From paul.bilokon at lehman.com  Wed Sep  3 10:08:36 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Wed, 3 Sep 2008 09:08:36 +0100
Subject: [C++-sig] Attempt to Return Dangling Reference Error --
	What'sthe solution?
In-Reply-To: <87hc93ho04.fsf@mcbain.luannocracy.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B8E@lopets1exms003.leh.lbcorp.lehman.com>
	<87hc93ho04.fsf@mcbain.luannocracy.com>
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1BA4@lopets1exms003.leh.lbcorp.lehman.com>

Hi Dave,

Thanks a lot. Changing clone's signature to...

    virtual boost::shared_ptr<Animal> clone() const

did the trick. However, since I'm wrapping an existing library, I'm
trying to find a non-intrusive solution. I'd rather keep the current
signature of clone. Are there any other options?

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
David Abrahams
Sent: Friday, August 29, 2008 3:53 PM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Attempt to Return Dangling Reference Error --
What'sthe solution?


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#dangl
> in
> 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
_______________________________________________
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 Sep  3 11:27:02 2008
From: paul.bilokon at lehman.com (Bilokon, Paul)
Date: Wed, 3 Sep 2008 10:27:02 +0100
Subject: [C++-sig] Attempt to Return Dangling Reference Error --
	What'sthe solution?
In-Reply-To: <BBABA8164420CB43ABEA9DC81FE7A884065D1BA4@lopets1exms003.leh.lbcorp.lehman.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B8E@lopets1exms003.leh.lbcorp.lehman.com>
	<87hc93ho04.fsf@mcbain.luannocracy.com>
	<BBABA8164420CB43ABEA9DC81FE7A884065D1BA4@lopets1exms003.leh.lbcorp.lehman.com>
Message-ID: <BBABA8164420CB43ABEA9DC81FE7A884065D1BA6@lopets1exms003.leh.lbcorp.lehman.com>

Hi Dave,

Thought I should give more background. The library I'm wrapping is using
a custom smart pointer, which is really unfortunate. It would be
difficult (though desirable) to replace these custom shared pointers
with shared_ptr. Therefore I'm trying to provide a mechanism for calling
the functions expecting CustomPtr with a CustomPtr-to-copy by calling
clone and wrapping the copy into CustomPtr, wherever this has reasonable
semantics. Therefore I cannot wrap the return of clone into shared_ptr.
Is there any way to cheat the dangling reference detection mechanism?
(Of course, shared_ptr does not do release().)

Many thanks,
Paul

-----Original Message-----
From: Bilokon, Paul 
Sent: Wednesday, September 03, 2008 9:09 AM
To: 'Development of Python/C++ integration'
Subject: RE: [C++-sig] Attempt to Return Dangling Reference Error --
What'sthe solution?

Hi Dave,

Thanks a lot. Changing clone's signature to...

    virtual boost::shared_ptr<Animal> clone() const

did the trick. However, since I'm wrapping an existing library, I'm
trying to find a non-intrusive solution. I'd rather keep the current
signature of clone. Are there any other options?

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
David Abrahams
Sent: Friday, August 29, 2008 3:53 PM
To: Development of Python/C++ integration
Subject: Re: [C++-sig] Attempt to Return Dangling Reference Error --
What'sthe solution?


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#dangl
> in
> 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
_______________________________________________
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 roman.yakovenko at gmail.com  Wed Sep  3 11:40:33 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 3 Sep 2008 12:40:33 +0300
Subject: [C++-sig] Attempt to Return Dangling Reference Error --
	What'sthe solution?
In-Reply-To: <BBABA8164420CB43ABEA9DC81FE7A884065D1BA6@lopets1exms003.leh.lbcorp.lehman.com>
References: <BBABA8164420CB43ABEA9DC81FE7A884065D1B8E@lopets1exms003.leh.lbcorp.lehman.com>
	<87hc93ho04.fsf@mcbain.luannocracy.com>
	<BBABA8164420CB43ABEA9DC81FE7A884065D1BA4@lopets1exms003.leh.lbcorp.lehman.com>
	<BBABA8164420CB43ABEA9DC81FE7A884065D1BA6@lopets1exms003.leh.lbcorp.lehman.com>
Message-ID: <7465b6170809030240k5458faa4rd6bbe8d801eda4b7@mail.gmail.com>

On Wed, Sep 3, 2008 at 12:27 PM, Bilokon, Paul <paul.bilokon at lehman.com> wrote:
> Hi Dave,
>
> Thought I should give more background. The library I'm wrapping is using
> a custom smart pointer, which is really unfortunate. It would be
> difficult (though desirable) to replace these custom shared pointers
> with shared_ptr. Therefore I'm trying to provide a mechanism for calling
> the functions expecting CustomPtr with a CustomPtr-to-copy by calling
> clone and wrapping the copy into CustomPtr, wherever this has reasonable
> semantics. Therefore I cannot wrap the return of clone into shared_ptr.
> Is there any way to cheat the dangling reference detection mechanism?
> (Of course, shared_ptr does not do release().)

May be my troubleshooting guide can help you:
http://language-binding.net/pyplusplus/troubleshooting_guide/smart_ptrs/smart_ptrs.html

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


From ope-devel at gmx.de  Wed Sep  3 21:14:00 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Wed, 03 Sep 2008 21:14:00 +0200
Subject: [C++-sig] embbed simple example failed
Message-ID: <48BEE1F8.1020102@gmx.de>

Hi,

I get mad on this snipped; I've got:

$ g++ -I /usr/include/python2.5 bpl.cpp -o bpl -l boost_python -l
python2.5 && ./bpl
  File "<string>", line 1
    print("Hello World\n")
        ^
SyntaxError: invalid syntax
something went wrong

What did I wrong???

Thanks,
Olaf

-------------- next part --------------
A non-text attachment was scrubbed...
Name: bpl.cpp
Type: text/x-c++src
Size: 1897 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080903/419ccf0f/attachment.cpp>

From seefeld at sympatico.ca  Wed Sep  3 21:40:58 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Wed, 03 Sep 2008 15:40:58 -0400
Subject: [C++-sig] embbed simple example failed
In-Reply-To: <48BEE1F8.1020102@gmx.de>
References: <48BEE1F8.1020102@gmx.de>
Message-ID: <48BEE84A.8020002@sympatico.ca>

Olaf Peter wrote:
> Hi,
>
> I get mad on this snipped; I've got:
>
> $ g++ -I /usr/include/python2.5 bpl.cpp -o bpl -l boost_python -l
> python2.5 && ./bpl
>   File "<string>", line 1
>     print("Hello World\n")
>         ^
> SyntaxError: invalid syntax
>   

The compiler is right. What you probably meant is 'printf', not 'print'.

It is usually a good idea to narrow down the error by removing bits from 
the sources until the problem disappears. Doing that you would have 
realized that your problem has nothing to do with boost, neither with 
python.

HTH,
       Stefan

-- 

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



From amohr at pixar.com  Wed Sep  3 22:51:10 2008
From: amohr at pixar.com (Alex Mohr)
Date: Wed, 03 Sep 2008 13:51:10 -0700
Subject: [C++-sig] embbed simple example failed
In-Reply-To: <48BEE1F8.1020102@gmx.de>
References: <48BEE1F8.1020102@gmx.de>
Message-ID: <48BEF8BE.9080501@pixar.com>

> I get mad on this snipped; I've got:
> 
> $ g++ -I /usr/include/python2.5 bpl.cpp -o bpl -l boost_python -l
> python2.5 && ./bpl
>   File "<string>", line 1
>     print("Hello World\n")
>         ^
> SyntaxError: invalid syntax
> something went wrong

The problem is that in python, 'print' is a statement, not an 
expression.  When you call PyRun_String, you pass Py_eval_input as the 
start token, which means python expects the string you pass to be an 
expression, not a statement.

I think you can fix this by passing Py_single_input to PyRun_String 
rather than Py_eval_input.

Check out the docs for PyRun_String, Py_eval_input, and Py_single_input 
here:

http://docs.python.org/api/veryhigh.html

Alex



From seefeld at sympatico.ca  Wed Sep  3 22:58:05 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Wed, 03 Sep 2008 16:58:05 -0400
Subject: [C++-sig] embbed simple example failed
In-Reply-To: <48BEF8BE.9080501@pixar.com>
References: <48BEE1F8.1020102@gmx.de> <48BEF8BE.9080501@pixar.com>
Message-ID: <48BEFA5D.8020209@sympatico.ca>

Alex Mohr wrote:
>> I get mad on this snipped; I've got:
>>
>> $ g++ -I /usr/include/python2.5 bpl.cpp -o bpl -l boost_python -l
>> python2.5 && ./bpl
>>   File "<string>", line 1
>>     print("Hello World\n")
>>         ^
>> SyntaxError: invalid syntax
>> something went wrong
>
> The problem is that in python, 'print' is a statement, not an 
> expression.  When you call PyRun_String, you pass Py_eval_input as the 
> start token, which means python expects the string you pass to be an 
> expression, not a statement.

Olaf,

I'm really sorry I replied after only a casual look at the error message.
Let me try to rectify that by suggesting that you use the boost.python 
embedding API instead. Something like

namespace bpl = boost::python;
bpl::object result = bpl::exec("print 'Hello World\n'");

Instead of what you originally wrote, and bpl::eval() for actual 
expressions.

HTH,
       Stefan


-- 

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



From ope-devel at gmx.de  Thu Sep  4 05:06:55 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Thu, 04 Sep 2008 05:06:55 +0200
Subject: [C++-sig] embbed simple example failed
In-Reply-To: <48BEF8BE.9080501@pixar.com>
References: <48BEE1F8.1020102@gmx.de> <48BEF8BE.9080501@pixar.com>
Message-ID: <48BF50CF.4050603@gmx.de>

> I think you can fix this by passing Py_single_input to PyRun_String
> rather than Py_eval_input.
> 
> Check out the docs for PyRun_String, Py_eval_input, and Py_single_input
> here:
> 
> http://docs.python.org/api/veryhigh.html
> 
> Alex

Thanks, that's. Some time ago I did introduce this bug and saw what I
did expect to see.

Thanks,
Olaf


From rogeeff at gmail.com  Thu Sep  4 08:29:15 2008
From: rogeeff at gmail.com (Gennadiy Rozental)
Date: Thu, 4 Sep 2008 06:29:15 +0000 (UTC)
Subject: [C++-sig] how t o specify and/or update global dictionary for
	bp::call call
Message-ID: <loom.20080904T061649-100@post.gmane.org>

Hi,

I've got class implemented in python, inherited from C++ class. I am using
wrapper/get_overload to invoke it's methods:

bp::override o = this->get_override( "update" );
bp::call< void >( o.ptr(), arg1, arg2 );

Now in update method I need to refer to the function foo that reside 
in module moo.

I've tried to add 
from moo import foo
and 
from moo import *

before class definition:

class MyPyDerivedClass( MyCPPBase ):
    def update( s, arg1, ARG2 ):
        foo()

Both attempts were futile. Looks like global dictionary during update call does
not contain foo.

So, now I am looking for a way for either update it somehow or specify my own.

Please advise.

Gennadiy

P.S. solution to add import method inside update method is unacceptable for
various (mostly political) reasons, though it obviously will work. I definitely
need this outside class definition

P.P.S. Instance of class MyPyDerivedClass is created on C++ side based on
pre-cached python type:

cpp_derived = m_py_type();






From seefeld at sympatico.ca  Thu Sep  4 15:00:54 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Thu, 04 Sep 2008 09:00:54 -0400
Subject: [C++-sig] how t o specify and/or update global dictionary for
 bp::call call
In-Reply-To: <loom.20080904T061649-100@post.gmane.org>
References: <loom.20080904T061649-100@post.gmane.org>
Message-ID: <48BFDC06.1010706@sympatico.ca>

Gennadiy Rozental wrote:
> Hi,
>
> I've got class implemented in python, inherited from C++ class. I am using
> wrapper/get_overload to invoke it's methods:
>
> bp::override o = this->get_override( "update" );
> bp::call< void >( o.ptr(), arg1, arg2 );
>
> Now in update method I need to refer to the function foo that reside 
> in module moo.
>
> I've tried to add 
> from moo import foo
> and 
> from moo import *
>
> before class definition:
>
> class MyPyDerivedClass( MyCPPBase ):
>     def update( s, arg1, ARG2 ):
>         foo()
>
> Both attempts were futile. Looks like global dictionary during update call does
> not contain foo.
>   

For avoidance of doubt: you are saying the import itself was successful, 
but there was no 'foo' symbol in the global namespace after that point ?
I just modified the exec.cpp test (libs/python/tests/exec.cpp), and it 
all works like a charm.
(For reference: go to libs/python/tests/exec.cpp and modify the 
'exec_test' function to simulate what you want. Here is the modified 
script I use:

  python::object main = python::import("__main__");
  python::object global(main.attr("__dict__"));
  python::object result = python::exec(
    "from embedded_hello import *        \n"
    "from sys import *                   \n"                             
//<-- added just to show how to populate the global namespace of the script
    "class PythonDerived(Base):          \n"
    "    def hello(self):                \n"
    "        print platform         \n"                                 
//<-- use it
    "        return 'Hello from Python!' \n",
    global, global);

  python::object PythonDerived = global["PythonDerived"];
  python::object py_base = PythonDerived();
  Base& py = python::extract<Base&>(py_base);
  assert(py.hello() == "Hello from Python!");

> P.S. solution to add import method inside update method is unacceptable for
> various (mostly political) reasons, though it obviously will work. I definitely
> need this outside class definition
>
> P.P.S. Instance of class MyPyDerivedClass is created on C++ side based on
> pre-cached python type:
>
> cpp_derived = m_py_type();
>   

Sounds like what you are doing is very similar to what the test I just 
exerimented with does. Right ?


HTH,
       Stefan



-- 

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



From rogeeff at gmail.com  Thu Sep  4 18:38:05 2008
From: rogeeff at gmail.com (Gennadiy Rozental)
Date: Thu, 4 Sep 2008 16:38:05 +0000 (UTC)
Subject: [C++-sig] how t o specify and/or update global dictionary for
	bp::call call
References: <loom.20080904T061649-100@post.gmane.org>
	<48BFDC06.1010706@sympatico.ca>
Message-ID: <loom.20080904T161514-940@post.gmane.org>

Stefan Seefeld <seefeld <at> sympatico.ca> writes:

> For avoidance of doubt: you are saying the import itself was successful, 
> but there was no 'foo' symbol in the global namespace after that point ?

Yes. That's true

> I just modified the exec.cpp test (libs/python/tests/exec.cpp), and it 
> all works like a charm.
> (For reference: go to libs/python/tests/exec.cpp and modify the 
> 'exec_test' function to simulate what you want. Here is the modified 
> script I use:
> 
>   python::object main = python::import("__main__");
>   python::object global(main.attr("__dict__"));
>   python::object result = python::exec(
>     "from embedded_hello import *        \n"
>     "from sys import *                   \n"                             
> //<-- added just to show how to populate the global namespace of the script
>     "class PythonDerived(Base):          \n"
>     "    def hello(self):                \n"
>     "        print platform         \n"                                 
> //<-- use it
>     "        return 'Hello from Python!' \n",
>     global, global);


This only works because you used global dictionary as local. I can't do this for
various reasons, main of which is that I am trying to mimic namespacing, which
is unavailable directly in python. So the local dictionary is the dictionary of
the module I am trying to put this class in.
 
>   python::object PythonDerived = global["PythonDerived"];
>   python::object py_base = PythonDerived();
>   Base& py = python::extract<Base&>(py_base);
>   assert(py.hello() == "Hello from Python!");

So what I need to update global dictionary before hello call here even if it was
not used as local during exec call.

Gennadiy




From william.marie at gmail.com  Fri Sep  5 16:01:21 2008
From: william.marie at gmail.com (=?UTF-8?Q?William_Mari=C3=A9?=)
Date: Fri, 5 Sep 2008 07:01:21 -0700 (PDT)
Subject: [C++-sig]  Calling Object virtual function from C++
Message-ID: <19331891.post@talk.nabble.com>


Hi,

I'm trying to call a object virtual function from the C++ side but if the
python function is not defined into the Python class it doesn't find the one
defined in the C++ object :

------- C++
class Base
{
public:
	Base() {};
	virtual ~Base() {};
	virtual void testVirtual()
	{ 
		itn a = 0;
	}
};

struct BaseWrap : Base, bp::wrapper<Base>
{
	void testVirtual()
	{
		if (bp::override fun = this->get_override("testVirtual"))
			fun(); // *note*
		Base::testVirtual();
	}
	void default_testVirtual() { this->Base::testVirtual(); }
};

BOOST_PYTHON_MODULE(MyModule)
{
	bp::class_<BaseWrap, boost::noncopyable>("Base")
		.def("testVirtual", &Base::testVirtual, &BaseWrap::default_testVirtual)
		;

	bp::class_<Base>("Base", bp::init<>())
		;
} 
------- Python ( testvirtual.py )
import MyModule

class TestVirtual( MyModule.Base ):

	## Constructor
	def __init__( self ) :
                pass

------------ Now from a C++ side :
bp::object module = bp::import("testvirtual");
bp::object modDict(module.attr("__dict__"));
bp::object bpClass = modDict["TestVirtual"];
bp::object bpTestVirtual = pbClass();

# Doesn't call the Base class virtual function and tell that the testVirtual
function doesn't exist
bp::object bpFun = bpTestVirtual.attr( "testVirtual" );
bpFun(); 

# Doesn't call the Base class virtual function and tell that the testVirtual
function doesn't exist bp::call<void>(bpFun.ptr());

# Doesn't call the Base class virtual function and tell that the testVirtual
function doesn't exist
bp::call_method<void>(bpTestVirtual.ptr(), "testVirtual" );

# Always call the Base::testVirtual even if the function is overidden in
Python
Base& pTestVirtual = BP_EXTRACT( bpTestVirtual, Base& );
pTestVirtual->testVirtual()


Here we go, so i'm sure there is an easy way to do that but i don't
understand what i'm doing wrong.
Could someone help me please ?
Thanks

William


-- 
View this message in context: http://www.nabble.com/Calling-Object-virtual-function-from-C%2B%2B-tp19331891p19331891.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From wladwig at wdtinc.com  Fri Sep  5 17:50:25 2008
From: wladwig at wdtinc.com (William Ladwig)
Date: Fri, 5 Sep 2008 10:50:25 -0500
Subject: [C++-sig] Calling Object virtual function from C++
In-Reply-To: <19331891.post@talk.nabble.com>
References: <19331891.post@talk.nabble.com>
Message-ID: <765CBD9053EA2B438625895F30D1856FCA025885@storm.wdtinc.com>

In your BOOST_PYTHON_MODULE you have "Base" defined twice.  Try removing:

bp::class_<Base>("Base", bp::init<>())
        ;

since you already exposed your "Base" class using BaseWrapper.

Also, in your python module, try adding the base class constructor:

MyModule.Base.__init__(self)

to the first line of your TestVirtual __init__ function.

These are the first two places I would start.  Hope this helps.

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 William Mari?
Sent: Friday, September 05, 2008 9:01 AM
To: c++-sig at python.org
Subject: [C++-sig] Calling Object virtual function from C++


Hi,

I'm trying to call a object virtual function from the C++ side but if the
python function is not defined into the Python class it doesn't find the one
defined in the C++ object :

------- C++
class Base
{
public:
        Base() {};
        virtual ~Base() {};
        virtual void testVirtual()
        {
                itn a = 0;
        }
};

struct BaseWrap : Base, bp::wrapper<Base>
{
        void testVirtual()
        {
                if (bp::override fun = this->get_override("testVirtual"))
                        fun(); // *note*
                Base::testVirtual();
        }
        void default_testVirtual() { this->Base::testVirtual(); }
};

BOOST_PYTHON_MODULE(MyModule)
{
        bp::class_<BaseWrap, boost::noncopyable>("Base")
                .def("testVirtual", &Base::testVirtual, &BaseWrap::default_testVirtual)
                ;

        bp::class_<Base>("Base", bp::init<>())
                ;
}
------- Python ( testvirtual.py )
import MyModule

class TestVirtual( MyModule.Base ):

        ## Constructor
        def __init__( self ) :
                pass

------------ Now from a C++ side :
bp::object module = bp::import("testvirtual");
bp::object modDict(module.attr("__dict__"));
bp::object bpClass = modDict["TestVirtual"];
bp::object bpTestVirtual = pbClass();

# Doesn't call the Base class virtual function and tell that the testVirtual
function doesn't exist
bp::object bpFun = bpTestVirtual.attr( "testVirtual" );
bpFun();

# Doesn't call the Base class virtual function and tell that the testVirtual
function doesn't exist bp::call<void>(bpFun.ptr());

# Doesn't call the Base class virtual function and tell that the testVirtual
function doesn't exist
bp::call_method<void>(bpTestVirtual.ptr(), "testVirtual" );

# Always call the Base::testVirtual even if the function is overidden in
Python
Base& pTestVirtual = BP_EXTRACT( bpTestVirtual, Base& );
pTestVirtual->testVirtual()


Here we go, so i'm sure there is an easy way to do that but i don't
understand what i'm doing wrong.
Could someone help me please ?
Thanks

William


--
View this message in context: http://www.nabble.com/Calling-Object-virtual-function-from-C%2B%2B-tp19331891p19331891.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.

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


From snemeth at houston.rr.com  Sat Sep  6 23:48:47 2008
From: snemeth at houston.rr.com (Rocketman@JSC)
Date: Sat, 6 Sep 2008 14:48:47 -0700 (PDT)
Subject: [C++-sig] Base class defined in other header file not
 includedin bp::bases<>
In-Reply-To: <7465b6170607140134r5534654dtb54cec53cd930fa7@mail.gmail.com>
References: <5383A81D5EB8F14FBDB487D9B5FBEA880152C02F@NAEX11.na.qualcomm.com>
	<7465b6170607122218j6aeb97c4y2d1c60cff878037f@mail.gmail.com>
	<5383A81D5EB8F14FBDB487D9B5FBEA880152C24D@NAEX11.na.qualcomm.com>
	<7465b6170607140134r5534654dtb54cec53cd930fa7@mail.gmail.com>
Message-ID: <19351675.post@talk.nabble.com>


I have a few questions about this thread.  Basically I have the same problem,
my bases for a binding are not getting picked up cause they are in a
separate library.

I've tried solution #2 and here's a snippet from the Py++ generator file:

<<Start
Snippet=================================================================

mb.build_code_creator( module_name='SmpSim' )

# This is an attempt at getting the bases included for Vehicle

baseModel = mb.class_('Model')

print type(mb.code_creator)

baseModel_creators = code_creators.creator_finder.find_by_declaration(
lambda decl: decl is baseModel, mb.code_creator, False )

vehicle = mb.class_('Vehicle') # This is the class I need to add the bases
to
vehicle = ( code_creators.class_t, code_creators.class_wrapper_t )

baseModel_creators = filter( lambda c: isinstance(c, vehicle),
baseModel_creators)

for c in baseModel_creators:
    c.create = lambda: ''
<<End
Snippet=================================================================

and I get the following error:

Traceback (most recent call last):
  File "/data/home/scott/eclipse/Smp_lib/generate_code.py", line 157, in
<module>
    baseModel_creators = code_creators.creator_finder.find_by_declaration(
lambda decl: decl is baseModel, mb.code_creator, False )
  File "/data/apps/oss/lib/python/pyplusplus/code_creators/algorithm.py",
line 87, in find_by_declaration
    , search_area )
TypeError: 'module_t' object is not iterable

I'm sure this is something I'm doing wrong, but I looked at the API's and
could not figure it out.  Does anyone have an idea what I'm doing wrong
here?

Thanks in advance for any help.

Scott


Roman Yakovenko wrote:
> 
> On 7/13/06, Haridev, Meghana <mharidev at qualcomm.com> wrote:
>> In other words, I want the generated code to look like this:
>>     /* bp::class_< baseB >( "baseB" );  - Do not want to generate */
>>
>>     bp::class_< baseA >( "baseA" );
>>
>>     bp::class_< derivedC, bp::bases< baseA, baseB > >( "derivedC" );
> 
> It is possible, but you will have to understand a little how code
> creators works.
> There are few ways to solve the problem.
> 
> 1. Exclude baseB from being generated, and add it manually to derivedC.
>     Take a look on pyplusplus/code_creators/class_declaration.py -
> class "class_t",
>     method - _generate_bases. You will have to redefine this method,
> something
>     like this:
> 
>     def my_generate_bases( self, base_creators ):
>         if self.declaration.Name != 'derivedC':
>             run current code
>         else:
>             your logic is going her
> 
>    pyplusplus.code_creators.class_t._generate_bases = my_generate_bases
> 
> 2. You can include baseB to be generated, but instead of generating code
> it
>     will generate empty string:
> 
>     mb = module_builder_t( ... )
>     mb.build_code_creators( ... )
> 
>     from pyplusplus import code_creators
> 
>     baseB = mb.class_( 'baseB' )
> 
>     baseB_creators \
>         = code_creators.creator_finder.find_by_declaration(
>                      lambda decl: decl is baseB #what to look
>                      , mb.code_creator ) #where
>     #now, we have to leave only class_t and class_wrapper_t code creators
>      relevant_clss = ( code_creators.class_t,
> code_creators.class_wrapper_t )
>      baseB_creators = filters( lambda c: isinstance( c, relevant_clss ) )
>      for c in baseB_creators:
>          c.create = lambda self: ''
> 
> Second approach is better, because in this case pyplusplus will take
> into account
> all data available from baseB, but will not generate code.
> 
> 
> -- 
> Roman Yakovenko
> C++ Python language binding
> http://www.language-binding.net/
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
> 
> 

-- 
View this message in context: http://www.nabble.com/-C%2B%2B-sig--Base-class-defined-in-other-header-file-not-included-in-bp%3A%3Abases%3C%3E-tp5299734p19351675.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From j.reid at mail.cryst.bbk.ac.uk  Wed Sep 10 13:51:18 2008
From: j.reid at mail.cryst.bbk.ac.uk (John Reid)
Date: Wed, 10 Sep 2008 12:51:18 +0100
Subject: [C++-sig] [Fwd: NumPy arrays that use memory allocated from other
 libraries or tools]
Message-ID: <ga8cc1$hso$2@ger.gmane.org>

I saw this on the rpy list and thought it might be of interest to some here.

John.

-------- Original Message --------
Subject: NumPy arrays that use memory allocated from other libraries or 
tools
Date: Tue, 09 Sep 2008 23:39:26 -0500
From: Travis Oliphant <oliphant.travis at ieee.org>
To: python-list at python.org
CC: rpy-list at lists.sourceforge.net, ctypes-users at lists.sourceforge.net
Newsgroups: 
gmane.comp.python.general,gmane.comp.python.rpy,gmane.comp.python.ctypes


I wanted to point anybody interested to a blog post that describes a
useful pattern for having a NumPy array that points to the memory
created by a different memory manager than the standard one used by
NumPy.   The pattern  shows how to create a NumPy array that points to
previously allocated memory and then shows how to construct an object
that allows the correct deallocator to be called when the NumPy array is
freed.

This may be useful if you are wrapping code that has it's own memory
management scheme.   Comments and feedback is welcome.

The post is

http://blog.enthought.com/?p=62


Best regards,

-Travis Oliphant

--
http://mail.python.org/mailman/listinfo/python-list



From william.marie at gmail.com  Thu Sep 11 09:32:04 2008
From: william.marie at gmail.com (=?ISO-8859-1?Q?William_Mari=E9?=)
Date: Thu, 11 Sep 2008 08:32:04 +0100
Subject: [C++-sig] Calling Object virtual function from C++
In-Reply-To: <765CBD9053EA2B438625895F30D1856FCA025885@storm.wdtinc.com>
References: <19331891.post@talk.nabble.com>
	<765CBD9053EA2B438625895F30D1856FCA025885@storm.wdtinc.com>
Message-ID: <48C8C974.1030603@gmail.com>

Hi Bill,

Thanks for your answer it was exactly that, no need to declare both the 
wrapper and the class ...
Have a good day and thank a lot
William

William Ladwig a ?crit :
> In your BOOST_PYTHON_MODULE you have "Base" defined twice.  Try removing:
>
> bp::class_<Base>("Base", bp::init<>())
>         ;
>
> since you already exposed your "Base" class using BaseWrapper.
>
> Also, in your python module, try adding the base class constructor:
>
> MyModule.Base.__init__(self)
>
> to the first line of your TestVirtual __init__ function.
>
> These are the first two places I would start.  Hope this helps.
>
> 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 William Mari?
> Sent: Friday, September 05, 2008 9:01 AM
> To: c++-sig at python.org
> Subject: [C++-sig] Calling Object virtual function from C++
>
>
> Hi,
>
> I'm trying to call a object virtual function from the C++ side but if the
> python function is not defined into the Python class it doesn't find the one
> defined in the C++ object :
>
> ------- C++
> class Base
> {
> public:
>         Base() {};
>         virtual ~Base() {};
>         virtual void testVirtual()
>         {
>                 itn a = 0;
>         }
> };
>
> struct BaseWrap : Base, bp::wrapper<Base>
> {
>         void testVirtual()
>         {
>                 if (bp::override fun = this->get_override("testVirtual"))
>                         fun(); // *note*
>                 Base::testVirtual();
>         }
>         void default_testVirtual() { this->Base::testVirtual(); }
> };
>
> BOOST_PYTHON_MODULE(MyModule)
> {
>         bp::class_<BaseWrap, boost::noncopyable>("Base")
>                 .def("testVirtual", &Base::testVirtual, &BaseWrap::default_testVirtual)
>                 ;
>
>         bp::class_<Base>("Base", bp::init<>())
>                 ;
> }
> ------- Python ( testvirtual.py )
> import MyModule
>
> class TestVirtual( MyModule.Base ):
>
>         ## Constructor
>         def __init__( self ) :
>                 pass
>
> ------------ Now from a C++ side :
> bp::object module = bp::import("testvirtual");
> bp::object modDict(module.attr("__dict__"));
> bp::object bpClass = modDict["TestVirtual"];
> bp::object bpTestVirtual = pbClass();
>
> # Doesn't call the Base class virtual function and tell that the testVirtual
> function doesn't exist
> bp::object bpFun = bpTestVirtual.attr( "testVirtual" );
> bpFun();
>
> # Doesn't call the Base class virtual function and tell that the testVirtual
> function doesn't exist bp::call<void>(bpFun.ptr());
>
> # Doesn't call the Base class virtual function and tell that the testVirtual
> function doesn't exist
> bp::call_method<void>(bpTestVirtual.ptr(), "testVirtual" );
>
> # Always call the Base::testVirtual even if the function is overidden in
> Python
> Base& pTestVirtual = BP_EXTRACT( bpTestVirtual, Base& );
> pTestVirtual->testVirtual()
>
>
> Here we go, so i'm sure there is an easy way to do that but i don't
> understand what i'm doing wrong.
> Could someone help me please ?
> Thanks
>
> William
>
>
> --
> View this message in context: http://www.nabble.com/Calling-Object-virtual-function-from-C%2B%2B-tp19331891p19331891.html
> Sent from the Python - c++-sig mailing list archive at Nabble.com.
>
> _______________________________________________
> 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 william.marie at gmail.com  Thu Sep 11 09:39:31 2008
From: william.marie at gmail.com (=?UTF-8?Q?William_Mari=C3=A9?=)
Date: Thu, 11 Sep 2008 00:39:31 -0700 (PDT)
Subject: [C++-sig]  Call Python functions generically
Message-ID: <19429482.post@talk.nabble.com>


Hi,

I would like to know if there is an easy way to create generic python
function calls using a tuple or something like that.

ex : 

void call( const char* szFunName, bp::tuple bpTuple )
{
      // Something like that
      bp::call( szFunName, bpTuple );
}

bpTuple being the args list...

In Python to do that we can use something like that :
arglist = Py_BuildValue("(i)", arg);
result = PyEval_CallObject(my_callback, arglist);

but i don't if a similar mechanism exists in boost python.

Thanks for your help

William
-- 
View this message in context: http://www.nabble.com/Call-Python-functions-generically-tp19429482p19429482.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From seefeld at sympatico.ca  Thu Sep 11 14:01:40 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Thu, 11 Sep 2008 08:01:40 -0400
Subject: [C++-sig] Call Python functions generically
In-Reply-To: <19429482.post@talk.nabble.com>
References: <19429482.post@talk.nabble.com>
Message-ID: <48C908A4.1070403@sympatico.ca>

William Mari? wrote:
> Hi,
>
> I would like to know if there is an easy way to create generic python
> function calls using a tuple or something like that.
>   

I have checked in support for that into trunk a couple of weeks ago. I'm 
not sure when this will make it into a release.
> ex : 
>
> void call( const char* szFunName, bp::tuple bpTuple )
> {
>       // Something like that
>       bp::call( szFunName, bpTuple );
> }
>
> bpTuple being the args list...
>   

Assuming you have a callable object, you simply invoke the call 
operator, passing an argument sequence, and optionally, a keyword 
dictionary, using the same syntax as in Python, i.e. unary '*' and '**' 
operators:

object args = ...;
object kwds = ...;

object callable = ...;

callable(*args, **kwds);


Regards,
       Stefan

-- 

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



From william.marie at gmail.com  Fri Sep 12 09:57:53 2008
From: william.marie at gmail.com (=?ISO-8859-1?Q?William_Mari=E9?=)
Date: Fri, 12 Sep 2008 08:57:53 +0100
Subject: [C++-sig] Call Python functions generically
In-Reply-To: <48C908A4.1070403@sympatico.ca>
References: <19429482.post@talk.nabble.com> <48C908A4.1070403@sympatico.ca>
Message-ID: <48CA2101.8010301@gmail.com>

Hi Stefan,

Ok this would be a very pretty way. Do you know when this feature will 
be commited ?

Stefan Seefeld a ?crit :
> William Mari? wrote:
>> Hi,
>>
>> I would like to know if there is an easy way to create generic python
>> function calls using a tuple or something like that.
>>   
>
> I have checked in support for that into trunk a couple of weeks ago. 
> I'm not sure when this will make it into a release.
>> ex :
>> void call( const char* szFunName, bp::tuple bpTuple )
>> {
>>       // Something like that
>>       bp::call( szFunName, bpTuple );
>> }
>>
>> bpTuple being the args list...
>>   
>
> Assuming you have a callable object, you simply invoke the call 
> operator, passing an argument sequence, and optionally, a keyword 
> dictionary, using the same syntax as in Python, i.e. unary '*' and 
> '**' operators:
>
> object args = ...;
> object kwds = ...;
>
> object callable = ...;
>
> callable(*args, **kwds);
>
>
> Regards,
>       Stefan
>



From seefeld at sympatico.ca  Fri Sep 12 12:59:08 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Fri, 12 Sep 2008 06:59:08 -0400
Subject: [C++-sig] Call Python functions generically
In-Reply-To: <48CA2101.8010301@gmail.com>
References: <19429482.post@talk.nabble.com> <48C908A4.1070403@sympatico.ca>
	<48CA2101.8010301@gmail.com>
Message-ID: <48CA4B7C.7010307@sympatico.ca>

William Mari? wrote:
> Hi Stefan,
>
> Ok this would be a very pretty way. Do you know when this feature will 
> be commited ?

As I indicated, it already is (to trunk).

Regards,
       Stefan

-- 

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



From surye80 at gmail.com  Sat Sep 13 08:56:59 2008
From: surye80 at gmail.com (Vincent Castellano)
Date: Fri, 12 Sep 2008 23:56:59 -0700
Subject: [C++-sig] Basic example segfaults in PyInt_FromLong
Message-ID: <8628ae450809122356w19a92b27o88539f3713946c17@mail.gmail.com>

I'm attempting to compile and run the following:

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

using namespace boost::python;
using namespace std;
int main(int argc, char* argv[])
{
    object result;
    int number;
    result = eval("5 ** 2");
    number = extract<int>(result);

    cout << number << endl;
    return 0;
}

g++ boostpythontest.cpp -o boostpython -I /usr/include/python2.5/
-lboost_python -lpython2.5

It's segfaulting with the following backtrace:

#0  0x00002aad3d27cb6c in PyInt_FromLong () from /usr/lib/libpython2.5.so.1.0
#1  0x00002aad3d2ca80b in ?? () from /usr/lib/libpython2.5.so.1.0
#2  0x00002aad3d2ccf01 in ?? () from /usr/lib/libpython2.5.so.1.0
#3  0x00002aad3d2cb702 in ?? () from /usr/lib/libpython2.5.so.1.0
#4  0x00002aad3d2cf835 in PyAST_FromNode () from /usr/lib/libpython2.5.so.1.0
#5  0x00002aad3d2fcb74 in PyParser_ASTFromString () from
/usr/lib/libpython2.5.so.1.0
#6  0x00002aad3d2fcd6c in PyRun_StringFlags () from /usr/lib/libpython2.5.so.1.0
#7  0x00002aad3d0135d3 in boost::python::eval ()
   from /usr/lib/libboost_python-gcc42-mt-1_34_1.so.1.34.1
#8  0x00000000004012a5 in main (argc=1, argv=0x7fff6dcee1a8)
    at boostpythontest.cpp:10


I set this up as a basic test of the library, taken pretty much from
the tutorial. boost and python on Hardy Ubuntu. Any pointers would be
really appreciated, I cannot figure out where this could be failing so
early.


From roman.yakovenko at gmail.com  Sun Sep 14 08:32:43 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sun, 14 Sep 2008 09:32:43 +0300
Subject: [C++-sig] Base class defined in other header file not
	includedin bp::bases<>
In-Reply-To: <19351675.post@talk.nabble.com>
References: <5383A81D5EB8F14FBDB487D9B5FBEA880152C02F@NAEX11.na.qualcomm.com>
	<7465b6170607122218j6aeb97c4y2d1c60cff878037f@mail.gmail.com>
	<5383A81D5EB8F14FBDB487D9B5FBEA880152C24D@NAEX11.na.qualcomm.com>
	<7465b6170607140134r5534654dtb54cec53cd930fa7@mail.gmail.com>
	<19351675.post@talk.nabble.com>
Message-ID: <7465b6170809132332m1b43469dsdf6083ceec5de0e2@mail.gmail.com>

On Sun, Sep 7, 2008 at 12:48 AM, Rocketman at JSC <snemeth at houston.rr.com> wrote:
> Roman Yakovenko wrote:
>>
>> On 7/13/06, Haridev, Meghana <mharidev at qualcomm.com> wrote:
>>> In other words, I want the generated code to look like this:
>>>     /* bp::class_< baseB >( "baseB" );  - Do not want to generate */
>>>
>>>     bp::class_< baseA >( "baseA" );
>>>
>>>     bp::class_< derivedC, bp::bases< baseA, baseB > >( "derivedC" );

Did you read the following document
http://language-binding.net/pyplusplus/documentation/multi_module_development.html
?

I think, it will give you easy and clear way to achieve what you want.

HTH

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


From mattias.brandstrom at todos.se  Wed Sep 17 09:58:14 2008
From: mattias.brandstrom at todos.se (=?iso-8859-1?Q?Mattias_Br=E4ndstr=F6m?=)
Date: Wed, 17 Sep 2008 09:58:14 +0200
Subject: [C++-sig] Attributes of type char[x]...
Message-ID: <B48E5D5E8E1DDE4D816167B851EC4435722AC1@exchange.mejl.ecode.se>

Hello!

I'm trying to expose a C++ (using Boost.Python) class containing public char[] members to python. I would have guessed that I would be able to do this using def_readwrite(). However, I get a compilation error when trying to do so. Here is a minimal example of what I'm trying to do:

#include <boost/python.hpp>

class A
{
public:
    char foo[10];
};

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

    class_<A>("A")
        .def_readwrite("foo", &A::foo)
    ;
}

This is the compilation error message:
c:\3pty\boost_1_35_0\boost\python\data_members.hpp(64) : error C2440: '=' : cannot convert from 'const char [10]' to 'char [10]'

I have not used the Boost.Python library for that long so there might be something obvious I'm missing. Any tips will be greatly appreciated!

Regards,
Mattias


From roman.yakovenko at gmail.com  Wed Sep 17 10:19:25 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 17 Sep 2008 11:19:25 +0300
Subject: [C++-sig] Attributes of type char[x]...
In-Reply-To: <B48E5D5E8E1DDE4D816167B851EC4435722AC1@exchange.mejl.ecode.se>
References: <B48E5D5E8E1DDE4D816167B851EC4435722AC1@exchange.mejl.ecode.se>
Message-ID: <7465b6170809170119u551eede2tec343e45703d533a@mail.gmail.com>

On Wed, Sep 17, 2008 at 10:58 AM, Mattias Br?ndstr?m
<mattias.brandstrom at todos.se> wrote:
> Hello!
>
> I'm trying to expose a C++ (using Boost.Python) class containing public char[] members to python. I would have guessed that I would be able to do this using def_readwrite(). However, I get a compilation error when trying to do so. Here is a minimal example of what I'm trying to do:
>
> #include <boost/python.hpp>
>
> class A
> {
> public:
>    char foo[10];
> };

You have 3 approaches to expose such code:

* to create a wrapper like this
  http://pygccxml.svn.sourceforge.net/viewvc/pygccxml/pyplusplus_dev/pyplusplus/code_repository/array_1.py?revision=1097&view=markup
  and the usage code attached

* to use ctypes module and expose "foo" address

* to use Py++ and have both possibilities :-)

HTH

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: usage.cpp
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080917/d62c67d1/attachment.txt>

From ope-devel at gmx.de  Thu Sep 18 07:50:03 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Thu, 18 Sep 2008 07:50:03 +0200
Subject: [C++-sig] [boost.python] Debug build on win32 runtime trouble
Message-ID: <48D1EC0B.5030405@gmx.de>

Hi,

I run into trouble on my debug build on windows, which works fine on 
linux. The interesting point is, that it does crash in python.dll and 
not python_d.dll.

Depency walker shows:

d:\work\...\build\debug\PYTHON25_D.DLL
d:\work\...\build\debug\BOOST_PYTHON-VC80-MT-GD-1_35.DLL
+ c:\windows\system32\PYTHON25.DLL

Here we are.

I've used cmake and compile flags:

/Od /I .... /D "WIN32" /D "_WINDOWS" /D "_DEBUG" /D 
"BOOST_ENABLE_ASSERT_HANDLER=1" /D "BOOST_ALL_DYN_LINK" /D 
"BOOST_LIB_DIAGNOSTIC" /D "BOOST_DEBUG_PYTHON" /D "QT_DLL" /D 
"QT_GUI_LIB" /D "QT_CORE_LIB" /D "QT_DEBUG" /D "CMAKE_INTDIR=\"Debug\"" 
/D "_MBCS" /FD /EHsc /RTC1 /MDd /Fo"eda.dir\Debug\\" 
/Fd"D:/work/.../build/Debug/eda.pdb" /W3 /nologo /c /Zi /TP 
/errorReport:prompt

Note the BOOST_DEBUG_PYTHON define. As linker flags I've used:

/OUT:"D:\work\...\build\Debug\eda.exe" /VERSION:0.0 /INCREMENTAL /NOLOGO 
/LIBPATH:"optimized\Debug" /LIBPATH:"optimized" 
/LIBPATH:"C:\Cpp\boost_1_35_0\lib\boost_iostreams-vc80-mt-1_35.lib\Debug" 
/LIBPATH:"C:\Cpp\boost_1_35_0\lib\boost_iostreams-vc80-mt-1_35.lib" 
/LIBPATH:"debug\Debug" /LIBPATH:"debug" 
/LIBPATH:"C:\Cpp\boost_1_35_0\lib\Debug" 
/LIBPATH:"C:\Cpp\boost_1_35_0\lib" 
/LIBPATH:"C:\Programme\Python25\libs\Debug" /MANIFEST 
/MANIFESTFILE:"eda.dir\Debug\eda.exe.intermediate.manifest" /DEBUG 
/PDB:"D:/work/.../build/Debug/eda.pdb" /SUBSYSTEM:WINDOWS 
/IMPLIB:"D:\work\...\build\Debug\eda.lib" /ERRORREPORT:PROMPT 
kernel32.lib user32.lib gdi32.lib winspool.lib shell32.lib ole32.lib 
oleaut32.lib uuid.lib comdlg32.lib advapi32.lib 
C:\Cpp\qt-win-opensource-4.4.1-msvc2005\lib\qtmaind.lib 
C:\Cpp\qt-win-opensource-4.4.1-msvc2005\lib\QtGuid4.lib imm32.lib 
winmm.lib C:\Cpp\qt-win-opensource-4.4.1-msvc2005\lib\QtCored4.lib 
ws2_32.lib python25_d.lib 
C:\Cpp\boost_1_35_0\lib\boost_python-vc80-mt-gd-1_35.lib 
C:\Cpp\boost_1_35_0\lib\boost_program_options-vc80-mt-gd-1_35.lib 
C:\Cpp\boost_1_35_0\lib\boost_thread-vc80-mt-gd-1_35.lib 
C:\Cpp\boost_1_35_0\lib\boost_filesystem-vc80-mt-gd-1_35.lib 
C:\Cpp\boost_1_35_0\lib\boost_iostreams-vc80-mt-gd-1_35.lib

for MS VS 2005 (MSVC 8.0)

Did I miss something on win32??

Thanks,
Olaf


From william.marie at gmail.com  Thu Sep 18 14:21:38 2008
From: william.marie at gmail.com (=?UTF-8?Q?William_Mari=C3=A9?=)
Date: Thu, 18 Sep 2008 05:21:38 -0700 (PDT)
Subject: [C++-sig]  How to use BOOST_PYTHON_OPAQUE_SPECIALIZED_TYPE_ID ?
Message-ID: <19551601.post@talk.nabble.com>


Hi,

I would need to create an opaque boost python type, i had a look on the
internet, here is what i found for instance to declare a void* opaque type :

struct void_ {};
BOOST_PYTHON_OPAQUE_SPECIALIZED_TYPE_ID(void_);

This works well if you have to call C++ function from Python but how do i do
if i want to call a python function from C++ with a void* parameter :

ex :
void* pFoo = 0x78985465;
object callable = ....;
callable( (void_*)pFoo );

I can't find out how to make this work properly i always have a runtime
error :
TypeError: No to_python (by-value) converter found for C++ type: struct
void_ *

Thanks for your help

William




-- 
View this message in context: http://www.nabble.com/How-to-use-BOOST_PYTHON_OPAQUE_SPECIALIZED_TYPE_ID---tp19551601p19551601.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From merlin66b at gmail.com  Thu Sep 18 20:44:03 2008
From: merlin66b at gmail.com (Thomas Berg)
Date: Thu, 18 Sep 2008 20:44:03 +0200
Subject: [C++-sig] [boost.python] Debug build on win32 runtime trouble
In-Reply-To: <48D1EC0B.5030405@gmx.de>
References: <48D1EC0B.5030405@gmx.de>
Message-ID: <c472dffc0809181144u435e0cd5wdc05fd087c9bcc3e@mail.gmail.com>

On Thu, Sep 18, 2008 at 7:50 AM, Olaf Peter <ope-devel at gmx.de> wrote:
> Hi,
>
> I run into trouble on my debug build on windows, which works fine on linux.
> The interesting point is, that it does crash in python.dll and not
> python_d.dll.
>
> Depency walker shows:
>
> d:\work\...\build\debug\PYTHON25_D.DLL
> d:\work\...\build\debug\BOOST_PYTHON-VC80-MT-GD-1_35.DLL
> + c:\windows\system32\PYTHON25.DLL
>
> Here we are.
>
> I've used cmake and compile flags:
>
> /Od /I .... /D "WIN32" /D "_WINDOWS" /D "_DEBUG" /D
> "BOOST_ENABLE_ASSERT_HANDLER=1" /D "BOOST_ALL_DYN_LINK" /D
> "BOOST_LIB_DIAGNOSTIC" /D "BOOST_DEBUG_PYTHON" /D "QT_DLL" /D "QT_GUI_LIB"
> /D "QT_CORE_LIB" /D "QT_DEBUG" /D "CMAKE_INTDIR=\"Debug\"" /D "_MBCS" /FD
> /EHsc /RTC1 /MDd /Fo"eda.dir\Debug\\" /Fd"D:/work/.../build/Debug/eda.pdb"
> /W3 /nologo /c /Zi /TP /errorReport:prompt
>
> Note the BOOST_DEBUG_PYTHON define. As linker flags I've used:
>
> /OUT:"D:\work\...\build\Debug\eda.exe" /VERSION:0.0 /INCREMENTAL /NOLOGO
> /LIBPATH:"optimized\Debug" /LIBPATH:"optimized"
> /LIBPATH:"C:\Cpp\boost_1_35_0\lib\boost_iostreams-vc80-mt-1_35.lib\Debug"
> /LIBPATH:"C:\Cpp\boost_1_35_0\lib\boost_iostreams-vc80-mt-1_35.lib"
> /LIBPATH:"debug\Debug" /LIBPATH:"debug"
> /LIBPATH:"C:\Cpp\boost_1_35_0\lib\Debug" /LIBPATH:"C:\Cpp\boost_1_35_0\lib"
> /LIBPATH:"C:\Programme\Python25\libs\Debug" /MANIFEST
> /MANIFESTFILE:"eda.dir\Debug\eda.exe.intermediate.manifest" /DEBUG
> /PDB:"D:/work/.../build/Debug/eda.pdb" /SUBSYSTEM:WINDOWS
> /IMPLIB:"D:\work\...\build\Debug\eda.lib" /ERRORREPORT:PROMPT kernel32.lib
> user32.lib gdi32.lib winspool.lib shell32.lib ole32.lib oleaut32.lib
> uuid.lib comdlg32.lib advapi32.lib
> C:\Cpp\qt-win-opensource-4.4.1-msvc2005\lib\qtmaind.lib
> C:\Cpp\qt-win-opensource-4.4.1-msvc2005\lib\QtGuid4.lib imm32.lib winmm.lib
> C:\Cpp\qt-win-opensource-4.4.1-msvc2005\lib\QtCored4.lib ws2_32.lib
> python25_d.lib C:\Cpp\boost_1_35_0\lib\boost_python-vc80-mt-gd-1_35.lib
> C:\Cpp\boost_1_35_0\lib\boost_program_options-vc80-mt-gd-1_35.lib
> C:\Cpp\boost_1_35_0\lib\boost_thread-vc80-mt-gd-1_35.lib
> C:\Cpp\boost_1_35_0\lib\boost_filesystem-vc80-mt-gd-1_35.lib
> C:\Cpp\boost_1_35_0\lib\boost_iostreams-vc80-mt-gd-1_35.lib
>
> for MS VS 2005 (MSVC 8.0)
>
> Did I miss something on win32??
>
> Thanks,
> Olaf
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>


To get a pure python-debug build you need to explicitly tell boost to
link to the debug version of python, by passing the option
python-debugging=on to bjam when building boost. You will ghen get
boost_python-vc80-mt-gyd-1_35.lib and .dll instead, and need to link
to that.

You also need to define BOOST_DEBUG_PYTHON when you build your
application, see this url:
http://www.boost.org/doc/libs/1_36_0/libs/python/doc/building.html#python-debugging-builds

Hope this helps!
Thomas


From s_sourceforge at nedprod.com  Thu Sep 18 21:24:59 2008
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Thu, 18 Sep 2008 20:24:59 +0100
Subject: [C++-sig] How to use BOOST_PYTHON_OPAQUE_SPECIALIZED_TYPE_ID ?
In-Reply-To: <19551601.post@talk.nabble.com>
References: <19551601.post@talk.nabble.com>
Message-ID: <48D2B91B.13559.16935278@s_sourceforge.nedprod.com>

On 18 Sep 2008 at 5:21, William Mari? wrote:

> I would need to create an opaque boost python type, i had a look on the
> internet, here is what i found for instance to declare a void* opaque type :
> 
> struct void_ {};
> BOOST_PYTHON_OPAQUE_SPECIALIZED_TYPE_ID(void_);

No, that's trickery from before my 2005 patch implementing native 
void * support.

> This works well if you have to call C++ function from Python but how do i do
> if i want to call a python function from C++ with a void* parameter :
> 
> ex :
> void* pFoo = 0x78985465;
> object callable = ....;
> callable( (void_*)pFoo );
> 
> I can't find out how to make this work properly i always have a runtime
> error :
> TypeError: No to_python (by-value) converter found for C++ type: struct
> void_ *

It just automatically works as an opaque pointer. Python cannot 
access the value, but can pass it into functions.

There should be an example in the regression suite.

Cheers,
Niall





From ope-devel at gmx.de  Fri Sep 19 07:21:30 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Fri, 19 Sep 2008 07:21:30 +0200
Subject: [C++-sig] [boost.python] Debug build on win32 runtime trouble
In-Reply-To: <c472dffc0809181144u435e0cd5wdc05fd087c9bcc3e@mail.gmail.com>
References: <48D1EC0B.5030405@gmx.de>
	<c472dffc0809181144u435e0cd5wdc05fd087c9bcc3e@mail.gmail.com>
Message-ID: <48D336DA.5000005@gmx.de>

>> Depency walker shows:
>>
>> d:\work\...\build\debug\PYTHON25_D.DLL
>> d:\work\...\build\debug\BOOST_PYTHON-VC80-MT-GD-1_35.DLL
>> + c:\windows\system32\PYTHON25.DLL
[...]
>> Note the BOOST_DEBUG_PYTHON define. As linker flags I've used:

> To get a pure python-debug build you need to explicitly tell boost to
> link to the debug version of python, by passing the option
> python-debugging=on to bjam when building boost. You will ghen get
> boost_python-vc80-mt-gyd-1_35.lib and .dll instead, and need to link
> to that.
> 
> You also need to define BOOST_DEBUG_PYTHON when you build your
> application, see this url:
> http://www.boost.org/doc/libs/1_36_0/libs/python/doc/building.html#python-debugging-builds

As first thank you for the answer. Anyway, I assume the debug build of 
boost_python dll ist wrong and points to the wrong python25.dll, should 
be python25_d.dll. Could it be? I've used the boostpro windows installer 
1.35.0.

Thanks,
Olaf

PS: My code works without defined BOOST_DEBUG_PYTHON; my crash results 
really from different memory layout of the python25{_d}.dll used by 
BOOST_PYTHON-VC80-MT-GD-1_35.DLL.


From dave at boostpro.com  Fri Sep 19 08:16:56 2008
From: dave at boostpro.com (David Abrahams)
Date: Thu, 18 Sep 2008 23:16:56 -0700
Subject: [C++-sig] [boost.python] Debug build on win32 runtime trouble
In-Reply-To: <48D336DA.5000005@gmx.de> (Olaf Peter's message of "Fri, 19 Sep
	2008 07:21:30 +0200")
References: <48D1EC0B.5030405@gmx.de>
	<c472dffc0809181144u435e0cd5wdc05fd087c9bcc3e@mail.gmail.com>
	<48D336DA.5000005@gmx.de>
Message-ID: <87zlm4mzhj.fsf@mcbain.luannocracy.com>


on Thu Sep 18 2008, Olaf Peter <ope-devel-AT-gmx.de> wrote:

>>> Depency walker shows:
>>>
>>> d:\work\...\build\debug\PYTHON25_D.DLL
>>> d:\work\...\build\debug\BOOST_PYTHON-VC80-MT-GD-1_35.DLL
>>> + c:\windows\system32\PYTHON25.DLL
> [...]
>>> Note the BOOST_DEBUG_PYTHON define. As linker flags I've used:
>
>> To get a pure python-debug build you need to explicitly tell boost to
>> link to the debug version of python, by passing the option
>> python-debugging=on to bjam when building boost. You will ghen get
>> boost_python-vc80-mt-gyd-1_35.lib and .dll instead, and need to link
>> to that.
>>
>> You also need to define BOOST_DEBUG_PYTHON when you build your
>> application, see this url:
>>
> http://www.boost.org/doc/libs/1_36_0/libs/python/doc/building.html#python-debugging-builds
>
> As first thank you for the answer. Anyway, I assume the debug build of
> boost_python dll ist wrong and points to the wrong python25.dll, should be
> python25_d.dll. Could it be? 

No, the build is right, i.e. intentional.  Please read the URL as
suggested by Olaf, it really does explain everything.  Also please
follow the advice in

http://www.boost.org/doc/libs/1_36_0/libs/python/doc/building.html#no-install-quickstart

for people like you who want to use build systems other than bjam.  It
will potentially save you a great deal of pain.

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


From renjipanicker at gmail.com  Fri Sep 19 12:35:14 2008
From: renjipanicker at gmail.com (Renji Panicker)
Date: Fri, 19 Sep 2008 16:05:14 +0530
Subject: [C++-sig] Returning None from __getitem__
Message-ID: <3c0065cc0809190335s7219b866hada4f7d71eb48458@mail.gmail.com>

Hi all,

I am using boost.python to wrap a map class. My get function looks
like this (edited):

std::map<std::wstring, std::wstring> _values;

class PropSet
{
    std::wstring get(const std::wstring& key)
    {
        if(_values.find(key) == _values.end())
            throw InvalidKeyException();
        return _values[key];
    }
};

In my python code, I would like to be able to do this:
if ps["ssl"] == "yes":
     print "ssl enabled"

My requirement in brief is, if the key 'ssl' exists in the map, and
its value is 'yes', turn on SSL. If the key does not exist, or has any
value other than 'yes', do not turn SSL on. To that extent, I would
like to return None from the get function if the key does not exist,
instead of throwing an exception.

My obvious workaround is to use an empty string, but I am expecting
use cases where the empty string can be a valid value for a key. So I
will have to consider using some other string which cannot be a
potential value, to indicate an empty value. But before I go with this
workaround, I would like to know if there is any standard way of
achieving this.

Thanks in advance.
-/renji


From meine at informatik.uni-hamburg.de  Fri Sep 19 13:11:40 2008
From: meine at informatik.uni-hamburg.de (Hans Meine)
Date: Fri, 19 Sep 2008 13:11:40 +0200
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: <200809191311.41721.meine@informatik.uni-hamburg.de>

On Thursday 21 August 2008 14:32:00 Andreas Eriksson wrote:
> 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.
That's a very common need and has been discussed on this list before.

I have done that in the past (and posted on it); basically I used SIP's 
functions to (un)wrap the C++ pointer within my BPL code.  Then, I simply 
returned the PyObject * and exported that using B::P.

Please try to find previous discussions, if you don't find anything I can try.

HTH,
  Hans


From seefeld at sympatico.ca  Fri Sep 19 14:07:46 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Fri, 19 Sep 2008 08:07:46 -0400
Subject: [C++-sig] Returning None from __getitem__
In-Reply-To: <3c0065cc0809190335s7219b866hada4f7d71eb48458@mail.gmail.com>
References: <3c0065cc0809190335s7219b866hada4f7d71eb48458@mail.gmail.com>
Message-ID: <48D39612.3060303@sympatico.ca>

Hi Renji,

 From a language / API design point of view, I think it would be best 
(i.e., least surprising) to follow the expected behavior. Python's dict 
type does have ways to handle what you want (and you can even set 
default values to be returned for non-existent keys).

As for your particular case, where your C++ map throws an exception, but 
you want the wrapper to return None, that's simple, too:

boost::python::object my_get_wrapper(PropSet &p, std::wstring const &k)
{
   try { return p->get(k);} // implicitly convert valid value to python 
object
   catch (InvalidKeyException const &) { return 
boost::python::object();} // on failure, return None
}

Then, you bind that as 'get':

class_<PropSet, ...> (...)
    .def("get", my_get_wrapper);


HTH,
       Stefan

-- 

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



From aeriksson at gmail.com  Fri Sep 19 17:01:45 2008
From: aeriksson at gmail.com (Andreas Eriksson)
Date: Fri, 19 Sep 2008 17:01:45 +0200
Subject: [C++-sig] pybindgen and PyQt
In-Reply-To: <200809191311.41721.meine@informatik.uni-hamburg.de>
References: <1f4b35e00808210006r66135dcfr1d52678adc0eac1@mail.gmail.com>
	<a467ca4f0808210447m6f4ee44fj7c47132131d5f082@mail.gmail.com>
	<1f4b35e00808210532g252d2df0mf4885bc972feda66@mail.gmail.com>
	<200809191311.41721.meine@informatik.uni-hamburg.de>
Message-ID: <1f4b35e00809190801s1309df26h2bc23f78dd6f25c3@mail.gmail.com>

> > 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.
> That's a very common need and has been discussed on this list before.
>
> I have done that in the past (and posted on it); basically I used SIP's
> functions to (un)wrap the C++ pointer within my BPL code.  Then, I simply
> returned the PyObject * and exported that using B::P.
>
> Please try to find previous discussions, if you don't find anything I can try.
Thanks,

are you referring to the following from
http://mail.python.org/pipermail/c++-sig/2005-November/009865.html?
(Included at bottom of post)

I'll have a look at this and probably come back with more questions...

Regards,

Andreas

> > > Can I tell boost to use qt objects from PyQt instead of generating it's
> > > own version of them?
> >
> > That's probably something you'd have to figure out yourself. A few months
> > ago someone asked a similar question regarding SWIG. It was surprisingly
> > easy to support passing SWIG-wrapped objects to Boost.Python wrapped
> > functions:
> >
> > http://cvs.sourceforge.net/viewcvs.py/cctbx/boost_adaptbx/swig_arg.hpp?view
> >=markup
> >
> > http://cvs.sourceforge.net/viewcvs.py/cctbx/boost_adaptbx/swig_args_ext.cpp
> >?view=markup
> >
> > If you are lucky, maybe something similar works for SIP. You'd have to find
> > out how the SIP wrapping mechanism works.
>
> Yes, it is possible; for example I did it with code fragments like these (this
> is also a hint for Jan Langer how such a thing could look like):
>
> #include <sipvigraqtDeclvigraqt.h> // my SIP module was called vigraqt
>
> PyObject *getViewer()
> {
>     return sipMapCppToSelfSubClass(
>         somehowGetViewer(), sipClass_QImageViewer);
> }
>
> BOOST_PYTHON_MODULE_INIT(sipbridge)
> {
>     def("viewer", &getViewer);
> }

Btw, the URLs above seem to have changed into the following:

http://cctbx.svn.sourceforge.net/viewvc/cctbx/trunk/boost_adaptbx/swig_arg.hpp?view=markup

http://cctbx.svn.sourceforge.net/viewvc/cctbx/trunk/boost_adaptbx/swig_args_ext.cpp?view=markup


From hans_meine at gmx.net  Fri Sep 19 20:20:48 2008
From: hans_meine at gmx.net (Hans Meine)
Date: Fri, 19 Sep 2008 20:20:48 +0200
Subject: [C++-sig] pybindgen and PyQt
In-Reply-To: <1f4b35e00809190801s1309df26h2bc23f78dd6f25c3@mail.gmail.com>
References: <1f4b35e00808210006r66135dcfr1d52678adc0eac1@mail.gmail.com>
	<200809191311.41721.meine@informatik.uni-hamburg.de>
	<1f4b35e00809190801s1309df26h2bc23f78dd6f25c3@mail.gmail.com>
Message-ID: <200809192020.58995.hans_meine@gmx.net>

On Freitag 19 September 2008, Andreas Eriksson wrote:
> are you referring to the following from
> http://mail.python.org/pipermail/c++-sig/2005-November/009865.html?
> [...]
> > #include <sipvigraqtDeclvigraqt.h> // my SIP module was called vigraqt
> >
> > PyObject *getViewer()
> > {
> >     return sipMapCppToSelfSubClass(
> >         somehowGetViewer(), sipClass_QImageViewer);
> > }
> >
> > BOOST_PYTHON_MODULE_INIT(sipbridge)
> > {
> >     def("viewer", &getViewer);
> > }

Correct, that's how I did the to-python conversion.  Of course, the C++ object 
must not be destroyed before the python one, but I am not so sure about the 
lifecycle management in SIP and at least this is a working start.  Probably 
you cannot even do much better.

-- 
Ciao, /  /                                                    .o.
     /--/                                                     ..o
    /  / ANS                                                  ooo
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080919/d3d0b51b/attachment.pgp>

From agrimstrup at gmail.com  Fri Sep 19 22:37:34 2008
From: agrimstrup at gmail.com (Arne Grimstrup)
Date: Fri, 19 Sep 2008 14:37:34 -0600
Subject: [C++-sig] .def( bp::self == bp::self );
 causes compile problems for Py++ generated wrapper
Message-ID: <48D40D8E.4040604@gmail.com>

I'm having some trouble with Py++ generating a wrapper for a class
that defines a == operator.  The class definition looks like this:

class MyFrac {
   private:
     int numerator;
     int denominator;
     friend std::ostream& operator<<(std::ostream&, MyFrac&);

   public:
     MyFrac(void) : numerator(1), denominator(1) {}
     MyFrac(int n, int d) : numerator(n), denominator(d) {}
     bool operator==(MyFrac& b) const {
        return numerator == b.numerator && denominator == b.denominator;
     }	
};

When the code is generated using pyplusplus_gui, the resulting C++ bindings
look like this:

#include "boost/python.hpp"

#include "MyFrac.h"

namespace bp = boost::python;

BOOST_PYTHON_MODULE(MyFrac){
     bp::class_< MyFrac >( "MyFrac" )
         .def( bp::init< >() )
         .def( bp::init< int, int >(( bp::arg("n"), bp::arg("d") )) )
         .def( bp::self == bp::self );
}

Attempting to compile the bindings gives:

operators.hpp:226: error: no match for ?operator==? in ?l == r?
MyFrac.h:14: note: candidates are: bool MyFrac::operator==(MyFrac&) const
function/function_base.hpp:577: note:
    bool boost::operator==(const boost::function_base&,
                           boost::detail::function::useless_clear_type*)
function/function_base.hpp:589: note:
    bool boost::operator==(boost::detail::function::useless_clear_type*,
                         const boost::function_base&)

Removing the line ".def( bp::self == bp::self );" from the bindings
allows it to compile cleanly and the resulting code works correctly
from what I can tell.

Could anyone shed light on what I've done wrong and how to fix the
error?

Thanks,

Arne


From wladwig at wdtinc.com  Fri Sep 19 23:37:09 2008
From: wladwig at wdtinc.com (William Ladwig)
Date: Fri, 19 Sep 2008 16:37:09 -0500
Subject: [C++-sig] .def( bp::self == bp::self );
 causes compile problems for Py++ generated wrapper
In-Reply-To: <48D40D8E.4040604@gmail.com>
References: <48D40D8E.4040604@gmail.com>
Message-ID: <765CBD9053EA2B438625895F30D1856FCBC2C537@storm.wdtinc.com>

If you can change the C++ code, try declaring the operator== argument as a const reference (should probably do this anyhow since you aren't modifying the object):

bool operator==(const MyFrac& b) const {
        return numerator == b.numerator && denominator == b.denominator;}


Hope this helps,
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 Arne Grimstrup
Sent: Friday, September 19, 2008 3:38 PM
To: c++-sig at python.org
Subject: [C++-sig] .def( bp::self == bp::self ); causes compile problems for Py++ generated wrapper

I'm having some trouble with Py++ generating a wrapper for a class
that defines a == operator.  The class definition looks like this:

class MyFrac {
   private:
     int numerator;
     int denominator;
     friend std::ostream& operator<<(std::ostream&, MyFrac&);

   public:
     MyFrac(void) : numerator(1), denominator(1) {}
     MyFrac(int n, int d) : numerator(n), denominator(d) {}
     bool operator==(MyFrac& b) const {
        return numerator == b.numerator && denominator == b.denominator;
     }
};

When the code is generated using pyplusplus_gui, the resulting C++ bindings
look like this:

#include "boost/python.hpp"

#include "MyFrac.h"

namespace bp = boost::python;

BOOST_PYTHON_MODULE(MyFrac){
     bp::class_< MyFrac >( "MyFrac" )
         .def( bp::init< >() )
         .def( bp::init< int, int >(( bp::arg("n"), bp::arg("d") )) )
         .def( bp::self == bp::self );
}

Attempting to compile the bindings gives:

operators.hpp:226: error: no match for 'operator==' in 'l == r'
MyFrac.h:14: note: candidates are: bool MyFrac::operator==(MyFrac&) const
function/function_base.hpp:577: note:
    bool boost::operator==(const boost::function_base&,
                           boost::detail::function::useless_clear_type*)
function/function_base.hpp:589: note:
    bool boost::operator==(boost::detail::function::useless_clear_type*,
                         const boost::function_base&)

Removing the line ".def( bp::self == bp::self );" from the bindings
allows it to compile cleanly and the resulting code works correctly
from what I can tell.

Could anyone shed light on what I've done wrong and how to fix the
error?

Thanks,

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


From jean-sebastien.guay at cm-labs.com  Sat Sep 20 06:24:14 2008
From: jean-sebastien.guay at cm-labs.com (=?ISO-8859-1?Q?Jean-S=E9bastien_Guay?=)
Date: Sat, 20 Sep 2008 00:24:14 -0400
Subject: [C++-sig] Wrapping a class whose constructor takes a pointer to an
 abstract class
Message-ID: <48D47AEE.6050405@cm-labs.com>

Hi all,

I've just started trying to use boost.python on a personal project of 
mine, and I decided to start with the hardest part :-)  I'm trying to 
wrap OpenSceneGraph... It's going really well for the most part, and I 
do mean really well. In little time I've got a good number of classes 
working and tested. But I've hit a roadblock.

I've done a lot of searching and have found lots of posts on this list 
dealing with pure virtual methods, but never anything about a method 
taking a pointer to an abstract class as a parameter.

I have to wrap the following structure:

     class osg::Shape : public osg::Object
     {
         // has some pure virtual methods, none of which I need to
         // expose to python at this point
     };

     class osg::Sphere : public osg::Shape
     {
         // fully concrete
     };

     class osg::ShapeDrawable
     {
         ShapeDrawable(osg::Shape*, TessellationHints) { /* ... */ }
     };

So the constructor to ShapeDrawable takes a pointer to the abstract base 
class. My attempt to wrap this (leaving out some details) is:

     // Abstract class, so noncopyable and no_init.
     class_<Shape, bases<Object>, ref_ptr<Shape>,
            boost::noncopyable >("Shape", no_init);

     // Concrete derived class.
     class_<Sphere, bases<Shape>, ref_ptr<Sphere> >("Sphere");

     // Details of this class not important for now.
     class_<TessellationHints, bases<Object>,
            ref_ptr<TessellationHints> >("TessellationHints");

     class_<ShapeDrawable, bases<Drawable>,
            ref_ptr<ShapeDrawable> >("ShapeDrawable")
         .def(init<Shape, TessellationHints>())
     ;

The first three classes wrap correctly, but the fourth fails with a 
compiler error saying it cannot instantiate abstract class Shape. OK, I 
understand that, but what can I do about it?

Thanks in advance for any help,

J-S
-- 
______________________________________________________
Jean-Sebastien Guay    jean-sebastien.guay at cm-labs.com
                                http://www.cm-labs.com/
                         http://whitestar02.webhop.org/


From renjipanicker at gmail.com  Sat Sep 20 13:08:03 2008
From: renjipanicker at gmail.com (Renji Panicker)
Date: Sat, 20 Sep 2008 16:38:03 +0530
Subject: [C++-sig] Returning None from __getitem__
In-Reply-To: <48D39612.3060303@sympatico.ca>
References: <3c0065cc0809190335s7219b866hada4f7d71eb48458@mail.gmail.com>
	<48D39612.3060303@sympatico.ca>
Message-ID: <3c0065cc0809200408q2fd7aa6dnaf8604cf11c57c30@mail.gmail.com>

Exactly what I wanted! Thanks a lot, Stefan!

On Fri, Sep 19, 2008 at 5:37 PM, Stefan Seefeld <seefeld at sympatico.ca> wrote:
> Hi Renji,
>
> From a language / API design point of view, I think it would be best (i.e.,
> least surprising) to follow the expected behavior. Python's dict type does
> have ways to handle what you want (and you can even set default values to be
> returned for non-existent keys).
>
> As for your particular case, where your C++ map throws an exception, but you
> want the wrapper to return None, that's simple, too:
>
> boost::python::object my_get_wrapper(PropSet &p, std::wstring const &k)
> {
>  try { return p->get(k);} // implicitly convert valid value to python object
>  catch (InvalidKeyException const &) { return boost::python::object();} //
> on failure, return None
> }
>
> Then, you bind that as 'get':
>
> class_<PropSet, ...> (...)
>   .def("get", my_get_wrapper);
>
>
> 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 merlin66b at gmail.com  Sat Sep 20 15:57:35 2008
From: merlin66b at gmail.com (Thomas Berg)
Date: Sat, 20 Sep 2008 15:57:35 +0200
Subject: [C++-sig] [boost.python] Debug build on win32 runtime trouble
In-Reply-To: <87zlm4mzhj.fsf@mcbain.luannocracy.com>
References: <48D1EC0B.5030405@gmx.de>
	<c472dffc0809181144u435e0cd5wdc05fd087c9bcc3e@mail.gmail.com>
	<48D336DA.5000005@gmx.de> <87zlm4mzhj.fsf@mcbain.luannocracy.com>
Message-ID: <c472dffc0809200657q338f3626qfd36f28b738d4a3f@mail.gmail.com>

On Fri, Sep 19, 2008 at 7:21 AM, Olaf Peter <ope-devel at gmx.de> wrote:
> As first thank you for the answer. Anyway, I assume the debug build of
> boost_python dll ist wrong and points to the wrong python25.dll, should be
> python25_d.dll. Could it be? I've used the boostpro windows installer
> 1.35.0.
>
> Thanks,
> Olaf
>
> PS: My code works without defined BOOST_DEBUG_PYTHON; my crash results
> really from different memory layout of the python25{_d}.dll used by
> BOOST_PYTHON-VC80-MT-GD-1_35.DLL.


Sorry Olaf, I assumed that you had built boost yourself, as I always
do... As David says this is the intentional default debug variant of
boost python. I don't know if the boostpro installer is capable of
installing the proper python-debug versions of the libraries. If not,
I guess you will have to build it yourself with bjam, passing
python-debugging=on. Notice that you get a different name: "gyd"
instead of "gd" in the lib / dll name.

Believe me, you do want to define BOOST_DEBUG_PYTHON when building
your application, otherwise one of the boost python headers will
undefine _DEBUG before including "python.h". See
boost/python/detail/wrap_python.hpp, or the documentation. It might
accidentaly work if you "illegaly" include python.h yourself.

You may also want to disable the autolinking of boost, since it
doesn't handle the python debug naming scheme (at least in earlier
versions of boost it doesn't), and pulls in the wrong library. Do that
by defining BOOST_ALL_NO_LIB in your project.

Thomas


From ope-devel at gmx.de  Sat Sep 20 19:50:22 2008
From: ope-devel at gmx.de (Olaf Peter)
Date: Sat, 20 Sep 2008 19:50:22 +0200
Subject: [C++-sig] [boost.python] Debug build on win32 runtime trouble
In-Reply-To: <c472dffc0809200657q338f3626qfd36f28b738d4a3f@mail.gmail.com>
References: <48D1EC0B.5030405@gmx.de>	<c472dffc0809181144u435e0cd5wdc05fd087c9bcc3e@mail.gmail.com>	<48D336DA.5000005@gmx.de>
	<87zlm4mzhj.fsf@mcbain.luannocracy.com>
	<c472dffc0809200657q338f3626qfd36f28b738d4a3f@mail.gmail.com>
Message-ID: <48D537DE.80506@gmx.de>

Thomas Berg schrieb:
> On Fri, Sep 19, 2008 at 7:21 AM, Olaf Peter <ope-devel at gmx.de> wrote:
>> As first thank you for the answer. Anyway, I assume the debug build of
>> boost_python dll ist wrong and points to the wrong python25.dll, should be
>> python25_d.dll. Could it be? I've used the boostpro windows installer
>> 1.35.0.
>>
>> Thanks,
>> Olaf
>>
>> PS: My code works without defined BOOST_DEBUG_PYTHON; my crash results
>> really from different memory layout of the python25{_d}.dll used by
>> BOOST_PYTHON-VC80-MT-GD-1_35.DLL.
> 
> 
> Sorry Olaf, I assumed that you had built boost yourself, as I always
> do... As David says this is the intentional default debug variant of
> boost python. I don't know if the boostpro installer is capable of
> installing the proper python-debug versions of the libraries. If not,
> I guess you will have to build it yourself with bjam, passing
> python-debugging=on. Notice that you get a different name: "gyd"
> instead of "gd" in the lib / dll name.
> 
> Believe me, you do want to define BOOST_DEBUG_PYTHON when building
> your application, otherwise one of the boost python headers will
> undefine _DEBUG before including "python.h". See
> boost/python/detail/wrap_python.hpp, or the documentation. It might
> accidentaly work if you "illegaly" include python.h yourself.
> 
> You may also want to disable the autolinking of boost, since it
> doesn't handle the python debug naming scheme (at least in earlier
> versions of boost it doesn't), and pulls in the wrong library. Do that
> by defining BOOST_ALL_NO_LIB in your project.

Thanks for your explanation and patience.

Regards,
Olaf


From surye80 at gmail.com  Mon Sep 22 02:49:18 2008
From: surye80 at gmail.com (Vincent Castellano)
Date: Sun, 21 Sep 2008 17:49:18 -0700
Subject: [C++-sig] Exporting Classes to Embedded Interpreter
Message-ID: <8628ae450809211749o703b1059gb02d05cd205a7f9b@mail.gmail.com>

I can create a shared library that exports C++ classes to python. And
I have the embedded interpreter working for boost:python::exec() to
call python code (as well as pulling out objects and function from
globals, and interacting with them in C.

What I am trying to is combine these techniques to enable classes
written in C++ to be accessible to the embedded interpreter, and more
to the point, instantiate instances of those classes in C++, and pass
them into the globals of python to interact with. I have tried to find
documentation on this but it eludes me. Any pointers on how to
accomplish this, or is it possible at all?

Thanks,
Vince Castellano


From seefeld at sympatico.ca  Mon Sep 22 03:24:15 2008
From: seefeld at sympatico.ca (Stefan Seefeld)
Date: Sun, 21 Sep 2008 21:24:15 -0400
Subject: [C++-sig] Exporting Classes to Embedded Interpreter
In-Reply-To: <8628ae450809211749o703b1059gb02d05cd205a7f9b@mail.gmail.com>
References: <8628ae450809211749o703b1059gb02d05cd205a7f9b@mail.gmail.com>
Message-ID: <48D6F3BF.4090404@sympatico.ca>

Vincent Castellano wrote:
> I can create a shared library that exports C++ classes to python. And
> I have the embedded interpreter working for boost:python::exec() to
> call python code (as well as pulling out objects and function from
> globals, and interacting with them in C.
>
> What I am trying to is combine these techniques to enable classes
> written in C++ to be accessible to the embedded interpreter, and more
> to the point, instantiate instances of those classes in C++, and pass
> them into the globals of python to interact with. I have tried to find
> documentation on this but it eludes me. Any pointers on how to
> accomplish this, or is it possible at all?
>   

It's easy, really: the boost::python::exec() function takes a 
dictionary, providing the environment that the script will see. You can 
push whatever you want into that dictionary, inclusively objects of 
types you previously exported to python. E.g.:

somewhere in your code:

class_<your_class>....;

somewhere else in your code:

your_class &an_object = ...;
python::object main = python::import("__main__");
python::object global(main.attr("__dict__"));
global["an_object"] = an_object; // <- this injects a python wrapper 
around 'an_object' into the 'global' namespace.
python::object::exec(your_script, global); // <- this runs a script, 
exposing 'an_object' to it.


HTH,
       Stefan

-- 

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



From renjipanicker at gmail.com  Mon Sep 22 07:20:29 2008
From: renjipanicker at gmail.com (Renji Panicker)
Date: Mon, 22 Sep 2008 10:50:29 +0530
Subject: [C++-sig] Exporting Classes to Embedded Interpreter
In-Reply-To: <8628ae450809211749o703b1059gb02d05cd205a7f9b@mail.gmail.com>
References: <8628ae450809211749o703b1059gb02d05cd205a7f9b@mail.gmail.com>
Message-ID: <3c0065cc0809212220rdea7008p21fe441985cc6149@mail.gmail.com>

You need to introduce an instance of the object into the namespace of
your main module.

1. Get reference to main module
boost::python::object
main_module((boost::python::handle<>(boost::python::borrowed(PyImport_AddModule("__main__")))));

2. Get reference to main namespace
boost::python::object main_namespace = main_module.attr("__dict__");

3. Create an instance of MyObj, which you want to expose to your Python code
MyObj myObj;

4. Insert reference to myObj into the main namespace
main_namespace["myObj"] = boost::python::ptr(&myObj);

5. Code to be executed. This code can access myObj
const char* code = "print 'hello'";

6. Execute the code
boost::python::handle<> ignored(( PyRun_String(code, Py_file_input,
main_namespace.ptr(), main_namespace.ptr() ) ));

HTH,
-/renji

On Mon, Sep 22, 2008 at 6:19 AM, Vincent Castellano <surye80 at gmail.com> wrote:
> I can create a shared library that exports C++ classes to python. And
> I have the embedded interpreter working for boost:python::exec() to
> call python code (as well as pulling out objects and function from
> globals, and interacting with them in C.
>
> What I am trying to is combine these techniques to enable classes
> written in C++ to be accessible to the embedded interpreter, and more
> to the point, instantiate instances of those classes in C++, and pass
> them into the globals of python to interact with. I have tried to find
> documentation on this but it eludes me. Any pointers on how to
> accomplish this, or is it possible at all?
>
> Thanks,
> Vince Castellano
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>


From arne at ras.ucalgary.ca  Mon Sep 22 18:28:32 2008
From: arne at ras.ucalgary.ca (Arne Grimstrup)
Date: Mon, 22 Sep 2008 10:28:32 -0600
Subject: [C++-sig] .def( bp::self == bp::self );
 causes compile problems for Py++ generated wrapper
In-Reply-To: <mailman.11111.1221884685.3486.c++-sig@python.org>
References: <mailman.11111.1221884685.3486.c++-sig@python.org>
Message-ID: <48D7C7B0.5000700@ras.ucalgary.ca>

On Friday, September 19th, William Ladwig wrote:
>
> If you can change the C++ code, try declaring the operator== argument
 > as a const reference (should probably do this anyhow since you aren't
 > modifying the object):
>
> bool operator==(const MyFrac& b) const {
>        return numerator == b.numerator && denominator == b.denominator;}
>
>
> Hope this helps,
> Bill

Thanks very much, Bill.  Adding the const reference to the argument fixed
the problem.

Regards,

Arne


From dave at boostpro.com  Mon Sep 22 19:59:49 2008
From: dave at boostpro.com (David Abrahams)
Date: Mon, 22 Sep 2008 10:59:49 -0700
Subject: [C++-sig] Wrapping a class whose constructor takes a pointer to
	an abstract class
In-Reply-To: <48D47AEE.6050405@cm-labs.com> (=?utf-8?Q?=22Jean-S=C3=A9bast?=
	=?utf-8?Q?ien?= Guay"'s message
	of "Sat, 20 Sep 2008 00:24:14 -0400")
References: <48D47AEE.6050405@cm-labs.com>
Message-ID: <87myi0f4dm.fsf@mcbain.luannocracy.com>


on Fri Sep 19 2008, Jean-S?bastien Guay <jean-sebastien.guay-AT-cm-labs.com> wrote:

>     class_<ShapeDrawable, bases<Drawable>,
>            ref_ptr<ShapeDrawable> >("ShapeDrawable")
>         .def(init<Shape, TessellationHints>())
>     ;
>
> The first three classes wrap correctly, but the fourth fails with a compiler
> error saying it cannot instantiate abstract class Shape. OK, I understand that,
> but what can I do about it?

I can't promise that this is enough to make everything work, but you
could start by making the init signature match your constructor
signature:

     class_<ShapeDrawable, bases<Drawable>,
            ref_ptr<ShapeDrawable> >("ShapeDrawable")
         .def(init<Shape*, TessellationHints>())
                        ^
here--------------------^

It will at least compile, then!

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


From jean-sebastien.guay at cm-labs.com  Mon Sep 22 20:06:59 2008
From: jean-sebastien.guay at cm-labs.com (=?UTF-8?B?SmVhbi1Tw6liYXN0aWVuIEd1YXk=?=)
Date: Mon, 22 Sep 2008 14:06:59 -0400
Subject: [C++-sig] Wrapping a class whose constructor takes a pointer
 to	an abstract class
In-Reply-To: <87myi0f4dm.fsf@mcbain.luannocracy.com>
References: <48D47AEE.6050405@cm-labs.com>
	<87myi0f4dm.fsf@mcbain.luannocracy.com>
Message-ID: <48D7DEC3.3070108@cm-labs.com>

Hi David,

> I can't promise that this is enough to make everything work, but you
> could start by making the init signature match your constructor
> signature:
[...]
> It will at least compile, then!

Hmmm, you're right of course. The only thing I can say in my defense is 
that I'm still confused as to when I need to use "python" syntax and 
when I need to use "C++" syntax... I bet there are a lot more places 
where I made the same mistake. :-)

Thanks a lot, and I'm sure I'll have many more such newbie questions in 
the near future.

J-S
-- 
______________________________________________________
Jean-Sebastien Guay    jean-sebastien.guay at cm-labs.com
                                http://www.cm-labs.com/
                         http://whitestar02.webhop.org/


From jean-sebastien.guay at cm-labs.com  Tue Sep 23 05:21:40 2008
From: jean-sebastien.guay at cm-labs.com (=?ISO-8859-1?Q?Jean-S=E9bastien_Guay?=)
Date: Mon, 22 Sep 2008 23:21:40 -0400
Subject: [C++-sig] Reference counted object return value policy
Message-ID: <48D860C4.6000300@cm-labs.com>

Hi all,

Thanks to David for pointing out my mistake earlier, I am progressing 
again on my OSG boost.python wrappers.

Now, you may or may not know that OSG uses its own reference counting 
mechanism. All ref counted objects inherit from an osg::Referenced base 
class which has the reference count, and a smart pointer template class 
ref_ptr<T> manages these counts. I've thus far been able to use pointee< 
ref_ptr<T> > as heldtype to get around the fact that destructors are 
protected, but now I want to expose the file loading method:

   osg::Node* osgDB::readNodeFile(const std::string& filename, /*...*/)

I've got this:

     def("readNodeFile", readNodeFile1, /* some return value policy */);

I'm stuck at the choice of return value policy. I guess the most logical 
would be return_value_policy<manage_new_object>(), but this gives a 
compilation error because the actual C++ return type of the function has 
a protected destructor. I also tried (in true newbie fashion) 
return_value_policy<return_by_value>(), 
return_value_policy<reference_existing_object>() and even 
return_internal_reference<>(), and all three cause a crash of the python 
interpreter (though in the second and third case I can understand why, 
after all no reference to that object is kept by the function itself).

a) Am I just looking at this the wrong way? What would be the right 
return value policy in this case?
b) I have a hunch that if I had a way to tell boost.python *how* to 
manage the new object, i.e. using a ref_ptr, that might solve my problem...
c) I imagine the only way to debug and find out *why* the python 
interpreter crashes in the other cases (other than just being not 
applicable) is to compile Python in debug and run it in the debugger? Do 
I have other options?


On a side note, there seem to be other people who have attempted to wrap 
OSG with boost.python, but if any of them succeeded, they haven't - to 
my knowledge - made their work public. This is just a "free time" 
project for me. So I won't be too disappointed if I don't end up with 
something usable, though until now I like what boost.python provides. It 
seems expressive enough to capture pretty much any C++ API, I just have 
to get to know it more.


Thanks in advance for any guidance you can provide,

J-S
-- 
______________________________________________________
Jean-Sebastien Guay    jean-sebastien.guay at cm-labs.com
                                http://www.cm-labs.com/
                         http://whitestar02.webhop.org/


From roman.yakovenko at gmail.com  Tue Sep 23 05:26:24 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Tue, 23 Sep 2008 06:26:24 +0300
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <48D860C4.6000300@cm-labs.com>
References: <48D860C4.6000300@cm-labs.com>
Message-ID: <7465b6170809222026i4d91a77bm26f1d6a6c33ffe3@mail.gmail.com>

On Tue, Sep 23, 2008 at 6:21 AM, Jean-S?bastien Guay
<jean-sebastien.guay at cm-labs.com> wrote:
> Hi all,
> ...
> On a side note, there seem to be other people who have attempted to wrap OSG
> with boost.python, but if any of them succeeded, they haven't - to my
> knowledge - made their work public. This is just a "free time" project for
> me. So I won't be too disappointed if I don't end up with something usable,
> though until now I like what boost.python provides. It seems expressive
> enough to capture pretty much any C++ API, I just have to get to know it
> more.

I am not sure whether OSG and OpenSG are same project, but working
Python wrapper for the latest one you can find here:
https://realityforge.vrsource.org/trac/pyopensg
(Of course it uses boost.python ).

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


From jean-sebastien.guay at cm-labs.com  Tue Sep 23 14:31:19 2008
From: jean-sebastien.guay at cm-labs.com (=?ISO-8859-1?Q?Jean-S=E9bastien_Guay?=)
Date: Tue, 23 Sep 2008 08:31:19 -0400
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <7465b6170809222026i4d91a77bm26f1d6a6c33ffe3@mail.gmail.com>
References: <48D860C4.6000300@cm-labs.com>
	<7465b6170809222026i4d91a77bm26f1d6a6c33ffe3@mail.gmail.com>
Message-ID: <48D8E197.60504@cm-labs.com>

Hello Roman,

> I am not sure whether OSG and OpenSG are same project, but working
> Python wrapper for the latest one you can find here:
> https://realityforge.vrsource.org/trac/pyopensg
> (Of course it uses boost.python ).

No, they are different projects, though of course they are often 
confused because of their names. OSG stands for OpenSceneGraph:

http://www.openscenegraph.org/

OpenSG is different:

http://www.opensg.org/

Though if people were able to make wrappers with boosy.python for the 
latter, it at least gives me hope that it should be possible to do the 
same for the former.

J-S
-- 
______________________________________________________
Jean-Sebastien Guay    jean-sebastien.guay at cm-labs.com
                                http://www.cm-labs.com/
                         http://whitestar02.webhop.org/


From dave at boostpro.com  Tue Sep 23 18:04:19 2008
From: dave at boostpro.com (David Abrahams)
Date: Tue, 23 Sep 2008 12:04:19 -0400
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <48D860C4.6000300@cm-labs.com> (=?utf-8?Q?=22Jean-S=C3=A9bast?=
	=?utf-8?Q?ien?= Guay"'s message
	of "Mon, 22 Sep 2008 23:21:40 -0400")
References: <48D860C4.6000300@cm-labs.com>
Message-ID: <87wsh2c0ho.fsf@mcbain.luannocracy.com>


on Mon Sep 22 2008, Jean-S?bastien Guay <jean-sebastien.guay-AT-cm-labs.com> wrote:

> Hi all,
>
> Thanks to David for pointing out my mistake earlier, I am progressing again on
> my OSG boost.python wrappers.
>
> Now, you may or may not know that OSG uses its own reference counting
> mechanism. All ref counted objects inherit from an osg::Referenced
> base class which has the reference count, and a smart pointer template
> class ref_ptr<T> manages these counts. I've thus far been able to use
> pointee< ref_ptr<T> > as heldtype to get around the fact that
> destructors are protected, but now I want to expose the file loading
> method:
>
>   osg::Node* osgDB::readNodeFile(const std::string& filename, /*...*/)
>
> I've got this:
>
>     def("readNodeFile", readNodeFile1, /* some return value policy */);
>
> I'm stuck at the choice of return value policy. I guess the most
> logical would be return_value_policy<manage_new_object>(), but this
> gives a compilation error because the actual C++ return type of the
> function has a protected destructor.

As long as it's protected, you can break in by:

   struct Hole : osg::Node
   {
       struct deleter
       {
            void operator()(osg::Node* n) const { delete n; }
       };
   };

and then expose a thin wrapper around readNodeFile that returns

    boost::shared_ptr<Node*>(readNodeFile(...), Hole::deleter());

Or use your own smart pointer if that's what you have.

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


From jean-sebastien.guay at cm-labs.com  Tue Sep 23 18:25:26 2008
From: jean-sebastien.guay at cm-labs.com (=?UTF-8?B?SmVhbi1Tw6liYXN0aWVuIEd1YXk=?=)
Date: Tue, 23 Sep 2008 12:25:26 -0400
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <87wsh2c0ho.fsf@mcbain.luannocracy.com>
References: <48D860C4.6000300@cm-labs.com>
	<87wsh2c0ho.fsf@mcbain.luannocracy.com>
Message-ID: <48D91876.1080401@cm-labs.com>

Hello David,

> Or use your own smart pointer if that's what you have.

So you're saying that exposing a thin wrapper that would return a 
ref_ptr<osg::Node> instead of osg::Node would work? What would be the 
return value policy for that? (since it's returning an object on the 
stack instead of one created with new, I would think 
return_value_policy<copy_non_const_reference>?

Thanks again for your help,

J-S
-- 
______________________________________________________
Jean-Sebastien Guay    jean-sebastien.guay at cm-labs.com
                                http://www.cm-labs.com/
                         http://whitestar02.webhop.org/


From rogeeff at gmail.com  Tue Sep 23 21:23:30 2008
From: rogeeff at gmail.com (Gennadiy Rozental)
Date: Tue, 23 Sep 2008 19:23:30 +0000 (UTC)
Subject: [C++-sig] export of dictionary like class
Message-ID: <loom.20080923T192145-911@post.gmane.org>

What is the best way to export dictionary like class so 
it support python dictionary API. Pretty much what I 
want is to inherit from UserDict.DictMixin in C++

Thanks,

Gennadiy



From rwgk at yahoo.com  Tue Sep 23 21:36:02 2008
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Tue, 23 Sep 2008 12:36:02 -0700 (PDT)
Subject: [C++-sig] export of dictionary like class
Message-ID: <118782.65438.qm@web31403.mail.mud.yahoo.com>

There is the map_indexing_suite that comes with Boost.Python

#include <boost/python/suite/indexing/map_indexing_suite.hpp>

or you can just roll the wrapper yourself like this:

http://cctbx.svn.sourceforge.net/viewvc/cctbx/trunk/scitbx/stl/map_wrapper.h?view=markup




----- Original Message ----
From: Gennadiy Rozental <rogeeff at gmail.com>
To: c++-sig at python.org
Sent: Tuesday, September 23, 2008 12:23:30 PM
Subject: [C++-sig] export of dictionary like class

What is the best way to export dictionary like class so 
it support python dictionary API. Pretty much what I 
want is to inherit from UserDict.DictMixin in C++

Thanks,

Gennadiy



From rogeeff at gmail.com  Tue Sep 23 22:00:43 2008
From: rogeeff at gmail.com (Gennadiy Rozental)
Date: Tue, 23 Sep 2008 20:00:43 +0000 (UTC)
Subject: [C++-sig] export of dictionary like class
References: <118782.65438.qm@web31403.mail.mud.yahoo.com>
Message-ID: <loom.20080923T195851-891@post.gmane.org>

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

> 
> There is the map_indexing_suite that comes with Boost.Python
> 
> #include <boost/python/suite/indexing/map_indexing_suite.hpp>

Class I have export is quite crippled. One of the problem is that value is
smart_ptr<T>, which seems to prevent above template usage.

Gennadiy



From rwgk at yahoo.com  Tue Sep 23 22:30:01 2008
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Tue, 23 Sep 2008 13:30:01 -0700 (PDT)
Subject: [C++-sig] export of dictionary like class
Message-ID: <87237.64852.qm@web31401.mail.mud.yahoo.com>

> Class I have export is quite crippled. One of the problem is that value is

> smart_ptr<T>, which seems to prevent above template usage.

In that case I'd go the hand-rolled route. Usually, it isn't much work.
Ralf


From rogeeff at gmail.com  Tue Sep 23 22:58:36 2008
From: rogeeff at gmail.com (Gennadiy Rozental)
Date: Tue, 23 Sep 2008 20:58:36 +0000 (UTC)
Subject: [C++-sig] export of dictionary like class
References: <87237.64852.qm@web31401.mail.mud.yahoo.com>
Message-ID: <loom.20080923T205128-955@post.gmane.org>

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

> 
> > Class I have export is quite crippled. One of the problem is that value is
> 
> > smart_ptr<T>, which seems to prevent above template usage.
> 
> In that case I'd go the hand-rolled route. Usually, it isn't much work.
> Ralf
> 

I've decided on mixed C++/python approach.

on c++ side:

class_<IMap,IMapSP, bases<IObject>, boost::noncopyable>( "Map", bp::no_init )
        .def( "keys", &object_op::get_map_keys )
        .def( "items", &object_op::get_map_items )
        ;
class_<IReadableMap,IReadableMapSP, bases<IMap>, boost::noncopyable>(
"ReadableMap", bp::no_init )
        .def( "__getitem__", &IReadableMap::get )
        ;
class_<IWriteableMap,IWriteableMapSP, bases<IMap>, boost::noncopyable>(
"WriteableMap", bp::no_init )
        .def( "__setitem__", &IWriteableMap::put )
        ;

on python side during library init phase
import UserDict
add_mixin( Map, UserDict.DictMixin )

where 
def add_mixin( clazz, mixin_clazz ):
    bases = inspect.getmro( clazz )
    if mixin_clazz not in bases:
        clazz.__bases__ = (mixin_clazz,) + clazz.__bases__

Gennadiy





From jean-sebastien.guay at cm-labs.com  Wed Sep 24 04:09:10 2008
From: jean-sebastien.guay at cm-labs.com (=?UTF-8?B?SmVhbi1Tw6liYXN0aWVuIEd1YXk=?=)
Date: Tue, 23 Sep 2008 22:09:10 -0400
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <48D91876.1080401@cm-labs.com>
References: <48D860C4.6000300@cm-labs.com>
	<87wsh2c0ho.fsf@mcbain.luannocracy.com>
	<48D91876.1080401@cm-labs.com>
Message-ID: <48D9A146.1040000@cm-labs.com>

Hello again,

> So you're saying that exposing a thin wrapper that would return a 
> ref_ptr<osg::Node> instead of osg::Node would work? 

Continuing my (mostly bind) experimentation, I tried this:

osg::ref_ptr<osg::Node> readNodeFileWrapper(const std::string& filename)
{
     return osg::ref_ptr<osg::Node>(osgDB::readNodeFile(filename));
}

BOOST_PYTHON_MODULE(_osgDB)
{
     def("readNodeFile", readNodeFileWrapper);
}

That gets me a python.exe crash like the other experiments did, running 
this way from the Python interpreter:

import osg
import osgDB
n = osgDB.readNodeFile("cow.osg")

Note that this works:

import osg
n = osg.Node()

so I assume my osg::Node wrapper is at least minimally functional... 
Though I have no way of knowing if n is actually properly constructed or 
if it's enclosed in a ref_ptr... My osg::Node wrapper starts like this, 
for reference:

class_<Node, bases<Object>, ref_ptr<Node> >("Node")
     /* ... */
;

Another weird thing, if I change the above thin wrapper to hard-code the 
filename:

osg::ref_ptr<osg::Node> readNodeFileWrapper()
{
     return osg::ref_ptr<osg::Node>(osgDB::readNodeFile("cow.osg"));
}

I don't get a crash, but OSG reports it can't load a file called "?" 
(lowercase greek alpha), which seems to indicate that somewhere, the 
string got mangled somehow? I don't understand how that can be since 
it's a direct call into OSG from C++ code, nothing else involved.

Has anyone seen something like that before? Thanks in advance,

J-S
-- 
______________________________________________________
Jean-Sebastien Guay    jean-sebastien.guay at cm-labs.com
                                http://www.cm-labs.com/
                         http://whitestar02.webhop.org/


From dave at boostpro.com  Wed Sep 24 05:27:50 2008
From: dave at boostpro.com (David Abrahams)
Date: Tue, 23 Sep 2008 23:27:50 -0400
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <48D9A146.1040000@cm-labs.com> (=?utf-8?Q?=22Jean-S=C3=A9bast?=
	=?utf-8?Q?ien?= Guay"'s message
	of "Tue, 23 Sep 2008 22:09:10 -0400")
References: <48D860C4.6000300@cm-labs.com>
	<87wsh2c0ho.fsf@mcbain.luannocracy.com> <48D91876.1080401@cm-labs.com>
	<48D9A146.1040000@cm-labs.com>
Message-ID: <87wsh2te89.fsf@mcbain.luannocracy.com>


on Tue Sep 23 2008, Jean-S?bastien Guay <jean-sebastien.guay-AT-cm-labs.com> wrote:

> Hello again,
>
>> So you're saying that exposing a thin wrapper that would return a
>> ref_ptr<osg::Node> instead of osg::Node would work? 

I don't think I said that.  I don't know anything about your ref_ptr,
its relationship to osg::Node, or the reference count that you might
have for your new Node, so I wouldn't make such an assertion.

> Continuing my (mostly bind) experimentation, I tried this:
>
> osg::ref_ptr<osg::Node> readNodeFileWrapper(const std::string& filename)
> {
>     return osg::ref_ptr<osg::Node>(osgDB::readNodeFile(filename));
> }
>
> BOOST_PYTHON_MODULE(_osgDB)
> {
>     def("readNodeFile", readNodeFileWrapper);
> }
>
> That gets me a python.exe crash like the other experiments did, running this way
> from the Python interpreter:
>
> import osg
> import osgDB
> n = osgDB.readNodeFile("cow.osg")
>
> Note that this works:
>
> import osg
> n = osg.Node()
>
> so I assume my osg::Node wrapper is at least minimally functional... Though I
> have no way of knowing if n is actually properly constructed or if it's enclosed
> in a ref_ptr... My osg::Node wrapper starts like this, for reference:
>
> class_<Node, bases<Object>, ref_ptr<Node> >("Node")
>     /* ... */
> ;
>
> Another weird thing, if I change the above thin wrapper to hard-code the
> filename:
>
> osg::ref_ptr<osg::Node> readNodeFileWrapper()
> {
>     return osg::ref_ptr<osg::Node>(osgDB::readNodeFile("cow.osg"));
> }
>
> I don't get a crash, but OSG reports it can't load a file called "?" (lowercase
> greek alpha), which seems to indicate that somewhere, the string got mangled
> somehow? I don't understand how that can be since it's a direct call into OSG
> from C++ code, nothing else involved.

It sounds like you are casting about for something that seems to work
without really understanding what you're doing.  I suggest you slow down
and analyze.  Remember that "seems to work" and "works" can be widely
separated ;-)

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


From rogeeff at gmail.com  Wed Sep 24 07:02:30 2008
From: rogeeff at gmail.com (Gennadiy Rozental)
Date: Wed, 24 Sep 2008 05:02:30 +0000 (UTC)
Subject: [C++-sig] once again about global dictionary replacement
Message-ID: <loom.20080924T043646-219@post.gmane.org>

Hi,

This is an example that illustrate problem similar to what I have.

Let's say I have c++ function 

bp::object s_save;
void foo( bp::object const& obj )
{
   s_save = obj;
}

exported with the same name in python.

Now let say in one place of my program I execute following string (this is an
input to my program for example)

"
import some_module_abc
def boo()
   .... do something here and use some_module_abc

foo( boo )
"

when I execute this string I use command like 
exec( code, my_module_dictionary )

I specify my_module_dictionary as local context because I want all the symbols
defined in string (functions and classes) to become members of my module.

Some time later I execute stored function like this

s_save();

Now here is a punchline: python complains it does not know about some_module_abc

The reason is that import occurred in local context of first exec call,
resulting in all symbols to be imported into my module dictionary, not global
dictionary. At execution point global dictionary is the same (not sure now, but
it seems bound to the python object). Local dictionary is temporary containing
function arguments (if any). Both know nothing about some_module_abc.

It would help if I can as the point of invocation do one of these
a) replace the bound global dictionary
b) update the bound global dictionary
c) update local dictionary that is going to be used

Can anyone suggest the solution?

Thanks,

Gennadiy




From meine at informatik.uni-hamburg.de  Wed Sep 24 11:07:15 2008
From: meine at informatik.uni-hamburg.de (Hans Meine)
Date: Wed, 24 Sep 2008 11:07:15 +0200
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <48D860C4.6000300@cm-labs.com>
References: <48D860C4.6000300@cm-labs.com>
Message-ID: <200809241107.16769.meine@informatik.uni-hamburg.de>

Am Dienstag, 23. September 2008 05:21:40 schrieb Jean-S?bastien Guay:
> Now, you may or may not know that OSG uses its own reference counting
> mechanism. All ref counted objects inherit from an osg::Referenced base
> class which has the reference count, and a smart pointer template class
> ref_ptr<T> manages these counts.

Do you know that people have more or less successfully used custom smart 
pointers with BP?  IIRC Roman has a nice web page about that.. Seems to be 
here:
http://www.language-binding.net/pyplusplus/troubleshooting_guide/smart_ptrs/smart_ptrs.html

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


From jean-sebastien.guay at cm-labs.com  Wed Sep 24 15:03:52 2008
From: jean-sebastien.guay at cm-labs.com (=?UTF-8?B?SmVhbi1Tw6liYXN0aWVuIEd1YXk=?=)
Date: Wed, 24 Sep 2008 09:03:52 -0400
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <87wsh2te89.fsf@mcbain.luannocracy.com>
References: <48D860C4.6000300@cm-labs.com>	<87wsh2c0ho.fsf@mcbain.luannocracy.com>
	<48D91876.1080401@cm-labs.com>	<48D9A146.1040000@cm-labs.com>
	<87wsh2te89.fsf@mcbain.luannocracy.com>
Message-ID: <48DA3AB8.4070309@cm-labs.com>

Hello David,

> I don't think I said that.  I don't know anything about your ref_ptr,
> its relationship to osg::Node, or the reference count that you might
> have for your new Node, so I wouldn't make such an assertion.

Of course, I was interpreting what you said (perhaps too 
enthusiastically). I've really been impressed at how boost.python seemed 
to just do the right thing automatically in many cases, so I thought 
this was another instance where I thought something complex was required 
but something simple actually worked better. Sorry for putting words in 
your mouth.

> It sounds like you are casting about for something that seems to work
> without really understanding what you're doing.  I suggest you slow down
> and analyze.  Remember that "seems to work" and "works" can be widely
> separated ;-)

Sure, that's a valid assessment. I'll do more reading and see if I can 
make a new start.

Thanks,

J-S
-- 
______________________________________________________
Jean-Sebastien Guay    jean-sebastien.guay at cm-labs.com
                                http://www.cm-labs.com/
                         http://whitestar02.webhop.org/


From jean-sebastien.guay at cm-labs.com  Wed Sep 24 15:08:59 2008
From: jean-sebastien.guay at cm-labs.com (=?ISO-8859-1?Q?Jean-S=E9bastien_Guay?=)
Date: Wed, 24 Sep 2008 09:08:59 -0400
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <200809241107.16769.meine@informatik.uni-hamburg.de>
References: <48D860C4.6000300@cm-labs.com>
	<200809241107.16769.meine@informatik.uni-hamburg.de>
Message-ID: <48DA3BEB.9030608@cm-labs.com>

Hello Hans,

> IIRC Roman has a nice web page about that.. Seems to be 
> here:
> http://www.language-binding.net/pyplusplus/troubleshooting_guide/smart_ptrs/smart_ptrs.html

I've seen that and will try to apply it. Thanks.

Actually, I think my problems with boost.python can be summed up thus:
* I don't know when such and such strategy is required, and when it's
   just useless code that adds to the maintenance overhead of my wrappers
* I don't know how to debug the wrappers if they don't give the intended
   results (other than blindly trying something else).

The first point, I can work on by reading the documentation more 
thoroughly. But I would appreciate tips on the second. I haven't seen 
any documentation on that.

In particular, if some code makes the python interpreter crash, what can 
I do to get a stack trace or go into it with a debugger (on windows in 
my case)? Do I need to compile Python and/or boost in debug?

Thanks in advance,

J-S
-- 
______________________________________________________
Jean-Sebastien Guay    jean-sebastien.guay at cm-labs.com
                                http://www.cm-labs.com/
                         http://whitestar02.webhop.org/


From gjcarneiro at gmail.com  Wed Sep 24 15:15:29 2008
From: gjcarneiro at gmail.com (Gustavo Carneiro)
Date: Wed, 24 Sep 2008 14:15:29 +0100
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <48DA3BEB.9030608@cm-labs.com>
References: <48D860C4.6000300@cm-labs.com>
	<200809241107.16769.meine@informatik.uni-hamburg.de>
	<48DA3BEB.9030608@cm-labs.com>
Message-ID: <a467ca4f0809240615k7e14a06dvd677f028346e1aeb@mail.gmail.com>

2008/9/24 Jean-S?bastien Guay <jean-sebastien.guay at cm-labs.com>

> Hello Hans,
>
>  IIRC Roman has a nice web page about that.. Seems to be here:
>>
>> http://www.language-binding.net/pyplusplus/troubleshooting_guide/smart_ptrs/smart_ptrs.html
>>
>
> I've seen that and will try to apply it. Thanks.
>
> Actually, I think my problems with boost.python can be summed up thus:
> * I don't know when such and such strategy is required, and when it's
>  just useless code that adds to the maintenance overhead of my wrappers
> * I don't know how to debug the wrappers if they don't give the intended
>  results (other than blindly trying something else).
>
> The first point, I can work on by reading the documentation more
> thoroughly. But I would appreciate tips on the second. I haven't seen any
> documentation on that.
>
> In particular, if some code makes the python interpreter crash, what can I
> do to get a stack trace or go into it with a debugger (on windows in my
> case)? Do I need to compile Python and/or boost in debug?


Shameless plug... :-)

I think PyBindGen is better in that respect.  It may not have all the
features of BP yet, or be as mature, but it generates reasonably clean C/C++
wrapping code without templates or macros or funny variable names, so you
can debug it better.

Regards,

-- 
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/20080924/3c97ea03/attachment.htm>

From meine at informatik.uni-hamburg.de  Wed Sep 24 15:17:15 2008
From: meine at informatik.uni-hamburg.de (Hans Meine)
Date: Wed, 24 Sep 2008 15:17:15 +0200
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <48DA3BEB.9030608@cm-labs.com>
References: <48D860C4.6000300@cm-labs.com>
	<200809241107.16769.meine@informatik.uni-hamburg.de>
	<48DA3BEB.9030608@cm-labs.com>
Message-ID: <200809241517.17305.meine@informatik.uni-hamburg.de>

Am Mittwoch, 24. September 2008 15:08:59 schrieb Jean-S?bastien Guay:
> Actually, I think my problems with boost.python can be summed up thus:
> * I don't know when such and such strategy is required, and when it's
>    just useless code that adds to the maintenance overhead of my wrappers

My strategy for preventing useless code usually is to leave out as much code 
as possible and see what breaks (not).

> * I don't know how to debug the wrappers if they don't give the intended
>    results (other than blindly trying something else).

That's tough indeed.

> In particular, if some code makes the python interpreter crash, what can
> I do to get a stack trace or go into it with a debugger (on windows in
> my case)? Do I need to compile Python and/or boost in debug?

AFAIK, yes.  But I have nearly zero experience with that.  I occasionally run 
python through valgrind (with the supplied suppressions-file) in order to 
find memleaks/similar memory errors though.  It's slow, but works fairly 
well.

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


From jean-sebastien.guay at cm-labs.com  Wed Sep 24 15:28:03 2008
From: jean-sebastien.guay at cm-labs.com (=?UTF-8?B?SmVhbi1Tw6liYXN0aWVuIEd1YXk=?=)
Date: Wed, 24 Sep 2008 09:28:03 -0400
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <a467ca4f0809240615k7e14a06dvd677f028346e1aeb@mail.gmail.com>
References: <48D860C4.6000300@cm-labs.com>	<200809241107.16769.meine@informatik.uni-hamburg.de>	<48DA3BEB.9030608@cm-labs.com>
	<a467ca4f0809240615k7e14a06dvd677f028346e1aeb@mail.gmail.com>
Message-ID: <48DA4063.8090101@cm-labs.com>

Hi Gustavo,

> I think PyBindGen is better in that respect.

Looks interesting, but from the features page, it doesn't support 
multiple inheritance, which OSG uses a lot. So it's probably a no go. 
I'll check it out though.

Thanks for the suggestion,

J-S
-- 
______________________________________________________
Jean-Sebastien Guay    jean-sebastien.guay at cm-labs.com
                                http://www.cm-labs.com/
                         http://whitestar02.webhop.org/


From jean-sebastien.guay at cm-labs.com  Wed Sep 24 15:32:12 2008
From: jean-sebastien.guay at cm-labs.com (=?ISO-8859-1?Q?Jean-S=E9bastien_Guay?=)
Date: Wed, 24 Sep 2008 09:32:12 -0400
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <200809241517.17305.meine@informatik.uni-hamburg.de>
References: <48D860C4.6000300@cm-labs.com>	<200809241107.16769.meine@informatik.uni-hamburg.de>	<48DA3BEB.9030608@cm-labs.com>
	<200809241517.17305.meine@informatik.uni-hamburg.de>
Message-ID: <48DA415C.2030107@cm-labs.com>

Hello Hans,

>> * I don't know when such and such strategy is required, and when it's
>>    just useless code that adds to the maintenance overhead of my wrappers
> 
> My strategy for preventing useless code usually is to leave out as much code 
> as possible and see what breaks (not).

That's what I've been doing until now, but it doesn't seem that successful.

> AFAIK, yes.  But I have nearly zero experience with that.  I occasionally run 
> python through valgrind (with the supplied suppressions-file) in order to 
> find memleaks/similar memory errors though.  It's slow, but works fairly 
> well.

I'm not talking about memory leaks per se, I'm talking about the actual 
wrapper code causing crashes of the interpreter. I've seen a few 
instances of choosing the wrong return value policy making the 
interpreter crash, and I don't see a way of finding out what code was 
being called to cause the crash.

Anyways, I'll read up on the actual low-level process that's going on 
and hopefully that will help me understand what I'm doing.

Thanks,

J-S
-- 
______________________________________________________
Jean-Sebastien Guay    jean-sebastien.guay at cm-labs.com
                                http://www.cm-labs.com/
                         http://whitestar02.webhop.org/


From meine at informatik.uni-hamburg.de  Wed Sep 24 15:42:08 2008
From: meine at informatik.uni-hamburg.de (Hans Meine)
Date: Wed, 24 Sep 2008 15:42:08 +0200
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <48DA415C.2030107@cm-labs.com>
References: <48D860C4.6000300@cm-labs.com>
	<200809241517.17305.meine@informatik.uni-hamburg.de>
	<48DA415C.2030107@cm-labs.com>
Message-ID: <200809241542.10101.meine@informatik.uni-hamburg.de>

Am Mittwoch, 24. September 2008 15:32:12 schrieb Jean-S?bastien Guay:
> I'm not talking about memory leaks per se, I'm talking about the actual
> wrapper code causing crashes of the interpreter.

That's what I meant with "similar memory errors".  My guess is that you're 
using a policy that allows the C++ object to be deleted although a python 
object still references it.

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


From dave at boostpro.com  Wed Sep 24 16:20:25 2008
From: dave at boostpro.com (David Abrahams)
Date: Wed, 24 Sep 2008 10:20:25 -0400
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <48DA3BEB.9030608@cm-labs.com> (=?utf-8?Q?=22Jean-S=C3=A9bast?=
	=?utf-8?Q?ien?= Guay"'s message
	of "Wed, 24 Sep 2008 09:08:59 -0400")
References: <48D860C4.6000300@cm-labs.com>
	<200809241107.16769.meine@informatik.uni-hamburg.de>
	<48DA3BEB.9030608@cm-labs.com>
Message-ID: <874p4562xi.fsf@mcbain.luannocracy.com>


on Wed Sep 24 2008, Jean-S?bastien Guay <jean-sebastien.guay-AT-cm-labs.com> wrote:

> Hello Hans,
>
>> IIRC Roman has a nice web page about that.. Seems to be here:
>>
> http://www.language-binding.net/pyplusplus/troubleshooting_guide/smart_ptrs/smart_ptrs.html
>
> I've seen that and will try to apply it. Thanks.
>
> Actually, I think my problems with boost.python can be summed up thus:
> * I don't know when such and such strategy is required, and when it's
>   just useless code that adds to the maintenance overhead of my wrappers
> * I don't know how to debug the wrappers if they don't give the intended
>   results (other than blindly trying something else).
>
> The first point, I can work on by reading the documentation more thoroughly. But
> I would appreciate tips on the second. I haven't seen any documentation on that.
>
> In particular, if some code makes the python interpreter crash, what can I do to
> get a stack trace or go into it with a debugger (on windows in my case)? Do I
> need to compile Python and/or boost in debug?

That's what I do.

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


From roman.yakovenko at gmail.com  Wed Sep 24 19:44:00 2008
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 24 Sep 2008 20:44:00 +0300
Subject: [C++-sig] Reference counted object return value policy
In-Reply-To: <48DA415C.2030107@cm-labs.com>
References: <48D860C4.6000300@cm-labs.com>
	<200809241107.16769.meine@informatik.uni-hamburg.de>
	<48DA3BEB.9030608@cm-labs.com>
	<200809241517.17305.meine@informatik.uni-hamburg.de>
	<48DA415C.2030107@cm-labs.com>
Message-ID: <7465b6170809241044v193d9148i3ac83195f507ac49@mail.gmail.com>

On Wed, Sep 24, 2008 at 4:32 PM, Jean-S?bastien Guay
<jean-sebastien.guay at cm-labs.com> wrote:
> I'm not talking about memory leaks per se, I'm talking about the actual
> wrapper code causing crashes of the interpreter. I've seen a few instances
> of choosing the wrong return value policy making the interpreter crash, and
> I don't see a way of finding out what code was being called to cause the
> crash.
>

>From my experience you can assume that "boost.python" just works.
After you accepted such assumption, it takes some time, - your life is
a little bit easier.

If your wrapper consists from "registration code" only, than simple
code review usually helps.
If you use visual studio you can build your wrapper in release mode
with debug information. In most cases this is enough to debug your
problems.
Also, building small test case that reproduce the problem and asking
for help here works too and quite well.

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


From william.marie at gmail.com  Wed Sep 24 20:26:33 2008
From: william.marie at gmail.com (=?UTF-8?Q?William_Mari=C3=A9?=)
Date: Wed, 24 Sep 2008 11:26:33 -0700 (PDT)
Subject: [C++-sig]  How to use an opaque pointer ?
Message-ID: <19655049.post@talk.nabble.com>


Hi,

I would like to pass a CObject from C++ to Python without exposing the class
i've seen the opaque pointer feature but i can't make it work.

Here is what i would like to do :

CObject* pObj = ..... ;
bp::object callable = .... ;
callable( (bpCObject*)pObj );

I tried to do like in the documentation :

typedef struct CObject_ *bpCObject;
BOOST_PYTHON_OPAQUE_SPECIALIZED_TYPE_ID(CObject_)

but i always have this error at runtime :

TypeError: No to_python (by-value) converter found for C++ type: struct
bpCObject *

and i don't understand, could someone tell me why i have that and what do i
have to do to fix the problem ? 

Thanks a lot
William


-- 
View this message in context: http://www.nabble.com/How-to-use-an-opaque-pointer---tp19655049p19655049.html
Sent from the Python - c++-sig mailing list archive at Nabble.com.



From renatox at gmail.com  Sun Sep 28 17:30:49 2008
From: renatox at gmail.com (Renato Araujo)
Date: Sun, 28 Sep 2008 12:30:49 -0300
Subject: [C++-sig] downcasting problems
Message-ID: <95291a80809280830y40a01a47l45f6b204212a70f1@mail.gmail.com>

Hi guys,
I'm looking for a solution to my problem with python and c++ objects.

I will show this little sample code to introduce my problem:

I have this c++ code:
========================

class Shape
{
public:
    Shape(Shape *parent=0)
    {
        this->_parent = parent;
    };

    Shape* parent()
    {
        return this->_parent;
    };

private:
    Shape *_parent;
};

class Circle : public Shape
{
public:

    Circle(Shape *parent=NULL)
        : Shape(parent)
    {
    };
};

======================
and this python.boost code:
======================

BOOST_PYTHON_MODULE(shapes)
{
class_<Shape, boost::noncopyable> ("Shape", init<optional<Shape*> > ())
    .def("parent", &Shape::parent, return_internal_reference<>())
;

class_<Circle, boost::noncopyable, bases<Shape>
>("Circle",init<optional<Shape*> >())
;
}

=======================================
the python test is this, and this always output  "FAIL":
=======================================

if __name__ == '__main__':
    shape = Shape()
    circle = Circle(shape)
    p = circle.parent()
    if (p == shape):
        print "OK"
    else:
        print "FAIL"


How I can fix this without change c++ code?

-- 
Renato Araujo Oliveira Filho


From entropydevice at gmail.com  Mon Sep 29 15:21:23 2008
From: entropydevice at gmail.com (Frederick Heckel)
Date: Mon, 29 Sep 2008 09:21:23 -0400
Subject: [C++-sig] static arrays to Python
Message-ID: <2d0df3380809290621y54463a62ib059a0ab1ab4b696@mail.gmail.com>

Hi,

I have a series of structs which contain static arrays of more custom
structs and some built-in types. One example:

struct connectivity
{
   int       connections[SSPS_MAX_GATEWAYS];
   gateway   gateways[SSPS_MAX_GATEWAYS];
   int       adjacents[SSPS_MAX_ADJACENTS];
};

In every case, the arrays are static, as in this case, where
SSPS_MAX_GATEWAYS is #defined to 16. Previously, these were moved
around via shared memory in C. We're now moving to Python, and find
ourselves having to use named pipes in windows, because of the general
weirdness of Vista. But we are still using some of our old services,
including the one that generates this struct (and several more like
it); it seemed as though it should be fairly straightforward to wrap
all the functionality in a class, and invoke that via Python.

But dealing with these static arrays is turning out to be an enormous
pain. I initially tried iterating over them to turn them into
boost::python lists, but we'd like to be able to use pickle to turn
them into strings to send over the pipe. Pickle is having none of it.

So my question is, what am I missing? These are contiguous blocks of
memory, boost::python has no trouble wrapping the individual structs,
so long as I use accessors for the static arrays, but it looks like
the task of getting them set up for pickle is going to be a massive
PITA. It seems like there should be some [relatively] easy, efficient
way to get static arrays into python (even discarding the use of
pickle for now, we'll assume they don't need to be mutable from
python), but I'm not finding it.

I'll rewrite the whole silly service in Python if I have to, but it
seems like there should be a Better Way.

-- 
(fwph)

Frederick Heckel


From rwgk at yahoo.com  Mon Sep 29 17:29:37 2008
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Mon, 29 Sep 2008 08:29:37 -0700 (PDT)
Subject: [C++-sig] static arrays to Python
Message-ID: <422250.12801.qm@web31402.mail.mud.yahoo.com>

>    gateway   gateways[SSPS_MAX_GATEWAYS];


> we'd like to be able to use pickle to turn
> them into strings to send over the pipe.

You have to provide pickle support for your gateway type.
See http://www.boost.org/doc/libs/1_36_0/libs/python/doc/v2/pickle.html

What are the data members of your gateway type?

I often ended up using the simple __getinitargs__ approach, where you just have
to construct a tuple that works as an argument list for one of your constructors.

Ralf


From renatox at gmail.com  Mon Sep 29 20:46:50 2008
From: renatox at gmail.com (Renato Araujo)
Date: Mon, 29 Sep 2008 15:46:50 -0300
Subject: [C++-sig] downcasting problems
In-Reply-To: <95291a80809280830y40a01a47l45f6b204212a70f1@mail.gmail.com>
References: <95291a80809280830y40a01a47l45f6b204212a70f1@mail.gmail.com>
Message-ID: <95291a80809291146j2885674eo618a959da5b3704b@mail.gmail.com>

Hi all,

After some test I got this working but with some extra code. I created
2 wrapper classes to Shape and Circle like that:

===========
wrapper code:
===========
class ShapeWrapper : public Shape, public wrapper<Shape>
{
....
};

class CircleWrapper : public Cirlce, public wrapper<Circle>
{
.....
};

and used this classes in boost code. And all works fine.
But is this the best way to do this?? I always need create theses
wrapper class??


BR




On Sun, Sep 28, 2008 at 12:30 PM, Renato Araujo <renatox at gmail.com> wrote:
> Hi guys,
> I'm looking for a solution to my problem with python and c++ objects.
>
> I will show this little sample code to introduce my problem:
>
> I have this c++ code:
> ========================
>
> class Shape
> {
> public:
>    Shape(Shape *parent=0)
>    {
>        this->_parent = parent;
>    };
>
>    Shape* parent()
>    {
>        return this->_parent;
>    };
>
> private:
>    Shape *_parent;
> };
>
> class Circle : public Shape
> {
> public:
>
>    Circle(Shape *parent=NULL)
>        : Shape(parent)
>    {
>    };
> };
>
> ======================
> and this python.boost code:
> ======================
>
> BOOST_PYTHON_MODULE(shapes)
> {
> class_<Shape, boost::noncopyable> ("Shape", init<optional<Shape*> > ())
>    .def("parent", &Shape::parent, return_internal_reference<>())
> ;
>
> class_<Circle, boost::noncopyable, bases<Shape>
>>("Circle",init<optional<Shape*> >())
> ;
> }
>
> =======================================
> the python test is this, and this always output  "FAIL":
> =======================================
>
> if __name__ == '__main__':
>    shape = Shape()
>    circle = Circle(shape)
>    p = circle.parent()
>    if (p == shape):
>        print "OK"
>    else:
>        print "FAIL"
>
>
> How I can fix this without change c++ code?
>
> --
> Renato Araujo Oliveira Filho
>



-- 
Renato Araujo Oliveira Filho


From lists_ravi at lavabit.com  Mon Sep 29 20:17:25 2008
From: lists_ravi at lavabit.com (Ravi)
Date: Mon, 29 Sep 2008 14:17:25 -0400
Subject: [C++-sig] Return value policies for objects with custom to-python
	converters
Message-ID: <200809291417.26620.lists_ravi@lavabit.com>

Hi,
  The return value policies from boost.python do not seem to handle objects 
with custom converters; they seem to handle only those types exposed via 
boost.python. Attached is some code that handles reference_existing_object for 
objects with custom converters. However, the interface is unsatisying:

1. The _1 subscripts need to be removed. However, I could not find a way to 
disambiguate types not exposed through boost.python and those with custom 
converters. From the perspective of a library user, both the 
reference_existing_object from boost.python and reference_existing_object_1 
from the attached file accomplish exactly the same thing, but the user needs 
to remember (and know) that certain classes need one vs the other. Could a 
guru tell me how to plug my converter into boost.python's 
reference_existing_object?

2. So far, I have used only reference_existing_object. Do the other return 
value policies need similar special handling for objects with custom to-python 
converters? How do I figure out which ones do and which ones don't?

3. If there is a satisfactory answer to 1, the resulting technique could be 
added to boost.python. No new documentation would be necessary, since the 
existing documentation would convey all the necessary information. In order to 
submit a patch though, I suppose that some tests are needed. How would one 
write tests for this functionality?

Regards,
Ravi

-------------- next part --------------
A non-text attachment was scrubbed...
Name: reference_existing_object_1.hpp
Type: text/x-c++hdr
Size: 1104 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080929/ca874a87/attachment.hpp>

From petela at gocougs.wsu.edu  Mon Sep 29 21:20:06 2008
From: petela at gocougs.wsu.edu (Peter LaDow)
Date: Mon, 29 Sep 2008 12:20:06 -0700
Subject: [C++-sig] Importing and internal modules
Message-ID: <9e264e780809291220i15f0557aq581af0ca58e9e0f4@mail.gmail.com>

I'm not sure if my subject is clear (probably not), but I'll try to
describe what I am trying to do here.

A commercial tool we are using is capable of loading a shared library
as an extension, and we'd like to use Python to script much of the
extension.  To that end, I am embedding Python in our own shared
library.  The scripted part of the extension is a collection of
callback functions from the shared library into Python.  I also want
to expose an API to the extension.

So, I have created an extension using Boost and the BOOST_PYTHON_MODULE, i.e.:

namespace bp = boost::python;

BOOST_PYTHON_MODULE(foo)
{
  bp::def("foo1", foo1);
  bp::def("foo2", foo2);
}

And then I make sure to append foo to the inittab before calling
Py_Initialize():

PyImport_AppendInittab("foo", initfoo);
Py_Initialize();

No problems so far.  In fact, I can use foo in the Python script as
long as I import it, i.e. ("from foo import *").  However, here's were
I am running into trouble.

The scriptable part is a module.  It exposes a set of functions, so
I'd really like to just import that module, and not execute it.  So, I
do the following:

bp::object main = bp::import("__main__");
bp::object globals(main.attr("__dict__"));

// Import "foo"
bp::exec("from foo import *", globals, globals);

// Import the scriptable code
bp::exec("from script import *", globals, globals);

And here's an example of a script I am trying to import:

def bar1():
  foo1()
  print "Called bar1!"

def bar2():
  foo2()
  print "Called bar2!"

Now, the import for this fails with:

NameError:  global name 'foo1" is not defined

Unless I put "from foo import *" at the top of the script.  I want the
'foo' module already imported into the script.  That's the point of
the first of the imports.

I'm sure I'm missing something obvious here.  Any help would be
greatly appreciated.

Thanks,
Pete
-- 
--
"To love for the sake of being loved is human;  to love for the sake
of loving is Angelic."  -- Alphonse de Lamartine


From lists_ravi at lavabit.com  Mon Sep 29 23:07:58 2008
From: lists_ravi at lavabit.com (Ravi)
Date: Mon, 29 Sep 2008 17:07:58 -0400
Subject: [C++-sig] Importing and internal modules
In-Reply-To: <9e264e780809291220i15f0557aq581af0ca58e9e0f4@mail.gmail.com>
References: <9e264e780809291220i15f0557aq581af0ca58e9e0f4@mail.gmail.com>
Message-ID: <200809291708.00340.lists_ravi@lavabit.com>

On Monday 29 September 2008 15:20:06 Peter LaDow wrote:
> The scriptable part is a module. ?It exposes a set of functions, so
> I'd really like to just import that module, and not execute it. ?So, I
> do the following:

[snip]

> // Import "foo"
> bp::exec("from foo import *", globals, globals);
>
> // Import the scriptable code
> bp::exec("from script import *", globals, globals);
>
> And here's an example of a script I am trying to import:
>
> def bar1():
> ? foo1()
> ? print "Called bar1!"

[snip]

> Now, the import for this fails with:
>
> NameError: ?global name 'foo1" is not defined
>
> Unless I put "from foo import *" at the top of the script. ?I want the
> 'foo' module already imported into the script. ?That's the point of
> the first of the imports.

Think about how this would work in pure python. When you import a module X, 
all modules/functions X depends on must be explicitly imported in X.py. You 
cannot import Y in the global interpreter namespace and somehow expect to 
inject the contents of X into Y even if X has already been imported into the 
global namespace. So what you are trying to do is impossible even in the pure 
python case. Why would you expect it to work here?

Example:
foo.py
def foo():
  print "foo"

bar.py
# import foo <-- note that this is commented out
def bar():
  foo()

main.py
from foo import *
import bar
bar.bar() # <-- you will get a NameError here!

That said, you could inject the contents of foo into the dictionary of bar, 
but you will run into subtle problems later. Try prototyping this in pure 
python before attempting this from an extension module.

Regards,
Ravi




From petela at gocougs.wsu.edu  Mon Sep 29 23:32:37 2008
From: petela at gocougs.wsu.edu (Peter LaDow)
Date: Mon, 29 Sep 2008 14:32:37 -0700
Subject: [C++-sig] Importing and internal modules
In-Reply-To: <200809291708.00340.lists_ravi@lavabit.com>
References: <9e264e780809291220i15f0557aq581af0ca58e9e0f4@mail.gmail.com>
	<200809291708.00340.lists_ravi@lavabit.com>
Message-ID: <9e264e780809291432h2144788fg22bbd13d483b8c20@mail.gmail.com>

On Mon, Sep 29, 2008 at 2:07 PM, Ravi <lists_ravi at lavabit.com> wrote:
> Think about how this would work in pure python. When you import a module X,
> all modules/functions X depends on must be explicitly imported in X.py. You
> cannot import Y in the global interpreter namespace and somehow expect to
> inject the contents of X into Y even if X has already been imported into the
> global namespace. So what you are trying to do is impossible even in the pure
> python case. Why would you expect it to work here?

I was thinking that by passing 'globals' to both calls to import(),
symbols would be injected into the global namespace.  The goal is to
make a set of functions exported by the extension available to Python
scripting modules without the need for an explicit import statement.
It isn't a killer if that cannot occur, I was just attempting to make
things easier for the users of the extension.

> That said, you could inject the contents of foo into the dictionary of bar,
> but you will run into subtle problems later. Try prototyping this in pure
> python before attempting this from an extension module.

Fair enough.  I tried your example.  I'm no Python expert, and chose
Python because it was already popular amongst many people here at
work.  The difference with the Boost method, I thought, was the
passing of the 'globals' object to each of the imports, thinking that
symbols were injected into the dictionary, and that later imports
would see any previous imports.  I see your point.

Thanks for the feedback.  It appears my goal of exporting a set of
global "built-in" functions (a la TCL) is not possible.

Pete
-- 
--
"To love for the sake of being loved is human;  to love for the sake
of loving is Angelic."  -- Alphonse de Lamartine


From lists_ravi at lavabit.com  Tue Sep 30 01:32:29 2008
From: lists_ravi at lavabit.com (Ravi)
Date: Mon, 29 Sep 2008 19:32:29 -0400
Subject: [C++-sig] Importing and internal modules
In-Reply-To: <9e264e780809291432h2144788fg22bbd13d483b8c20@mail.gmail.com>
References: <9e264e780809291220i15f0557aq581af0ca58e9e0f4@mail.gmail.com>
	<200809291708.00340.lists_ravi@lavabit.com>
	<9e264e780809291432h2144788fg22bbd13d483b8c20@mail.gmail.com>
Message-ID: <200809291932.30996.lists_ravi@lavabit.com>

On Monday 29 September 2008 17:32:37 Peter LaDow wrote:
> Fair enough. ?I tried your example. ?I'm no Python expert, and chose
> Python because it was already popular amongst many people here at
> work. ?The difference with the Boost method, I thought, was the
> passing of the 'globals' object to each of the imports, thinking that
> symbols were injected into the dictionary, and that later imports
> would see any previous imports. ?I see your point.

I see. The trouble is that importing modules in a namespace adds that module 
to the namespace rather than providing a namespace for the module being 
imported. I can't think of a way to accomplish what you want easily.

> Thanks for the feedback. 

I re-read my email and now apologize for sounding less than helpful. My 
frustration from battling all day long with boost.python trying to get custom 
lvalue from-python conversions working shows :-)

> It appears my goal of exporting a set of
> global "built-in" functions (a la TCL) is not possible.

The possibility I alluded to in my previous email is as follows. Using python, 
parse the file you would have imported and convert it to an AST. Then, modify 
the AST in order to inject stuff. Alternatively, you could simply read in the 
file as a string and prepend the import statement. After that you could run it 
using bp::exec or PyRun_SimpleString. However, that still has the drawback 
that script code is still executed; the drawback has a workaround: don't let 
your module files contain statements for immediate execution.

Regards,
Ravi




From troy at resophonic.com  Tue Sep 30 02:21:34 2008
From: troy at resophonic.com (troy d. straszheim)
Date: Mon, 29 Sep 2008 20:21:34 -0400
Subject: [C++-sig] Importing and internal modules
In-Reply-To: <200809291932.30996.lists_ravi@lavabit.com>
References: <9e264e780809291220i15f0557aq581af0ca58e9e0f4@mail.gmail.com>	<200809291708.00340.lists_ravi@lavabit.com>	<9e264e780809291432h2144788fg22bbd13d483b8c20@mail.gmail.com>
	<200809291932.30996.lists_ravi@lavabit.com>
Message-ID: <48E1710E.40908@resophonic.com>

Ravi wrote:
> On Monday 29 September 2008 17:32:37 Peter LaDow wrote:
>> Fair enough.  I tried your example.  I'm no Python expert, and chose
>> Python because it was already popular amongst many people here at
>> work.  The difference with the Boost method, I thought, was the
>> passing of the 'globals' object to each of the imports, thinking that
>> symbols were injected into the dictionary, and that later imports
>> would see any previous imports.  I see your point.
> 
> I see. The trouble is that importing modules in a namespace adds that module 
> to the namespace rather than providing a namespace for the module being 
> imported. I can't think of a way to accomplish what you want easily.

How about this:

####  foo.py  ####
def foo1():
     print "foo1()"

def foo2():
     print "foo2()"

####  bar.py ####
def bar1():
     foo1()
     print "called bar1"

def bar2():
     foo2()
     print "called bar2"

#### driver.py ####
mod = __import__("foo")

for symbol in mod.__dict__:
     if symbol not in __builtins__.__dict__.keys() + ['__builtins__', '__file__']:
         __builtins__.__dict__[symbol] = mod.__dict__[symbol]
         print 'Injecting symbol', symbol

import bar

bar.bar1()

#### at the command line:

% ./driver.py
Injecting symbol foo1
Injecting symbol foo2
foo1()
called bar1



From ralvek at yahoo.it  Tue Sep 30 14:17:56 2008
From: ralvek at yahoo.it (k k)
Date: Tue, 30 Sep 2008 12:17:56 +0000 (GMT)
Subject: [C++-sig] Building "hello world" with Boost.Python
Message-ID: <161806.85421.qm@web27803.mail.ukl.yahoo.com>

Hi all,
i want to use a C++ dll (containing overloaded functions) with Python.
I'm trying to use Boost.Python.
I dowloaded it and now I'm doing the hello world tutorial situated in the boost installation folder in \libs\python\example\tutorial
I got also bjam and configured it to work with gcc and python 2.5.
This is the hello.cpp file to export:

#include <boost/python/module.hpp>
#include <boost/python/def.hpp>

char const* greet()
{
   return "hello, world";
}

BOOST_PYTHON_MODULE(hello_ext)
{
    using namespace boost::python;
    def("greet", greet);
}

I use this Jamroot file (that is in the same folder as hello.cpp):

# Copyright David Abrahams 2006. Distributed under the Boost
# Software License, Version 1.0. (See accompanying
# file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)

# Specify the path to the Boost project.  If you move this project
# adjust this path to refer to the Boost root directory.

use-project boost
    : ../../../.. ;

# Set up the project-wide requirements that everything uses the
# boost_python library from the project whose global ID is
# /boost/python.

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

# Declare the three extension modules.  You can specify multiple
# source files after the colon separated by spaces.

python-extension hello_ext : hello.cpp ;

# A little "rule" (function) to clean up the syntax of declaring tests
# of these extension modules.

local rule run-test ( test-name : sources + )
    {
        import testing ;
        testing.make-test run-pyd : $(sources) : : $(test-name) ;
    }

# Declare test targets

run-test hello : hello_ext hello.py ;

In the Jamroot file i changed only this:

use-project boost
  : ../../../.. ;with this(the boost root folder on on my disk):


use-project boost
  : C:/boost/boost_1_35_0 ;When i run bjam it seams that it builds well, only with some warning, but without any error.
In the tutorial they say that if all is well i should now have built the DLLs.
But the only things in there are 2 files: hello.o and hello.lib
I can't use these as python modules, how can i get the DLL? and the .pyd?

Thanks in advance.
Stefano

__________________________________________________
Do You Yahoo!?
Poco spazio e tanto spam? Yahoo! Mail ti protegge dallo spam e ti da tanto spazio gratuito per i tuoi file e i messaggi 
http://mail.yahoo.it 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080930/75c156a7/attachment.htm>

From ralvek at yahoo.it  Tue Sep 30 14:20:25 2008
From: ralvek at yahoo.it (k k)
Date: Tue, 30 Sep 2008 12:20:25 +0000 (GMT)
Subject: [C++-sig] Building "hello world" with Boost.Python
Message-ID: <189120.72518.qm@web27805.mail.ukl.yahoo.com>

Correction: the 2 output files are hello.o and hello_ext.lib


----- Messaggio originale -----
Da: k k <ralvek at yahoo.it>
A: c++-sig at python.org
Inviato: Marted? 30 settembre 2008, 14:17:56
Oggetto: Building "hello world" with Boost.Python


Hi all,
i want to use a C++ dll (containing overloaded functions) with Python.
I'm trying to use Boost.Python.
I dowloaded it and now I'm doing the hello world tutorial situated in the boost installation folder in \libs\python\example\tutorial
I got also bjam and configured it to work with gcc and python 2.5.
This is the hello.cpp file to export:

#include <boost/python/module.hpp>
#include <boost/python/def.hpp>

char const* greet()
{
   return "hello, world";
}

BOOST_PYTHON_MODULE(hello_ext)
{
    using namespace boost::python;
    def("greet", greet);
}

I use this Jamroot file (that is in the same folder as hello.cpp):

# Copyright David Abrahams 2006. Distributed under the Boost
# Software License, Version 1.0. (See accompanying
# file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)

# Specify the path to the Boost project.  If you move this project
# adjust this path to refer to the Boost root directory.

use-project boost
    : ../../../.. ;

# Set up the project-wide requirements that everything uses the
# boost_python library from the project whose global ID is
# /boost/python.

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

# Declare the three extension modules.  You can specify multiple
# source files after the colon separated by spaces.

python-extension hello_ext : hello.cpp ;

# A little "rule" (function) to clean up the syntax of declaring tests
# of these extension modules.

local rule run-test ( test-name : sources + )
    {
        import testing ;
        testing.make-test run-pyd : $(sources) : : $(test-name) ;
    }

# Declare test targets

run-test hello : hello_ext hello.py ;

In the Jamroot file i changed only this:

use-project boost
  : ../../../.. ;with this(the boost root folder on on my disk):


use-project boost
  : C:/boost/boost_1_35_0 ;When i run bjam it seams that it builds well, only with some warning, but without any error.
In the tutorial they say that if all is well i should now have built the DLLs.
But the only things in there are 2 files: hello.o and hello.lib
I can't use these as python modules, how can i get the DLL? and the .pyd?

Thanks in advance.
Stefano




__________________________________________________
Do You Yahoo!?
Poco spazio e tanto spam? Yahoo! Mail ti protegge dallo spam e ti da tanto spazio gratuito per i tuoi file e i messaggi 
http://mail.yahoo.it 


      Scopri il blog di Yahoo! Mail:
Trucchi, novit? e la tua opinione.
http://www.ymailblogit.com/blog
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080930/34933da7/attachment.htm>

From dave at boostpro.com  Tue Sep 30 16:44:44 2008
From: dave at boostpro.com (David Abrahams)
Date: Tue, 30 Sep 2008 10:44:44 -0400
Subject: [C++-sig] Building "hello world" with Boost.Python
In-Reply-To: <189120.72518.qm@web27805.mail.ukl.yahoo.com> (k. k.'s message of
	"Tue, 30 Sep 2008 12:20:25 +0000 (GMT)")
References: <189120.72518.qm@web27805.mail.ukl.yahoo.com>
Message-ID: <871vz1bsmb.fsf@mcbain.luannocracy.com>


on Tue Sep 30 2008, k k <ralvek-AT-yahoo.it> wrote:

> When i run bjam it seams that it builds well, only with some warning, but without
> any error.
> In the tutorial they say that if all is well i should now have built the DLLs.
> But the only things in there are 2 files: hello.o and hello.lib
> I can't use these as python modules, how can i get the DLL? and the .pyd?

Try running bjam with -n -a.  The output will be all the build commands
it used.  Can you see where it created the files you are looking for?

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


From ralvek at yahoo.it  Tue Sep 30 17:40:06 2008
From: ralvek at yahoo.it (k k)
Date: Tue, 30 Sep 2008 15:40:06 +0000 (GMT)
Subject: [C++-sig] Building "hello world" with Boost.Python
Message-ID: <43701.24155.qm@web27805.mail.ukl.yahoo.com>

Yes, running bjam -n -a i can see that it really creates hello_ext.pyd:

gcc.link.dll bin\gcc-mingw-3.4.5\debug\hello_ext.lib

    "g++" -L"C:\Python25\libs" -Wl,-R -Wl,"C:\Python25" -Wl,-R -Wl,"C:\Python25\
libs" "-Wl,--out-implib,bin\gcc-mingw-3.4.5\debug\hello_ext.lib" -o "bin\gcc-min
gw-3.4.5\debug\hello_ext.pyd"  -shared -Wl,--start-group "bin\gcc-mingw-3.4.5\de
bug\hello.o" "..\..\..\..\bin.v2\libs\python\build\gcc-mingw-3.4.5\debug\boost_p
ython-mgw34-d-1_35.lib"  -Wl,-Bstatic  -Wl,-Bdynamic -lpython25 -Wl,--end-group
-g

but after some other operations it deletes this file(as i understand it):

RmTemps bin\hello.test\gcc-mingw-3.4.5\debug\hello

    del /f /q "bin\gcc-mingw-3.4.5\debug\hello_ext.pyd"

this is strange, isn't it? how can i tell it to do not delete it?





----- Messaggio originale -----
Da: David Abrahams <dave at boostpro.com>
A: Development of Python/C++ integration <c++-sig at python.org>
Inviato: Marted? 30 settembre 2008, 16:44:44
Oggetto: Re: [C++-sig] Building "hello world" with Boost.Python


on Tue Sep 30 2008, k k <ralvek-AT-yahoo.it> wrote:

> When i run bjam it seams that it builds well, only with some warning, but without
> any error.
> In the tutorial they say that if all is well i should now have built the DLLs.
> But the only things in there are 2 files: hello.o and hello.lib
> I can't use these as python modules, how can i get the DLL? and the .pyd?

Try running bjam with -n -a.  The output will be all the build commands
it used.  Can you see where it created the files you are looking for?

-- 
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



      Scopri il blog di Yahoo! Mail:
Trucchi, novit? e la tua opinione.
http://www.ymailblogit.com/blog
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080930/cc3210cb/attachment.htm>

From dave at boostpro.com  Tue Sep 30 17:50:30 2008
From: dave at boostpro.com (David Abrahams)
Date: Tue, 30 Sep 2008 11:50:30 -0400
Subject: [C++-sig] Building "hello world" with Boost.Python
In-Reply-To: <43701.24155.qm@web27805.mail.ukl.yahoo.com> (k. k.'s message of
	"Tue, 30 Sep 2008 15:40:06 +0000 (GMT)")
References: <43701.24155.qm@web27805.mail.ukl.yahoo.com>
Message-ID: <87skrhab09.fsf@mcbain.luannocracy.com>


on Tue Sep 30 2008, k k <ralvek-AT-yahoo.it> wrote:

> Yes, running bjam -n -a i can see that it really creates hello_ext.pyd:
>
> gcc.link.dll bin\gcc-mingw-3.4.5\debug\hello_ext.lib
>
>     "g++" -L"C:\Python25\libs" -Wl,-R -Wl,"C:\Python25" -Wl,-R -Wl,"C:\Python25\
> libs" "-Wl,--out-implib,bin\gcc-mingw-3.4.5\debug\hello_ext.lib" -o "bin\gcc-min
> gw-3.4.5\debug\hello_ext.pyd"  -shared -Wl,--start-group "bin\gcc-mingw-3.4.5\de
> bug\hello.o" "..\..\..\..\bin.v2\libs\python\build\gcc-mingw-3.4.5\debug\boost_p
> ython-mgw34-d-1_35.lib"  -Wl,-Bstatic  -Wl,-Bdynamic -lpython25 -Wl,--end-group
> -g
>
> but after some other operations it deletes this file(as i understand it):
>
> RmTemps bin\hello.test\gcc-mingw-3.4.5\debug\hello
>
>     del /f /q "bin\gcc-mingw-3.4.5\debug\hello_ext.pyd"
>
> this is strange, isn't it? how can i tell it to do not delete it?

Pass --preserve-test-targets to bjam, and/or remove the test target from
the Jamfile.

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


From ralvek at yahoo.it  Tue Sep 30 18:04:33 2008
From: ralvek at yahoo.it (k k)
Date: Tue, 30 Sep 2008 16:04:33 +0000 (GMT)
Subject: [C++-sig] Building "hello world" with Boost.Python
Message-ID: <21133.40564.qm@web27805.mail.ukl.yahoo.com>

I passed --preserve-test-targets to bjam and i removed test target from Jamfile,
and now it works !!

Thank you very much !



----- Messaggio originale -----
Da: David Abrahams <dave at boostpro.com>
A: Development of Python/C++ integration <c++-sig at python.org>
Inviato: Marted? 30 settembre 2008, 17:50:30
Oggetto: Re: [C++-sig] Building "hello world" with Boost.Python


on Tue Sep 30 2008, k k <ralvek-AT-yahoo.it> wrote:

> Yes, running bjam -n -a i can see that it really creates hello_ext.pyd:
>
> gcc.link.dll bin\gcc-mingw-3.4.5\debug\hello_ext.lib
>
>     "g++" -L"C:\Python25\libs" -Wl,-R -Wl,"C:\Python25" -Wl,-R -Wl,"C:\Python25\
> libs" "-Wl,--out-implib,bin\gcc-mingw-3.4.5\debug\hello_ext.lib" -o "bin\gcc-min
> gw-3.4.5\debug\hello_ext.pyd"  -shared -Wl,--start-group "bin\gcc-mingw-3.4.5\de
> bug\hello.o" "..\..\..\..\bin.v2\libs\python\build\gcc-mingw-3.4.5\debug\boost_p
> ython-mgw34-d-1_35.lib"  -Wl,-Bstatic  -Wl,-Bdynamic -lpython25 -Wl,--end-group
> -g
>
> but after some other operations it deletes this file(as i understand it):
>
> RmTemps bin\hello.test\gcc-mingw-3.4.5\debug\hello
>
>     del /f /q "bin\gcc-mingw-3.4.5\debug\hello_ext.pyd"
>
> this is strange, isn't it? how can i tell it to do not delete it?

Pass --preserve-test-targets to bjam, and/or remove the test target from
the Jamfile.

-- 
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


__________________________________________________
Do You Yahoo!?
Poco spazio e tanto spam? Yahoo! Mail ti protegge dallo spam e ti da tanto spazio gratuito per i tuoi file e i messaggi 
http://mail.yahoo.it 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20080930/720ab8f3/attachment.htm>