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Fri, 15 Aug 2008 16:13:19 +0100
Subject: [C++-sig] Overloading member function problem
Message-ID: <BAY114-W55D1DD80C9C14C18F78CC9BE6D0@phx.gbl>




>
>
> Looks to me like OP just forgot to pass a type argument to class_<>
>
>> class_("logInterface")
>> .def("Log", &LogS)
>> .def("Log", &LogF)
Apologies.

Code actually reads:

class_("logInterface")
.def("Log", &LogS)
.def("Log", &LogF)
>
>
>

-- 
Linux Counter: User# 424693 



_________________________________________________________________
Win a voice over part with Kung Fu Panda & Live Search?? and?? 100?s of Kung Fu Panda prizes to win with Live Search
http://clk.atdmt.com/UKM/go/107571439/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080815/48443ca1/attachment.htm>

From seefeld at sympatico.ca  Fri Aug 15 17:25:58 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Fri, 15 Aug 2008 11:25:58 -0400
Subject: [C++-sig] Overloading member function problem
In-Reply-To: <BAY114-W55D1DD80C9C14C18F78CC9BE6D0@phx.gbl>
References: <BAY114-W55D1DD80C9C14C18F78CC9BE6D0@phx.gbl>
Message-ID: <48A5A006.7090606@sympatico.ca>

Simon Pickles wrote:
>
>
> >
> >
> > Looks to me like OP just forgot to pass a type argument to class_<>
> >
> >> class_("logInterface")
> >> .def("Log", &LogS)
> >> .def("Log", &LogF)
> Apologies.
>
> Code actually reads:
>
> class_("logInterface")
> .def("Log", &LogS)
> .def("Log", &LogF)

I don't quite believe you. :-)

As Troy pointed out, class_ is a class template, so you need to provide 
a template argument list to make this code valid. For example:

class_<CLog>("logInterface")
....

HTH,
       Stefan

-- 

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



From sipickles at hotmail.com  Fri Aug 15 17:34:51 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Fri, 15 Aug 2008 16:34:51 +0100
Subject: [C++-sig] Overloading member function problem
Message-ID: <BAY114-W48433A0CE68A6FB92B9AA2BE6D0@phx.gbl>




Stefan Seefeld wrote:
> Simon Pickles wrote:
>>
>>
>> Stefan Seefeld wrote:
>>> Simon Pickles wrote:
>>>>
>>>> How come I am getting these errors:
>>>>
>>>> /boost/python/class.hpp:536: error: no matching function for call to
>>>> ?get_signature(void (CLog::**&)(const std::string&), CLog*)?
>>>> /boost/python/class.hpp:536: error: no matching function for call to
>>>> ?get_signature(void (CLog::**&)(const std::string&), CLog*)?
>>>
>>> While boost.python knows (defines) how to convert std::string, it
>>> doesn't seem to know about boost::format. Try wrapping boost::format
>>> first,
>>
>> What does 'wrapping' mean? :)
> class_ format("format");
>
Sorry, no effect. My module now looks like this:

class CLog
{
public:
    void Log(const std::string&);
    void Log(const boost::format&);
}

void    (CLog::*pLogString)(const std::string&)        = &CLog::Log;
void    (CLog::*pLogFormat)(const boost::format&)    = &CLog::Log;

BOOST_PYTHON_MODULE(Hybrid)
{
    using namespace boost::python;
   
    class_ format("format");
   
    class_("CLog")
        .def("Log", &pLogString)
        .def("Log", &pLogFormat)
       ;
       
}


No change in the error tho:

boost/python/class.hpp:536: error: no matching function for call to 
?get_signature(void (CLog::**&)(const std::string&), CLog*)?

Thanks

Si


-- 
Linux Counter: User# 424693 



_________________________________________________________________
Make a mini you on Windows Live Messenger!
http://clk.atdmt.com/UKM/go/107571437/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080815/098114dd/attachment.htm>

From sipickles at hotmail.com  Sat Aug 16 15:35:40 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Sat, 16 Aug 2008 14:35:40 +0100
Subject: [C++-sig] R_X86_64_PC32 error and -fPIC
Message-ID: <BAY114-DAV12B2D493D93AA27300E2EBBE6C0@phx.gbl>

Hi

I hope this is on topic. It may be a gcc problem, but its certainly 
caused by using bjam to link a c++ lib for boost::python. I should point 
out this whole area is new to me!

Here's my error:

...updating 1 target...
gcc.link.dll HybridBulletObject.so
/usr/bin/ld: HybridModule.o: relocation R_X86_64_PC32 against 
`(anonymous 
namespace)::CHybridBulletObject::Init(boost::shared_ptr<CLog>, 
std::basic_string<char, std::char_traits<char>, std::allocator<char> > 
const&, unsigned int const&, float const&, float const&, float const&, 
float const&, float const&, float const&)' can not be used when making a 
shared object; recompile with -fPIC
/usr/bin/ld: final link failed: Bad value
collect2: ld returned 1 exit status


Compile what with -fPIC? I have an AMD64 (ubuntu 8.04) system. I thought 
only shared libraries needed to be compiled with PIC, and I am only 
linking to static libraries in HybridBulletObject.so.

So, assuming its talking about the shared library output by bjam, I 
tried modifying my user-config.jam to:

using gcc : : : <linkflags>-fPIC ;
using python : 2.5 : /usr/local ; #stackless python location


but I can't see the -fPIC added to the command line produce by bjam:

    "g++"    -o "HybridBulletObject.so" -Wl,-h -Wl,HybridBulletObject.so 
-shared -Wl,--start-group "HybridModule.o" "HybridBulletObject.o" 
"Log.o" 
"/home/simon/Source/bullet-2.70/out/linux/optimize/libs/libbulletdynamics.a" 
"/home/simon/Source/boost_1_35_0/bin.v2/libs/python/build/gcc-4.2.3/debug/libboost_python-gcc42-d-1_35.so.1.35.0"  
-Wl,-Bstatic  -Wl,-Bdynamic -lutil -lpthread -ldl -Wl,--end-group -g

...failed gcc.link.dll HybridBulletObject.so...
...failed updating 1 target...

Thanks

Simon

-- 
Linux User# 424693 






From sipickles at hotmail.com  Sat Aug 16 16:10:01 2008
From: sipickles at hotmail.com (Simon Pickles)
Date: Sat, 16 Aug 2008 15:10:01 +0100
Subject: [C++-sig] R_X86_64_PC32 error and -fPIC
In-Reply-To: <BAY114-DAV12B2D493D93AA27300E2EBBE6C0@phx.gbl>
References: <BAY114-DAV12B2D493D93AA27300E2EBBE6C0@phx.gbl>
Message-ID: <BAY114-DAV5E6473D5ECC605DA27F4EBE6C0@phx.gbl>

sorry, that s

Simon Pickles wrote:
> Hi
>
> I hope this is on topic. It may be a gcc problem, but its certainly 
> caused by using bjam to link a c++ lib for boost::python. I should 
> point out this whole area is new to me!
>
> Here's my error:
>
> ...updating 1 target...
> gcc.link.dll HybridBulletObject.so
> /usr/bin/ld: HybridModule.o: relocation R_X86_64_PC32 against 
> `(anonymous 
> namespace)::CHybridBulletObject::Init(boost::shared_ptr<CLog>, 
> std::basic_string<char, std::char_traits<char>, std::allocator<char> > 
> const&, unsigned int const&, float const&, float const&, float const&, 
> float const&, float const&, float const&)' can not be used when making 
> a shared object; recompile with -fPIC
> /usr/bin/ld: final link failed: Bad value
> collect2: ld returned 1 exit status
>
>
> Compile what with -fPIC? I have an AMD64 (ubuntu 8.04) system. I 
> thought only shared libraries needed to be compiled with PIC, and I am 
> only linking to static libraries in HybridBulletObject.so.
>
> So, assuming its talking about the shared library output by bjam, I 
> tried modifying my user-config.jam to:
>
> using gcc : : : <linkflags>-fPIC ;
> using python : 2.5 : /usr/local ; #stackless python location
sorry, should have been:

using gcc : : : <cxxflags>-fPIC ;
using python : 2.5 : /usr/local ;

Now, I can see -fPIC on the g++ compile command line. No effect on the 
error tho.

Do I need to rebuild all my static libs with -fPIC too?

I notice that the boost python module will compile and link if I only 
add const functions. Add a normal function and it baulks. Weird.

Simon








From gjcarneiro at gmail.com  Sat Aug 16 16:38:28 2008
From: gjcarneiro at gmail.com (Gustavo Carneiro)
Date: Sat, 16 Aug 2008 15:38:28 +0100
Subject: [C++-sig] R_X86_64_PC32 error and -fPIC
In-Reply-To: <BAY114-DAV5E6473D5ECC605DA27F4EBE6C0@phx.gbl>
References: <BAY114-DAV12B2D493D93AA27300E2EBBE6C0@phx.gbl>
	<BAY114-DAV5E6473D5ECC605DA27F4EBE6C0@phx.gbl>
Message-ID: <a467ca4f0808160738g675334e4xdb9487492001274@mail.gmail.com>

2008/8/16 Simon Pickles <sipickles at hotmail.com>

> sorry, that s
>
> Simon Pickles wrote:
>
>> Hi
>>
>> I hope this is on topic. It may be a gcc problem, but its certainly caused
>> by using bjam to link a c++ lib for boost::python. I should point out this
>> whole area is new to me!
>>
>> Here's my error:
>>
>> ...updating 1 target...
>> gcc.link.dll HybridBulletObject.so
>> /usr/bin/ld: HybridModule.o: relocation R_X86_64_PC32 against `(anonymous
>> namespace)::CHybridBulletObject::Init(boost::shared_ptr<CLog>,
>> std::basic_string<char, std::char_traits<char>, std::allocator<char> >
>> const&, unsigned int const&, float const&, float const&, float const&, float
>> const&, float const&, float const&)' can not be used when making a shared
>> object; recompile with -fPIC
>> /usr/bin/ld: final link failed: Bad value
>> collect2: ld returned 1 exit status
>>
>>
>> Compile what with -fPIC? I have an AMD64 (ubuntu 8.04) system. I thought
>> only shared libraries needed to be compiled with PIC, and I am only linking
>> to static libraries in HybridBulletObject.so.
>>
>> So, assuming its talking about the shared library output by bjam, I tried
>> modifying my user-config.jam to:
>>
>> using gcc : : : <linkflags>-fPIC ;
>> using python : 2.5 : /usr/local ; #stackless python location
>>
> sorry, should have been:
>
> using gcc : : : <cxxflags>-fPIC ;
> using python : 2.5 : /usr/local ;
>
> Now, I can see -fPIC on the g++ compile command line. No effect on the
> error tho.
>
> Do I need to rebuild all my static libs with -fPIC too?


If you are building a shared library then all the libraries that you depend
on need to be shared as well, i.e. they all need to be compiled with -fPIC.
You can't link static libraries into shared libraries (the difference
between '.a archive' and 'static library' should be noted here, since you
could have a .a archive containing only objects compiled with -fPIC).


>
>
> I notice that the boost python module will compile and link if I only add
> const functions. Add a normal function and it baulks. Weird.


Not sure what "const function" means.  I only know the concepts of "const
method", and "static function".

-- 
Gustavo J. A. M. Carneiro
INESC Porto, Telecommunications and Multimedia Unit
"The universe is always one step beyond logic." -- Frank Herbert
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080816/8412be07/attachment.htm>

From troy at resophonic.com  Sat Aug 16 17:28:41 2008
From: troy at resophonic.com (troy d. straszheim)
Date: Sat, 16 Aug 2008 11:28:41 -0400
Subject: [C++-sig] error while formatting signature for <wrapped_class> arg
 is not a Python function
Message-ID: <48A6F229.3040201@resophonic.com>


I'm playing with Sphinx 0.4.3, trying to make some docs for wrapped classes.
I'm getting the error

   error while formatting signature for <wrapped_class> arg is not a Python function

where 'arg is not a Python function' is the TypeError you get when you call
inspect.getargspec() of a wrapped member function.
Appears to relate to this thread about doctests:

   http://mail.python.org/pipermail/c++-sig/2007-June/012481.html

Has anybody looked further at this?

-t



From dave at boostpro.com  Sat Aug 16 22:51:52 2008
From: dave at boostpro.com (David Abrahams)
Date: Sat, 16 Aug 2008 12:51:52 -0800
Subject: [C++-sig] error while formatting signature for <wrapped_class>
	arg is not a Python function
In-Reply-To: <48A6F229.3040201@resophonic.com> (troy d. straszheim's message
	of "Sat, 16 Aug 2008 11:28:41 -0400")
References: <48A6F229.3040201@resophonic.com>
Message-ID: <87k5egptsn.fsf@mcbain.luannocracy.com>


on Sat Aug 16 2008, "troy d. straszheim" <troy-AT-resophonic.com> wrote:

> I'm playing with Sphinx 0.4.3, trying to make some docs for wrapped classes.
> I'm getting the error
>
>   error while formatting signature for <wrapped_class> arg is not a Python function
>
> where 'arg is not a Python function' is the TypeError you get when you call
> inspect.getargspec() of a wrapped member function.
> Appears to relate to this thread about doctests:
>
>   http://mail.python.org/pipermail/c++-sig/2007-June/012481.html
>
> Has anybody looked further at this?

I'm afraid I never did look at it further.  I'd be very happy if the
the problem were solved, though :-)

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From william.marie at gmail.com  Mon Aug 18 12:01:48 2008
From: william.marie at gmail.com (=?UTF-8?Q?William_Mari=C3=A9?=)
Date: Mon, 18 Aug 2008 03:01:48 -0700 (PDT)
Subject: [C++-sig]  How to wrap a python function pointer ?
Message-ID: <19028640.post@talk.nabble.com>


Hi,

I would like to wrap a python function pointer to a void (*c++fun)() pointer
:

typedef void (*fct_void_void)( void );
void addTimer( double dStartIn, double dTickTime, u32 iTickCount,
fct_void_void bpFctCallback )
{
 .....
       
} 
BOOST_PYTHON_MODULE(ksys)
{
	using namespace boost::python;
	def("addTimer", &addTimer);
} 

here is my output :
1>        ..\..\libs\boost\boost/python/converter/registered.hpp(41) : see
reference to class template instantiation 'boost::add_cv<T>' being compiled
1>        with
1>        [
1>            T=void (void)
1>        ]
1>        ..\..\libs\boost\boost/python/converter/registered_pointee.hpp(27)
: see reference to class template instantiation
'boost::python::converter::registered<T>' being compiled
1>        with
1>        [
1>            T=void (void)
1>        ]
1>        ..\..\libs\boost\boost/python/converter/arg_from_python.hpp(269) :
see reference to class template instantiation
'boost::python::converter::registered_pointee<T>' being compiled
1>        with
1>        [
1>            T=void (__cdecl *)(void)
1>        ]
1>        ..\..\libs\boost\boost/python/converter/arg_from_python.hpp(268) :
while compiling class template member function
'boost::python::converter::pointer_arg_from_python<T>::pointer_arg_from_python(PyObject
*)'
1>        with
1>        [
1>            T=void (__cdecl *)(void)
1>        ]
1>        ..\..\libs\boost\boost/python/arg_from_python.hpp(70) : while
compiling class template member function
'boost::python::arg_from_python<T>::arg_from_python(PyObject *)'
1>        with
1>        [
1>            T=void (__cdecl *)(void)
1>        ]
1>        ..\..\libs\boost\boost/preprocessor/iteration/detail/local.hpp(43)
: see reference to class template instantiation
'boost::python::arg_from_python<T>' being compiled
1>        with
1>        [
1>            T=void (__cdecl *)(void)
1>        ]
1>        ..\..\libs\boost\boost/python/detail/caller.hpp(200) : while
compiling class template member function 'PyObject
*boost::python::detail::caller_arity<4>::impl<F,Policies,Sig>::operator
()(PyObject *,PyObject *)'
1>        with
1>        [
1>            F=void (__cdecl *)(double,double,u32,fct_void_void),
1>            Policies=boost::python::default_call_policies,
1>            Sig=boost::mpl::vector5<void,double,double,u32,fct_void_void>
1>        ]
1>        ..\..\libs\boost\boost/python/detail/caller.hpp(169) : see
reference to class template instantiation
'boost::python::detail::caller_arity<4>::impl<F,Policies,Sig>' being
compiled
1>        with
1>        [
1>            F=void (__cdecl *)(double,double,u32,fct_void_void),
1>            Policies=boost::python::default_call_policies,
1>            Sig=boost::mpl::vector5<void,double,double,u32,fct_void_void>
1>        ]
1>        ..\..\libs\boost\boost/python/make_function.hpp(37) : see
reference to class template instantiation
'boost::python::detail::caller<F,CallPolicies,Sig>' being compiled
1>        with
1>        [
1>            F=void (__cdecl *)(double,double,u32,fct_void_void),
1>            CallPolicies=boost::python::default_call_policies,
1>            Sig=boost::mpl::vector5<void,double,double,u32,fct_void_void>
1>        ]
1>        ..\..\libs\boost\boost/python/make_function.hpp(104) : see
reference to function template instantiation 'boost::python::api::object
boost::python::detail::make_function_aux<void(__cdecl
*)(double,double,u32,fct_void_void),boost::python::default_call_policies,boost::mpl::vector5<T0,T1,T2,T3,T4>>(F,const
CallPolicies &,const Sig &)' being compiled
1>        with
1>        [
1>            T0=void,
1>            T1=double,
1>            T2=double,
1>            T3=u32,
1>            T4=fct_void_void,
1>            F=void (__cdecl *)(double,double,u32,fct_void_void),
1>            CallPolicies=boost::python::default_call_policies,
1>            Sig=boost::mpl::vector5<void,double,double,u32,fct_void_void>
1>        ]
1>        ..\..\libs\boost\boost/python/def.hpp(82) : see reference to
function template instantiation 'boost::python::api::object
boost::python::make_function<void(__cdecl
*)(double,double,u32,fct_void_void)>(F)' being compiled
1>        with
1>        [
1>            F=void (__cdecl *)(double,double,u32,fct_void_void)
1>        ]
1>        ..\..\libs\boost\boost/python/def.hpp(91) : see reference to
function template instantiation 'boost::python::api::object
boost::python::detail::make_function1<void(__cdecl
*)(double,double,u32,fct_void_void)>(T,...)' being compiled
1>        with
1>        [
1>            T=void (__cdecl *)(double,double,u32,fct_void_void)
1>        ]
1>        ..\..\libs\script\src\ksysmodule.cpp(35) : see reference to
function template instantiation 'void boost::python::def<void(__cdecl
*)(double,double,u32,fct_void_void)>(const char *,Fn)' being compiled
1>        with
1>        [
1>            Fn=void (__cdecl *)(double,double,u32,fct_void_void)
1>        ]
1>..\..\libs\boost\boost/python/converter/registered.hpp(86) : error C2784:
'void boost::python::converter::detail::register_shared_ptr1(volatile const
T *)' : could not deduce template argument for 'volatile const T *' from
'void (__cdecl *)(void)'
1>        ..\..\libs\boost\boost/python/converter/registered.hpp(77) : see
declaration of 'boost::python::converter::detail::register_shared_ptr1'
1>        ..\..\libs\boost\boost/python/converter/registered.hpp(94) : see
reference to function template instantiation 'const
boost::python::converter::registration
&boost::python::converter::detail::registry_lookup2<T>(T (__cdecl &(__cdecl
*)(void)))' being compiled
1>        with
1>        [
1>            T=void (void)
1>        ]
1>        ..\..\libs\boost\boost/python/converter/registered.hpp(105) : see
reference to function template instantiation 'const
boost::python::converter::registration
&boost::python::converter::detail::registry_lookup1<T>(boost::type<T>)'
being compiled
1>        with
1>        [
1>            T=void (__cdecl &)(void)
1>        ]
1>        ..\..\libs\boost\boost/python/converter/registered.hpp(105) :
while compiling class template static data member 'const
boost::python::converter::registration
&boost::python::converter::detail::registered_base<T>::converters'
1>        with
1>        [
1>            T=void (__cdecl &)(void)
1>        ]
1>        ..\..\libs\boost\boost/python/converter/registered.hpp(44) : see
reference to class template instantiation
'boost::python::converter::detail::registered_base<T>' being compiled
1>        with
1>        [
1>            T=void (__cdecl &)(void)
1>        ]
1>        ..\..\libs\boost\boost/python/converter/registered_pointee.hpp(27)
: see reference to class template instantiation
'boost::python::converter::registered<T>' being compiled
1>        with
1>        [
1>            T=void (void)
1>        ]
1>        ..\..\libs\boost\boost/python/converter/arg_from_python.hpp(269) :
see reference to class template instantiation
'boost::python::converter::registered_pointee<T>' being compiled
1>        with
1>        [
1>            T=void (__cdecl *)(void)
1>        ]
1>        ..\..\libs\boost\boost/python/converter/arg_from_python.hpp(268) :
while compiling class template member function
'boost::python::converter::pointer_arg_from_python<T>::pointer_arg_from_python(PyObject
*)'
1>        with
1>        [
1>            T=void (__cdecl *)(void)
1>        ]

Could someone tell me how i can do that please ?
Thank you

William
-- 
View this message in context: http://www.nabble.com/How-to-wrap-a-python-function-pointer---tp19028640p19028640.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From paul.bilokon at lehman.com  Mon Aug 18 16:41:37 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Mon, 18 Aug 2008 15:41:37 +0100
Subject: [C++-sig] Multiple Constructors
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B2E@lopets1exms003.leh.lbcorp.lehman.com>

Hi,

I'm wondering if there is an example that demonstrates how one should
deal with classes with multiple constructors with different signatures,
e.g.

    scope in_Term = class_<Term>("Term", init<>())
        .def(init<const Term::Foo, Bar>())
        .def(init<const Term::Foo, Bar, bool>())
        .def(init<const std::string &>())

I don't think this approach is valid.

Many thanks,
Paul
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From paul.bilokon at lehman.com  Mon Aug 18 16:45:46 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Mon, 18 Aug 2008 15:45:46 +0100
Subject: [C++-sig] "Returning" to Scope
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B2F@lopets1exms003.leh.lbcorp.lehman.com>

Hi,

I'm trying to figure out how I can "return" to a previously defined
scope. Since the current scope is normally set via construction, this is
the only method I could think of:

#include <iostream>

#include <boost/python.hpp>

using std::cout;
using std::endl;

struct Foo
{
    struct Bar
    {
        int barVar1;
        static const int barVar2 = 22;

        Bar() : barVar1(18)
        {
        }
    };

    int fooVar1;
    int fooVar2;
    static const int fooVar3 = 7;

    Bar fooBar;

    enum FooEnum
    {
        RED,
        GREEN,
        BLUE,
    };

    Foo() : fooVar1(2), fooVar2(3)
    {
    }
};

struct Baz
{
    int bazVar1;
    static const int bazVar2 = 13;

    Baz() : bazVar1(11)
    {
    }
};

using namespace boost::python;

BOOST_PYTHON_MODULE(scoping)
{
    scope in_scoping;

    scope in_Foo = class_<Foo>("Foo")
        .def_readwrite("fooVar1", &Foo::fooVar1)
        .def_readwrite("fooVar2", &Foo::fooVar2)
        .def_readonly("fooVar3", &Foo::fooVar3)
        ;

    // The following line is not strictly speaking necessary, since
"scope
    // in_Foo = ..." would have set the current scope to in_Foo.
However,
    // we prefer explicit to implicit
    scope in_Foo_1(in_Foo);

    class_<Foo::Bar>("Bar")
        .def_readwrite("barVar1", &Foo::Bar::barVar1)
        .def_readonly("barVar2", &Foo::Bar::barVar2)
        ;

    scope in_scoping_1(in_scoping);

    class_<Baz>("Baz")
        .def_readwrite("bazVar1", &Baz::bazVar1)
        .def_readonly("bazVar2", &Baz::bazVar2)
        ;

    scope in_Foo_2(in_Foo);

    enum_<Foo::FooEnum>("FooEnum")
        .value("RED", Foo::RED)
        .value("GREEN", Foo::GREEN)
        .value("BLUE", Foo::BLUE)
        ;
}


However, this approach is pretty messy, particularly keeping track of
..._1, ..._2, ..._3, etc.

Is there a neater approach?

Many thanks,
Paul

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

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From troy at resophonic.com  Mon Aug 18 17:07:45 2008
From: troy at resophonic.com (troy d. straszheim)
Date: Mon, 18 Aug 2008 11:07:45 -0400
Subject: [C++-sig] error while formatting signature for <wrapped_class>
 arg is not a Python function
In-Reply-To: <87k5egptsn.fsf@mcbain.luannocracy.com>
References: <48A6F229.3040201@resophonic.com>
	<87k5egptsn.fsf@mcbain.luannocracy.com>
Message-ID: <48A99041.1080605@resophonic.com>

David Abrahams wrote:
> on Sat Aug 16 2008, "troy d. straszheim" <troy-AT-resophonic.com> wrote:
> 
>> I'm playing with Sphinx 0.4.3, trying to make some docs for wrapped classes.
>> I'm getting the error
>>
>>   error while formatting signature for <wrapped_class> arg is not a Python function
>>
>> where 'arg is not a Python function' is the TypeError you get when you call
>> inspect.getargspec() of a wrapped member function.
>> Appears to relate to this thread about doctests:
>>
>>   http://mail.python.org/pipermail/c++-sig/2007-June/012481.html
>>
>> Has anybody looked further at this?
> 
> I'm afraid I never did look at it further.  I'd be very happy if the
> the problem were solved, though :-)
> 

This has recently been worked around / fixed on the trunk of sphinx.  No more big red
boxes in the generated html.  Yay.  Sphinx is some Really Good Stuff.

I did spend some time hacking at this, and at this point am not sure there
exists a problem on the boost.python side.   The behavior of
inspect.getargspec() is the same for builtin C functions that come packaged
with python as for boost.python wrapped functions.  Take the standard C
package 'audioop':

 >>> import audioop, inspect
 >>> def f(x): pass
...
 >>> inspect.getargspec(f)
(['x'], None, None, None)
 >>> inspect.getargspec(audioop.bias)
Traceback (most recent call last):
   File "<stdin>", line 1, in <module>
   File "/usr/lib/python2.5/inspect.py", line 742, in getargspec
     raise TypeError('arg is not a Python function')
TypeError: arg is not a Python function

So apparently when getargspec says 'not a Python function' it does mean
'Python' and not just 'function'.

I don't know how this impacts the doctest stuff in the post mentioned above.
Anyway, I have to get back to writing docs....

-t


From troy at resophonic.com  Mon Aug 18 17:14:30 2008
From: troy at resophonic.com (troy d. straszheim)
Date: Mon, 18 Aug 2008 11:14:30 -0400
Subject: [C++-sig] Multiple Constructors
In-Reply-To: <BBABA8164420CB43ABEA9DC81FE7A884065D1B2E@lopets1exms003.leh.lbcorp.lehman.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B2E@lopets1exms003.leh.lbcorp.lehman.com>
Message-ID: <48A991D6.9030702@resophonic.com>

Bilokon, Paul wrote:
> Hi,
> 
> I'm wondering if there is an example that demonstrates how one should
> deal with classes with multiple constructors with different signatures,
> e.g.

There is.  It is right in the docs.

class_<World>("World", init<std::string>())
     .def(init<double, double>())
     .def("greet", &World::greet)
     .def("set", &World::set)
;

>     scope in_Term = class_<Term>("Term", init<>())
>         .def(init<const Term::Foo, Bar>())
>         .def(init<const Term::Foo, Bar, bool>())
>         .def(init<const std::string &>())
> 
> I don't think this approach is valid.

You're going to have to be more specific than that.

-t




From paul.bilokon at lehman.com  Mon Aug 18 17:43:41 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Mon, 18 Aug 2008 16:43:41 +0100
Subject: [C++-sig] Multiple Constructors
In-Reply-To: <48A991D6.9030702@resophonic.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B2E@lopets1exms003.leh.lbcorp.lehman.com>
	<48A991D6.9030702@resophonic.com>
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B30@lopets1exms003.leh.lbcorp.lehman.com>

Hi Troy,

My apologies, thanks a lot for pointing this out.

Kind regards,
Paul 

-----Original Message-----
From: c++-sig-bounces at python.org [mailto:c++-sig-bounces at python.org] On
Behalf Of troy d. straszheim
Sent: Monday, August 18, 2008 4:15 PM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Multiple Constructors

Bilokon, Paul wrote:
> Hi,
> 
> I'm wondering if there is an example that demonstrates how one should 
> deal with classes with multiple constructors with different 
> signatures, e.g.

There is.  It is right in the docs.

class_<World>("World", init<std::string>())
     .def(init<double, double>())
     .def("greet", &World::greet)
     .def("set", &World::set)
;

>     scope in_Term = class_<Term>("Term", init<>())
>         .def(init<const Term::Foo, Bar>())
>         .def(init<const Term::Foo, Bar, bool>())
>         .def(init<const std::string &>())
> 
> I don't think this approach is valid.

You're going to have to be more specific than that.

-t


_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From troy at resophonic.com  Mon Aug 18 18:07:48 2008
From: troy at resophonic.com (troy d. straszheim)
Date: Mon, 18 Aug 2008 12:07:48 -0400
Subject: [C++-sig] "Returning" to Scope
In-Reply-To: <BBABA8164420CB43ABEA9DC81FE7A884065D1B2F@lopets1exms003.leh.lbcorp.lehman.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B2F@lopets1exms003.leh.lbcorp.lehman.com>
Message-ID: <48A99E54.7000205@resophonic.com>

Bilokon, Paul wrote:
> Hi,
> 
> I'm trying to figure out how I can "return" to a previously defined
> scope. 

Scopes essentially push and pop themselves off a static stack of scopes
as they are created/destroyed, so you can generally just use... scopes:

BOOST_PYTHON_MODULE(scope)
{
   {
     scope s = class_<Foo> ...
     enum_<Foo::Something> ...	
   }

   {
      scope s = class_<Baz> ...
      enum_<Baz::Something> ...
   }

   class_<Quux> ...
}



-t


From paul.bilokon at lehman.com  Mon Aug 18 18:11:45 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Mon, 18 Aug 2008 17:11:45 +0100
Subject: [C++-sig] "Returning" to Scope
In-Reply-To: <48A99E54.7000205@resophonic.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B2F@lopets1exms003.leh.lbcorp.lehman.com>
	<48A99E54.7000205@resophonic.com>
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B31@lopets1exms003.leh.lbcorp.lehman.com>

This is great, thanks a lot, Troy!

Regards,
Paul 

-----Original Message-----
From: c++-sig-bounces at python.org [mailto:c++-sig-bounces at python.org] On
Behalf Of troy d. straszheim
Sent: Monday, August 18, 2008 5:08 PM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] "Returning" to Scope

Bilokon, Paul wrote:
> Hi,
> 
> I'm trying to figure out how I can "return" to a previously defined 
> scope.

Scopes essentially push and pop themselves off a static stack of scopes
as they are created/destroyed, so you can generally just use... scopes:

BOOST_PYTHON_MODULE(scope)
{
   {
     scope s = class_<Foo> ...
     enum_<Foo::Something> ...	
   }

   {
      scope s = class_<Baz> ...
      enum_<Baz::Something> ...
   }

   class_<Quux> ...
}



-t
_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From troy at resophonic.com  Mon Aug 18 19:24:51 2008
From: troy at resophonic.com (troy d. straszheim)
Date: Mon, 18 Aug 2008 13:24:51 -0400
Subject: [C++-sig] Alternate packaging technique for bindings + pure python
Message-ID: <48A9B063.8070503@resophonic.com>


I'm in the process of trying to organize a large project that is about
half-and-half python and c++.  There are many components, each of which
consists of both python and c++, and I'm trying to hide this from the user.
These components should appear to be internally seamless.

The docs:

   http://www.boost.org/doc/libs/1_36_0/libs/python/doc/tutorial/doc/html/python/techniques.html#python.creating_packages

recommend putting the cpp bindings in nested libraries, like so (BTW, the directory
listing in the docs is buggy):

  sounds/
     __init__.py
     core/
         __init__.py
         _core.so
     filters/
         __init__.py
         _filters.so
     io/
         __init__.py
         _io.so

where e.g.  core/__init__.py contains

   from _core import *

this works, but it is leaky, the user still sees the underscores in the package
names and docs, cpp wrapper authors have to remember these underscores....
the interface not as clean as it could be.

If you organize things like this, instead:

  sounds/
     __init__.py

     core.so
     core/
         __init__.py
         foo.py
         bar.py

     filters.so
     filters/
         __init__.py

     io.so
     io/
         __init__.py


You can use the following function:

def load_pybindings(name, path):
     """
     Merges python bindings from shared library 'name' into module 'name'.
     Use when you have a directory structure::

       lib/
          foo.so
          foo/
            __init__.py
            something.py

     Here, inside ``foo/__init__.py`` call ``load_pybindings(__name__, __path__)``

     this assumes that the first entry in list ``__path__`` is where
     you want the wrapped classes to merge to.

     """
     import imp, sys
     m = imp.load_dynamic(name, __path__[0] + ".so")
     thismod = sys.modules[name]

     for (k,v) in m.__dict__.items():
         if not k.startswith("_"):
             thismod.__dict__[k] = v

Wherein you just use the __name__ and __path__ of the current module being imported to
locate and load the associated shared library.  Then you copy the symbols into the
current library.

So the call to it in each __init__.py would be:

   load_pybindings(__name__, __path__)

Nice thing is that everything looks quite homogenous to the user.   It appears to work
quite well, but I just now came up with this and thought I ask if anybody
sees any problems.

-t








From troy at resophonic.com  Mon Aug 18 19:41:43 2008
From: troy at resophonic.com (troy d. straszheim)
Date: Mon, 18 Aug 2008 13:41:43 -0400
Subject: [C++-sig] Alternate packaging technique for bindings + pure
	python
In-Reply-To: <48A9B063.8070503@resophonic.com>
References: <48A9B063.8070503@resophonic.com>
Message-ID: <48A9B457.7080707@resophonic.com>

troy d. straszheim wrote:

>     m = imp.load_dynamic(name, __path__[0] + ".so")

whoops, that should have been

   m = imp.load_dynamic(name, path[0] + ".so")

sorry for the noise.

-t


From wladwig at wdtinc.com  Mon Aug 18 22:05:37 2008
From: wladwig at wdtinc.com (William Ladwig)
Date: Mon, 18 Aug 2008 15:05:37 -0500
Subject: [C++-sig] Overriding c++ abstract class in python
Message-ID: <765CBD9053EA2B438625895F30D1856FC9E56A66@storm.wdtinc.com>

I'm having problem overriding a c++ abstract class in python (when passing the python object to a c++ function).  Here is a really simplified version of my code.:

c++ code:

class GenericInput
{
public:
        virtual void foo(const std::string& filename) = 0;
};

void processInput(GenericInput& input, const std::string& filename)
{
        input.foo(filename);
}

wrapping code:

struct GenericInputWrapper : GenericInput, wrapper<GenericInput>
{
        this->get_override("foo")();
};

BOOST_PYTHON_MODULE(_core)
{
        class_<GenericInputWrapper, boost::noncopyable>("GenericInput", no_init)
                .def("foo", pure_virtual(&GenericInput::foo))
                ;

        def ("processInput", processInput);
}


python code:

import _core
class TestInput(_core.GenericInput):
        def __init__(self):
                print("init called")

        def foo(self, filename):
                print("foo called")

>>> A = TestInput()
>>> processInput(A, "test")

This fails with:

Boost.Python.ArgumentError: Python argument types in
    _core.processInput(TestInput, str)
did not match C++ signature:
    processInput(GenericInput {lvalue}, std::string)


However, if I remove the "no_init" from my GenericInput wrapper and also remove the __init__ function from my python TestInput class, I don't get the error.  What am I doing wrong here?  Thanks.

Bill



From seefeld at sympatico.ca  Mon Aug 18 22:17:33 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Mon, 18 Aug 2008 16:17:33 -0400
Subject: [C++-sig] Overriding c++ abstract class in python
In-Reply-To: <765CBD9053EA2B438625895F30D1856FC9E56A66@storm.wdtinc.com>
References: <765CBD9053EA2B438625895F30D1856FC9E56A66@storm.wdtinc.com>
Message-ID: <48A9D8DD.70506@sympatico.ca>

William Ladwig wrote:
> python code:
>
> import _core
> class TestInput(_core.GenericInput):
>         def __init__(self):
>                 print("init called")
>   

Make sure to call the base class' __init__ method explicitely. Python, 
unlike C++, won't do that for you automatically:

class TestInput(_core.GenericInput):
    def __init__(self):
        super(TestInput, self).__init__()
        ...

HTH,
       Stefan

-- 

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



From wladwig at wdtinc.com  Mon Aug 18 22:50:55 2008
From: wladwig at wdtinc.com (William Ladwig)
Date: Mon, 18 Aug 2008 15:50:55 -0500
Subject: [C++-sig] Overriding c++ abstract class in python
In-Reply-To: <48A9D8DD.70506@sympatico.ca>
References: <765CBD9053EA2B438625895F30D1856FC9E56A66@storm.wdtinc.com>
	<48A9D8DD.70506@sympatico.ca>
Message-ID: <765CBD9053EA2B438625895F30D1856FC9E56A87@storm.wdtinc.com>

The base class is abstract in c++ and has no constructor (it will throw an exception if called in python).  The problem I am having is more related to how I extract the derived python object properly in a c++ function.  If I don't use an __init__ function in the derived python class (and remove the no_init in the c++ wrapper), it happens automatically.  However, as soon as I add an __init__ function for the derived class, I get the error:

Boost.Python.ArgumentError: Python argument types in
    _core.processInput(TestInput, str)
did not match C++ signature:
    processInput(GenericInput {lvalue}, std::string)

I suspect I need to write a thin wrapper around the processInput() function in order to extract the c++ object from python, but I'm not sure how to do it and preserve the overridden virtual functions.  I tried extracting a GenericInputWrapper object, but I get the same error.  Any ideas?

Bill



-----Original Message-----
From: c++-sig-bounces+wladwig=wdtinc.com at python.org [mailto:c++-sig-bounces+wladwig=wdtinc.com at python.org] On Behalf Of Stefan Seefeld
Sent: Monday, August 18, 2008 3:18 PM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Overriding c++ abstract class in python

William Ladwig wrote:
> python code:
>
> import _core
> class TestInput(_core.GenericInput):
>         def __init__(self):
>                 print("init called")
>

Make sure to call the base class' __init__ method explicitely. Python,
unlike C++, won't do that for you automatically:

class TestInput(_core.GenericInput):
    def __init__(self):
        super(TestInput, self).__init__()
        ...

HTH,
       Stefan

--

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

_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig


From wladwig at wdtinc.com  Tue Aug 19 00:30:14 2008
From: wladwig at wdtinc.com (William Ladwig)
Date: Mon, 18 Aug 2008 17:30:14 -0500
Subject: [C++-sig] Overriding c++ abstract class in python
In-Reply-To: <765CBD9053EA2B438625895F30D1856FC9E56A87@storm.wdtinc.com>
References: <765CBD9053EA2B438625895F30D1856FC9E56A66@storm.wdtinc.com>
	<48A9D8DD.70506@sympatico.ca>
	<765CBD9053EA2B438625895F30D1856FC9E56A87@storm.wdtinc.com>
Message-ID: <765CBD9053EA2B438625895F30D1856FC9E56ACC@storm.wdtinc.com>

Correction:  If I try to extract a reference to the GenericInputWrapper, I get:

TypeError: No registered converter was able to extract a C++ reference to type GenericInputWrapper from this Python object of type TestInput

Is this description on the boost python Wiki page the correct way to handle this situation or is it out of date?

http://wiki.python.org/moin/boost.python/HowTo#head-5665b70c155b2eb814a3d39898d9b9fae04c3dc3

Thanks.

Bill

-----Original Message-----
From: c++-sig-bounces at python.org [mailto:c++-sig-bounces at python.org] On Behalf Of William Ladwig
Sent: Monday, August 18, 2008 3:51 PM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Overriding c++ abstract class in python

The base class is abstract in c++ and has no constructor (it will throw an exception if called in python).  The problem I am having is more related to how I extract the derived python object properly in a c++ function.  If I don't use an __init__ function in the derived python class (and remove the no_init in the c++ wrapper), it happens automatically.  However, as soon as I add an __init__ function for the derived class, I get the error:

Boost.Python.ArgumentError: Python argument types in
    _core.processInput(TestInput, str)
did not match C++ signature:
    processInput(GenericInput {lvalue}, std::string)

I suspect I need to write a thin wrapper around the processInput() function in order to extract the c++ object from python, but I'm not sure how to do it and preserve the overridden virtual functions.  I tried extracting a GenericInputWrapper object, but I get the same error.  Any ideas?

Bill



-----Original Message-----
From: c++-sig-bounces+wladwig=wdtinc.com at python.org [mailto:c++-sig-bounces+wladwig=wdtinc.com at python.org] On Behalf Of Stefan Seefeld
Sent: Monday, August 18, 2008 3:18 PM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Overriding c++ abstract class in python

William Ladwig wrote:
> python code:
>
> import _core
> class TestInput(_core.GenericInput):
>         def __init__(self):
>                 print("init called")
>

Make sure to call the base class' __init__ method explicitely. Python,
unlike C++, won't do that for you automatically:

class TestInput(_core.GenericInput):
    def __init__(self):
        super(TestInput, self).__init__()
        ...

HTH,
       Stefan

--

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

_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig
_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig


From greg.landrum at gmail.com  Tue Aug 19 07:07:22 2008
From: greg.landrum at gmail.com (Greg Landrum)
Date: Tue, 19 Aug 2008 07:07:22 +0200
Subject: [C++-sig] Accessing python configuration data when building
	extensions using bjam and boost.python
Message-ID: <60825b0f0808182207k62f65c5aq1771aafb9d412836@mail.gmail.com>

Dear all,

I've recently switched the code for one of my projects from Numeric to
numpy. The project includes a number of extension modules built using
boost.python. So far everything seems to work without problems except
for building the modules in a "pleasant" way.

Here's the problem:
Numeric python would, by default, install itself in such a way that
its header files were in a subdirectory of the python include
directory. So things like "#include <Numeric/arrayobject.h>" would
work without any special handling in extension files built using bjam.
Numpy, on the other hand, installs its header files with its Python
files in site-packages. So I need to add something like this to my
Jamroot :
          <toolset>gcc:<include>/usr/lib/python2.5/site-packages/numpy/core/include
This is, obviously, dependent on operating system, python version and
python installation-location. So the user will have to edit the
Jamroot based on their details.  One of the things I really like about
boost.build is that this kind of stuff is normally not required.

I could solve this problem, I think, there were a variable (?) in
python.jam defining the location of python's "lib" directory that I
could access from my Jamroot. Then I could have something like:
  <include>$(python-lib-dir)/numpy/core/include
that would work on every machine.

I've spent some time looking through the boost.build documentation and
python.jam, but I was unable to figure out how to expose a new
variable that would be accessible from other Jamfiles.

Does anyone have a recipe for doing this?

Thanks,
-greg


From roman.yakovenko at gmail.com  Tue Aug 19 08:30:57 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Tue, 19 Aug 2008 09:30:57 +0300
Subject: [C++-sig] Overriding c++ abstract class in python
In-Reply-To: <765CBD9053EA2B438625895F30D1856FC9E56ACC@storm.wdtinc.com>
References: <765CBD9053EA2B438625895F30D1856FC9E56A66@storm.wdtinc.com>
	<48A9D8DD.70506@sympatico.ca>
	<765CBD9053EA2B438625895F30D1856FC9E56A87@storm.wdtinc.com>
	<765CBD9053EA2B438625895F30D1856FC9E56ACC@storm.wdtinc.com>
Message-ID: <7465b6170808182330t40f7f10du4bbf0d8ca9574339@mail.gmail.com>

On Tue, Aug 19, 2008 at 1:30 AM, William Ladwig <wladwig at wdtinc.com> wrote:
> Correction:  If I try to extract a reference to the GenericInputWrapper, I get:
>
> TypeError: No registered converter was able to extract a C++ reference to type GenericInputWrapper from this Python object of type TestInput
>
> Is this description on the boost python Wiki page the correct way to handle this situation or is it out of date?
>
> http://wiki.python.org/moin/boost.python/HowTo#head-5665b70c155b2eb814a3d39898d9b9fae04c3dc3

The code you build for your classes is incorrect.

GenericInput class do have default constructor.

Next code was generated by Py++, for GenericInput class

struct GenericInput_wrapper : GenericInput, bp::wrapper< GenericInput > {

    GenericInput_wrapper()
    : GenericInput()
      , bp::wrapper< GenericInput >(){
        // null constructor

    }

    virtual void foo( ::std::string const & filename ){
        bp::override func_foo = this->get_override( "foo" );
        func_foo( filename );
    }

};

BOOST_PYTHON_MODULE(pyplusplus){
    bp::class_< GenericInput_wrapper, boost::noncopyable >( "GenericInput" )
        .def(
            "foo"
            , bp::pure_virtual( (void ( ::GenericInput::* )(
::std::string const & ) )(&::GenericInput::foo) )
            , ( bp::arg("filename") ) );
}


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


From meine at informatik.uni-hamburg.de  Tue Aug 19 10:42:39 2008
From: meine at informatik.uni-hamburg.de (Hans Meine)
Date: Tue, 19 Aug 2008 10:42:39 +0200
Subject: [C++-sig] Alternate packaging technique for bindings + pure
	python
In-Reply-To: <48A9B063.8070503@resophonic.com>
References: <48A9B063.8070503@resophonic.com>
Message-ID: <200808191042.40134.meine@informatik.uni-hamburg.de>

On Monday 18 August 2008 19:24:51 troy d. straszheim wrote:
> If you organize things like this, instead:
>
>      core.so
>      core/
>          __init__.py
>          foo.py
>          bar.py
>
> You can use the following function: [...]
>
> Nice thing is that everything looks quite homogenous to the user.   It
> appears to work quite well, but I just now came up with this and thought I
> ask if anybody sees any problems.

I would check with pickle, ipython, epydoc, sphinx, etc. whether your approach 
makes any problems.  (The documentation tools probably have ways to work 
around the underscores etc.)

Also, not everybody might want to rely on imp, __name__, and __path__.

Otherwise, it looks like a neat trick to me.

Nice greetings,
  Hans


From paul.bilokon at lehman.com  Tue Aug 19 15:10:23 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Tue, 19 Aug 2008 14:10:23 +0100
Subject: [C++-sig] Static and non-static functions with the same name
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B38@lopets1exms003.leh.lbcorp.lehman.com>

Hi,

In C++ it's possible to have the following situation:

class Foo
{
public:

    string garply(int x, int y)
    {
        return "I am the non-static Foo::garply(int, int)";
    }

    static string garply(string z)
    {
        return "I am the static Foo::garly(string)";
    }
};

Now I want to expose both the static and non-static garply to Python,
and here I'm stuck:

using namespace boost::python;

BOOST_PYTHON_MODULE(overloading)
{
    class_<Foo>("Foo")
        .def("garply", (string (*)(string)) &Foo::garply)
        .def("garply", (string (Foo::*)(int, int)) &Foo::garply)
        .staticmethod("garply")
            // ??? Which garply is that ???
    ;
}

staticmethod takes a single string as its parameter, the function name.
But I have both static and non-static variants. How do I specify which
of them is static? [Technically it's possible. We already differentiate
between overloaded functions on the Python side, so we just need to
check if the first argument is "self", e.g. by looking at its type --
hardly foolproof though, e.g. consider a "copy-constructor"
signature...]

Of course, one may ask why don't I give the Python functions different
names. But what if I'm trying to replicate the C++ API as faithfully as
I can? I would like to avoid name changes.

What's the best thing to do?

Many thanks!

Regards,
Paul


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

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From lists_ravi at lavabit.com  Tue Aug 19 16:52:35 2008
From: lists_ravi at lavabit.com (Ravi)
Date: Tue, 19 Aug 2008 10:52:35 -0400
Subject: [C++-sig] Static and non-static functions with the same name
In-Reply-To: <BBABA8164420CB43ABEA9DC81FE7A884065D1B38@lopets1exms003.leh.lbcorp.lehman.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B38@lopets1exms003.leh.lbcorp.lehman.com>
Message-ID: <200808191052.36559.lists_ravi@lavabit.com>

On Tuesday 19 August 2008 09:10:23 Bilokon, Paul wrote:
> using namespace boost::python;
>
> BOOST_PYTHON_MODULE(overloading)
> {
> ? ? class_<Foo>("Foo")
> ? ? ? ? .def("garply", (string (*)(string)) &Foo::garply)
> ? ? ? ? .def("garply", (string (Foo::*)(int, int)) &Foo::garply)
> ? ? ? ? .staticmethod("garply")
> ? ? ? ? ? ? // ??? Which garply is that ???
> ? ? ;
> }

Try an extra level of indirection:

using namespace boost::python;

string foo_pygarply( string z ) { return Foo::garply(z); }

BOOST_PYTHON_MODULE(overloading)
{
? ? class_<Foo>("Foo")
? ? ? ? .def("garply", (string (*)(string)) &Foo::garply)
? ? ? ? .def("garply", (string (Foo::*)(int, int)) &Foo::garply)
? ? ? ? .staticmethod("foo_pygarply")
? ? ;
}

Regards,
Ravi

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080819/c85f5e1f/attachment.htm>

From wladwig at wdtinc.com  Tue Aug 19 16:54:40 2008
From: wladwig at wdtinc.com (William Ladwig)
Date: Tue, 19 Aug 2008 09:54:40 -0500
Subject: [C++-sig] Overriding c++ abstract class in python
In-Reply-To: <7465b6170808182330t40f7f10du4bbf0d8ca9574339@mail.gmail.com>
References: <765CBD9053EA2B438625895F30D1856FC9E56A66@storm.wdtinc.com>
	<48A9D8DD.70506@sympatico.ca>
	<765CBD9053EA2B438625895F30D1856FC9E56A87@storm.wdtinc.com>
	<765CBD9053EA2B438625895F30D1856FC9E56ACC@storm.wdtinc.com>
	<7465b6170808182330t40f7f10du4bbf0d8ca9574339@mail.gmail.com>
Message-ID: <765CBD9053EA2B438625895F30D1856FC9E56B20@storm.wdtinc.com>


Thanks for your help!  I was actually going to play around with Py++ to see what it came up with, but haven't had a chance yet to try it.  Now I see where I messed up.  My c++ base class is abstract, but the wrapper around it is not, so I probably don't want to be using "no_init" when I create the bp::class_ definition.



From nickm at sitius.com  Tue Aug 19 19:08:32 2008
From: nickm at sitius.com (Nikolay Mladenov)
Date: Tue, 19 Aug 2008 13:08:32 -0400
Subject: [C++-sig] Static and non-static functions with the same name
In-Reply-To: <BBABA8164420CB43ABEA9DC81FE7A884065D1B38@lopets1exms003.leh.lbcorp.lehman.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B38@lopets1exms003.leh.lbcorp.lehman.com>
Message-ID: <g8eumj$4mk$1@ger.gmane.org>

Please remind me, what happens if you do not put .staticmethod("garply"),
and then try :
 >>> Foo().garply(0,0)
 >>> Foo.garply("")

?

 >     class_<Foo>("Foo")
 >         .def("garply", (string (*)(string)) &Foo::garply)
 >         .def("garply", (string (Foo::*)(int, int)) &Foo::garply)
 >         .staticmethod("garply")


Bilokon, Paul wrote:
> Hi,
> 
> In C++ it's possible to have the following situation:
> 
> class Foo
> {
> public:
> 
>     string garply(int x, int y)
>     {
>         return "I am the non-static Foo::garply(int, int)";
>     }
> 
>     static string garply(string z)
>     {
>         return "I am the static Foo::garly(string)";
>     }
> };
> 
> Now I want to expose both the static and non-static garply to Python,
> and here I'm stuck:
> 
> using namespace boost::python;
> 
> BOOST_PYTHON_MODULE(overloading)
> {
>     class_<Foo>("Foo")
>         .def("garply", (string (*)(string)) &Foo::garply)
>         .def("garply", (string (Foo::*)(int, int)) &Foo::garply)
>         .staticmethod("garply")
>             // ??? Which garply is that ???
>     ;
> }
> 
> staticmethod takes a single string as its parameter, the function name.
> But I have both static and non-static variants. How do I specify which
> of them is static? [Technically it's possible. We already differentiate
> between overloaded functions on the Python side, so we just need to
> check if the first argument is "self", e.g. by looking at its type --
> hardly foolproof though, e.g. consider a "copy-constructor"
> signature...]
> 
> Of course, one may ask why don't I give the Python functions different
> names. But what if I'm trying to replicate the C++ API as faithfully as
> I can? I would like to avoid name changes.
> 
> What's the best thing to do?
> 
> Many thanks!
> 
> Regards,
> Paul
> 
> 
> - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
> 
> This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From wladwig at wdtinc.com  Tue Aug 19 21:16:08 2008
From: wladwig at wdtinc.com (William Ladwig)
Date: Tue, 19 Aug 2008 14:16:08 -0500
Subject: [C++-sig] Extracting and returning a c++ object from python and
 garbage collection
Message-ID: <765CBD9053EA2B438625895F30D1856FC9E56BAF@storm.wdtinc.com>



I'm having a problem overriding a pure C++ virtual function in python which returns an auto_ptr to a C++ type (which has been exposed to python).  It looks like I am extracting the class correctly in my base class wrapper, but when the function returns to python, it throws a segmentation fault if I try to access the returned object.  I suspect I have a garbage collection problem, but I'm not sure what I'm doing wrong.  How do I convert an object from python to C++ and return the C++ object without it being garbage collected?  Here is a sample to illustrate the problem:

C++ code:

using namespace boost::python;

// Generic Input abstract base class
class GenericInput
{
public:
        // Pure virtual function to convert to the required type
        std::auto_ptr<ReqInput> void convToReqFormat() = 0;
};

// Required input format
class ReqInput
{
        public:
                ReqInput(unsigned int rows, unsigned int cols)
                :       _Rows(rows),
                        _Cols(cols)

                unsigned int getRows() const {return _Rows;}
                unsigned int getCols() const {return _Cols;}

        private:
                unsigned int _Rows;
                unsigned int _Cols;
};

// Function which converts GenericInput to ReqInput
std::auto_ptr<ReqInput> processInput(GenericInput& input)
{
        return input.convToReqFormat();
}


boost python wrapping code:

struct GenericInputWrapper : GenericInput, wrapper<GenericInput>
{
        GenericInputWrapper()
        :       GenericInput(), wrapper<GenericInput>() {}

        // ********** Here is where my problem is *********************
        std::auto_ptr<ReqInput> void convToReqFormat()
        {
                // This will return a python object of the exposed ReqInput type
                override func = this->get_override("convToReqFormat");
                object py_obj = func();

                ReqInput& req_in = extract<ReqInput&>(py_obj);

                // Testing extraction
                std::cout << "Testing extraction: " << req_in.getRows();
                std::cout << "," << req_in.getCols() << std::endl;

                // Store in an auto_ptr
                std::auto_ptr<ReqInput> r;
                r.reset(&req_in);

                return r;

};

BOOST_PYTHON_MODULE(_core)
{
        class_<ReqInput>("ReqInput", init<unsigned int, unsigned int>)
                .add_property("Rows", &ReqInput::getRows)
                .add_property("Cols", &ReqInput::getCols)

        class_<GenericInputWrapper, boost::noncopyable>("GenericInput", no_init)
                .def("foo", pure_virtual(&GenericInput::foo))
                ;

        def ("processInput", processInput);
}


Python test code:

import _core
class TestInput(_core.GenericInput):
        def __init__(self, rows, cols):
                _core.GenericInput.__init__(self)
                self.Rows = rows
                self.Cols = cols
        def convToReqFormat(self):
                return ReqInput(10,10)

>>> A = TestInput(10,10)
>>> B = processInput(A)
Testing extraction: 10,10

>>> B.Rows
Segmentation Fault


Thank you for your help,
Bill



From roman.yakovenko at gmail.com  Tue Aug 19 21:39:39 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Tue, 19 Aug 2008 22:39:39 +0300
Subject: [C++-sig] Extracting and returning a c++ object from python and
	garbage collection
In-Reply-To: <765CBD9053EA2B438625895F30D1856FC9E56BAF@storm.wdtinc.com>
References: <765CBD9053EA2B438625895F30D1856FC9E56BAF@storm.wdtinc.com>
Message-ID: <7465b6170808191239w28719f5bra5f40eea0e5ec5f4@mail.gmail.com>

On Tue, Aug 19, 2008 at 10:16 PM, William Ladwig <wladwig at wdtinc.com> wrote:
>
>
> I'm having a problem overriding a pure C++ virtual function in python which returns an auto_ptr to a C++ type (which has been exposed to python).  It looks like I am extracting the class correctly in my base class wrapper, but when the function returns to python, it throws a segmentation fault if I try to access the returned object.  I suspect I have a garbage collection problem, but I'm not sure what I'm doing wrong.  How do I convert an object from python to C++ and return the C++ object without it being garbage collected?  Here is a sample to illustrate the problem:
>
> C++ code:
> ...
>        // ********** Here is where my problem is *********************
>        std::auto_ptr<ReqInput> void convToReqFormat()
>        {
>                // This will return a python object of the exposed ReqInput type
>                override func = this->get_override("convToReqFormat");
>                object py_obj = func();
>
>                ReqInput& req_in = extract<ReqInput&>(py_obj);
>                std::auto_ptr<ReqInput> r;
>                r.reset(&req_in);

At some point you have same pointer, managed by 2 objects.

Also why do you need to write such code?

I attach the code, generated by Py++. I am pretty sure it will work as-is.


-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 1.h
Type: text/x-chdr
Size: 2753 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080819/c3b04e79/attachment.h>

From wladwig at wdtinc.com  Tue Aug 19 22:13:08 2008
From: wladwig at wdtinc.com (William Ladwig)
Date: Tue, 19 Aug 2008 15:13:08 -0500
Subject: [C++-sig] Extracting and returning a c++ object from python
	and	garbage collection
In-Reply-To: <7465b6170808191239w28719f5bra5f40eea0e5ec5f4@mail.gmail.com>
References: <765CBD9053EA2B438625895F30D1856FC9E56BAF@storm.wdtinc.com>
	<7465b6170808191239w28719f5bra5f40eea0e5ec5f4@mail.gmail.com>
Message-ID: <765CBD9053EA2B438625895F30D1856FC9E56BDE@storm.wdtinc.com>

My actual code is a little different than what is posted here, but the basic idea is that we use many different input formats that are similar (all represent "2D" arrays of data), but need to be converted to a common "2D" array format before being processed.  For this, I am using an abstract base class where one of the virtual functions handles the conversion and returns an auto_ptr to the required input type.  This is easy to do in C++, but if I want to override that virtual function in python, I'm not quite sure how to return an auto_ptr to a C++ object.  (I'm trying to allow for the development of the input portion of the application to be done in either c++ or python).

When I initially tried to do this, I wrapped the GenericInput class as:

struct GenericInputWrapper : GenericInput, wrapper<GenericInput>
{
        GenericInputWrapper()
        :       GenericInput(), wrapper<GenericInput>() {}

        std::auto_ptr<ReqInput> void convToReqFormat()
        {
                std::auto_ptr<ReqInput> r;
                r.reset(this->get_override("convToReqFormat"));

                return r;
        }

};

However, this threw the "dangling reference" error.  So, I figured I needed to extract the python object first, and then convert it to an auto_ptr.

-----Original Message-----
From: c++-sig-bounces at python.org [mailto:c++-sig-bounces at python.org] On Behalf Of Roman Yakovenko
Sent: Tuesday, August 19, 2008 2:40 PM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Extracting and returning a c++ object from python and garbage collection

On Tue, Aug 19, 2008 at 10:16 PM, William Ladwig <wladwig at wdtinc.com> wrote:
>
>
> I'm having a problem overriding a pure C++ virtual function in python which returns an auto_ptr to a C++ type (which has been exposed to python).  It looks like I am extracting the class correctly in my base class wrapper, but when the function returns to python, it throws a segmentation fault if I try to access the returned object.  I suspect I have a garbage collection problem, but I'm not sure what I'm doing wrong.  How do I convert an object from python to C++ and return the C++ object without it being garbage collected?  Here is a sample to illustrate the problem:
>
> C++ code:
> ...
>        // ********** Here is where my problem is *********************
>        std::auto_ptr<ReqInput> void convToReqFormat()
>        {
>                // This will return a python object of the exposed ReqInput type
>                override func = this->get_override("convToReqFormat");
>                object py_obj = func();
>
>                ReqInput& req_in = extract<ReqInput&>(py_obj);
>                std::auto_ptr<ReqInput> r;
>                r.reset(&req_in);

At some point you have same pointer, managed by 2 objects.

Also why do you need to write such code?

I attach the code, generated by Py++. I am pretty sure it will work as-is.


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


From wladwig at wdtinc.com  Wed Aug 20 00:03:03 2008
From: wladwig at wdtinc.com (William Ladwig)
Date: Tue, 19 Aug 2008 17:03:03 -0500
Subject: [C++-sig] Extracting and returning a c++ object from
	python	and	garbage collection
In-Reply-To: <765CBD9053EA2B438625895F30D1856FC9E56BDE@storm.wdtinc.com>
References: <765CBD9053EA2B438625895F30D1856FC9E56BAF@storm.wdtinc.com>
	<7465b6170808191239w28719f5bra5f40eea0e5ec5f4@mail.gmail.com>
	<765CBD9053EA2B438625895F30D1856FC9E56BDE@storm.wdtinc.com>
Message-ID: <765CBD9053EA2B438625895F30D1856FC9E56C1C@storm.wdtinc.com>

I looked at the Py++ code and tried it out (it's similar to how I tried it originally) and it also throws a "dangling pointer" exception from the interpreter.  I was able to keep it from crashing by extracting a ReqInput reference and then creating a new ReqInput by copying the extracted one.  It's not very efficient unfortunately, but it might be the only way to do this.  Thanks again for your help.


std::auto_ptr<ReqInput> convToReqInput()
{
        object py_ob = this->get_override("toMapleInputArray")();

        ReqInput& r_arr = extract<ReqInput&>(py_ob);

        std::auto_ptr<ReqInput> r(new ReqInput(r_arr));

        return r;
}



Bill



From dave at boostpro.com  Wed Aug 20 04:35:41 2008
From: dave at boostpro.com (David Abrahams)
Date: Tue, 19 Aug 2008 18:35:41 -0800
Subject: [C++-sig] How to wrap a python function pointer ?
In-Reply-To: <19028640.post@talk.nabble.com> ("William =?utf-8?Q?Mari?=
	=?utf-8?Q?=C3=A9=22's?= message of
	"Mon, 18 Aug 2008 03:01:48 -0700 (PDT)")
References: <19028640.post@talk.nabble.com>
Message-ID: <87hc9g4dmq.fsf@mcbain.luannocracy.com>


on Mon Aug 18 2008, William Mari? <william.marie-AT-gmail.com> wrote:

> I would like to wrap a python function pointer to a void (*c++fun)() pointer
> :
>
> typedef void (*fct_void_void)( void );
> void addTimer( double dStartIn, double dTickTime, u32 iTickCount,
> fct_void_void bpFctCallback )
> {
>  .....
>        
> } 
> BOOST_PYTHON_MODULE(ksys)
> {
> 	using namespace boost::python;
> 	def("addTimer", &addTimer);
> } 

Hi,

Sorry, but right now we don't have a good solution for that use case.
The best I can offer immediately is that you define a small struct that
holds a fct_void_void, expose that to Python as usual, and then create a
thin wrapper for addTimer that takes the struct instead of the function
pointer.

This could be fixed, though.  Would you care to open a trac ticket at
http://svn.boost.org


-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From william.marie at gmail.com  Wed Aug 20 09:37:27 2008
From: william.marie at gmail.com (=?ISO-8859-1?Q?william_mari=E9?=)
Date: Wed, 20 Aug 2008 08:37:27 +0100
Subject: [C++-sig] How to wrap a python function pointer ?
In-Reply-To: <87hc9g4dmq.fsf@mcbain.luannocracy.com>
References: <19028640.post@talk.nabble.com>
	<87hc9g4dmq.fsf@mcbain.luannocracy.com>
Message-ID: <168485c00808200037t2ad3a1cav4d097090e5289d4@mail.gmail.com>

Hi,

Ok ticket created : http://svn.boost.org/trac/boost/ticket/2230
Thanks
bye
William

2008/8/20 David Abrahams <dave at boostpro.com>

>
> on Mon Aug 18 2008, William Mari? <william.marie-AT-gmail.com> wrote:
>
> > I would like to wrap a python function pointer to a void (*c++fun)()
> pointer
> > :
> >
> > typedef void (*fct_void_void)( void );
> > void addTimer( double dStartIn, double dTickTime, u32 iTickCount,
> > fct_void_void bpFctCallback )
> > {
> >  .....
> >
> > }
> > BOOST_PYTHON_MODULE(ksys)
> > {
> >       using namespace boost::python;
> >       def("addTimer", &addTimer);
> > }
>
> Hi,
>
> Sorry, but right now we don't have a good solution for that use case.
> The best I can offer immediately is that you define a small struct that
> holds a fct_void_void, expose that to Python as usual, and then create a
> thin wrapper for addTimer that takes the struct instead of the function
> pointer.
>
> This could be fixed, though.  Would you care to open a trac ticket at
> http://svn.boost.org
>
>
> --
> Dave Abrahams
> BoostPro Computing
> http://www.boostpro.com
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org <C%2B%2B-sig at python.org>
> http://mail.python.org/mailman/listinfo/c++-sig
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080820/da5b68ee/attachment.htm>

From paul.bilokon at lehman.com  Wed Aug 20 10:32:56 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Wed, 20 Aug 2008 09:32:56 +0100
Subject: [C++-sig] Static and non-static functions with the same name
In-Reply-To: <g8eumj$4mk$1@ger.gmane.org>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B38@lopets1exms003.leh.lbcorp.lehman.com>
	<g8eumj$4mk$1@ger.gmane.org>
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B44@lopets1exms003.leh.lbcorp.lehman.com>

 Hi Nikolay,

This is what you get. Say you have

        .def("staticGarply", (string (*)(string)) &Foo::garply)

And you comment out...

        // .staticmethod("staticGarply")

In Python, I attempt the following:

        print Foo.staticGarply('yo')

And get...

        TypeError: unbound method Boost.Python.function object must be
called with Foo instance as first argument (got str instance instead)

Regards,
Paul

-----Original Message-----
From: c++-sig-bounces at python.org [mailto:c++-sig-bounces at python.org] On
Behalf Of Nikolay Mladenov
Sent: Tuesday, August 19, 2008 6:09 PM
To: c++-sig at python.org
Subject: Re: [C++-sig] Static and non-static functions with the same
name

Please remind me, what happens if you do not put
.staticmethod("garply"), and then try :
 >>> Foo().garply(0,0)
 >>> Foo.garply("")

?

 >     class_<Foo>("Foo")
 >         .def("garply", (string (*)(string)) &Foo::garply)
 >         .def("garply", (string (Foo::*)(int, int)) &Foo::garply)
 >         .staticmethod("garply")


Bilokon, Paul wrote:
> Hi,
> 
> In C++ it's possible to have the following situation:
> 
> class Foo
> {
> public:
> 
>     string garply(int x, int y)
>     {
>         return "I am the non-static Foo::garply(int, int)";
>     }
> 
>     static string garply(string z)
>     {
>         return "I am the static Foo::garly(string)";
>     }
> };
> 
> Now I want to expose both the static and non-static garply to Python, 
> and here I'm stuck:
> 
> using namespace boost::python;
> 
> BOOST_PYTHON_MODULE(overloading)
> {
>     class_<Foo>("Foo")
>         .def("garply", (string (*)(string)) &Foo::garply)
>         .def("garply", (string (Foo::*)(int, int)) &Foo::garply)
>         .staticmethod("garply")
>             // ??? Which garply is that ???
>     ;
> }
> 
> staticmethod takes a single string as its parameter, the function
name.
> But I have both static and non-static variants. How do I specify which

> of them is static? [Technically it's possible. We already 
> differentiate between overloaded functions on the Python side, so we 
> just need to check if the first argument is "self", e.g. by looking at

> its type -- hardly foolproof though, e.g. consider a
"copy-constructor"
> signature...]
> 
> Of course, one may ask why don't I give the Python functions different

> names. But what if I'm trying to replicate the C++ API as faithfully 
> as I can? I would like to avoid name changes.
> 
> What's the best thing to do?
> 
> Many thanks!
> 
> Regards,
> Paul
> 
> 
> - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 
> - - - - - - - - -
> 
> This message is intended only for the personal and confidential use of
the designated recipient(s) named above.  If you are not the intended
recipient of this message you are hereby notified that any review,
dissemination, distribution or copying of this message is strictly
prohibited.  This communication is for information purposes only and
should not be regarded as an offer to sell or as a solicitation of an
offer to buy any financial product, an official confirmation of any
transaction, or as an official statement of Lehman Brothers.  Email
transmission cannot be guaranteed to be secure or error-free.
Therefore, we do not represent that this information is complete or
accurate and it should not be relied upon as such.  All information is
subject to change without notice.

_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From paul.bilokon at lehman.com  Wed Aug 20 10:40:04 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Wed, 20 Aug 2008 09:40:04 +0100
Subject: [C++-sig] Static and non-static functions with the same name
In-Reply-To: <200808191052.36559.lists_ravi@lavabit.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B38@lopets1exms003.leh.lbcorp.lehman.com>
	<200808191052.36559.lists_ravi@lavabit.com>
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B46@lopets1exms003.leh.lbcorp.lehman.com>

Hi Ravi,
 
If my understanding is correct, this method has the same deficiency as
simply doing this:
 
        // OK, explicitly specifying the static garply(string):
        .def("staticGarply", (string (*)(string)) &Foo::garply)
        .staticmethod("staticGarply")
        // OK, explicitly specifying the non-static garply(int, int):
        .def("garply", (string (Foo::*)(int, int)) &Foo::garply)
 
        // NB: The syntax: (*) for static, (Foo::*) for non-static
 
        // NB: Since staticmethod does not distinguish between the
method
        // signatures, we must give the static and non-static variants
        // distinct names in Python
 
This works, but the static and non-static variants end up having
different names on the Python side.

Kind regards,
Paul

________________________________

From: c++-sig-bounces at python.org [mailto:c++-sig-bounces at python.org] On
Behalf Of Ravi
Sent: Tuesday, August 19, 2008 3:53 PM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Static and non-static functions with the same
name



On Tuesday 19 August 2008 09:10:23 Bilokon, Paul wrote:

> using namespace boost::python;

>

> BOOST_PYTHON_MODULE(overloading)

> {

>     class_<Foo>("Foo")

>         .def("garply", (string (*)(string)) &Foo::garply)

>         .def("garply", (string (Foo::*)(int, int)) &Foo::garply)

>         .staticmethod("garply")

>             // ??? Which garply is that ???

>     ;

> }

Try an extra level of indirection:

using namespace boost::python;

string foo_pygarply( string z ) { return Foo::garply(z); }

BOOST_PYTHON_MODULE(overloading)

{

    class_<Foo>("Foo")

        .def("garply", (string (*)(string)) &Foo::garply)

        .def("garply", (string (Foo::*)(int, int)) &Foo::garply)

        .staticmethod("foo_pygarply")

    ;

}

Regards,

Ravi

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

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080820/6ef81214/attachment.htm>

From ivanov.maxim at gmail.com  Wed Aug 20 20:13:13 2008
From: ivanov.maxim at gmail.com (Max Ivanov)
Date: Wed, 20 Aug 2008 22:13:13 +0400
Subject: [C++-sig] copy_const_reference on add_property
Message-ID: <fbfa7b0c0808201113s65023c50o95f2f7c03b44a14d@mail.gmail.com>

Hi all!
If C++ method returns std::strings I'm forced to add
"return_value_policy<bp::copy_const_reference>()" as third argument to
"def". How to make similair  for  add_property?


From roman.yakovenko at gmail.com  Wed Aug 20 20:18:15 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 20 Aug 2008 21:18:15 +0300
Subject: [C++-sig] Extracting and returning a c++ object from python and
	garbage collection
In-Reply-To: <765CBD9053EA2B438625895F30D1856FC9E56C1C@storm.wdtinc.com>
References: <765CBD9053EA2B438625895F30D1856FC9E56BAF@storm.wdtinc.com>
	<7465b6170808191239w28719f5bra5f40eea0e5ec5f4@mail.gmail.com>
	<765CBD9053EA2B438625895F30D1856FC9E56BDE@storm.wdtinc.com>
	<765CBD9053EA2B438625895F30D1856FC9E56C1C@storm.wdtinc.com>
Message-ID: <7465b6170808201118j1f6dd0b8q78741d8f7b8c627@mail.gmail.com>

On Wed, Aug 20, 2008 at 1:03 AM, William Ladwig <wladwig at wdtinc.com> wrote:
> I looked at the Py++ code and tried it out (it's similar to how I tried it originally) and it also throws a "dangling pointer" exception from the interpreter.

Interesting, I've got other errors.

Using Visual Studio 2003, I got warning during compilation and
TypeError during runtime, which says, that Boost.Python is not able to
construct rvalue, std::auto_ptr< your class >, from the Python object.

Using gcc 4.2.3, on Ubuntu the generated code didn't compiled and when
I fixed it, than everything was fine and worked as expected
I attached source code, "fixed" generated code and small tester.

What is your environment?

Any way, it seems that I found solution. The only change was to
extract std::auto_ptr<> as is, by value, from Python object.

Let me know, if this really solves your problem, I will fix Py++.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: return_auto_ptr_to_be_exported.hpp
Type: text/x-csrc
Size: 1265 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080820/1a8af2d2/attachment.c>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: return_auto_ptr.cpp
Type: text/x-c++src
Size: 3964 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080820/1a8af2d2/attachment.cpp>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: x.py
Type: text/x-python
Size: 573 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080820/1a8af2d2/attachment.py>

From roman.yakovenko at gmail.com  Wed Aug 20 20:22:50 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 20 Aug 2008 21:22:50 +0300
Subject: [C++-sig] copy_const_reference on add_property
In-Reply-To: <fbfa7b0c0808201113s65023c50o95f2f7c03b44a14d@mail.gmail.com>
References: <fbfa7b0c0808201113s65023c50o95f2f7c03b44a14d@mail.gmail.com>
Message-ID: <7465b6170808201122x72a9710ev59d7232278beafdb@mail.gmail.com>

On Wed, Aug 20, 2008 at 9:13 PM, Max Ivanov <ivanov.maxim at gmail.com> wrote:
> Hi all!
> If C++ method returns std::strings I'm forced to add
> "return_value_policy<bp::copy_const_reference>()" as third argument to
> "def". How to make similair  for  add_property?

You should use boost::python::make_function

namespace bp = boost::python;

.add_property(
"name"

                        , bp::make_function(
<ref to function>, <call  policies> ) );

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


From wladwig at wdtinc.com  Wed Aug 20 21:11:25 2008
From: wladwig at wdtinc.com (William Ladwig)
Date: Wed, 20 Aug 2008 14:11:25 -0500
Subject: [C++-sig] Extracting and returning a c++ object from python
	and	garbage collection
In-Reply-To: <7465b6170808201118j1f6dd0b8q78741d8f7b8c627@mail.gmail.com>
References: <765CBD9053EA2B438625895F30D1856FC9E56BAF@storm.wdtinc.com>
	<7465b6170808191239w28719f5bra5f40eea0e5ec5f4@mail.gmail.com>
	<765CBD9053EA2B438625895F30D1856FC9E56BDE@storm.wdtinc.com>
	<765CBD9053EA2B438625895F30D1856FC9E56C1C@storm.wdtinc.com>
	<7465b6170808201118j1f6dd0b8q78741d8f7b8c627@mail.gmail.com>
Message-ID: <765CBD9053EA2B438625895F30D1856FC9E56D3B@storm.wdtinc.com>

I am using an older version of gcc (3.4.6) on Redhat Linux.  My python interpreter is v2.5.2

Upon doing some digging using google, I found this thread where the poster is doing something very similar to what I am trying to do:

http://mail.python.org/pipermail/c++-sig/2005-May/008867.html

The compiling errors are due to the way the auto_ptr is defined in the STL.  I don't think the suggested fix in the thread compiled for me either due to "ambiguous" overloaded function errors.  So, I ended up using (before I posted my thread yesterday):

std::auto_ptr<ReqInput> void convToReqFormat()
{
        std::auto_ptr<ReqInput> r;
        r.reset(this->get_override("convToReqInput")());
        return r;
}

This code will compile, but if I try to use the returned object, it crashes with a segmentation fault, which I suspect is because the python exposed "c++" object returned by this->get_override() is being garbage collected when the wrapper function returns.  One solution that fixed the problem for me was to extract to the ReqInput object, then create a new ReqInput object which copies the extracted object, and returns an auto_ptr to the new object.

std::auto_ptr<ReqInput> void convToReqFormat()
{
        object py_input = this->get_override("convToReqInput")();
        ReqInput& r_in = extract<ReqInput&>(py_arr);
        std::auto_ptr<ReqInput> r(new ReqInput(r_in));
        return r;
}

This is horribly inefficient (considering the object contained in my real code contains a 6000x6000 array), but doesn't crash on me.  I never tried extracting an auto_ptr directly however, which is probably what I should have been doing all along, since that is the object that the virtual function is returning.  I'll give it a try and see how it works out.  Thanks!




-----Original Message-----
From: c++-sig-bounces at python.org [mailto:c++-sig-bounces at python.org] On Behalf Of Roman Yakovenko
Sent: Wednesday, August 20, 2008 1:18 PM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Extracting and returning a c++ object from python and garbage collection

On Wed, Aug 20, 2008 at 1:03 AM, William Ladwig <wladwig at wdtinc.com> wrote:
> I looked at the Py++ code and tried it out (it's similar to how I tried it originally) and it also throws a "dangling pointer" exception from the interpreter.

Interesting, I've got other errors.

Using Visual Studio 2003, I got warning during compilation and
TypeError during runtime, which says, that Boost.Python is not able to
construct rvalue, std::auto_ptr< your class >, from the Python object.

Using gcc 4.2.3, on Ubuntu the generated code didn't compiled and when
I fixed it, than everything was fine and worked as expected
I attached source code, "fixed" generated code and small tester.

What is your environment?

Any way, it seems that I found solution. The only change was to
extract std::auto_ptr<> as is, by value, from Python object.

Let me know, if this really solves your problem, I will fix Py++.

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


From ope-devel at gmx.de  Wed Aug 20 21:27:48 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Wed, 20 Aug 2008 21:27:48 +0200
Subject: [C++-sig] [pybindgen] Problems on PyObject* param
Message-ID: <48AC7034.8080802@gmx.de>

Hi,

I try out some binding generator tools and start with pybindgen :)

Anyway, here my script:


import pybindgen
import sys

# convinience
param=pybindgen.param
retval=pybindgen.retval

mod = pybindgen.Module('redirector')
mod.add_include('"redirector.hpp"')

stdout_py = mod.add_class('StdoutPy')
stdout_py.add_constructor([])
stdout_py.add_method('write', None, [param('PyObject*', 'value')])
stdout_py.add_method('flush', None, [])

stderr_py = mod.add_class('StderrPy')
stderr_py.add_constructor([])
stderr_py.add_method('write', None, [param('PyObject*', 'value')])
stderr_py.add_method('flush', None, [])

stdin_py  = mod.add_class('StdinPy')
stdin_py.add_constructor([])
stdin_py.add_method('readline', retval('PyObject*'), [param('PyObject*',
'value')])

mod.generate(sys.stdout)

for this classes:

	class PyConsole;

	class StdoutPy
		: public boost::noncopyable
	{
	public:
		StdoutPy(PyConsole* console);
		void write(PyObject* args);
		void flush();
	private:
		PyConsole*			m_console;
	};


	class StderrPy
		: public boost::noncopyable
	{
	public:
		StderrPy(PyObject* console);
		void write(PyObject* args);
		void flush();
	private:
		PyConsole*			m_console;
	};


	class StdinPy
		: public boost::noncopyable
	{
	public:
		StdinPy(PyConsole* console);
		PyObject* readline(PyObject* args);
	private:
		PyConsole*			m_console;
	};

as you assume, topic is redirection of the sys write function to a gui.
Anyway, I've got errors. Any help?

$ python redirector.py
/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py:1173:
UserWarning: Exception '__init__() takes at least 4 arguments (3 given)'
in type handler <class
'pybindgen.typehandlers.pyobjecttype.PyObjectParam'> constructor
  warnings.warn("Exception %r in type handler %s constructor" %
(str(ex), type_handler_class))
Traceback (most recent call last):
  File "redirector.py", line 13, in <module>
    stdout_py.add_method('write', None, [param('PyObject*', 'value')])
  File "/usr/lib/python2.5/site-packages/pybindgen/cppclass.py", line
1127, in add_method
    meth = CppMethod(*args, **kwargs)
  File "/usr/lib/python2.5/site-packages/pybindgen/cppmethod.py", line
104, in __init__
    parameters = [utils.eval_param(param, self) for param in parameters]
  File "/usr/lib/python2.5/site-packages/pybindgen/utils.py", line 205,
in eval_param
    TypeLookupError))
  File "/usr/lib/python2.5/site-packages/pybindgen/utils.py", line 115,
in call_with_error_handling
    return callable(*args, **kwargs)
  File
"/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py", line
1171, in new
    return type_handler_class(*args, **kwargs)
TypeError: __init__() takes at least 4 arguments (3 given)

The idea is to create the object in C++ and then set it to python,
inside PyConsole class like:

stdoutPy = new StdoutPy(this);
...
PySys_SetObject("stdout", stdoutPy);

and the redirect the member calls from python into C++.

Thanks,
Olaf






From wladwig at wdtinc.com  Wed Aug 20 22:32:43 2008
From: wladwig at wdtinc.com (William Ladwig)
Date: Wed, 20 Aug 2008 15:32:43 -0500
Subject: [C++-sig] Extracting and returning a c++ object from python
	and	garbage collection
In-Reply-To: <7465b6170808201118j1f6dd0b8q78741d8f7b8c627@mail.gmail.com>
References: <765CBD9053EA2B438625895F30D1856FC9E56BAF@storm.wdtinc.com>
	<7465b6170808191239w28719f5bra5f40eea0e5ec5f4@mail.gmail.com>
	<765CBD9053EA2B438625895F30D1856FC9E56BDE@storm.wdtinc.com>
	<765CBD9053EA2B438625895F30D1856FC9E56C1C@storm.wdtinc.com>
	<7465b6170808201118j1f6dd0b8q78741d8f7b8c627@mail.gmail.com>
Message-ID: <765CBD9053EA2B438625895F30D1856FC9E56D76@storm.wdtinc.com>

Woohoo!  It worked.  It turns out that I also forgot to hold the ReqInput in an auto_ptr in the ReqInput "class_" wrapper definition (along with extracting the auto_ptr object by value in the virtual function wrapper).  All of this is being done correctly by your Py++ code, so I think your Py++ change is ready for use.  Thanks for your help!




From gjcarneiro at gmail.com  Wed Aug 20 22:41:21 2008
From: gjcarneiro at gmail.com (Gustavo Carneiro)
Date: Wed, 20 Aug 2008 21:41:21 +0100
Subject: [C++-sig] [pybindgen] Problems on PyObject* param
In-Reply-To: <48AC7034.8080802@gmx.de>
References: <48AC7034.8080802@gmx.de>
Message-ID: <a467ca4f0808201341p4c0b25c8y4e668c5f8d634fbf@mail.gmail.com>

2008/8/20 Olaf Peter <ope-devel at gmx.de>

> Hi,
>
> I try out some binding generator tools and start with pybindgen :)
>
> Anyway, here my script:
>
>
> import pybindgen
> import sys
>
> # convinience
> param=pybindgen.param
> retval=pybindgen.retval
>
> mod = pybindgen.Module('redirector')
> mod.add_include('"redirector.hpp"')
>
> stdout_py = mod.add_class('StdoutPy')
> stdout_py.add_constructor([])
> stdout_py.add_method('write', None, [param('PyObject*', 'value')])
> stdout_py.add_method('flush', None, [])
>
> stderr_py = mod.add_class('StderrPy')
> stderr_py.add_constructor([])
> stderr_py.add_method('write', None, [param('PyObject*', 'value')])
> stderr_py.add_method('flush', None, [])
>
> stdin_py  = mod.add_class('StdinPy')
> stdin_py.add_constructor([])
> stdin_py.add_method('readline', retval('PyObject*'), [param('PyObject*',
> 'value')])
>
> mod.generate(sys.stdout)
>
> for this classes:
>
>        class PyConsole;
>
>        class StdoutPy
>                : public boost::noncopyable
>        {
>        public:
>                StdoutPy(PyConsole* console);
>                void write(PyObject* args);
>                void flush();
>        private:
>                PyConsole*                      m_console;
>        };
>
>
>        class StderrPy
>                : public boost::noncopyable
>        {
>        public:
>                StderrPy(PyObject* console);
>                void write(PyObject* args);
>                void flush();
>        private:
>                PyConsole*                      m_console;
>        };
>
>
>        class StdinPy
>                : public boost::noncopyable
>        {
>        public:
>                StdinPy(PyConsole* console);
>                PyObject* readline(PyObject* args);
>        private:
>                PyConsole*                      m_console;
>        };
>
> as you assume, topic is redirection of the sys write function to a gui.
> Anyway, I've got errors. Any help?
>
> $ python redirector.py
> /usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py:1173:
> UserWarning: Exception '__init__() takes at least 4 arguments (3 given)'
> in type handler <class
> 'pybindgen.typehandlers.pyobjecttype.PyObjectParam'> constructor
>  warnings.warn("Exception %r in type handler %s constructor" %
> (str(ex), type_handler_class))
> Traceback (most recent call last):
>  File "redirector.py", line 13, in <module>
>    stdout_py.add_method('write', None, [param('PyObject*', 'value')])
>  File "/usr/lib/python2.5/site-packages/pybindgen/cppclass.py", line
> 1127, in add_method
>    meth = CppMethod(*args, **kwargs)
>  File "/usr/lib/python2.5/site-packages/pybindgen/cppmethod.py", line
> 104, in __init__
>    parameters = [utils.eval_param(param, self) for param in parameters]
>  File "/usr/lib/python2.5/site-packages/pybindgen/utils.py", line 205,
> in eval_param
>    TypeLookupError))
>  File "/usr/lib/python2.5/site-packages/pybindgen/utils.py", line 115,
> in call_with_error_handling
>    return callable(*args, **kwargs)
>  File
> "/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py", line
> 1171, in new
>    return type_handler_class(*args, **kwargs)
> TypeError: __init__() takes at least 4 arguments (3 given)


I know the exception message isn't terribly informative, but you are missing
an argument.

should be:

   retval('PyObject*', caller_owns_return=True|False)

and:

   param('PyObject*', 'name', transfer_ownership=True|False)

caller_owns_return indicates whether the caller owns the returned PyObject*
reference or not.  tranfer_ownership tells whether the callee expects to own
the reference passed in.

I am unable to say, in your example, what are the correct values.  Only the
implementation file can tell.

Hope it helps.

-- 
Gustavo J. A. M. Carneiro
INESC Porto, Telecommunications and Multimedia Unit
"The universe is always one step beyond logic." -- Frank Herbert
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080820/9f98b0c4/attachment.htm>

From aeriksson at gmail.com  Thu Aug 21 09:06:57 2008
From: aeriksson at gmail.com (Andreas Eriksson)
Date: Thu, 21 Aug 2008 09:06:57 +0200
Subject: [C++-sig] pybindgen and PyQt
Message-ID: <1f4b35e00808210006r66135dcfr1d52678adc0eac1@mail.gmail.com>

Hi,

does anyone know if pybindgen plays nicely with PyQt (which uses SIP
for wrapping)? Can C++ class wrapped with pybindgen be passed to PyQt?

Regards,

Andreas


From ivanov.maxim at gmail.com  Thu Aug 21 09:35:59 2008
From: ivanov.maxim at gmail.com (Max Ivanov)
Date: Thu, 21 Aug 2008 11:35:59 +0400
Subject: [C++-sig] copy_const_reference on add_property
In-Reply-To: <7465b6170808201122x72a9710ev59d7232278beafdb@mail.gmail.com>
References: <fbfa7b0c0808201113s65023c50o95f2f7c03b44a14d@mail.gmail.com>
	<7465b6170808201122x72a9710ev59d7232278beafdb@mail.gmail.com>
Message-ID: <fbfa7b0c0808210035l3d42a0d9se17842a33b7ee3c9@mail.gmail.com>

> You should use boost::python::make_function
>
> namespace bp = boost::python;
>
> .add_property(
> "name"
>
>                        , bp::make_function(
> <ref to function>, <call  policies> ) );
>

Works great. Thank you.


From ivanov.maxim at gmail.com  Thu Aug 21 11:30:23 2008
From: ivanov.maxim at gmail.com (Max Ivanov)
Date: Thu, 21 Aug 2008 13:30:23 +0400
Subject: [C++-sig] non const std::string* as argument
Message-ID: <fbfa7b0c0808210230x24864212tffde91585e858a6e@mail.gmail.com>

Hi all!
How to export to python class members like "bool
SerializeToString(string* output) const"? In python strings are
immutables, so there is no way to export this method directly. So I
tried to write a wrapper:

std::string&    wrap_SerializeToString(const gp::Message& msg) {
        std::string tmp;
        msg.SerializeToString(&tmp);
        return tmp;
};

and then  how I use it. Message class is base class for Link which I
try to export to python, but Link never overrides SerializeToString.

bp::class_<gp::Message, boost::noncopyable>("Message", bp::no_init)
            .def("SerializeToString", wrap_SerializeToString,
                            bp::return_value_policy<bp::copy_const_reference())

gcc errors are following:
bindings.cpp: In function 'std::string& wrap_SerializeToString(const
google::protobuf::Message&)':
bindings.cpp:21: warning: reference to local variable 'tmp' returned
bindings.cpp: In function 'void init_module__spidermsg()':
bindings.cpp:46: error: a call to a constructor cannot appear in a
constant-expression
bindings.cpp:46: error: template argument 1 is invalid


From roman.yakovenko at gmail.com  Thu Aug 21 11:38:13 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 21 Aug 2008 12:38:13 +0300
Subject: [C++-sig] non const std::string* as argument
In-Reply-To: <fbfa7b0c0808210230x24864212tffde91585e858a6e@mail.gmail.com>
References: <fbfa7b0c0808210230x24864212tffde91585e858a6e@mail.gmail.com>
Message-ID: <7465b6170808210238m6675c83btfeb2520dc7560078@mail.gmail.com>

On Thu, Aug 21, 2008 at 12:30 PM, Max Ivanov <ivanov.maxim at gmail.com> wrote:
> Hi all!
> How to export to python class members like "bool
> SerializeToString(string* output) const"? In python strings are
> immutables, so there is no way to export this method directly. So I
> tried to write a wrapper:
>
> std::string&    wrap_SerializeToString(const gp::Message& msg) {
>        std::string tmp;
>        msg.SerializeToString(&tmp);
>        return tmp;
> };

Just remove reference from return type and you should be fine.

> and then  how I use it.

Like any other class method.

> Message class is base class for Link which I
> try to export to python, but Link never overrides SerializeToString.

What?

> bp::class_<gp::Message, boost::noncopyable>("Message", bp::no_init)
>            .def("SerializeToString", wrap_SerializeToString,
>                            bp::return_value_policy<bp::copy_const_reference())
>
> gcc errors are following:
> bindings.cpp: In function 'std::string& wrap_SerializeToString(const
> google::protobuf::Message&)':
> bindings.cpp:21: warning: reference to local variable 'tmp' returned
> bindings.cpp: In function 'void init_module__spidermsg()':
> bindings.cpp:46: error: a call to a constructor cannot appear in a
> constant-expression
> bindings.cpp:46: error: template argument 1 is invalid

Take a look on Py++ "Function Transformation" feature:
http://language-binding.net/pyplusplus/documentation/functions/transformation/transformation.html

It deals with such "wrappers".

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


From ivanov.maxim at gmail.com  Thu Aug 21 12:31:31 2008
From: ivanov.maxim at gmail.com (Max Ivanov)
Date: Thu, 21 Aug 2008 14:31:31 +0400
Subject: [C++-sig] non const std::string* as argument
In-Reply-To: <7465b6170808210238m6675c83btfeb2520dc7560078@mail.gmail.com>
References: <fbfa7b0c0808210230x24864212tffde91585e858a6e@mail.gmail.com>
	<7465b6170808210238m6675c83btfeb2520dc7560078@mail.gmail.com>
Message-ID: <fbfa7b0c0808210331u513d2386v5e38d3a2117b15bf@mail.gmail.com>

>> std::string&    wrap_SerializeToString(const gp::Message& msg) {
>>        std::string tmp;
>>        msg.SerializeToString(&tmp);
>>        return tmp;
>> };
>
> Just remove reference from return type and you should be fine.
I've remove reference and bp::return_value_policy... from .def and
everything goes fine now, thanks again.

>> Message class is base class for Link which I
>> try to export to python, but Link never overrides SerializeToString.
>
> What?
I've Link class which I tried to export to Python Initially, which is
subclass of Message.

> Take a look on Py++ "Function Transformation" feature:
> http://language-binding.net/pyplusplus/documentation/functions/transformation/transformation.html
>
> It deals with such "wrappers".

Lot's of nice docs there, nice.


From ivanov.maxim at gmail.com  Thu Aug 21 12:31:31 2008
From: ivanov.maxim at gmail.com (Max Ivanov)
Date: Thu, 21 Aug 2008 14:31:31 +0400
Subject: [C++-sig] non const std::string* as argument
In-Reply-To: <7465b6170808210238m6675c83btfeb2520dc7560078@mail.gmail.com>
References: <fbfa7b0c0808210230x24864212tffde91585e858a6e@mail.gmail.com>
	<7465b6170808210238m6675c83btfeb2520dc7560078@mail.gmail.com>
Message-ID: <fbfa7b0c0808210331u513d2386v5e38d3a2117b15bf@mail.gmail.com>

>> std::string&    wrap_SerializeToString(const gp::Message& msg) {
>>        std::string tmp;
>>        msg.SerializeToString(&tmp);
>>        return tmp;
>> };
>
> Just remove reference from return type and you should be fine.
I've remove reference and bp::return_value_policy... from .def and
everything goes fine now, thanks again.

>> Message class is base class for Link which I
>> try to export to python, but Link never overrides SerializeToString.
>
> What?
I've Link class which I tried to export to Python Initially, which is
subclass of Message.

> Take a look on Py++ "Function Transformation" feature:
> http://language-binding.net/pyplusplus/documentation/functions/transformation/transformation.html
>
> It deals with such "wrappers".

Lot's of nice docs there, nice.


From gjcarneiro at gmail.com  Thu Aug 21 13:47:27 2008
From: gjcarneiro at gmail.com (Gustavo Carneiro)
Date: Thu, 21 Aug 2008 12:47:27 +0100
Subject: [C++-sig] pybindgen and PyQt
In-Reply-To: <1f4b35e00808210006r66135dcfr1d52678adc0eac1@mail.gmail.com>
References: <1f4b35e00808210006r66135dcfr1d52678adc0eac1@mail.gmail.com>
Message-ID: <a467ca4f0808210447m6f4ee44fj7c47132131d5f082@mail.gmail.com>

2008/8/21 Andreas Eriksson <aeriksson at gmail.com>

> Hi,
>
> does anyone know if pybindgen plays nicely with PyQt (which uses SIP
> for wrapping)? Can C++ class wrapped with pybindgen be passed to PyQt?


For one thing, I find it hard to believe that PyQt would have any API that
accepts an arbitrary class.  At most I would believe in PyQt accepting an
arbitrary subclass of a well known C++ base class wrapped by PyQt.  So I am
guessing you mean pybindgen wrapping a subclass of a PyQt-wrapped class.
That I'm afraid will not work.

Pybindgen does not allow subclassing classes defined in foreign wrapping
tools.  They would need to agree on many subtle details (instance structure
layout and C++ object memory management protocol) to make it work, and it's
very complicated.  So, not going to happen.  Sorry.

-- 
Gustavo J. A. M. Carneiro
INESC Porto, Telecommunications and Multimedia Unit
"The universe is always one step beyond logic." -- Frank Herbert
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080821/ad14e396/attachment.htm>

From paul.bilokon at lehman.com  Thu Aug 21 14:00:58 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Thu, 21 Aug 2008 13:00:58 +0100
Subject: [C++-sig] Container Conversions: Wrapping Functions Expecting
	Non-const References
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B56@lopets1exms003.leh.lbcorp.lehman.com>

Hi C++-sig,

I am looking at scitbx/boost_python/container_conversions.h, which I
think was written by Ralf Grosse-Kunstleve:
http://cci.lbl.gov/~rwgk/shortcuts/htdocs/current/c_plus_plus/container_
_conversions_8h-source.html
It's a great tool for something like displayVector below, which accepts
its parameter by value:
template <typename _Element>
void displayVector(vector<_Element> v)
{
    typedef vector<_Element> Vector;
    typedef Vector::iterator VectorIterator;

    VectorIterator vend = v.end();
    for (VectorIterator iter = v.begin(); iter != vend; ++iter)
    {
        cout << " - " << *iter << endl;
    }
}
I can wrap something like this into Python:

class_<ContainersExample2>("ContainersExample2", init<>())
    .def("displayVectorOfInts", &ContainersExample2::displayVector<int>)
    .def("displayVectorOfStrings",
&ContainersExample2::displayVector<string>)
    ...

No problem, as long as I also have this:

container_conversions::from_python_sequence<vector<int>,
container_conversions::variable_capacity_policy>();
container_conversions::from_python_sequence<vector<string>,
container_conversions::variable_capacity_policy>();

But what about the following method:

template <typename _Element>
void doubleVector(vector<_Element> & v)
{
    typedef vector<_Element> Vector;
    typedef Vector::iterator VectorIterator;

    VectorIterator vend = v.end();
    for (VectorIterator iter = v.begin(); iter != vend; ++iter)
    {
        *iter *= 2;
    }
}

Here I have difficulties:

Boost.Python.ArgumentError: Python argument types in
    ContainersExample2.doubleVectorOfInts(ContainersExample2, list)
did not match C++ signature:
    doubleVectorOfInts(class ContainersExample2 {lvalue}, class
std::vector<int,class std::allocator<int> > {lvalue})

What's the recommended technique(s) for this situation?

Many thanks,
Paul
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From aeriksson at gmail.com  Thu Aug 21 14:32:00 2008
From: aeriksson at gmail.com (Andreas Eriksson)
Date: Thu, 21 Aug 2008 14:32:00 +0200
Subject: [C++-sig] pybindgen and PyQt
In-Reply-To: <a467ca4f0808210447m6f4ee44fj7c47132131d5f082@mail.gmail.com>
References: <1f4b35e00808210006r66135dcfr1d52678adc0eac1@mail.gmail.com>
	<a467ca4f0808210447m6f4ee44fj7c47132131d5f082@mail.gmail.com>
Message-ID: <1f4b35e00808210532g252d2df0mf4885bc972feda66@mail.gmail.com>

Thanks for the reply Gustavo.

>> does anyone know if pybindgen plays nicely with PyQt (which uses SIP
>> for wrapping)? Can C++ class wrapped with pybindgen be passed to PyQt?
>
> For one thing, I find it hard to believe that PyQt would have any API that
> accepts an arbitrary class.  At most I would believe in PyQt accepting an
> arbitrary subclass of a well known C++ base class wrapped by PyQt.  So I am
> guessing you mean pybindgen wrapping a subclass of a PyQt-wrapped class.
> That I'm afraid will not work.

Actually, it is the first scenario that I am after. Let me describe in
some detail what I am after (slightly simplified):

We have a C++ class that derives from the C++ Qt class QWidget. Lets
call it OwnWidget.  Several Qt methods expects pointers to subclasses
of QWidget, such as QLayout::addWidget(QWidget*). Using the PyQt
bindings I can create a python QLayout object.  Now, what I want to do
is create a pybindgen wrapped OwnWidget instance and pass it to the
addWidget method of the PyQt QLayout object.

The following pseudo-code illustrates what I want to do.  I tried it
using boost.python but without any luck.

C++:

    #include <qt/QWidget.h>
    class OwnWidget : public QWidget { ... }

python:

    import pyqt
    import ownwidget    # wrapped using pybindgen

    myWidget = ownwidget.OwnWidget()
    layout = pyqt.QLayout()
    layout.addWidget(myWidget)

Regards,

Andreas


From gjcarneiro at gmail.com  Thu Aug 21 15:47:40 2008
From: gjcarneiro at gmail.com (Gustavo Carneiro)
Date: Thu, 21 Aug 2008 14:47:40 +0100
Subject: [C++-sig] pybindgen and PyQt
In-Reply-To: <1f4b35e00808210532g252d2df0mf4885bc972feda66@mail.gmail.com>
References: <1f4b35e00808210006r66135dcfr1d52678adc0eac1@mail.gmail.com>
	<a467ca4f0808210447m6f4ee44fj7c47132131d5f082@mail.gmail.com>
	<1f4b35e00808210532g252d2df0mf4885bc972feda66@mail.gmail.com>
Message-ID: <a467ca4f0808210647w6a837dbepa0acb0c065cddfe6@mail.gmail.com>

2008/8/21 Andreas Eriksson <aeriksson at gmail.com>

> Thanks for the reply Gustavo.
>
> >> does anyone know if pybindgen plays nicely with PyQt (which uses SIP
> >> for wrapping)? Can C++ class wrapped with pybindgen be passed to PyQt?
> >
> > For one thing, I find it hard to believe that PyQt would have any API
> that
> > accepts an arbitrary class.  At most I would believe in PyQt accepting an
> > arbitrary subclass of a well known C++ base class wrapped by PyQt.  So I
> am
> > guessing you mean pybindgen wrapping a subclass of a PyQt-wrapped class.
> > That I'm afraid will not work.
>
> Actually, it is the first scenario that I am after. Let me describe in
> some detail what I am after (slightly simplified):
>
> We have a C++ class that derives from the C++ Qt class QWidget. Lets
> call it OwnWidget.  Several Qt methods expects pointers to subclasses
> of QWidget, such as QLayout::addWidget(QWidget*). Using the PyQt
> bindings I can create a python QLayout object.  Now, what I want to do
> is create a pybindgen wrapped OwnWidget instance and pass it to the
> addWidget method of the PyQt QLayout object.
>
> The following pseudo-code illustrates what I want to do.  I tried it
> using boost.python but without any luck.


I'm sure that boost::python would have had the exact same problem.  There is
no 'standard' for the layout and policy of python object instances wrapping
underlying C++ objects, therefore different wrapping tools cannot subclass
each other's classes.  Moreover, if you wrap in pybindgen both QWidget and
OwnWidget, it may work on its own "world", but then if you pass a OwnWidget
instance into PyQt it will not recognize it as being a pyqt.Widget type.
It's a parallel type hierarchy from the point of view of python.

I'm afraid I really don't have any simple solution for you.  The easiest way
this could work would be via double wrappers.  The same OwnWidget would have
to have an initial wrapper created by pybindgen, then an utility function
that you add to the module would create a pyqt wrapper for the pybindgen
wrapped object, which you can then use in pyqt.  The utility conversion
function needs to be written manually, with knowledge of the SIP API, and
added with Module.add_custom_function_wrapper().  And this only works if you
do not subclass OwnWidget in Python, because pybindgen then assumes it has
the only wrapper for an object; with two wrappers reference counting fails
and you get memory leaks.

It is probably easier just to wrap everything in SIP... :P


>
> C++:
>
>    #include <qt/QWidget.h>
>    class OwnWidget : public QWidget { ... }
>
> python:
>
>    import pyqt
>    import ownwidget    # wrapped using pybindgen
>
>    myWidget = ownwidget.OwnWidget()
>    layout = pyqt.QLayout()
>    layout.addWidget(myWidget)
>
> Regards,
>
> Andreas
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org <C%2B%2B-sig at python.org>
> http://mail.python.org/mailman/listinfo/c++-sig
>



-- 
Gustavo J. A. M. Carneiro
INESC Porto, Telecommunications and Multimedia Unit
"The universe is always one step beyond logic." -- Frank Herbert
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080821/764cb15d/attachment.htm>

From william.marie at gmail.com  Thu Aug 21 19:22:31 2008
From: william.marie at gmail.com (=?UTF-8?Q?William_Mari=C3=A9?=)
Date: Thu, 21 Aug 2008 10:22:31 -0700 (PDT)
Subject: [C++-sig]  Calling overidden pure virtual function from C++ side
Message-ID: <19093207.post@talk.nabble.com>


Hi, 

I have a little problem using a pure virtual function exposed to the python
and called from the c++ :

In C++ :
--------------------------
class A
{
	virtual void foo() = 0;
	void bar()
        {
              foo();
        }
}

struct AWrap : A, bp::wrapper 
{
	void foo()
        {
		this->get_override( "foo" );
	}
};

BOOST_PYTHON_MODULE(dum)
{
	bp::class_<AWrap,boost::noncopyable>( "A" )
		.def( "foo", bp::pure_virtual( &A::foo ) )
		.def( "bar", &A::bar )
		;
} 

In Python :
--------------------------
import dum
class B( dum.A ) :

    def foo :
        print "Foo from B"

b = B()
b.bar()

So here b.bar() never prints "Foo from B"

Does anyone could tell me what am i doing wrong ?
Thanks
William

-- 
View this message in context: http://www.nabble.com/Calling-overidden-pure-virtual-function-from-C%2B%2B-side-tp19093207p19093207.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From roman.yakovenko at gmail.com  Thu Aug 21 19:27:43 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 21 Aug 2008 20:27:43 +0300
Subject: [C++-sig] Calling overidden pure virtual function from C++ side
In-Reply-To: <19093207.post@talk.nabble.com>
References: <19093207.post@talk.nabble.com>
Message-ID: <7465b6170808211027p4b978d9dua57fee622d792e27@mail.gmail.com>

On Thu, Aug 21, 2008 at 8:22 PM, William Mari? <william.marie at gmail.com> wrote:
>
> Hi,
>
> I have a little problem using a pure virtual function exposed to the python
> and called from the c++ :
>
> In C++ :
> --------------------------
> class A
> {
>        virtual void foo() = 0;
>        void bar()
>        {
>              foo();
>        }
> }
>
> struct AWrap : A, bp::wrapper
> {
>        void foo()
>        {
>                this->get_override( "foo" );
>        }
> };
>
> BOOST_PYTHON_MODULE(dum)
> {
>        bp::class_<AWrap,boost::noncopyable>( "A" )
>                .def( "foo", bp::pure_virtual( &A::foo ) )
>                .def( "bar", &A::bar )
>                ;
> }
>
> In Python :
> --------------------------
> import dum
> class B( dum.A ) :

Try to add __init__ method

def __init__( self ):
    dum.A.__init__( self )


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


From ope-devel at gmx.de  Thu Aug 21 19:31:04 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Thu, 21 Aug 2008 19:31:04 +0200
Subject: [C++-sig] [pybindgen] Problems on class constructor
Message-ID: <48ADA658.6080709@gmx.de>

Hi,

given the following header:

---8<---
class PyConsole;

class StdoutPy
{
public:
    StdoutPy(PyConsole* console);
    void write(PyObject* args);
    void flush();
private:
    PyConsole*			m_console;
};

class PyConsole
{
    void writeOut(PyObject* args);
    void writeErr(PyObject* args);
    void flush();
};
--->8---


and the python script:

---8<---
import pybindgen
import sys

# convenience
param=pybindgen.param
retval=pybindgen.retval

pybindgen.settings.min_python_version = (2,5)

mod = pybindgen.Module('redirector')
mod.add_include('"redirector.hpp"')

stdout_py = mod.add_class('StdoutPy')
stdout_py.add_constructor([param('PyConsole*', 'cpp_parent',
transfer_ownership=False)])
stdout_py.add_method('write', None, [param('PyObject*', 'value',
transfer_ownership=False)])
stdout_py.add_method('flush', None, [])

mod.generate(sys.stdout)
--->8---

I get the error:

$ python redirector.py
Traceback (most recent call last):
  File "redirector.py", line 14, in <module>
    stdout_py.add_constructor([param('PyConsole*', 'cpp_parent',
transfer_ownership=False)])
  File "/usr/lib/python2.5/site-packages/pybindgen/cppclass.py", line
1228, in add_constructor
    constructor = CppConstructor(*args, **kwargs)
  File "/usr/lib/python2.5/site-packages/pybindgen/cppmethod.py", line
415, in __init__
    parameters = [utils.eval_param(param, self) for param in parameters]
  File "/usr/lib/python2.5/site-packages/pybindgen/utils.py", line 205,
in eval_param
    TypeLookupError))
  File "/usr/lib/python2.5/site-packages/pybindgen/utils.py", line 115,
in call_with_error_handling
    return callable(*args, **kwargs)
  File
"/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py", line
1167, in new
    param_type_matcher.lookup(ctype)
  File
"/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py", line
1300, in lookup
    raise TypeLookupError(name)
pybindgen.typehandlers.base.TypeLookupError: PyConsole *


It seems, that I have to register my own types but I have no idea how to
do this. For Testing I've used the xmlgcc approach:

---8<---
import sys

from pybindgen import FileCodeSink
from pybindgen.gccxmlparser import ModuleParser

def my_module_gen():
    module_parser = ModuleParser('redirector', '::')
    module_parser.parse([sys.argv[1]],
include_paths=['/usr/include/python2.5'],includes=['"redirector.hpp"'],
pygen_sink=FileCodeSink(sys.stdout))

if __name__ == '__main__':
    my_module_gen()
--->8---

which results in errors too:

/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py:580:
DeprecationWarning: Parameter include_paths is deprecated, use
gccxml_options instead
  self.parse_init(header_files, include_paths, whitelist_paths,
includes, pygen_sink, pygen_classifier)

INFO Parsing source file "redirector.hpp" ...

INFO gccxml cmd: /usr/bin/gccxml  -I"." -I"/usr/include/python2.5"
"redirector.hpp" -fxml="/tmp/tmp5ORqw3.xml"

INFO GCCXML version - 0.7
/usr/lib/python2.5/site-packages/pygccxml/parser/scanner.py:335:
UserWarning: unable to find out array size from expression ""
  warnings.warn( msg )
from pybindgen import Module, FileCodeSink, param, retval, cppclass


import pybindgen.settings
import warnings

class ErrorHandler(pybindgen.settings.ErrorHandler):
    def handle_error(self, wrapper, exception, traceback_):
        warnings.warn("exception %r in wrapper %s" % (exception, wrapper))
        return True
pybindgen.settings.error_handler = ErrorHandler()


import sys

def module_init():
    root_module = Module('redirector', cpp_namespace='::')
    root_module.add_include('"redirector.hpp"')
    return root_module

def register_types(module):
    root_module = module.get_root()

    module.add_class('StdoutPy')
    module.add_class('PyConsole')

    ## Register a nested module for the namespace std

    nested_module = module.add_cpp_namespace('std')
    register_types_std(nested_module)


def register_types_std(module):
    root_module = module.get_root()


def register_methods(root_module):
    register_StdoutPy_methods(root_module, root_module['StdoutPy'])
    register_PyConsole_methods(root_module, root_module['PyConsole'])
    return

def register_StdoutPy_methods(root_module, cls):
/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py:215:
UserWarning: multiple &'s not handled
  warnings.warn("multiple &'s not handled")
/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py:211:
UserWarning: multiple consts not handled
  warnings.warn("multiple consts not handled")
    cls.add_constructor([param('StdoutPy&', '_ctor_arg', is_const=True)])
    cls.add_constructor([param('PyConsole *', 'console')])
/var/tmp/redirector.hpp:12: WrapperWarning: Parameter '::PyConsole *
console' error (used in StdoutPy::StdoutPy(PyConsole * console)
[constructor]): TypeConfigurationError('transfer_ownership parameter
missing',)
  StdoutPy(PyConsole* console);
    cls.add_method('write',
                   'void',
                   [param('PyObject *', 'args')])
/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py:1173:
UserWarning: Exception '__init__() takes at least 4 arguments (3 given)'
in type handler <class
'pybindgen.typehandlers.pyobjecttype.PyObjectParam'> constructor
  warnings.warn("Exception %r in type handler %s constructor" %
(str(ex), type_handler_class))
Traceback (most recent call last):
  File "redirector_py.py", line 13, in <module>
    my_module_gen()
  File "redirector_py.py", line 10, in my_module_gen
    module_parser.parse([sys.argv[1]],
include_paths=['/usr/include/python2.5'],includes=['"redirector.hpp"'],
pygen_sink=FileCodeSink(sys.stdout))
  File "/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py",
line 582, in parse
    self.scan_methods()
  File "/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py",
line 809, in scan_methods
    self._scan_class_methods(class_wrapper.gccxml_definition,
class_wrapper, pygen_sink)
  File "/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py",
line 1420, in _scan_class_methods
    arguments.append(Parameter.new(*arg[0], **arg[1]))
  File
"/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py", line
1171, in new
    return type_handler_class(*args, **kwargs)
TypeError: __init__() takes at least 4 arguments (3 given)

BTW, how can I prevent the construction of the StdoutPy class from
python? It has to be created from c++ side.

Thanks,
Olaf


From william.marie at gmail.com  Thu Aug 21 19:50:59 2008
From: william.marie at gmail.com (=?ISO-8859-1?Q?william_mari=E9?=)
Date: Thu, 21 Aug 2008 18:50:59 +0100
Subject: [C++-sig] Calling overidden pure virtual function from C++ side
In-Reply-To: <7465b6170808211027p4b978d9dua57fee622d792e27@mail.gmail.com>
References: <19093207.post@talk.nabble.com>
	<7465b6170808211027p4b978d9dua57fee622d792e27@mail.gmail.com>
Message-ID: <168485c00808211050l6f04f83cqf6bb3ae133308a4a@mail.gmail.com>

Hi,

It's already done sorry, i didn't write the __init__ function in the post :
    def __init__( self ) :
          dum.A.__init__( self )

Even with that it's not working, can we normally call virtual function from
the C++ side ?


2008/8/21 Roman Yakovenko <roman.yakovenko at gmail.com>

> On Thu, Aug 21, 2008 at 8:22 PM, William Mari? <william.marie at gmail.com>
> wrote:
> >
> > Hi,
> >
> > I have a little problem using a pure virtual function exposed to the
> python
> > and called from the c++ :
> >
> > In C++ :
> > --------------------------
> > class A
> > {
> >        virtual void foo() = 0;
> >        void bar()
> >        {
> >              foo();
> >        }
> > }
> >
> > struct AWrap : A, bp::wrapper
> > {
> >        void foo()
> >        {
> >                this->get_override( "foo" );
> >        }
> > };
> >
> > BOOST_PYTHON_MODULE(dum)
> > {
> >        bp::class_<AWrap,boost::noncopyable>( "A" )
> >                .def( "foo", bp::pure_virtual( &A::foo ) )
> >                .def( "bar", &A::bar )
> >                ;
> > }
> >
> > In Python :
> > --------------------------
> > import dum
> > class B( dum.A ) :
>
> Try to add __init__ method
>
> def __init__( self ):
>    dum.A.__init__( self )
>
>
> --
> Roman Yakovenko
> C++ Python language binding
> http://www.language-binding.net/
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org <C%2B%2B-sig at python.org>
> http://mail.python.org/mailman/listinfo/c++-sig
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080821/7d26cd9c/attachment.htm>

From gjcarneiro at gmail.com  Thu Aug 21 20:00:07 2008
From: gjcarneiro at gmail.com (Gustavo Carneiro)
Date: Thu, 21 Aug 2008 19:00:07 +0100
Subject: [C++-sig] [pybindgen] Problems on class constructor
In-Reply-To: <48ADA658.6080709@gmx.de>
References: <48ADA658.6080709@gmx.de>
Message-ID: <a467ca4f0808211100s262bed2ekf85d887751609513@mail.gmail.com>

2008/8/21 Olaf Peter <ope-devel at gmx.de>

> Hi,
>
> given the following header:
>
> ---8<---
> class PyConsole;
>
> class StdoutPy
> {
> public:
>    StdoutPy(PyConsole* console);
>    void write(PyObject* args);
>    void flush();
> private:
>    PyConsole*                  m_console;
> };
>
> class PyConsole
> {
>    void writeOut(PyObject* args);
>    void writeErr(PyObject* args);
>    void flush();
> };
> --->8---
>
>
> and the python script:
>
> ---8<---
> import pybindgen
> import sys
>
> # convenience
> param=pybindgen.param
> retval=pybindgen.retval
>
> pybindgen.settings.min_python_version = (2,5)
>
> mod = pybindgen.Module('redirector')
> mod.add_include('"redirector.hpp"')
>
> stdout_py = mod.add_class('StdoutPy')
> stdout_py.add_constructor([param('PyConsole*', 'cpp_parent',
> transfer_ownership=False)])
> stdout_py.add_method('write', None, [param('PyObject*', 'value',
> transfer_ownership=False)])
> stdout_py.add_method('flush', None, [])
>
> mod.generate(sys.stdout)
> --->8---
>
> I get the error:
>
> $ python redirector.py
> Traceback (most recent call last):
>  File "redirector.py", line 14, in <module>
>    stdout_py.add_constructor([param('PyConsole*', 'cpp_parent',
> transfer_ownership=False)])
>  File "/usr/lib/python2.5/site-packages/pybindgen/cppclass.py", line
> 1228, in add_constructor
>    constructor = CppConstructor(*args, **kwargs)
>  File "/usr/lib/python2.5/site-packages/pybindgen/cppmethod.py", line
> 415, in __init__
>    parameters = [utils.eval_param(param, self) for param in parameters]
>  File "/usr/lib/python2.5/site-packages/pybindgen/utils.py", line 205,
> in eval_param
>    TypeLookupError))
>  File "/usr/lib/python2.5/site-packages/pybindgen/utils.py", line 115,
> in call_with_error_handling
>    return callable(*args, **kwargs)
>  File
> "/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py", line
> 1167, in new
>    param_type_matcher.lookup(ctype)
>  File
> "/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py", line
> 1300, in lookup
>    raise TypeLookupError(name)
> pybindgen.typehandlers.base.TypeLookupError: PyConsole *
>
>
> It seems, that I have to register my own types but I have no idea how to
> do this.


You just have to declare the PyConsole class, with:
mod.add_class('PyConsole')

For Testing I've used the xmlgcc approach:
>
> ---8<---
> import sys
>
> from pybindgen import FileCodeSink
> from pybindgen.gccxmlparser import ModuleParser
>
> def my_module_gen():
>    module_parser = ModuleParser('redirector', '::')
>    module_parser.parse([sys.argv[1]],
> include_paths=['/usr/include/python2.5'],includes=['"redirector.hpp"'],
> pygen_sink=FileCodeSink(sys.stdout))
>
> if __name__ == '__main__':
>    my_module_gen()
> --->8---
>
> which results in errors too:
>
> /usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py:580:
> DeprecationWarning: Parameter include_paths is deprecated, use
> gccxml_options instead
>  self.parse_init(header_files, include_paths, whitelist_paths,
> includes, pygen_sink, pygen_classifier)
>
> INFO Parsing source file "redirector.hpp" ...
>
> INFO gccxml cmd: /usr/bin/gccxml  -I"." -I"/usr/include/python2.5"
> "redirector.hpp" -fxml="/tmp/tmp5ORqw3.xml"
>
> INFO GCCXML version - 0.7
> /usr/lib/python2.5/site-packages/pygccxml/parser/scanner.py:335:
> UserWarning: unable to find out array size from expression ""
>  warnings.warn( msg )
> from pybindgen import Module, FileCodeSink, param, retval, cppclass
>
>
> import pybindgen.settings
> import warnings
>
> class ErrorHandler(pybindgen.settings.ErrorHandler):
>    def handle_error(self, wrapper, exception, traceback_):
>        warnings.warn("exception %r in wrapper %s" % (exception, wrapper))
>        return True
> pybindgen.settings.error_handler = ErrorHandler()
>
>
> import sys
>
> def module_init():
>    root_module = Module('redirector', cpp_namespace='::')
>    root_module.add_include('"redirector.hpp"')
>    return root_module
>
> def register_types(module):
>    root_module = module.get_root()
>
>    module.add_class('StdoutPy')
>    module.add_class('PyConsole')
>
>    ## Register a nested module for the namespace std
>
>    nested_module = module.add_cpp_namespace('std')
>    register_types_std(nested_module)
>
>
> def register_types_std(module):
>    root_module = module.get_root()
>
>
> def register_methods(root_module):
>    register_StdoutPy_methods(root_module, root_module['StdoutPy'])
>    register_PyConsole_methods(root_module, root_module['PyConsole'])
>    return
>
> def register_StdoutPy_methods(root_module, cls):
> /usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py:215:
> UserWarning: multiple &'s not handled
>  warnings.warn("multiple &'s not handled")
> /usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py:211:
> UserWarning: multiple consts not handled
>  warnings.warn("multiple consts not handled")
>    cls.add_constructor([param('StdoutPy&', '_ctor_arg', is_const=True)])
>    cls.add_constructor([param('PyConsole *', 'console')])
> /var/tmp/redirector.hpp:12: WrapperWarning: Parameter '::PyConsole *
> console' error (used in StdoutPy::StdoutPy(PyConsole * console)
> [constructor]): TypeConfigurationError('transfer_ownership parameter
> missing',)
>  StdoutPy(PyConsole* console);
>    cls.add_method('write',
>                   'void',
>                   [param('PyObject *', 'args')])
> /usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py:1173:
> UserWarning: Exception '__init__() takes at least 4 arguments (3 given)'
> in type handler <class
> 'pybindgen.typehandlers.pyobjecttype.PyObjectParam'> constructor
>  warnings.warn("Exception %r in type handler %s constructor" %
> (str(ex), type_handler_class))
> Traceback (most recent call last):
>  File "redirector_py.py", line 13, in <module>
>    my_module_gen()
>  File "redirector_py.py", line 10, in my_module_gen
>    module_parser.parse([sys.argv[1]],
> include_paths=['/usr/include/python2.5'],includes=['"redirector.hpp"'],
> pygen_sink=FileCodeSink(sys.stdout))
>  File "/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py",
> line 582, in parse
>    self.scan_methods()
>  File "/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py",
> line 809, in scan_methods
>    self._scan_class_methods(class_wrapper.gccxml_definition,
> class_wrapper, pygen_sink)
>  File "/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py",
> line 1420, in _scan_class_methods
>    arguments.append(Parameter.new(*arg[0], **arg[1]))
>  File
> "/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py", line
> 1171, in new
>    return type_handler_class(*args, **kwargs)
> TypeError: __init__() takes at least 4 arguments (3 given)


I am guessing you need to add a transfer_ownership=false annotation in some
places.  I know, annotations are not properly documented, so I don't blame
you for the mistake :P

I think your header file would need something like this (check the -#-
comments):

class PyConsole;

class StdoutPy
{
public:
   // -#- @console(transfer_ownership=false) -#-
   StdoutPy(PyConsole* console);
   // -#- @args(transfer_ownership=false) -#-
   void write(PyObject* args);
   void flush();
private:
   PyConsole*                  m_console;
};

class PyConsole
{
   // -#- @args(transfer_ownership=false) -#-
   void writeOut(PyObject* args);
   // -#- @args(transfer_ownership=false) -#-
   void writeErr(PyObject* args);
   void flush();
};


>
> BTW, how can I prevent the construction of the StdoutPy class from
> python? It has to be created from c++ side.


Using gccxml scanner, there's no easy way to do it.  I guess that, after
scanning, you can reset the list of constructors to the empty list:
module['StdoutPy'].constructors = []

Using the python API interface you just don't declare to pybindgen that a
class has a public constructor, and pybindgen will generate a special
constructor that raises an exception saying the class cannot be constructed.


-- 
Gustavo J. A. M. Carneiro
INESC Porto, Telecommunications and Multimedia Unit
"The universe is always one step beyond logic." -- Frank Herbert
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080821/cf6bad3b/attachment.htm>

From roman.yakovenko at gmail.com  Thu Aug 21 20:01:27 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 21 Aug 2008 21:01:27 +0300
Subject: [C++-sig] Calling overidden pure virtual function from C++ side
In-Reply-To: <168485c00808211050l6f04f83cqf6bb3ae133308a4a@mail.gmail.com>
References: <19093207.post@talk.nabble.com>
	<7465b6170808211027p4b978d9dua57fee622d792e27@mail.gmail.com>
	<168485c00808211050l6f04f83cqf6bb3ae133308a4a@mail.gmail.com>
Message-ID: <7465b6170808211101v3217a569g1d8f5b6a1348fa10@mail.gmail.com>

2008/8/21 william mari? <william.marie at gmail.com>:
> Hi,
>
> It's already done sorry, i didn't write the __init__ function in the post :
>     def __init__( self ) :
>           dum.A.__init__( self )
>
> Even with that it's not working, can we normally call virtual function from
> the C++ side ?

Here is an exercise for you: find the difference between your code and
the one, that were generated by Py++:

#include "boost/python.hpp"

#include "1.h"

namespace bp = boost::python;

struct A_wrapper : A, bp::wrapper< A > {

    A_wrapper()
    : A()
      , bp::wrapper< A >(){
        // null constructor

    }

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

};

BOOST_PYTHON_MODULE(pyplusplus){
    bp::class_< A_wrapper, boost::noncopyable >( "A" )
        .def(
            "bar"
            , (void ( ::A::* )(  ) )( &::A::bar ) )
        .def(
            "foo"
            , bp::pure_virtual( (void ( ::A::* )(  ) )(&::A::foo) ) );
}

P.S. Next time try to post example, which contains "the whole picture".

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


From seefeld at sympatico.ca  Thu Aug 21 20:01:58 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Thu, 21 Aug 2008 14:01:58 -0400
Subject: [C++-sig] Calling overidden pure virtual function from C++ side
In-Reply-To: <168485c00808211050l6f04f83cqf6bb3ae133308a4a@mail.gmail.com>
References: <19093207.post@talk.nabble.com>	<7465b6170808211027p4b978d9dua57fee622d792e27@mail.gmail.com>
	<168485c00808211050l6f04f83cqf6bb3ae133308a4a@mail.gmail.com>
Message-ID: <48ADAD96.5000102@sympatico.ca>

william mari? wrote:
>
>
>     > struct AWrap : A, bp::wrapper
>     > {
>     >        void foo()
>     >        {
>     >                this->get_override( "foo" );
>     >        }
>     > };
>

You retrieve the override, but you don't call it. See
http://www.boost.org/doc/libs/1_36_0/libs/python/doc/v2/wrapper.html#examples

HTH,
       Stefan

-- 

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



From gjcarneiro at gmail.com  Thu Aug 21 20:05:10 2008
From: gjcarneiro at gmail.com (Gustavo Carneiro)
Date: Thu, 21 Aug 2008 19:05:10 +0100
Subject: [C++-sig] [pybindgen] Problems on class constructor
In-Reply-To: <a467ca4f0808211100s262bed2ekf85d887751609513@mail.gmail.com>
References: <48ADA658.6080709@gmx.de>
	<a467ca4f0808211100s262bed2ekf85d887751609513@mail.gmail.com>
Message-ID: <a467ca4f0808211105w4d6ea2b1m1846c52cb58879ea@mail.gmail.com>

2008/8/21 Gustavo Carneiro <gjcarneiro at gmail.com>

>
>
> 2008/8/21 Olaf Peter <ope-devel at gmx.de>
>
> Hi,
>>
>> given the following header:
>>
>> ---8<---
>> class PyConsole;
>>
>> class StdoutPy
>> {
>> public:
>>    StdoutPy(PyConsole* console);
>>    void write(PyObject* args);
>>    void flush();
>> private:
>>    PyConsole*                  m_console;
>> };
>>
>> class PyConsole
>> {
>>    void writeOut(PyObject* args);
>>    void writeErr(PyObject* args);
>>    void flush();
>> };
>> --->8---
>>
>>
>> and the python script:
>>
>> ---8<---
>> import pybindgen
>> import sys
>>
>> # convenience
>> param=pybindgen.param
>> retval=pybindgen.retval
>>
>> pybindgen.settings.min_python_version = (2,5)
>>
>> mod = pybindgen.Module('redirector')
>> mod.add_include('"redirector.hpp"')
>>
>> stdout_py = mod.add_class('StdoutPy')
>> stdout_py.add_constructor([param('PyConsole*', 'cpp_parent',
>> transfer_ownership=False)])
>> stdout_py.add_method('write', None, [param('PyObject*', 'value',
>> transfer_ownership=False)])
>> stdout_py.add_method('flush', None, [])
>>
>> mod.generate(sys.stdout)
>> --->8---
>>
>> I get the error:
>>
>> $ python redirector.py
>> Traceback (most recent call last):
>>  File "redirector.py", line 14, in <module>
>>    stdout_py.add_constructor([param('PyConsole*', 'cpp_parent',
>> transfer_ownership=False)])
>>  File "/usr/lib/python2.5/site-packages/pybindgen/cppclass.py", line
>> 1228, in add_constructor
>>    constructor = CppConstructor(*args, **kwargs)
>>  File "/usr/lib/python2.5/site-packages/pybindgen/cppmethod.py", line
>> 415, in __init__
>>    parameters = [utils.eval_param(param, self) for param in parameters]
>>  File "/usr/lib/python2.5/site-packages/pybindgen/utils.py", line 205,
>> in eval_param
>>    TypeLookupError))
>>  File "/usr/lib/python2.5/site-packages/pybindgen/utils.py", line 115,
>> in call_with_error_handling
>>    return callable(*args, **kwargs)
>>  File
>> "/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py", line
>> 1167, in new
>>    param_type_matcher.lookup(ctype)
>>  File
>> "/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py", line
>> 1300, in lookup
>>    raise TypeLookupError(name)
>> pybindgen.typehandlers.base.TypeLookupError: PyConsole *
>>
>>
>> It seems, that I have to register my own types but I have no idea how to
>> do this.
>
>
> You just have to declare the PyConsole class, with:
> mod.add_class('PyConsole')
>
> For Testing I've used the xmlgcc approach:
>>
>> ---8<---
>> import sys
>>
>> from pybindgen import FileCodeSink
>> from pybindgen.gccxmlparser import ModuleParser
>>
>> def my_module_gen():
>>    module_parser = ModuleParser('redirector', '::')
>>    module_parser.parse([sys.argv[1]],
>> include_paths=['/usr/include/python2.5'],includes=['"redirector.hpp"'],
>> pygen_sink=FileCodeSink(sys.stdout))
>>
>> if __name__ == '__main__':
>>    my_module_gen()
>> --->8---
>>
>> which results in errors too:
>>
>> /usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py:580:
>> DeprecationWarning: Parameter include_paths is deprecated, use
>> gccxml_options instead
>>  self.parse_init(header_files, include_paths, whitelist_paths,
>> includes, pygen_sink, pygen_classifier)
>>
>> INFO Parsing source file "redirector.hpp" ...
>>
>> INFO gccxml cmd: /usr/bin/gccxml  -I"." -I"/usr/include/python2.5"
>> "redirector.hpp" -fxml="/tmp/tmp5ORqw3.xml"
>>
>> INFO GCCXML version - 0.7
>> /usr/lib/python2.5/site-packages/pygccxml/parser/scanner.py:335:
>> UserWarning: unable to find out array size from expression ""
>>  warnings.warn( msg )
>> from pybindgen import Module, FileCodeSink, param, retval, cppclass
>>
>>
>> import pybindgen.settings
>> import warnings
>>
>> class ErrorHandler(pybindgen.settings.ErrorHandler):
>>    def handle_error(self, wrapper, exception, traceback_):
>>        warnings.warn("exception %r in wrapper %s" % (exception, wrapper))
>>        return True
>> pybindgen.settings.error_handler = ErrorHandler()
>>
>>
>> import sys
>>
>> def module_init():
>>    root_module = Module('redirector', cpp_namespace='::')
>>    root_module.add_include('"redirector.hpp"')
>>    return root_module
>>
>> def register_types(module):
>>    root_module = module.get_root()
>>
>>    module.add_class('StdoutPy')
>>    module.add_class('PyConsole')
>>
>>    ## Register a nested module for the namespace std
>>
>>    nested_module = module.add_cpp_namespace('std')
>>    register_types_std(nested_module)
>>
>>
>> def register_types_std(module):
>>    root_module = module.get_root()
>>
>>
>> def register_methods(root_module):
>>    register_StdoutPy_methods(root_module, root_module['StdoutPy'])
>>    register_PyConsole_methods(root_module, root_module['PyConsole'])
>>    return
>>
>> def register_StdoutPy_methods(root_module, cls):
>> /usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py:215:
>> UserWarning: multiple &'s not handled
>>  warnings.warn("multiple &'s not handled")
>> /usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py:211:
>> UserWarning: multiple consts not handled
>>  warnings.warn("multiple consts not handled")
>>    cls.add_constructor([param('StdoutPy&', '_ctor_arg', is_const=True)])
>>    cls.add_constructor([param('PyConsole *', 'console')])
>> /var/tmp/redirector.hpp:12: WrapperWarning: Parameter '::PyConsole *
>> console' error (used in StdoutPy::StdoutPy(PyConsole * console)
>> [constructor]): TypeConfigurationError('transfer_ownership parameter
>> missing',)
>>  StdoutPy(PyConsole* console);
>>    cls.add_method('write',
>>                   'void',
>>                   [param('PyObject *', 'args')])
>> /usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py:1173:
>> UserWarning: Exception '__init__() takes at least 4 arguments (3 given)'
>> in type handler <class
>> 'pybindgen.typehandlers.pyobjecttype.PyObjectParam'> constructor
>>  warnings.warn("Exception %r in type handler %s constructor" %
>> (str(ex), type_handler_class))
>> Traceback (most recent call last):
>>  File "redirector_py.py", line 13, in <module>
>>    my_module_gen()
>>  File "redirector_py.py", line 10, in my_module_gen
>>    module_parser.parse([sys.argv[1]],
>> include_paths=['/usr/include/python2.5'],includes=['"redirector.hpp"'],
>> pygen_sink=FileCodeSink(sys.stdout))
>>  File "/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py",
>> line 582, in parse
>>    self.scan_methods()
>>  File "/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py",
>> line 809, in scan_methods
>>    self._scan_class_methods(class_wrapper.gccxml_definition,
>> class_wrapper, pygen_sink)
>>  File "/usr/lib/python2.5/site-packages/pybindgen/gccxmlparser.py",
>> line 1420, in _scan_class_methods
>>    arguments.append(Parameter.new(*arg[0], **arg[1]))
>>  File
>> "/usr/lib/python2.5/site-packages/pybindgen/typehandlers/base.py", line
>> 1171, in new
>>    return type_handler_class(*args, **kwargs)
>> TypeError: __init__() takes at least 4 arguments (3 given)
>
>
> I am guessing you need to add a transfer_ownership=false annotation in some
> places.  I know, annotations are not properly documented, so I don't blame
> you for the mistake :P
>
> I think your header file would need something like this (check the -#-
> comments):
>
> class PyConsole;
>
> class StdoutPy
> {
> public:
>    // -#- @console(transfer_ownership=false) -#-
>    StdoutPy(PyConsole* console);
>    // -#- @args(transfer_ownership=false) -#-
>    void write(PyObject* args);
>    void flush();
> private:
>    PyConsole*                  m_console;
> };
>
> class PyConsole
> {
>    // -#- @args(transfer_ownership=false) -#-
>    void writeOut(PyObject* args);
>    // -#- @args(transfer_ownership=false) -#-
>    void writeErr(PyObject* args);
>    void flush();
> };
>
>
>>
>> BTW, how can I prevent the construction of the StdoutPy class from
>> python? It has to be created from c++ side.
>>
>
> Using gccxml scanner, there's no easy way to do it.  I guess that, after
> scanning, you can reset the list of constructors to the empty list:
> module['StdoutPy'].constructors = []
>

After hitting send I remembered...  A less hackish way to do it with the
gccxml scanner is to add 'ignore' annotations to the constructors.

class PyConsole
{
public:

  // -#- ignore -#-
   PyConsole () {} // need to make the constructor explicit for this to
work...

[...]
};

P. S.: the annotations system is rather fragile.  Annotations are looked for
in the line above the line where GCC-XML reports that the declaration is
defined.  But GCC-XML reports weird line numbers on multi-line declarations,
so beware.

-- 
Gustavo J. A. M. Carneiro
INESC Porto, Telecommunications and Multimedia Unit
"The universe is always one step beyond logic." -- Frank Herbert
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080821/8e33b6c4/attachment.htm>

From william.marie at gmail.com  Thu Aug 21 20:15:22 2008
From: william.marie at gmail.com (=?ISO-8859-1?Q?william_mari=E9?=)
Date: Thu, 21 Aug 2008 19:15:22 +0100
Subject: [C++-sig] Calling overidden pure virtual function from C++ side
In-Reply-To: <48ADAD96.5000102@sympatico.ca>
References: <19093207.post@talk.nabble.com>
	<7465b6170808211027p4b978d9dua57fee622d792e27@mail.gmail.com>
	<168485c00808211050l6f04f83cqf6bb3ae133308a4a@mail.gmail.com>
	<48ADAD96.5000102@sympatico.ca>
Message-ID: <168485c00808211115o1cb4d9cevd87b643ab1867791@mail.gmail.com>

Thanks Stefan, sorry... silly mistake...
William

2008/8/21 Stefan Seefeld <seefeld at sympatico.ca>

> william mari? wrote:
>
>>
>>
>>    > struct AWrap : A, bp::wrapper
>>    > {
>>    >        void foo()
>>    >        {
>>    >                this->get_override( "foo" );
>>    >        }
>>    > };
>>
>>
> You retrieve the override, but you don't call it. See
>
> http://www.boost.org/doc/libs/1_36_0/libs/python/doc/v2/wrapper.html#examples
>
> HTH,
>      Stefan
>
> --
>
>     ...ich hab' noch einen Koffer in Berlin...
>
>
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org <C%2B%2B-sig at python.org>
> http://mail.python.org/mailman/listinfo/c++-sig
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080821/afa95eac/attachment.htm>

From troy at resophonic.com  Thu Aug 21 20:10:43 2008
From: troy at resophonic.com (troy d. straszheim)
Date: Thu, 21 Aug 2008 14:10:43 -0400
Subject: [C++-sig] Calling overidden pure virtual function from C++ side
In-Reply-To: <48ADAD96.5000102@sympatico.ca>
References: <19093207.post@talk.nabble.com>	<7465b6170808211027p4b978d9dua57fee622d792e27@mail.gmail.com>	<168485c00808211050l6f04f83cqf6bb3ae133308a4a@mail.gmail.com>
	<48ADAD96.5000102@sympatico.ca>
Message-ID: <48ADAFA3.2070504@resophonic.com>

Stefan Seefeld wrote:
> william mari? wrote:

>>     > struct AWrap : A, bp::wrapper

and you're low on CRTP

struct AWrap : A, bp::wrapper<A> {


-t


From seefeld at sympatico.ca  Thu Aug 21 20:35:40 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Thu, 21 Aug 2008 14:35:40 -0400
Subject: [C++-sig] Calling overidden pure virtual function from C++ side
In-Reply-To: <48ADAFA3.2070504@resophonic.com>
References: <19093207.post@talk.nabble.com>	<7465b6170808211027p4b978d9dua57fee622d792e27@mail.gmail.com>	<168485c00808211050l6f04f83cqf6bb3ae133308a4a@mail.gmail.com>	<48ADAD96.5000102@sympatico.ca>
	<48ADAFA3.2070504@resophonic.com>
Message-ID: <48ADB57C.1000502@sympatico.ca>

troy d. straszheim wrote:
> Stefan Seefeld wrote:
>> william mari? wrote:
>
>>>     > struct AWrap : A, bp::wrapper
>
> and you're low on CRTP
>
> struct AWrap : A, bp::wrapper<A> {

I suspect that might be an issue with his mailer client, eating angular 
brackets.

Regards,
       Stefan

-- 

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



From ivanov.maxim at gmail.com  Thu Aug 21 21:51:09 2008
From: ivanov.maxim at gmail.com (Max Ivanov)
Date: Thu, 21 Aug 2008 23:51:09 +0400
Subject: [C++-sig] returning pointer to instance and object ownership
Message-ID: <fbfa7b0c0808211251v60d49569ree43bf70653fff6e@mail.gmail.com>

Hi all! Some newbie questions again.

class B {}

class A {
  B* add();
 }

method "add" of class A creates a new instance of class B and stores
reference to it somewhere inside.

I'm trying to expose it to python this way:
".def("add", &A::add, bp::return_value_policy<bp::manage_new_object>())"

It works ok. If I change returned B instance I'm able to see changes
accessing them from A instance - I hope that means that returned
pointer and one stored inside A are related to same instance. Problems
accour when I try to delete objects:
a = A()
b = a.add()
del b
del a
Segmentation fault.

It's the simpliest case I could imagine, since add() method should
always return new object, but still couldn't understand how to do it
right.

ps. I've no control over A and B sources.


From dave at boostpro.com  Fri Aug 22 02:51:14 2008
From: dave at boostpro.com (David Abrahams)
Date: Thu, 21 Aug 2008 16:51:14 -0800
Subject: [C++-sig] Alternate packaging technique for bindings + pure
	python
In-Reply-To: <48A9B063.8070503@resophonic.com> (troy d. straszheim's message
	of "Mon, 18 Aug 2008 13:24:51 -0400")
References: <48A9B063.8070503@resophonic.com>
Message-ID: <87vdxt3m9p.fsf@mcbain.luannocracy.com>


on Mon Aug 18 2008, "troy d. straszheim" <troy-AT-resophonic.com> wrote:

> def load_pybindings(name, path):
>     """
>     Merges python bindings from shared library 'name' into module 'name'.
>     Use when you have a directory structure::
>
>       lib/
>          foo.so
>          foo/
>            __init__.py
>            something.py
>
>     Here, inside ``foo/__init__.py`` call ``load_pybindings(__name__, __path__)``
>
>     this assumes that the first entry in list ``__path__`` is where
>     you want the wrapped classes to merge to.
>
>     """
>     import imp, sys
>     m = imp.load_dynamic(name, __path__[0] + ".so")
>     thismod = sys.modules[name]
>
>     for (k,v) in m.__dict__.items():
>         if not k.startswith("_"):
>             thismod.__dict__[k] = v
>
> Wherein you just use the __name__ and __path__ of the current module being imported to
> locate and load the associated shared library.  Then you copy the symbols into the
> current library.
>
> So the call to it in each __init__.py would be:
>
>   load_pybindings(__name__, __path__)
>
> Nice thing is that everything looks quite homogenous to the user.   It appears to work
> quite well, but I just now came up with this and thought I ask if anybody
> sees any problems.

As Hans said, I'd want to check that it interacts well with everything
else before recommending it, but also I'd want to check that it works
properly on Windows as well as *nix.

If it does continue to work well, I would be very happy if you would
volunteer to rewrite that part of the tutorial :-)

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From dave at boostpro.com  Fri Aug 22 02:56:47 2008
From: dave at boostpro.com (David Abrahams)
Date: Thu, 21 Aug 2008 16:56:47 -0800
Subject: [C++-sig] Accessing python configuration data when building
	extensions using bjam and boost.python
In-Reply-To: <60825b0f0808182207k62f65c5aq1771aafb9d412836@mail.gmail.com>
	(Greg Landrum's message of "Tue, 19 Aug 2008 07:07:22 +0200")
References: <60825b0f0808182207k62f65c5aq1771aafb9d412836@mail.gmail.com>
Message-ID: <87k5e93m0g.fsf@mcbain.luannocracy.com>


on Mon Aug 18 2008, "Greg Landrum" <greg.landrum-AT-gmail.com> wrote:

> So I need to add something like this to my
> Jamroot :
>           <toolset>gcc:<include>/usr/lib/python2.5/site-packages/numpy/core/include
> This is, obviously, dependent on operating system, python version and
> python installation-location. So the user will have to edit the
> Jamroot based on their details.  One of the things I really like about
> boost.build is that this kind of stuff is normally not required.

I think that sort of thing is what site-config.jam is designed for.

> I could solve this problem, I think, there were a variable (?) in
> python.jam defining the location of python's "lib" directory that I
> could access from my Jamroot. Then I could have something like:
>   <include>$(python-lib-dir)/numpy/core/include
> that would work on every machine.
>
> I've spent some time looking through the boost.build documentation and
> python.jam, but I was unable to figure out how to expose a new
> variable that would be accessible from other Jamfiles.
>
> Does anyone have a recipe for doing this?

If you find you need to do something like this, I strongly suggest
asking about it on the boost-build list.

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From dave at boostpro.com  Fri Aug 22 05:28:37 2008
From: dave at boostpro.com (David Abrahams)
Date: Thu, 21 Aug 2008 19:28:37 -0800
Subject: [C++-sig] Static and non-static functions with the same name
In-Reply-To: <BBABA8164420CB43ABEA9DC81FE7A884065D1B46@lopets1exms003.leh.lbcorp.lehman.com>
	(Paul Bilokon's message of "Wed, 20 Aug 2008 09:40:04 +0100")
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B38@lopets1exms003.leh.lbcorp.lehman.com>
	<200808191052.36559.lists_ravi@lavabit.com>
	<BBABA8164420CB43ABEA9DC81FE7A884065D1B46@lopets1exms003.leh.lbcorp.lehman.com>
Message-ID: <877ia920ey.fsf@mcbain.luannocracy.com>


on Wed Aug 20 2008, "Bilokon, Paul" <paul.bilokon-AT-lehman.com> wrote:

> Hi Ravi,
>  
> If my understanding is correct, this method has the same deficiency as
> simply doing this:
>  
>         // OK, explicitly specifying the static garply(string):
>         .def("staticGarply", (string (*)(string)) &Foo::garply)
>         .staticmethod("staticGarply")
>         // OK, explicitly specifying the non-static garply(int, int):
>         .def("garply", (string (Foo::*)(int, int)) &Foo::garply)
>  
>         // NB: The syntax: (*) for static, (Foo::*) for non-static
>  
>         // NB: Since staticmethod does not distinguish between the method
>         // signatures, we must give the static and non-static variants
>         // distinct names in Python
>  
> This works, but the static and non-static variants end up having different
> names on the Python side.

There's not much you can really do about that, since there's no real
overloading in Python; it's just simulated by Boost.Python.

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From roman.yakovenko at gmail.com  Fri Aug 22 11:27:10 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 22 Aug 2008 12:27:10 +0300
Subject: [C++-sig] returning pointer to instance and object ownership
In-Reply-To: <fbfa7b0c0808211251v60d49569ree43bf70653fff6e@mail.gmail.com>
References: <fbfa7b0c0808211251v60d49569ree43bf70653fff6e@mail.gmail.com>
Message-ID: <7465b6170808220227x152919f6r72c5ea3b79e4918c@mail.gmail.com>

On Thu, Aug 21, 2008 at 10:51 PM, Max Ivanov <ivanov.maxim at gmail.com> wrote:
> Hi all! Some newbie questions again.
>
> class B {}
>
> class A {
>  B* add();
>  }
>
> method "add" of class A creates a new instance of class B and stores
> reference to it somewhere inside.

Class A is responcible for the deletion, right?

> I'm trying to expose it to python this way:
> ".def("add", &A::add, bp::return_value_policy<bp::manage_new_object>())"

If the answer to my question is "yes", that you should use
"return_internal_reference" call policy.
http://www.boost.org/doc/libs/1_36_0/libs/python/doc/v2/return_internal_reference.html#return_internal_reference-spec

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


From paul.bilokon at lehman.com  Fri Aug 22 14:21:04 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Fri, 22 Aug 2008 13:21:04 +0100
Subject: [C++-sig] Static and non-static functions with the same name
In-Reply-To: <877ia920ey.fsf@mcbain.luannocracy.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B38@lopets1exms003.leh.lbcorp.lehman.com><200808191052.36559.lists_ravi@lavabit.com><BBABA8164420CB43ABEA9DC81FE7A884065D1B46@lopets1exms003.leh.lbcorp.lehman.com>
	<877ia920ey.fsf@mcbain.luannocracy.com>
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B59@lopets1exms003.leh.lbcorp.lehman.com>

Thanks Dave,

Having different names for the static / non-static methods on the Python
side is not too big a problem.

Many thanks,
Paul

-----Original Message-----
From: c++-sig-bounces at python.org [mailto:c++-sig-bounces at python.org] On
Behalf Of David Abrahams
Sent: Friday, August 22, 2008 4:29 AM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Static and non-static functions with the same
name


on Wed Aug 20 2008, "Bilokon, Paul" <paul.bilokon-AT-lehman.com> wrote:

> Hi Ravi,
>  
> If my understanding is correct, this method has the same deficiency as

> simply doing this:
>  
>         // OK, explicitly specifying the static garply(string):
>         .def("staticGarply", (string (*)(string)) &Foo::garply)
>         .staticmethod("staticGarply")
>         // OK, explicitly specifying the non-static garply(int, int):
>         .def("garply", (string (Foo::*)(int, int)) &Foo::garply)
>  
>         // NB: The syntax: (*) for static, (Foo::*) for non-static
>  
>         // NB: Since staticmethod does not distinguish between the
method
>         // signatures, we must give the static and non-static variants
>         // distinct names in Python
>  
> This works, but the static and non-static variants end up having 
> different names on the Python side.

There's not much you can really do about that, since there's no real
overloading in Python; it's just simulated by Boost.Python.

--
Dave Abrahams
BoostPro Computing
http://www.boostpro.com
_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From ivanov.maxim at gmail.com  Fri Aug 22 23:16:13 2008
From: ivanov.maxim at gmail.com (Max Ivanov)
Date: Sat, 23 Aug 2008 01:16:13 +0400
Subject: [C++-sig] returning pointer to instance and object ownership
In-Reply-To: <7465b6170808220227x152919f6r72c5ea3b79e4918c@mail.gmail.com>
References: <fbfa7b0c0808211251v60d49569ree43bf70653fff6e@mail.gmail.com>
	<7465b6170808220227x152919f6r72c5ea3b79e4918c@mail.gmail.com>
Message-ID: <fbfa7b0c0808221416y4157a36ub38c2170782a786a@mail.gmail.com>

> If the answer to my question is "yes", that you should use
> "return_internal_reference" call policy.
> http://www.boost.org/doc/libs/1_36_0/libs/python/doc/v2/return_internal_reference.html#return_internal_reference-spec
I see now, that Boost::python is all about wrappers :)


From ope-devel at gmx.de  Sat Aug 23 19:01:49 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Sat, 23 Aug 2008 19:01:49 +0200
Subject: [C++-sig] [pybindgen] parameter defaults
Message-ID: <48B0427D.8070706@gmx.de>

Hello,

how can I emulate/code a py function that excepts a default type, e.g.

stdin.read(size=None)

in C/pybindgen which can called as stdin.read() and stdin.read(100).
This has to configure in the pybindgen's python script, but I did not
found this.

stdin_py.add_method('read', retval('PyObject*',
caller_owns_return=False), [param('int', 'size=None')])

doesn't generate code as expected. BTW, retval should be a python string
( Py_BuildValue((char *) "s", retval); ). How to get this?

Thanks,
Olaf



From ope-devel at gmx.de  Sat Aug 23 19:34:17 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Sat, 23 Aug 2008 19:34:17 +0200
Subject: [C++-sig] [pybindgen] parameter defaults
In-Reply-To: <48B0427D.8070706@gmx.de>
References: <48B0427D.8070706@gmx.de>
Message-ID: <48B04A19.1030006@gmx.de>


> stdin_py.add_method('read', retval('PyObject*',
> caller_owns_return=False), [param('int', 'size=None')])

therefore the C code must like shown something like:


PyObject *
PyRedirectIn_read(PyRedirectIn *self, PyObject *args, PyObject *kwargs)
{
    int size = std::numeric_limits<int>::max();
    const char *keywords[] = {"size", NULL};

    if (!PyArg_ParseTupleAndKeywords(args, kwargs, (char *) "|i", (char
**) keywords, &size)) {
        return NULL;
    }

    QString retval = self->obj->read_fn(size);
    PyObject *py_retval = Py_BuildValue((char *) "s", static_cast<const
char*>(retval.toUtf8()));

    return py_retval;
}



From gjcarneiro at gmail.com  Sat Aug 23 23:10:39 2008
From: gjcarneiro at gmail.com (Gustavo Carneiro)
Date: Sat, 23 Aug 2008 22:10:39 +0100
Subject: [C++-sig] [pybindgen] parameter defaults
In-Reply-To: <48B04A19.1030006@gmx.de>
References: <48B0427D.8070706@gmx.de> <48B04A19.1030006@gmx.de>
Message-ID: <a467ca4f0808231410q5a4990adt682a1a6f6952d080@mail.gmail.com>

2008/8/23 Olaf Peter <ope-devel at gmx.de>

>
> > stdin_py.add_method('read', retval('PyObject*',
> > caller_owns_return=False), [param('int', 'size=None')])
>
> therefore the C code must like shown something like:
>
>
> PyObject *
> PyRedirectIn_read(PyRedirectIn *self, PyObject *args, PyObject *kwargs)
> {
>    int size = std::numeric_limits<int>::max();
>    const char *keywords[] = {"size", NULL};
>
>    if (!PyArg_ParseTupleAndKeywords(args, kwargs, (char *) "|i", (char
> **) keywords, &size)) {
>        return NULL;
>    }
>
>    QString retval = self->obj->read_fn(size);
>    PyObject *py_retval = Py_BuildValue((char *) "s", static_cast<const
> char*>(retval.toUtf8()));
>
>    return py_retval;
> }
>

To support the default value, use: param('int', 'size',
default_value='std::numeric_limits<int>::max()').

To handle QString return value you'd have to write your own return type
handler for QString.  But it doesn't sound complicated.  You just have to
declare a class similar to StdStringReturn in
pybindgen/typehandlers/stringtype.py.  Untested code:

class QStringReturnValue(ReturnValue):

    CTYPES = ['QString']

    def get_c_error_return(self):
        return "return QString();"

    def convert_python_to_c(self, wrapper):
              raise NotImplementedError # TODO (needed only for virtual
methods where C calls Python code)

    def convert_c_to_python(self, wrapper):
        wrapper.build_params.add_parameter("s", ['%s.toUtf8()' % self.value
],
                                           prepend=True)

After the class is declared, you can use retval('QString') as return value.

Hope this helps.

-- 
Gustavo J. A. M. Carneiro
INESC Porto, Telecommunications and Multimedia Unit
"The universe is always one step beyond logic." -- Frank Herbert
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080823/e45f48d6/attachment.htm>

From gupta.sandeep at gmail.com  Sun Aug 24 20:58:06 2008
From: gupta.sandeep at gmail.com (Sandeep Gupta)
Date: Sun, 24 Aug 2008 11:58:06 -0700
Subject: [C++-sig] [BGL] python bindings for intel compiler
Message-ID: <98a412be0808241158t1794ac38y6ac854cc401b06cb@mail.gmail.com>

Hi,
 I haven't been able to create correct python bindings for BGL for intel
compiler. Will greatly appreciate any help as it has been more than four
months since I first encountered the problem.

The python bindings in the tutorial and examples work fine. The bgl_python
code from the sandbox compiles smoothly. I believe I have set the PYTHONPATH
and LD_LIBRARY_PATH correctly. But  the import of boost.graph fails with
following error:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "...../bgl-python/lib/python/boost/graph/__init__.py", line 16, in
<module>
    from boost.graph._graph import *
ImportError: "...../bgl-python/lib/python/boost/graph/_graph.so: undefined
symbol: _ZN5boost5graph6python11basic_graphINS_11undirectedSEED1Ev"

Thanks,
Sandeep
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080824/975ba1e6/attachment.htm>

From srmandrake at yahoo.com  Mon Aug 25 04:09:06 2008
From: srmandrake at yahoo.com (SrMandrake ElMago)
Date: Sun, 24 Aug 2008 19:09:06 -0700 (PDT)
Subject: [C++-sig] Py_Finalize() crashes when PyImport_Import() fails
Message-ID: <637465.86127.qm@web56204.mail.re3.yahoo.com>

Hello,
I am triying to learn some Python to use it with C++. After some time I managed to do what I was attempting, but I just realized a couple of days ago that there is a problem when I call Py_Finalize() just before my program exits.
The problem is as follows:
First of all, I am using VC Express 2005, on Windows XP.
My program is able to do everything perfectly as long as the script to load is found.
When the script is not found, the program keeps going perfectly, but crashes when calling Py_Finalize() raising an exception and poping the following message.
Exception exceptions.ImportError: 'No module named test_11' in 'garbage collection' ignored
Fatal Python error: unexpected exception during garbage collection
Now, I know the easiest solution to the problem is to put the file where it is supposed to be :-), but what I'm actually looking for is a way to prevent the error from?crashing the program?because, at the end, I would like the user to type?the name of the file (script)?to load (and there?may be a typo or the entered file name may not?exist )?Thanks in advance for any help you can give me... 
Rod-
Here's the code...
#include
?
int main()
{
Py_Initialize ();
//"test_1" is the actual name of the file. I changed it to "test_11" on purpose to cause the crash...
const char* s = "test_11";
PyObject* pName = PyString_FromString ( s );
PyObject* pModule = PyImport_Import ( pName );
if ( !pModule )
{??? 
??? std::cout << "Unable to open script " << s << std::endl;
??? getchar ();
}
Py_XDECREF( pModule );
Py_XDECREF( pName );
Py_Finalize ();
getchar ();
return0;
} <Python.h>


      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080824/3e837939/attachment.htm>

From seefeld at sympatico.ca  Mon Aug 25 04:31:42 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Sun, 24 Aug 2008 22:31:42 -0400
Subject: [C++-sig] Py_Finalize() crashes when PyImport_Import() fails
In-Reply-To: <637465.86127.qm@web56204.mail.re3.yahoo.com>
References: <637465.86127.qm@web56204.mail.re3.yahoo.com>
Message-ID: <48B2198E.1030306@sympatico.ca>

SrMandrake ElMago wrote:
>
> Exception exceptions.ImportError: 'No module named test_11' in 
> 'garbage collection' ignored
> Fatal Python error: unexpected exception during garbage collection
The Python runtime may expect you to call handle Python exceptions 
explicitely, prior to finalizing.

> if ( !pModule )
>
> {
>

If you got 0 here, Python has set an exception. Before moving on you 
should handle it, for example by explicitely clearing it:

PyErr_Clear();

>     std::cout << "Unable to open script " << s << std::endl;
>
>     getchar ();
>
> }
>
>  
>
> Py_XDECREF( pModule );
>
> Py_XDECREF( pName );
>
>  
>
> Py_Finalize ();
>

With the above addition, the Python runtime should finalize fine.

HTH,
       Stefan


-- 

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



From srmandrake at yahoo.com  Mon Aug 25 05:29:04 2008
From: srmandrake at yahoo.com (SrMandrake ElMago)
Date: Sun, 24 Aug 2008 20:29:04 -0700 (PDT)
Subject: [C++-sig] Py_Finalize() crashes when PyImport_Import() fails
Message-ID: <373824.3843.qm@web56213.mail.re3.yahoo.com>

That was it! Thanks Stefan...
If ever in Costa Rica...email me. I'll buy u a beer! :-D
Pura Vida (Y)



----- Original Message ----
From: Stefan Seefeld <seefeld at sympatico.ca>
To: Development of Python/C++ integration <c++-sig at python.org>
Sent: Sunday, August 24, 2008 8:31:42 PM
Subject: Re: [C++-sig] Py_Finalize() crashes when PyImport_Import() fails

SrMandrake ElMago wrote:
>
> Exception exceptions.ImportError: 'No module named test_11' in 
> 'garbage collection' ignored
> Fatal Python error: unexpected exception during garbage collection
The Python runtime may expect you to call handle Python exceptions 
explicitely, prior to finalizing.

> if ( !pModule )
>
> {
>

If you got 0 here, Python has set an exception. Before moving on you 
should handle it, for example by explicitely clearing it:

PyErr_Clear();

>? ? std::cout << "Unable to open script " << s << std::endl;
>
>? ? getchar ();
>
> }
>
>? 
>
> Py_XDECREF( pModule );
>
> Py_XDECREF( pName );
>
>? 
>
> Py_Finalize ();
>

With the above addition, the Python runtime should finalize fine.

HTH,
? ? ? Stefan


-- 

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

_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig



      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080824/3cb5b9af/attachment.htm>

From rwgk at yahoo.com  Mon Aug 25 11:48:54 2008
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Mon, 25 Aug 2008 02:48:54 -0700 (PDT)
Subject: [C++-sig] Container Conversions: Wrapping Functions Expecting
	Non-const References
Message-ID: <487287.80500.qm@web31403.mail.mud.yahoo.com>

Hi Paul,




> Here I have difficulties:
> 
> Boost.Python.ArgumentError: Python argument types in
>     ContainersExample2.doubleVectorOfInts(ContainersExample2, list)
> did not match C++ signature:
>     doubleVectorOfInts(class ContainersExample2 {lvalue}, class
> std::vector<int,class std::allocator<int> > {lvalue})
>
> What's the recommended technique(s) for this situation?

Boost.Python doesn't have "on-the-fly converters with write-back" (long story).

You have to wrap std::vector<int> "properly" if you want to pass non-const references to C++.
The vector-indexing suite is an option that should work well. Or you can follow this more
minimalistic example:


http://cctbx.svn.sourceforge.net/viewvc/cctbx/trunk/scitbx/stl/vector_wrapper.h?view=markup

Ignore the ref_from_array code near the bottom.

Ralf


From paul.bilokon at lehman.com  Wed Aug 27 13:01:36 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Wed, 27 Aug 2008 12:01:36 +0100
Subject: [C++-sig] Container Conversions: Wrapping Functions
	ExpectingNon-const References
In-Reply-To: <487287.80500.qm@web31403.mail.mud.yahoo.com>
References: <487287.80500.qm@web31403.mail.mud.yahoo.com>
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B77@lopets1exms003.leh.lbcorp.lehman.com>

Hi Ralf,

Many thanks for your explanation. I see there are two options:
on-the-fly converters without write-back and wrappers (a la the vector
indexing suite, or perhaps the set wrapper).

In essence, the former can be used for passing values, the latter for
passing references.

I have also tried adding conversions for a few simple types (e.g.
std::pairs and boost::tuples) to see how they work. Looks like they do
the trick.

Many thanks,
Paul

-----Original Message-----
From: c++-sig-bounces+paul.bilokon=lehman.com at python.org
[mailto:c++-sig-bounces+paul.bilokon=lehman.com at python.org] On Behalf Of
Ralf W. Grosse-Kunstleve
Sent: Monday, August 25, 2008 10:49 AM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Container Conversions: Wrapping Functions
ExpectingNon-const References

Hi Paul,




> Here I have difficulties:
> 
> Boost.Python.ArgumentError: Python argument types in
>     ContainersExample2.doubleVectorOfInts(ContainersExample2, list) 
> did not match C++ signature:
>     doubleVectorOfInts(class ContainersExample2 {lvalue}, class 
> std::vector<int,class std::allocator<int> > {lvalue})
>
> What's the recommended technique(s) for this situation?

Boost.Python doesn't have "on-the-fly converters with write-back" (long
story).

You have to wrap std::vector<int> "properly" if you want to pass
non-const references to C++.
The vector-indexing suite is an option that should work well. Or you can
follow this more minimalistic example:


http://cctbx.svn.sourceforge.net/viewvc/cctbx/trunk/scitbx/stl/vector_wr
apper.h?view=markup

Ignore the ref_from_array code near the bottom.

Ralf
_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: StdPairConversions.h
Type: application/octet-stream
Size: 3270 bytes
Desc: StdPairConversions.h
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080827/2f87e02b/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: BoostTupleConversions.h
Type: application/octet-stream
Size: 22972 bytes
Desc: BoostTupleConversions.h
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080827/2f87e02b/attachment-0001.obj>

From ope-devel at gmx.de  Wed Aug 27 21:22:20 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Wed, 27 Aug 2008 21:22:20 +0200
Subject: [C++-sig] [boost.python] PyErr_Fetch and object
Message-ID: <48B5A96C.3040403@gmx.de>

Hi,

following http://www.python.org/doc/api/exceptionHandling.html,
PyErr_Fetch returns objects and I own these (if!= NULL).

I wrote:

bpl::object exception, value, tb;
PyErr_Fetch(&exception.ptr(), &value.ptr(), &tb.ptr());

which doesn't compile:

error: lvalue required as unary '&' operand

is there a way for this using bpl:object? Maybe I have to wrap this
after calling PyErr_Fetch?

Thanks,
Olaf


From seefeld at sympatico.ca  Wed Aug 27 21:45:08 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Wed, 27 Aug 2008 15:45:08 -0400
Subject: [C++-sig] [boost.python] PyErr_Fetch and object
In-Reply-To: <48B5A96C.3040403@gmx.de>
References: <48B5A96C.3040403@gmx.de>
Message-ID: <48B5AEC4.7020409@sympatico.ca>

Olaf Peter wrote:
> Hi,
>
> following http://www.python.org/doc/api/exceptionHandling.html,
> PyErr_Fetch returns objects and I own these (if!= NULL).
>
> I wrote:
>
> bpl::object exception, value, tb;
> PyErr_Fetch(&exception.ptr(), &value.ptr(), &tb.ptr());
>   

I wouldn't expect this to work: You can't assign to the return value of 
bpl::object::ptr().

> is there a way for this using bpl:object? 

Not at this time.

> Maybe I have to wrap this
> after calling PyErr_Fetch?
>   
Yes. Something like this should work:

namespace boost
{
namespace python
{
void fetch_error(object &type, object &value, object &traceback)
{
  PyObject *t, *v, *b;
  PyErr_Fetch(&t, &v, &b);
  type = detail::new_reference(t);
  value = v ? detail::new_reference(v) : object();
  traceback = b ? detail::new_reference(b) : object();
}
}
}

(There are more functions like this that could be added. Some day I'll 
submit a new patch...

HTH,
       Stefan


-- 

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



From ope-devel at gmx.de  Thu Aug 28 07:25:23 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Thu, 28 Aug 2008 07:25:23 +0200
Subject: [C++-sig] [boost.python] PyErr_Fetch and object
In-Reply-To: <48B5AEC4.7020409@sympatico.ca>
References: <48B5A96C.3040403@gmx.de> <48B5AEC4.7020409@sympatico.ca>
Message-ID: <48B636C3.5030505@gmx.de>

>> Maybe I have to wrap this
>> after calling PyErr_Fetch?
>>   
> Yes. Something like this should work:
> 
> namespace boost
> {
> namespace python
> {
> void fetch_error(object &type, object &value, object &traceback)
> {
>  PyObject *t, *v, *b;
>  PyErr_Fetch(&t, &v, &b);
>  type = detail::new_reference(t);
>  value = v ? detail::new_reference(v) : object();
>  traceback = b ? detail::new_reference(b) : object();
> }
> }
> }
> 
> (There are more functions like this that could be added. Some day I'll 
> submit a new patch...

Thanks, would be great. Anyway, In the source of boost.python I saw a 
lot of used detail::.. stuff. The raw:pyobject header says this are to 
hide to the user. What are the correspondent user functions for
new_reference, borrowed_reference and new_non_null_reference which are 
more in common sense of the python C/API documentation.

Thanks,
Olaf




From amohr at pixar.com  Thu Aug 28 19:03:12 2008
From: amohr at pixar.com (Alex Mohr)
Date: Thu, 28 Aug 2008 10:03:12 -0700
Subject: [C++-sig] [boost.python] PyErr_Fetch and object
In-Reply-To: <48B636C3.5030505@gmx.de>
References: <48B5A96C.3040403@gmx.de> <48B5AEC4.7020409@sympatico.ca>
	<48B636C3.5030505@gmx.de>
Message-ID: <48B6DA50.20005@pixar.com>

>> Yes. Something like this should work:
>>
>> void fetch_error(object &type, object &value, object &traceback)
>> {
>>  PyObject *t, *v, *b;
>>  PyErr_Fetch(&t, &v, &b);
>>  type = detail::new_reference(t);
>>  value = v ? detail::new_reference(v) : object();
>>  traceback = b ? detail::new_reference(b) : object();
>> }
>>
>> (There are more functions like this that could be added. Some day I'll 
>> submit a new patch...
> 
> Thanks, would be great. Anyway, In the source of boost.python I saw a 
> lot of used detail::.. stuff. The raw:pyobject header says this are to 
> hide to the user. What are the correspondent user functions for
> new_reference, borrowed_reference and new_non_null_reference which are 
> more in common sense of the python C/API documentation.

I tried to rewrite the above using more public-api.  Also I think the 
original function does not check 't' for NULL, and so may only be safely 
called if there is a python exception.  I tried to change it so this 
code will set all three objects to None if there is no exception.

I haven't tried compiling or testing this.

void fetch_error(object &type, object &value, object &traceback) {
     PyObject *t, *v, *b;
     PyErr_Fetch(&t, &v, &b);
     type = object(handle<>(allow_null(t)));
     value = object(handle<>(allow_null(v)));
     traceback = object(handle<>(allow_null(b));
}

Alex



From seefeld at sympatico.ca  Thu Aug 28 19:14:15 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Thu, 28 Aug 2008 13:14:15 -0400
Subject: [C++-sig] [boost.python] PyErr_Fetch and object
In-Reply-To: <48B6DA50.20005@pixar.com>
References: <48B5A96C.3040403@gmx.de>
	<48B5AEC4.7020409@sympatico.ca>	<48B636C3.5030505@gmx.de>
	<48B6DA50.20005@pixar.com>
Message-ID: <48B6DCE7.5000807@sympatico.ca>

Alex Mohr wrote:
>>
>> Thanks, would be great. Anyway, In the source of boost.python I saw a 
>> lot of used detail::.. stuff. The raw:pyobject header says this are 
>> to hide to the user. What are the correspondent user functions for
>> new_reference, borrowed_reference and new_non_null_reference which 
>> are more in common sense of the python C/API documentation.
>
> I tried to rewrite the above using more public-api.  Also I think the 
> original function does not check 't' for NULL, and so may only be 
> safely called if there is a python exception.  I tried to change it so 
> this code will set all three objects to None if there is no exception.
>
> I haven't tried compiling or testing this.
>
> void fetch_error(object &type, object &value, object &traceback) {
>     PyObject *t, *v, *b;
>     PyErr_Fetch(&t, &v, &b);
>     type = object(handle<>(allow_null(t)));
>     value = object(handle<>(allow_null(v)));
>     traceback = object(handle<>(allow_null(b));
> }

Thanks ! Actually, the lack of a check for 'type' was intentional: Even 
if an error is set, 'value' and 'traceback' may be set to NULL.
'type' can only be NULL if no error is set, and thus has to be handled 
special. (For example, we could throw an exception in that case, as 
calling fetch_error if no error is set is an error (or may be, if we 
make it so).

Of course, with that then come other function wrappers for 'restore', 
'clear', etc.

I'll think about a patch, if no-one else beats me to it <wink/>

Regards,
       Stefan

-- 

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



From amohr at pixar.com  Thu Aug 28 19:47:00 2008
From: amohr at pixar.com (Alex Mohr)
Date: Thu, 28 Aug 2008 10:47:00 -0700
Subject: [C++-sig] [boost.python] PyErr_Fetch and object
In-Reply-To: <48B6DCE7.5000807@sympatico.ca>
References: <48B5A96C.3040403@gmx.de> <48B5AEC4.7020409@sympatico.ca>
	<48B636C3.5030505@gmx.de> <48B6DA50.20005@pixar.com>
	<48B6DCE7.5000807@sympatico.ca>
Message-ID: <48B6E494.6060908@pixar.com>

> Thanks ! Actually, the lack of a check for 'type' was intentional: Even 
> if an error is set, 'value' and 'traceback' may be set to NULL.
> 'type' can only be NULL if no error is set, and thus has to be handled 
> special. (For example, we could throw an exception in that case, as 
> calling fetch_error if no error is set is an error (or may be, if we 
> make it so).

Hmm -- well, my personal preference is for it to mimic the python c api, 
so if there is no error set, then all three are set to None, if there is 
an error set, then type is not None, and value and traceback may or may 
not be None.  That's how I tried to write it.

> Of course, with that then come other function wrappers for 'restore', 
> 'clear', etc.

Indeed.  It would be nice to do a project that tried to wrap up all (or 
as much as possible) of the python C api.  That would benefit and appeal 
to a wide audience, I think.  Even including folks who don't use 
boost.python to generate bindings.

> I'll think about a patch, if no-one else beats me to it <wink/>

:)

Alex



From azan00 at yahoo.com  Thu Aug 28 23:32:53 2008
From: azan00 at yahoo.com (Brian Dawn)
Date: Thu, 28 Aug 2008 14:32:53 -0700 (PDT)
Subject: [C++-sig] [boost.python] Assertion Fails with return_value_policy
Message-ID: <580094.27248.qm@web54409.mail.yahoo.com>

Hi, I currently have a function that returns an object pointer. I wish to expose this function to python. However when I set a return_value_policy to this member function I get the following assertion failure.

Assertion failed!
Program: ...
File: libs\python\src\converter\registry.cpp
Line: 207
Expression: slot->m_to_python == 0

I have been searching for a fix for quite a while. I have also tried many different return_value_policies. Below is the line causing the assertion failure.

namespace["pyEmitter"] = boost::python::class_<pyParticle>("pyEmitter")
            .def("grab", &pyEmitter::getParticle, boost::python::return_value_policy<boost::python::reference_existing_object>());

And here is my function.

pyParticle& pyEmitter::getParticle()
{
    return *particleList[0];
}

I am using MS Visual Express 2008 if it matters, I have seen many examples and my code appears to be correct. Any help would be appreciated.



      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080828/dce4264b/attachment.htm>

From seefeld at sympatico.ca  Thu Aug 28 23:44:19 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Thu, 28 Aug 2008 17:44:19 -0400
Subject: [C++-sig] [boost.python] Assertion Fails with
	return_value_policy
In-Reply-To: <580094.27248.qm@web54409.mail.yahoo.com>
References: <580094.27248.qm@web54409.mail.yahoo.com>
Message-ID: <48B71C33.4060308@sympatico.ca>

Brian Dawn wrote:
> Hi, I currently have a function that returns an object pointer. I wish 
> to expose this function to python. However when I set a 
> return_value_policy to this member function I get the following 
> assertion failure.

This is a bug in boost.python. (Even if you do something wrong, the 
library shouldn't abort, but raise an exception, in that case.)
Please submit an issue for this.

Thanks,
       Stefan


-- 

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



From ope-devel at gmx.de  Fri Aug 29 07:38:34 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Fri, 29 Aug 2008 07:38:34 +0200
Subject: [C++-sig] [boost.python] PyErr_Fetch and object
In-Reply-To: <48B6E494.6060908@pixar.com>
References: <48B5A96C.3040403@gmx.de>
	<48B5AEC4.7020409@sympatico.ca>	<48B636C3.5030505@gmx.de>
	<48B6DA50.20005@pixar.com>	<48B6DCE7.5000807@sympatico.ca>
	<48B6E494.6060908@pixar.com>
Message-ID: <48B78B5A.3090105@gmx.de>

Alex Mohr schrieb:
>> Thanks ! Actually, the lack of a check for 'type' was intentional: 
>> Even if an error is set, 'value' and 'traceback' may be set to NULL.
>> 'type' can only be NULL if no error is set, and thus has to be handled 
>> special. (For example, we could throw an exception in that case, as 
>> calling fetch_error if no error is set is an error (or may be, if we 
>> make it so).
> 
> Hmm -- well, my personal preference is for it to mimic the python c api, 
> so if there is no error set, then all three are set to None, if there is 
> an error set, then type is not None, and value and traceback may or may 
> not be None.  That's how I tried to write it.

which is intuitive to C/API. To check the case mentioned by Stefan one 
has to test on PyErr_Occured() and than decide the individual error 
handling.

>> Of course, with that then come other function wrappers for 'restore', 
>> 'clear', etc.
> 
> Indeed.  It would be nice to do a project that tried to wrap up all (or 
> as much as possible) of the python C api.  That would benefit and appeal 
> to a wide audience, I think.  Even including folks who don't use 
> boost.python to generate bindings.
> 
>> I'll think about a patch, if no-one else beats me to it <wink/>

I will do it too ;-) At this time I wrote my own extension to 
boost.python like the above.

Anyway, what are the appropriate user functions for
detail { new_reference, borrowed_reference and new_non_null_reference }?
Why does detail::borrow owns a reference and do not decide the user what 
to do?

Thanks,
Olaf



From dave at boostpro.com  Fri Aug 29 12:18:11 2008
From: dave at boostpro.com (David Abrahams)
Date: Fri, 29 Aug 2008 06:18:11 -0400
Subject: [C++-sig] [boost.python] Assertion Fails with
	return_value_policy
In-Reply-To: <580094.27248.qm@web54409.mail.yahoo.com> (Brian Dawn's message
	of "Thu, 28 Aug 2008 14:32:53 -0700 (PDT)")
References: <580094.27248.qm@web54409.mail.yahoo.com>
Message-ID: <877ia0ruos.fsf@mcbain.luannocracy.com>


on Thu Aug 28 2008, Brian Dawn <azan00-AT-yahoo.com> wrote:

> Hi, I currently have a function that returns an object pointer. I wish to expose
> this function to python. However when I set a return_value_policy to this member
> function I get the following assertion failure.
>
> Assertion failed!
> Program: ...
> File: libs\python\src\converter\registry.cpp
> Line: 207
> Expression: slot->m_to_python == 0

It's hard for me to see how that could be the case.  That assertion
means you've tried to register to-python conversions for the same type
twice.  But the code you show below doesn't look like it should be
registering any type converters, so it seems like the assertion probably
comes from elsewhere (possibly a bug in your code).

If you continue past that assertion (or skip over it in the debugger)
you should observe the Python interpreter giving you the name of the
offending C++ type, but given what I wrote above, I don't expect that to
illuminate much.  More likely it would be very useful to reduce your
code to a minimal, complete example that reproduces the problem.  If you
don't discover a problem in your own code along the way, it will give us
what we need to fix Boost.Python.

I do agree with Stefan that there is a bug in Boost.Python here.  The
problem is that I never decided whether the condition detected by the
assertion should be considered a precondition violation, and documented
it accordingly.  The response to the condition is suboptimal as a
result.

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From paul.bilokon at lehman.com  Fri Aug 29 16:01:38 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Fri, 29 Aug 2008 15:01:38 +0100
Subject: [C++-sig] Attempt to Return Dangling Reference Error -- What's the
	solution?
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1B8E@lopets1exms003.leh.lbcorp.lehman.com>

Hi,

I'm looking at this:

http://www.boost.org/doc/libs/1_36_0/libs/python/doc/v2/faq.html#danglin
g

('I'm getting the "attempt to return dangling reference" error. What am
I doing wrong?')

I understand the statement of the problem. However, suppose I need to
hold on to the new Python object and the C++ object contained therein.
How do I achieve that?

Let me give you the specific context.


	void takesAnimalRawPtr(Animal * animalRawPtr)
	{
	    cout << "This is " << animalRawPtr->whatIsIt() << endl;
	    Animal * clone = animalRawPtr->clone();
	    cout << "Now this is " << clone->whatIsIt() << endl;
	    delete clone;
	}


This works very well for a C++ child of Animal, Cat, which overrides the
virtual method clone.

This does not work for a Python child of Animal:


	class Dog(smartptr_step.Animal):
	    def __init__(self):
	        super(Dog, self).__init__()

	    def whatIsIt(self):
	        return "dog"

	    def clone(self):
	        dogClone = Dog()
	        return dogClone


Here I'm getting the aforementioned dangling reference.

Can you please help?

Many thanks,
Paul
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

This message is intended only for the personal and confidential use of the designated recipient(s) named above.  If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited.  This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers.  Email transmission cannot be guaranteed to be secure or error-free.  Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such.  All information is subject to change without notice.



From dave at boostpro.com  Fri Aug 29 16:52:43 2008
From: dave at boostpro.com (David Abrahams)
Date: Fri, 29 Aug 2008 10:52:43 -0400
Subject: [C++-sig] Attempt to Return Dangling Reference Error -- What's
	the solution?
In-Reply-To: <BBABA8164420CB43ABEA9DC81FE7A884065D1B8E@lopets1exms003.leh.lbcorp.lehman.com>
	(Paul Bilokon's message of "Fri, 29 Aug 2008 15:01:38 +0100")
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B8E@lopets1exms003.leh.lbcorp.lehman.com>
Message-ID: <87hc93ho04.fsf@mcbain.luannocracy.com>


on Fri Aug 29 2008, "Bilokon, Paul" <paul.bilokon-AT-lehman.com> wrote:

> Hi,
>
> I'm looking at this:
>
> http://www.boost.org/doc/libs/1_36_0/libs/python/doc/v2/faq.html#danglin
> g
>
> ('I'm getting the "attempt to return dangling reference" error. What am
> I doing wrong?')
>
> I understand the statement of the problem. However, suppose I need to
> hold on to the new Python object and the C++ object contained therein.
> How do I achieve that?
>
> Let me give you the specific context.
>
>
> 	void takesAnimalRawPtr(Animal * animalRawPtr)
> 	{
> 	    cout << "This is " << animalRawPtr->whatIsIt() << endl;
> 	    Animal * clone = animalRawPtr->clone();
> 	    cout << "Now this is " << clone->whatIsIt() << endl;
> 	    delete clone;
> 	}

I suggest you make your clone() functions return a
boost::shared_ptr<Animal>.  That would make all your problems disappear.

HTH,

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com


From amohr at pixar.com  Fri Aug 29 18:39:23 2008
From: amohr at pixar.com (Alex Mohr)
Date: Fri, 29 Aug 2008 09:39:23 -0700
Subject: [C++-sig] [boost.python] PyErr_Fetch and object
In-Reply-To: <48B78B5A.3090105@gmx.de>
References: <48B5A96C.3040403@gmx.de> <48B5AEC4.7020409@sympatico.ca>
	<48B636C3.5030505@gmx.de> <48B6DA50.20005@pixar.com>
	<48B6DCE7.5000807@sympatico.ca> <48B6E494.6060908@pixar.com>
	<48B78B5A.3090105@gmx.de>
Message-ID: <48B8263B.6010605@pixar.com>

> Anyway, what are the appropriate user functions for
> detail { new_reference, borrowed_reference and new_non_null_reference }?
> Why does detail::borrow owns a reference and do not decide the user what 
> to do?

I'd say just use handle<> if you can.  It's a lightweight wrapper around 
PyObject * that manages the refcount for you.  When constructing a 
handle<>, it by default assumes that the PyObject * you're passing to it 
is not null, and a "new reference" (meaning you already own a reference 
to it).  But you can tell it otherwise by using allow_null() and 
borrowed(), either alone or in combination.  Using borrowed() tells 
boost.python that the PyObject * is a "borrowed reference" meaning that 
you do not own a reference to it.  In that case, handle<> will bump the 
reference count.

Check out the docs:

http://www.boost.org/doc/libs/1_36_0/libs/python/doc/v2/handle.html

Alex


From azan00 at yahoo.com  Fri Aug 29 18:59:40 2008
From: azan00 at yahoo.com (Brian Dawn)
Date: Fri, 29 Aug 2008 09:59:40 -0700 (PDT)
Subject: [C++-sig] [boost.python] Assertion Fails with
	return_value_policy
Message-ID: <774156.21199.qm@web54403.mail.yahoo.com>

Thanks everyone for their replies. David you are correct, after ignoring the assertion failure it gave me the information I needed to find the flaw within my code. It was a typo, I had copied and pasted a class exposure line but forgot to actually change the name of the class (such a stupid mistake, but eh better than a major flaw).

The assertion failure can be recreated (on my system at least) with the following two lines of code.
global["class1"] = class_<someClass>("class1");
global["class2"] = class_<someClass>("class2");



----- Original Message ----
From: David Abrahams <dave at boostpro.com>
To: Development of Python/C++ integration <c++-sig at python.org>
Sent: Friday, August 29, 2008 5:18:11 AM
Subject: Re: [C++-sig] [boost.python] Assertion Fails with return_value_policy


on Thu Aug 28 2008, Brian Dawn <azan00-AT-yahoo.com> wrote:

> Hi, I currently have a function that returns an object pointer. I wish to expose
> this function to python. However when I set a return_value_policy to this member
> function I get the following assertion failure.
>
> Assertion failed!
> Program: ...
> File: libs\python\src\converter\registry.cpp
> Line: 207
> Expression: slot->m_to_python == 0

It's hard for me to see how that could be the case.  That assertion
means you've tried to register to-python conversions for the same type
twice.  But the code you show below doesn't look like it should be
registering any type converters, so it seems like the assertion probably
comes from elsewhere (possibly a bug in your code).

If you continue past that assertion (or skip over it in the debugger)
you should observe the Python interpreter giving you the name of the
offending C++ type, but given what I wrote above, I don't expect that to
illuminate much.  More likely it would be very useful to reduce your
code to a minimal, complete example that reproduces the problem.  If you
don't discover a problem in your own code along the way, it will give us
what we need to fix Boost.Python.

I do agree with Stefan that there is a bug in Boost.Python here.  The
problem is that I never decided whether the condition detected by the
assertion should be considered a precondition violation, and documented
it accordingly.  The response to the condition is suboptimal as a
result.

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com
_______________________________________________
C++-sig mailing list
C++-sig at python.org
http://mail.python.org/mailman/listinfo/c++-sig



      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080829/f49f42d7/attachment.htm>

From william.marie at gmail.com  Sat Aug 30 16:35:52 2008
From: william.marie at gmail.com (=?UTF-8?Q?William_Mari=C3=A9?=)
Date: Sat, 30 Aug 2008 07:35:52 -0700 (PDT)
Subject: [C++-sig]  How to link Python statically to boost python ?
Message-ID: <19233904.post@talk.nabble.com>


Hi,

I would like to link statically python25.lib to boost-python.lib and then
link statically boost-python.lib to my script.lib. i can't find a way to do
that.

Could someone tell me if it's possible and if yes how to do it ?
Thanks

William
-- 
View this message in context: http://www.nabble.com/How-to-link-Python-statically-to-boost-python---tp19233904p19233904.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From jlc at kergis.com  Sun Aug 31 15:35:37 2008
From: jlc at kergis.com (jeanluc.c)
Date: Sun, 31 Aug 2008 06:35:37 -0700 (PDT)
Subject: [C++-sig] how to link project with a specified libpython in static?
 (using boost.Python)
Message-ID: <19242281.post@talk.nabble.com>


Hi,

We would need help to solve a link problem on our project.
We are developing a project using boost, boost.Python, and bjam, but we
would like to embedded our own python library in static form.

I'm currently developing on darwin (macosx 10.5).

Python sources have been integrated in our project
(/.../project_path/libs/Python), and we building the static library thanks
to the Makefile (resulting in:
/.../project_path/libs/Python/libpython2.5.a).

Our project is using bjam, and in the Jamroot we define:
...
using python  : # 2.5
       : /.../project_path/libs/Python/python # cmd-or-prefix
       : [ getPythonIncludePaths ] # includes (return:
/.../project_path/libs/Python/Include ...)        
       : [ getPythonLibPath ] # libraries (return:
/.../project_path/libs/Python)
       : [ getPythonConditions ] #conditions (return: <variant>debug
<python-debugging>on <link>static)
       ;
...

The project Jamfile is:
...
exe testscript : [ glob *.cpp ] 
      /otherlib//otherlib
      /boost/python//boost_python/<link>static
      /python//python/<link>static
      ;   

The problem appear during the link stage, the generated g++ command looks
like:
g++ -L"/.../project_path/libs/Python"
-I"/.../project_path/libs/Python/Include" ... -o testscript
"/.../project_path/libs/sys/libotherlib.a"
"/.../project_path/libs/boost/libs/python/.../boost_python.a"  main.o ...
-lpython2.5 -g

This failed due to unresolved symbols with python library:
Undefined symbols:
  "__Py_NegativeRefcount", referenced from:
      boost::python::detail::decref_guard::~decref_guard()in main.o
...

Manually, I've tried:
a) replace "-lpython2.5" by /.../project_path/libs/Python/libpython2.5.a :
this is working just fine - the exe is linked and everything work

b) rename the system python library "/usr/lib/libpython2.5.dylib" (so g++
cannot find it at link stage) : this is also working fine.

So, by using -lpython2.5, we encourage g++ to find libraries in its standard
paths, which I would like to avoid since we would like to statically build
our own libpython.

Is there any nice method to tell bjam / boost.python to use our own python
library (in static)?

A good way would be that bjam use the full path to the static version
(project_path/libs/Python/libpython2.5.a) of it when linking exe to it, is
there a way to do so?

Thanks in advance for your help!

Best regards,

jeanluc

-- 
View this message in context: http://www.nabble.com/how-to-link-project-with-a-specified-libpython-in-static--%28using-boost.Python%29-tp19242281p19242281.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.