Nabble - swig-user

Passing structure pointings from a library to the other

2009-12-10T02:53:49Z

Dear all,

I have a C library (std_comp) with a header (std_comp.h) that defines a
structure like this:

typedef struct pset_impl PSet;

a function to get a new instance of it:

PSet * pset_new(char * name);

and a getter function:

char * get_name(PSet * pset);

The related code (std_comp.c) implements this structure as:

struct pset_impl {
char * name;
};
typedef struct pset_impl pset_impl;

and the functions as well:

PSet * pset_new(char * name) {
/* code omitted */
}

char * get_name(PSet * pset) [
return pset->name;
}

(I have simplified a bit).

If I compile dynamically this library and I link and use it inside a C
`main` it works perfectly.

Then I have create the Python bindings (as an extension module) through
SWIG using this interface:

%module std_comp
%{
#include "std_comp.h"
%}
%include "std_comp.h"

It works without problems and produces some Python bindings that work
perfectly as well:

>>> from std_comp import *
>>>
>>> p = pset_new("hello world")
>>> print get_name(p)
hello world

So far so good.

Now I have another C library (cont_cpl) that uses the structures and
functions above linking std_comp, example:

-- cont_cpl.h --

#include <std_comp.h>

char * get_pset_name(PSet * pset);

-- cont_cpl.c --

#include <std_comp.h>

char * get_pset_name(PSet * pset) {
return get_name(pset);
}

(the above is a simplification of the real library).

Again, if I compile this library linking std_comp and afterwords I
use/link it in a C `main`, I have no problems, it works.

Here comes the problem. I have created the SWIG Python bindings (as an
extension module) to this library using the the following interface:

%module cont_cpl

%{
#include "cont_cpl.h"
%}
%include "cont_cpl.h"

Apparently the bindings are created without problems, but if I try to
get a new instance of a PSet using the std_comp module and I pass the
pointer to the cont_cpl function I get an error:

>>> import std_comp
>>> import cont_cpl
>>>
>>> p = std_comp.pset_new("hello world")
>>> print p
<Swig Object of type 'PSet *' at 0x82841a0>
>>> print cont_cpl.get_pset_name(p)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: in method 'get_pset_name', argument 1 of type 'PSet *'

It seems like there is a type casting problem, but I don't know how and
I don't know how to work around this issue.

Any help is really appreciated.

Thanks in advance.

Luigi

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Python C++ extension (multithreadedenvironment)(Somethign useful ? )

2009-12-09T17:38:37Z

I SOLVED IT ...

thanks for people who provided me help .. but I believe solved it myself. .haha .. took me long time..

so .. the solution is something like this..

in my c function ..which calls python call ..

[code]
PyObject *argobj = SWIG_NewPointerObj(SWIG_as_voidptr(&msg2), SWIGTYPE_p_RFAMessageWrapper, SWIG_POINTER_NEW | 0 );
try {
  arglist = Py_BuildValue("(O)",argobj);             // Build argument list
  PyEval_CallObject(func,arglist);     // Call Python
} catch (...) {
  cout<<"Unknown exception..."<<endl;
}
[\code]

I am using

PyObject *argobj = SWIG_NewPointerObj(SWIG_as_voidptr(&msg2), SWIGTYPE_p_RFAMessageWrapper, SWIG_POINTER_NEW | 0 );

This was incorrect... i guess SWIG_POINTER_NEW is used to create pointer to the actual c++ object where as if I pass 0

PyObject *argobj = SWIG_NewPointerObj(SWIG_as_voidptr(&msg2), SWIGTYPE_p_RFAMessageWrapper, 0 | 0 );

This actually creates a pointer to the wrapped python object :)

I could access my object in my python code now.

I had to include std_string.i to enable string typemaps :)

Thanks,

Alok

From: Jadhav, Alok
Sent: Wednesday, December 09, 2009 3:42 PM
To: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multithreadedenvironment)(Somethign useful ? )

So basically, I have a C++ pointer object available.

From this I want to get (or create) a new object which is proxy to my RFAMessageWrapper class.

Example:

def callback_fun(msg):
print "RFAMessage received for service:"+msg.service

In python function above

msg is <Swig Object of type 'RFAMessageWrapper *' at 0x00DC93C0>

where as when I create a normal proxy object for some other class eg

client is <RFAClient.RFAClient; proxy of <Swig Object of type 'RFAClient *' at 0x00DC98C0>>

So from Swig Object of type 'RFAMessageWrapper' I want to get a new object which is something like

msg2 is <RFAClient.RFAMessageWrapper; proxy of <Swig Object of type 'RFAMessageWrapper *' at 0x00DC93C0>>

Can someone please tell me how to create a new proxy object using the given C++ pointer object.

I believe this should solve my issue.

Regards, Alok

From: Jadhav, Alok
Sent: Wednesday, December 09, 2009 1:56 PM
To: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threadedenvironment)(Somethign useful ? )

Everything boils down to one question now:

When I pass a C++ object to Python function, I want to be able to use that object using its Proxy class in the python code. How can I achieve that?

Regards, Alok

From: Jadhav, Alok
Sent: Tuesday, December 08, 2009 6:25 PM
To: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)(Somethign useful ? )

I used SWIG_NewPointerObj method to create a PyObject. It removed all my exceptions and seems to be working fine.

Callback is called properly.

But now, in Python code

Access to the passed object raises exception. I am expecting RFAMessageWrapper object in python code. But using SWIG_NewPointerObj passes SwigPyObject to python instead.

Python call back function is
[code]
def callback_fun(msg):
    try:
        print "RFAMessage received for service:"+msg.service
    except Exception:
        print "Exception handled"
        print type(msg)
        traceback.print_exc(file=sys.stdout)
[/code]

Result:

[code]
Exception handled
<type 'SwigPyObject'>
Traceback (most recent call last):
File "<stdin>", line 3, in callback_fun
AttributeError: 'SwigPyObject' object has no attribute 'service'
[/code]

I am expecting the object type as RFAMessageWrapper here. Even when the PyObject created in C code uses SWIGTYPE_p_RFAMessageWrapper as the typeinfo, still the object created is not RFAMessageWrapper.

What should I pass to callback PyEval_CallObject(func,arglist); so that I can access RFAMessageWrapper object in Python code.

RFAMessageWrapper class is wrapped using SWIG properly.

Any help is appreciated.

Regards, Alok

From: Jadhav, Alok
Sent: Tuesday, December 08, 2009 9:42 AM
To: Jadhav, Alok; 'Brian Cole'
Cc: 'swig-user@...'
Subject: RE: [Swig-user] Python C++ extension (multi threaded environment) (Somethign useful ? )

Someone from Python-C++ mailing list told me that the way I create argument list is incorrect. They did not have knowledge of SWIG so they couldn't help more.

In the function below

RFAMessageWrapper msg2(msg);
try {
  arglist = Py_BuildValue("(O)",msg2);             // Build argument list
  //arglist = Py_BuildValue("(i)", blah);
  PyEval_CallObject(func,arglist);     // Call Python
} catch (...) {
  cout<<"Unknown exception..."<<endl;
}

The code arglist = Py_BuildValue("(O)",msg2);

Here I create msg2 object as

RFAMessageWrapper msg2(msg);

which is on stack.

Other person told me that this is incorrect. I should create this object as PyObject and then call Py_BuildValue function. It should be created using PyObject_New.

Since RFAMessageWrapper is already wrapped using SWIG, I believe there should already be a function which would help me in creating a new PyObject for RFAMessageWrapper.

Is this correct?

Can someone tell me which function I should use to create a new PyObject type object for RFAMessageWrapper?

Regards,

Alok

From: Jadhav, Alok
Sent: Tuesday, December 08, 2009 8:21 AM
To: Jadhav, Alok; Brian Cole
Cc: swig-user@...
Subject: RE: [Swig-user] Python C++ extension (multi threaded environment)

I still don't have any progress. Funny thing is that, I am not even able to catch this exception !!!!

I put a try catch block in C code like below

While debugging, PyEval_CallObject(func,arglist); is executed and it raises exception but it is not cought in this code.

I put a try catch block in my python callback function as well ...

def callback_fun(msg):
    try:
        print "RFAMessage received for service:"+msg.service
    except Exception:
        print "Exception handled"

Here again exception is not handled.

msg.service accesss raises Access violation error. The memory at some address could not be read. Any access to msg object (of type RFAMessageWrapper) which is passed from C++ code raises this exception. It doesn't occurs for basic data types. I wonder what I am missing here.

Regards, Alok

From: Jadhav, Alok
Sent: Monday, December 07, 2009 11:02 AM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

In the example below, the example of int is working fine. The callback is called properly in Python and the int value is displayed. All good.

But when I pass the object of type RFAMessageWrapper, even type(msg) raises an exception. I am not able to send user defined c++ object to Python. Not sure what is wrong here.

Also the usage of

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

doesn't raise any exception. But I wonder if it succeeds in acquiring the GIL.

Please advice.

Regards, Alok

From: Jadhav, Alok
Sent: Saturday, December 05, 2009 12:32 PM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

I will try to use that but before using the SWIG macros I should be able to build the arguments list from the C objects.

arglist = Py_BuildValue("(O)",msg2); // Build argument list

Above line raises an Access violation exception.

msg2 is non nul C class object.

where as

int blah =1;

arglist = Py_BuildValue("i",blah); // Build argument list

this line doesn't raise any exception.

So it means it has to do with creating a PyObject from C object. (Works fine for basic datatypes.)

What is the issue in this case? How can this be solved?

Regards, Alok

From: Brian Cole [mailto:coleb@...]
Sent: Saturday, December 05, 2009 3:52 AM
To: Jadhav, Alok
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

This is a shot in the dark, but have you tried using SWIG's macros for acquiring and releasing the GIL?

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

-Brian

On Dec 3, 2009, at 11:57 PM, Jadhav, Alok wrote:

Guys,

I created a new wrapper class RFAMesageWrapper which is defined in SWIG
interface file.

Now the new code looks like

RFAMessageWrapper msg2(msg);
  arglist = Py_BuildValue("(O)",msg2);             // Build argument
list

Where as msg2 is identified by Python. Still this function raises
access violation exception. So my thinking that it was because
RFAMessage is not identified as RFAMessage was incorrect.

Now back to square one. . What is wrong here? Why does this line of code
raises exception?

Still couldn't identify why state = PyGILState_Ensure(); blocks the
thread.

Regards,
Alok

-----Original Message-----
From: Jadhav, Alok
Sent: Friday, December 04, 2009 3:05 PM
To: swig-user@...
Subject: [Swig-user] Python C++ extension (multi threaded environment)

Hello,

I have made some progress in using C library to call from Python. In
next step I have to register a callback function so that C can call
ProcessMessage() function from which is written in python. My code is
not working properly and I have couple of questions regarding this.

Backgroun
----------

Python creates RFAClient object. This happens in Python main thread. Now
Python calls a function to start postoffice. Postoffice spawns a new C++
thread and it runs in a while loop listening to some data. Now when a
message arrives on this thread it should call Python function to process
this data. Looks simple .. But multi threaded environment, C++ and
python all together makes it a bit messy :)

Issues
------

My callback registed in C looks like this

static void PythonCallBack(RFAMessage msg, void *clientdata) { PyObject
*func, *arglist;

PyGILState_STATE state;
state = PyGILState_Ensure();

  func = (PyObject *) clientdata;               // Get Python function
  arglist = Py_BuildValue("(O)",msg);             // Build argument
list
  PyEval_CallObject(func,arglist);     // Call Python
  Py_DECREF(arglist);                           // Trash arglist

PyGILState_Release(state);
}

Here 2 lines are giving me exceptions.. (Remember this is called from
postoffice thread.. which is not the main Python thread. .. I have
called
PyEval_InitThreads() function; at the start of main thread)

-> arglist = Py_BuildValue("(O)",msg);
   - Raises Access Violation error. I have no idea why this is
occuring.
My reasoning : msg is of type RFAMessage. RFAMessage is not part of my
swig interface file, but it is part of the library used by my C++
wrapper project. Can RAFMessage object be converted to PyObject if it is
not part of SWIG interface file?
What are your suggestions on this?

-> state = PyGILState_Ensure();
    The thread just gets blocked on this call. It doesn't do anything
after this. After searching on internet I couldn't find anything which
can describe why postOffice thread may get blocked here. Any reasons?
solutions?

Please advice. I have spent more than 2 days trying to tackle this
single issue.

Regards,
Alok
--
View this message in context:
http://old.nabble.com/Python-C%2B%2B-extension-%28multi-threaded-environ
ment%29-tp26635775p26635775.html
Sent from the swig-user mailing list archive at Nabble.com.

------------------------------------------------------------------------
------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

===============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
===============================================================================

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

==============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
==============================================================================

------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev

_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Python C++ extension (multi threadedenvironment)(Somethign useful ? )

2009-12-08T22:41:59Z

So basically, I have a C++ pointer object available.

From this I want to get (or create) a new object which is proxy to my RFAMessageWrapper class.

Example:

def callback_fun(msg):
print "RFAMessage received for service:"+msg.service

In python function above

msg is <Swig Object of type 'RFAMessageWrapper *' at 0x00DC93C0>

where as when I create a normal proxy object for some other class eg

client is <RFAClient.RFAClient; proxy of <Swig Object of type 'RFAClient *' at 0x00DC98C0>>

So from Swig Object of type 'RFAMessageWrapper' I want to get a new object which is something like

msg2 is <RFAClient.RFAMessageWrapper; proxy of <Swig Object of type 'RFAMessageWrapper *' at 0x00DC93C0>>

Can someone please tell me how to create a new proxy object using the given C++ pointer object.

I believe this should solve my issue.

Regards, Alok

From: Jadhav, Alok
Sent: Wednesday, December 09, 2009 1:56 PM
To: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threadedenvironment)(Somethign useful ? )

Everything boils down to one question now:

When I pass a C++ object to Python function, I want to be able to use that object using its Proxy class in the python code. How can I achieve that?

Regards, Alok

From: Jadhav, Alok
Sent: Tuesday, December 08, 2009 6:25 PM
To: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)(Somethign useful ? )

I used SWIG_NewPointerObj method to create a PyObject. It removed all my exceptions and seems to be working fine.

Callback is called properly.

But now, in Python code

Access to the passed object raises exception. I am expecting RFAMessageWrapper object in python code. But using SWIG_NewPointerObj passes SwigPyObject to python instead.

Result:

[code]
Exception handled
<type 'SwigPyObject'>
Traceback (most recent call last):
File "<stdin>", line 3, in callback_fun
AttributeError: 'SwigPyObject' object has no attribute 'service'
[/code]

What should I pass to callback PyEval_CallObject(func,arglist); so that I can access RFAMessageWrapper object in Python code.

RFAMessageWrapper class is wrapped using SWIG properly.

Any help is appreciated.

Regards, Alok

Someone from Python-C++ mailing list told me that the way I create argument list is incorrect. They did not have knowledge of SWIG so they couldn't help more.

In the function below

The code arglist = Py_BuildValue("(O)",msg2);

Here I create msg2 object as

RFAMessageWrapper msg2(msg);

which is on stack.

Other person told me that this is incorrect. I should create this object as PyObject and then call Py_BuildValue function. It should be created using PyObject_New.

Since RFAMessageWrapper is already wrapped using SWIG, I believe there should already be a function which would help me in creating a new PyObject for RFAMessageWrapper.

Is this correct?

Can someone tell me which function I should use to create a new PyObject type object for RFAMessageWrapper?

Regards,

Alok

I still don't have any progress. Funny thing is that, I am not even able to catch this exception !!!!

I put a try catch block in C code like below

While debugging, PyEval_CallObject(func,arglist); is executed and it raises exception but it is not cought in this code.

I put a try catch block in my python callback function as well ...

def callback_fun(msg):
    try:
        print "RFAMessage received for service:"+msg.service
    except Exception:
        print "Exception handled"

Here again exception is not handled.

Regards, Alok

From: Jadhav, Alok
Sent: Monday, December 07, 2009 11:02 AM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

In the example below, the example of int is working fine. The callback is called properly in Python and the int value is displayed. All good.

But when I pass the object of type RFAMessageWrapper, even type(msg) raises an exception. I am not able to send user defined c++ object to Python. Not sure what is wrong here.

Also the usage of

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

doesn't raise any exception. But I wonder if it succeeds in acquiring the GIL.

Please advice.

Regards, Alok

From: Jadhav, Alok
Sent: Saturday, December 05, 2009 12:32 PM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

I will try to use that but before using the SWIG macros I should be able to build the arguments list from the C objects.

arglist = Py_BuildValue("(O)",msg2); // Build argument list

Above line raises an Access violation exception.

msg2 is non nul C class object.

where as

int blah =1;

arglist = Py_BuildValue("i",blah); // Build argument list

this line doesn't raise any exception.

So it means it has to do with creating a PyObject from C object. (Works fine for basic datatypes.)

What is the issue in this case? How can this be solved?

Regards, Alok

This is a shot in the dark, but have you tried using SWIG's macros for acquiring and releasing the GIL?

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

-Brian

On Dec 3, 2009, at 11:57 PM, Jadhav, Alok wrote:

Guys,

I created a new wrapper class RFAMesageWrapper which is defined in SWIG
interface file.

Now the new code looks like

RFAMessageWrapper msg2(msg);
  arglist = Py_BuildValue("(O)",msg2);             // Build argument
list

Where as msg2 is identified by Python. Still this function raises
access violation exception. So my thinking that it was because
RFAMessage is not identified as RFAMessage was incorrect.

Now back to square one. . What is wrong here? Why does this line of code
raises exception?

Still couldn't identify why state = PyGILState_Ensure(); blocks the
thread.

Regards,
Alok

-----Original Message-----
From: Jadhav, Alok
Sent: Friday, December 04, 2009 3:05 PM
To: swig-user@...
Subject: [Swig-user] Python C++ extension (multi threaded environment)

Hello,

I have made some progress in using C library to call from Python. In
next step I have to register a callback function so that C can call
ProcessMessage() function from which is written in python. My code is
not working properly and I have couple of questions regarding this.

Backgroun
----------

Python creates RFAClient object. This happens in Python main thread. Now
Python calls a function to start postoffice. Postoffice spawns a new C++
thread and it runs in a while loop listening to some data. Now when a
message arrives on this thread it should call Python function to process
this data. Looks simple .. But multi threaded environment, C++ and
python all together makes it a bit messy :)

Issues
------

My callback registed in C looks like this

static void PythonCallBack(RFAMessage msg, void *clientdata) { PyObject
*func, *arglist;

PyGILState_STATE state;
state = PyGILState_Ensure();

  func = (PyObject *) clientdata;               // Get Python function
  arglist = Py_BuildValue("(O)",msg);             // Build argument
list
  PyEval_CallObject(func,arglist);     // Call Python
  Py_DECREF(arglist);                           // Trash arglist

PyGILState_Release(state);
}

Here 2 lines are giving me exceptions.. (Remember this is called from
postoffice thread.. which is not the main Python thread. .. I have
called
PyEval_InitThreads() function; at the start of main thread)

-> arglist = Py_BuildValue("(O)",msg);
   - Raises Access Violation error. I have no idea why this is
occuring.
My reasoning : msg is of type RFAMessage. RFAMessage is not part of my
swig interface file, but it is part of the library used by my C++
wrapper project. Can RAFMessage object be converted to PyObject if it is
not part of SWIG interface file?
What are your suggestions on this?

-> state = PyGILState_Ensure();
    The thread just gets blocked on this call. It doesn't do anything
after this. After searching on internet I couldn't find anything which
can describe why postOffice thread may get blocked here. Any reasons?
solutions?

Please advice. I have spent more than 2 days trying to tackle this
single issue.

Regards,
Alok
--
View this message in context:
http://old.nabble.com/Python-C%2B%2B-extension-%28multi-threaded-environ
ment%29-tp26635775p26635775.html
Sent from the swig-user mailing list archive at Nabble.com.

------------------------------------------------------------------------
------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

===============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
===============================================================================

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Python C++ extension (multi threaded environment)(Somethign useful ? )

2009-12-08T20:55:58Z

Everything boils down to one question now:

When I pass a C++ object to Python function, I want to be able to use that object using its Proxy class in the python code. How can I achieve that?

Regards, Alok

From: Jadhav, Alok
Sent: Tuesday, December 08, 2009 6:25 PM
To: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)(Somethign useful ? )

I used SWIG_NewPointerObj method to create a PyObject. It removed all my exceptions and seems to be working fine.

Callback is called properly.

But now, in Python code

Access to the passed object raises exception. I am expecting RFAMessageWrapper object in python code. But using SWIG_NewPointerObj passes SwigPyObject to python instead.

Result:

[code]
Exception handled
<type 'SwigPyObject'>
Traceback (most recent call last):
File "<stdin>", line 3, in callback_fun
AttributeError: 'SwigPyObject' object has no attribute 'service'
[/code]

What should I pass to callback PyEval_CallObject(func,arglist); so that I can access RFAMessageWrapper object in Python code.

RFAMessageWrapper class is wrapped using SWIG properly.

Any help is appreciated.

Regards, Alok

Someone from Python-C++ mailing list told me that the way I create argument list is incorrect. They did not have knowledge of SWIG so they couldn't help more.

In the function below

The code arglist = Py_BuildValue("(O)",msg2);

Here I create msg2 object as

RFAMessageWrapper msg2(msg);

which is on stack.

Other person told me that this is incorrect. I should create this object as PyObject and then call Py_BuildValue function. It should be created using PyObject_New.

Since RFAMessageWrapper is already wrapped using SWIG, I believe there should already be a function which would help me in creating a new PyObject for RFAMessageWrapper.

Is this correct?

Can someone tell me which function I should use to create a new PyObject type object for RFAMessageWrapper?

Regards,

Alok

I still don't have any progress. Funny thing is that, I am not even able to catch this exception !!!!

I put a try catch block in C code like below

While debugging, PyEval_CallObject(func,arglist); is executed and it raises exception but it is not cought in this code.

I put a try catch block in my python callback function as well ...

def callback_fun(msg):
    try:
        print "RFAMessage received for service:"+msg.service
    except Exception:
        print "Exception handled"

Here again exception is not handled.

Regards, Alok

From: Jadhav, Alok
Sent: Monday, December 07, 2009 11:02 AM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

In the example below, the example of int is working fine. The callback is called properly in Python and the int value is displayed. All good.

But when I pass the object of type RFAMessageWrapper, even type(msg) raises an exception. I am not able to send user defined c++ object to Python. Not sure what is wrong here.

Also the usage of

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

doesn't raise any exception. But I wonder if it succeeds in acquiring the GIL.

Please advice.

Regards, Alok

From: Jadhav, Alok
Sent: Saturday, December 05, 2009 12:32 PM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

I will try to use that but before using the SWIG macros I should be able to build the arguments list from the C objects.

arglist = Py_BuildValue("(O)",msg2); // Build argument list

Above line raises an Access violation exception.

msg2 is non nul C class object.

where as

int blah =1;

arglist = Py_BuildValue("i",blah); // Build argument list

this line doesn't raise any exception.

So it means it has to do with creating a PyObject from C object. (Works fine for basic datatypes.)

What is the issue in this case? How can this be solved?

Regards, Alok

This is a shot in the dark, but have you tried using SWIG's macros for acquiring and releasing the GIL?

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

-Brian

On Dec 3, 2009, at 11:57 PM, Jadhav, Alok wrote:

Guys,

I created a new wrapper class RFAMesageWrapper which is defined in SWIG
interface file.

Now the new code looks like

RFAMessageWrapper msg2(msg);
  arglist = Py_BuildValue("(O)",msg2);             // Build argument
list

Where as msg2 is identified by Python. Still this function raises
access violation exception. So my thinking that it was because
RFAMessage is not identified as RFAMessage was incorrect.

Now back to square one. . What is wrong here? Why does this line of code
raises exception?

Still couldn't identify why state = PyGILState_Ensure(); blocks the
thread.

Regards,
Alok

-----Original Message-----
From: Jadhav, Alok
Sent: Friday, December 04, 2009 3:05 PM
To: swig-user@...
Subject: [Swig-user] Python C++ extension (multi threaded environment)

Hello,

I have made some progress in using C library to call from Python. In
next step I have to register a callback function so that C can call
ProcessMessage() function from which is written in python. My code is
not working properly and I have couple of questions regarding this.

Backgroun
----------

Python creates RFAClient object. This happens in Python main thread. Now
Python calls a function to start postoffice. Postoffice spawns a new C++
thread and it runs in a while loop listening to some data. Now when a
message arrives on this thread it should call Python function to process
this data. Looks simple .. But multi threaded environment, C++ and
python all together makes it a bit messy :)

Issues
------

My callback registed in C looks like this

static void PythonCallBack(RFAMessage msg, void *clientdata) { PyObject
*func, *arglist;

PyGILState_STATE state;
state = PyGILState_Ensure();

  func = (PyObject *) clientdata;               // Get Python function
  arglist = Py_BuildValue("(O)",msg);             // Build argument
list
  PyEval_CallObject(func,arglist);     // Call Python
  Py_DECREF(arglist);                           // Trash arglist

PyGILState_Release(state);
}

Here 2 lines are giving me exceptions.. (Remember this is called from
postoffice thread.. which is not the main Python thread. .. I have
called
PyEval_InitThreads() function; at the start of main thread)

-> arglist = Py_BuildValue("(O)",msg);
   - Raises Access Violation error. I have no idea why this is
occuring.
My reasoning : msg is of type RFAMessage. RFAMessage is not part of my
swig interface file, but it is part of the library used by my C++
wrapper project. Can RAFMessage object be converted to PyObject if it is
not part of SWIG interface file?
What are your suggestions on this?

-> state = PyGILState_Ensure();
    The thread just gets blocked on this call. It doesn't do anything
after this. After searching on internet I couldn't find anything which
can describe why postOffice thread may get blocked here. Any reasons?
solutions?

Please advice. I have spent more than 2 days trying to tackle this
single issue.

Regards,
Alok
--
View this message in context:
http://old.nabble.com/Python-C%2B%2B-extension-%28multi-threaded-environ
ment%29-tp26635775p26635775.html
Sent from the swig-user mailing list archive at Nabble.com.

------------------------------------------------------------------------
------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

===============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
===============================================================================

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Converting c++ object to Python object

2009-12-08T17:07:21Z

Converting c++ object to Python object

In my C++ callback function below

[code]

static void PythonCallBack(RFAMessage *msg, void *clientdata)
{
   PyObject *func, *arglist;
   PyGILState_STATE state;
   state = PyGILState_Ensure();
   func = (PyObject *) clientdata;               // Get Python function
   RFAMessageWrapper msg2(msg);

   PyObject *argobj = SWIG_NewPointerObj(SWIG_as_voidptr(&msg2), SWIGTYPE_p_RFAMessageWrapper , SWIG_POINTER_NEW | 0 );

   arglist = Py_BuildValue("(O)",argobj);             // Build argument list
   PyEval_CallObject(func,arglist);     // Call Python
Py_XDECREF(arglist);                           // Trash arglist
   //SWIG_PYTHON_THREAD_END_BLOCK;
   PyGILState_Release(state);
}

[\code]

Can you tell me if the usage of SWIG_NewPointerObj is correct? I want to convert a C++ class object into python equivalent object. In the python code, the objetc is identified as SwigPyObject and not RFAMessageWrapper which I am expecting.

How can I convert RFAMessageWrapper object into its Python equivalent to that in my Python function I can use the passed object as RFAMessageWrapper.

Regards,
Alok

Re: Python C++ extension (multi threaded environment) (Somethign useful ? )

2009-12-08T01:25:19Z

I used SWIG_NewPointerObj method to create a PyObject. It removed all my exceptions and seems to be working fine.

Callback is called properly.

But now, in Python code

Access to the passed object raises exception. I am expecting RFAMessageWrapper object in python code. But using SWIG_NewPointerObj passes SwigPyObject to python instead.

Result:

[code]
Exception handled
<type 'SwigPyObject'>
Traceback (most recent call last):
File "<stdin>", line 3, in callback_fun
AttributeError: 'SwigPyObject' object has no attribute 'service'
[/code]

What should I pass to callback PyEval_CallObject(func,arglist); so that I can access RFAMessageWrapper object in Python code.

RFAMessageWrapper class is wrapped using SWIG properly.

Any help is appreciated.

Regards, Alok

Someone from Python-C++ mailing list told me that the way I create argument list is incorrect. They did not have knowledge of SWIG so they couldn't help more.

In the function below

The code arglist = Py_BuildValue("(O)",msg2);

Here I create msg2 object as

RFAMessageWrapper msg2(msg);

which is on stack.

Other person told me that this is incorrect. I should create this object as PyObject and then call Py_BuildValue function. It should be created using PyObject_New.

Since RFAMessageWrapper is already wrapped using SWIG, I believe there should already be a function which would help me in creating a new PyObject for RFAMessageWrapper.

Is this correct?

Can someone tell me which function I should use to create a new PyObject type object for RFAMessageWrapper?

Regards,

Alok

I still don't have any progress. Funny thing is that, I am not even able to catch this exception !!!!

I put a try catch block in C code like below

While debugging, PyEval_CallObject(func,arglist); is executed and it raises exception but it is not cought in this code.

I put a try catch block in my python callback function as well ...

def callback_fun(msg):
    try:
        print "RFAMessage received for service:"+msg.service
    except Exception:
        print "Exception handled"

Here again exception is not handled.

Regards, Alok

From: Jadhav, Alok
Sent: Monday, December 07, 2009 11:02 AM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

In the example below, the example of int is working fine. The callback is called properly in Python and the int value is displayed. All good.

But when I pass the object of type RFAMessageWrapper, even type(msg) raises an exception. I am not able to send user defined c++ object to Python. Not sure what is wrong here.

Also the usage of

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

doesn't raise any exception. But I wonder if it succeeds in acquiring the GIL.

Please advice.

Regards, Alok

From: Jadhav, Alok
Sent: Saturday, December 05, 2009 12:32 PM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

I will try to use that but before using the SWIG macros I should be able to build the arguments list from the C objects.

arglist = Py_BuildValue("(O)",msg2); // Build argument list

Above line raises an Access violation exception.

msg2 is non nul C class object.

where as

int blah =1;

arglist = Py_BuildValue("i",blah); // Build argument list

this line doesn't raise any exception.

So it means it has to do with creating a PyObject from C object. (Works fine for basic datatypes.)

What is the issue in this case? How can this be solved?

Regards, Alok

This is a shot in the dark, but have you tried using SWIG's macros for acquiring and releasing the GIL?

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

-Brian

On Dec 3, 2009, at 11:57 PM, Jadhav, Alok wrote:

Guys,

I created a new wrapper class RFAMesageWrapper which is defined in SWIG
interface file.

Now the new code looks like

RFAMessageWrapper msg2(msg);
  arglist = Py_BuildValue("(O)",msg2);             // Build argument
list

Where as msg2 is identified by Python. Still this function raises
access violation exception. So my thinking that it was because
RFAMessage is not identified as RFAMessage was incorrect.

Now back to square one. . What is wrong here? Why does this line of code
raises exception?

Still couldn't identify why state = PyGILState_Ensure(); blocks the
thread.

Regards,
Alok

-----Original Message-----
From: Jadhav, Alok
Sent: Friday, December 04, 2009 3:05 PM
To: swig-user@...
Subject: [Swig-user] Python C++ extension (multi threaded environment)

Hello,

I have made some progress in using C library to call from Python. In
next step I have to register a callback function so that C can call
ProcessMessage() function from which is written in python. My code is
not working properly and I have couple of questions regarding this.

Backgroun
----------

Python creates RFAClient object. This happens in Python main thread. Now
Python calls a function to start postoffice. Postoffice spawns a new C++
thread and it runs in a while loop listening to some data. Now when a
message arrives on this thread it should call Python function to process
this data. Looks simple .. But multi threaded environment, C++ and
python all together makes it a bit messy :)

Issues
------

My callback registed in C looks like this

static void PythonCallBack(RFAMessage msg, void *clientdata) { PyObject
*func, *arglist;

PyGILState_STATE state;
state = PyGILState_Ensure();

  func = (PyObject *) clientdata;               // Get Python function
  arglist = Py_BuildValue("(O)",msg);             // Build argument
list
  PyEval_CallObject(func,arglist);     // Call Python
  Py_DECREF(arglist);                           // Trash arglist

PyGILState_Release(state);
}

Here 2 lines are giving me exceptions.. (Remember this is called from
postoffice thread.. which is not the main Python thread. .. I have
called
PyEval_InitThreads() function; at the start of main thread)

-> arglist = Py_BuildValue("(O)",msg);
   - Raises Access Violation error. I have no idea why this is
occuring.
My reasoning : msg is of type RFAMessage. RFAMessage is not part of my
swig interface file, but it is part of the library used by my C++
wrapper project. Can RAFMessage object be converted to PyObject if it is
not part of SWIG interface file?
What are your suggestions on this?

-> state = PyGILState_Ensure();
    The thread just gets blocked on this call. It doesn't do anything
after this. After searching on internet I couldn't find anything which
can describe why postOffice thread may get blocked here. Any reasons?
solutions?

Please advice. I have spent more than 2 days trying to tackle this
single issue.

Regards,
Alok
--
View this message in context:
http://old.nabble.com/Python-C%2B%2B-extension-%28multi-threaded-environ
ment%29-tp26635775p26635775.html
Sent from the swig-user mailing list archive at Nabble.com.

------------------------------------------------------------------------
------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

===============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
===============================================================================

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Python C++ extension (multi threaded environment) (Somethign useful ? )

2009-12-07T16:41:53Z

Someone from Python-C++ mailing list told me that the way I create argument list is incorrect. They did not have knowledge of SWIG so they couldn't help more.

In the function below

The code arglist = Py_BuildValue("(O)",msg2);

Here I create msg2 object as

RFAMessageWrapper msg2(msg);

which is on stack.

Other person told me that this is incorrect. I should create this object as PyObject and then call Py_BuildValue function. It should be created using PyObject_New.

Since RFAMessageWrapper is already wrapped using SWIG, I believe there should already be a function which would help me in creating a new PyObject for RFAMessageWrapper.

Is this correct?

Can someone tell me which function I should use to create a new PyObject type object for RFAMessageWrapper?

Regards,

Alok

I still don't have any progress. Funny thing is that, I am not even able to catch this exception !!!!

I put a try catch block in C code like below

While debugging, PyEval_CallObject(func,arglist); is executed and it raises exception but it is not cought in this code.

I put a try catch block in my python callback function as well ...

def callback_fun(msg):
    try:
        print "RFAMessage received for service:"+msg.service
    except Exception:
        print "Exception handled"

Here again exception is not handled.

Regards, Alok

From: Jadhav, Alok
Sent: Monday, December 07, 2009 11:02 AM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

In the example below, the example of int is working fine. The callback is called properly in Python and the int value is displayed. All good.

But when I pass the object of type RFAMessageWrapper, even type(msg) raises an exception. I am not able to send user defined c++ object to Python. Not sure what is wrong here.

Also the usage of

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

doesn't raise any exception. But I wonder if it succeeds in acquiring the GIL.

Please advice.

Regards, Alok

From: Jadhav, Alok
Sent: Saturday, December 05, 2009 12:32 PM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

I will try to use that but before using the SWIG macros I should be able to build the arguments list from the C objects.

arglist = Py_BuildValue("(O)",msg2); // Build argument list

Above line raises an Access violation exception.

msg2 is non nul C class object.

where as

int blah =1;

arglist = Py_BuildValue("i",blah); // Build argument list

this line doesn't raise any exception.

So it means it has to do with creating a PyObject from C object. (Works fine for basic datatypes.)

What is the issue in this case? How can this be solved?

Regards, Alok

This is a shot in the dark, but have you tried using SWIG's macros for acquiring and releasing the GIL?

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

-Brian

On Dec 3, 2009, at 11:57 PM, Jadhav, Alok wrote:

Guys,

I created a new wrapper class RFAMesageWrapper which is defined in SWIG
interface file.

Now the new code looks like

RFAMessageWrapper msg2(msg);
  arglist = Py_BuildValue("(O)",msg2);             // Build argument
list

Where as msg2 is identified by Python. Still this function raises
access violation exception. So my thinking that it was because
RFAMessage is not identified as RFAMessage was incorrect.

Now back to square one. . What is wrong here? Why does this line of code
raises exception?

Still couldn't identify why state = PyGILState_Ensure(); blocks the
thread.

Regards,
Alok

-----Original Message-----
From: Jadhav, Alok
Sent: Friday, December 04, 2009 3:05 PM
To: swig-user@...
Subject: [Swig-user] Python C++ extension (multi threaded environment)

Hello,

I have made some progress in using C library to call from Python. In
next step I have to register a callback function so that C can call
ProcessMessage() function from which is written in python. My code is
not working properly and I have couple of questions regarding this.

Backgroun
----------

Python creates RFAClient object. This happens in Python main thread. Now
Python calls a function to start postoffice. Postoffice spawns a new C++
thread and it runs in a while loop listening to some data. Now when a
message arrives on this thread it should call Python function to process
this data. Looks simple .. But multi threaded environment, C++ and
python all together makes it a bit messy :)

Issues
------

My callback registed in C looks like this

static void PythonCallBack(RFAMessage msg, void *clientdata) { PyObject
*func, *arglist;

PyGILState_STATE state;
state = PyGILState_Ensure();

  func = (PyObject *) clientdata;               // Get Python function
  arglist = Py_BuildValue("(O)",msg);             // Build argument
list
  PyEval_CallObject(func,arglist);     // Call Python
  Py_DECREF(arglist);                           // Trash arglist

PyGILState_Release(state);
}

Here 2 lines are giving me exceptions.. (Remember this is called from
postoffice thread.. which is not the main Python thread. .. I have
called
PyEval_InitThreads() function; at the start of main thread)

-> arglist = Py_BuildValue("(O)",msg);
   - Raises Access Violation error. I have no idea why this is
occuring.
My reasoning : msg is of type RFAMessage. RFAMessage is not part of my
swig interface file, but it is part of the library used by my C++
wrapper project. Can RAFMessage object be converted to PyObject if it is
not part of SWIG interface file?
What are your suggestions on this?

-> state = PyGILState_Ensure();
    The thread just gets blocked on this call. It doesn't do anything
after this. After searching on internet I couldn't find anything which
can describe why postOffice thread may get blocked here. Any reasons?
solutions?

Please advice. I have spent more than 2 days trying to tackle this
single issue.

Regards,
Alok
--
View this message in context:
http://old.nabble.com/Python-C%2B%2B-extension-%28multi-threaded-environ
ment%29-tp26635775p26635775.html
Sent from the swig-user mailing list archive at Nabble.com.

------------------------------------------------------------------------
------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

===============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
===============================================================================

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Python C++ extension (multi threaded environment)

2009-12-07T15:20:36Z

I still don't have any progress. Funny thing is that, I am not even able to catch this exception !!!!

I put a try catch block in C code like below

While debugging, PyEval_CallObject(func,arglist); is executed and it raises exception but it is not cought in this code.

I put a try catch block in my python callback function as well ...

def callback_fun(msg):
    try:
        print "RFAMessage received for service:"+msg.service
    except Exception:
        print "Exception handled"

Here again exception is not handled.

Regards, Alok

From: Jadhav, Alok
Sent: Monday, December 07, 2009 11:02 AM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

In the example below, the example of int is working fine. The callback is called properly in Python and the int value is displayed. All good.

But when I pass the object of type RFAMessageWrapper, even type(msg) raises an exception. I am not able to send user defined c++ object to Python. Not sure what is wrong here.

Also the usage of

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

doesn't raise any exception. But I wonder if it succeeds in acquiring the GIL.

Please advice.

Regards, Alok

From: Jadhav, Alok
Sent: Saturday, December 05, 2009 12:32 PM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

I will try to use that but before using the SWIG macros I should be able to build the arguments list from the C objects.

arglist = Py_BuildValue("(O)",msg2); // Build argument list

Above line raises an Access violation exception.

msg2 is non nul C class object.

where as

int blah =1;

arglist = Py_BuildValue("i",blah); // Build argument list

this line doesn't raise any exception.

So it means it has to do with creating a PyObject from C object. (Works fine for basic datatypes.)

What is the issue in this case? How can this be solved?

Regards, Alok

This is a shot in the dark, but have you tried using SWIG's macros for acquiring and releasing the GIL?

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

-Brian

On Dec 3, 2009, at 11:57 PM, Jadhav, Alok wrote:

Guys,

I created a new wrapper class RFAMesageWrapper which is defined in SWIG
interface file.

Now the new code looks like

RFAMessageWrapper msg2(msg);
  arglist = Py_BuildValue("(O)",msg2);             // Build argument
list

Where as msg2 is identified by Python. Still this function raises
access violation exception. So my thinking that it was because
RFAMessage is not identified as RFAMessage was incorrect.

Now back to square one. . What is wrong here? Why does this line of code
raises exception?

Still couldn't identify why state = PyGILState_Ensure(); blocks the
thread.

Regards,
Alok

-----Original Message-----
From: Jadhav, Alok
Sent: Friday, December 04, 2009 3:05 PM
To: swig-user@...
Subject: [Swig-user] Python C++ extension (multi threaded environment)

Hello,

I have made some progress in using C library to call from Python. In
next step I have to register a callback function so that C can call
ProcessMessage() function from which is written in python. My code is
not working properly and I have couple of questions regarding this.

Backgroun
----------

Python creates RFAClient object. This happens in Python main thread. Now
Python calls a function to start postoffice. Postoffice spawns a new C++
thread and it runs in a while loop listening to some data. Now when a
message arrives on this thread it should call Python function to process
this data. Looks simple .. But multi threaded environment, C++ and
python all together makes it a bit messy :)

Issues
------

My callback registed in C looks like this

static void PythonCallBack(RFAMessage msg, void *clientdata) { PyObject
*func, *arglist;

PyGILState_STATE state;
state = PyGILState_Ensure();

  func = (PyObject *) clientdata;               // Get Python function
  arglist = Py_BuildValue("(O)",msg);             // Build argument
list
  PyEval_CallObject(func,arglist);     // Call Python
  Py_DECREF(arglist);                           // Trash arglist

PyGILState_Release(state);
}

Here 2 lines are giving me exceptions.. (Remember this is called from
postoffice thread.. which is not the main Python thread. .. I have
called
PyEval_InitThreads() function; at the start of main thread)

-> arglist = Py_BuildValue("(O)",msg);
   - Raises Access Violation error. I have no idea why this is
occuring.
My reasoning : msg is of type RFAMessage. RFAMessage is not part of my
swig interface file, but it is part of the library used by my C++
wrapper project. Can RAFMessage object be converted to PyObject if it is
not part of SWIG interface file?
What are your suggestions on this?

-> state = PyGILState_Ensure();
    The thread just gets blocked on this call. It doesn't do anything
after this. After searching on internet I couldn't find anything which
can describe why postOffice thread may get blocked here. Any reasons?
solutions?

Please advice. I have spent more than 2 days trying to tackle this
single issue.

Regards,
Alok
--
View this message in context:
http://old.nabble.com/Python-C%2B%2B-extension-%28multi-threaded-environ
ment%29-tp26635775p26635775.html
Sent from the swig-user mailing list archive at Nabble.com.

------------------------------------------------------------------------
------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

===============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
===============================================================================

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: C->Perl : converting C array into perl string.

2009-12-07T02:14:44Z

Ok. Found that my typemap(out) had a fairly small but critical error: I was missing "argvi++;"

In retrospec, this was clearly demonstrated in "28.7.2 Perl5 typemaps", but somehow I missed it & had to discover it the hardway.

My revised typemap works, in case it helps anyone in future.

%typemap(out) (NAME) {
/* printf("Name is %s\n", $1); */
$result = newSVpvn($1, strlen($1));
/*
if (SvOK($result) && SvPOKp($result))
printf("Name is non-null. %s\n", SvPVX($result));
*/
argvi++;
}

--- On Fri, 12/4/09, Sri <w2learn@...> wrote:

> From: Sri <w2learn@...>
> Subject: C->Perl : converting C array into perl string.
> To: swig-user@...
> Date: Friday, December 4, 2009, 5:18 AM
> Hello,
>
> I am trying to map C array into perl a string and pass it
> as an *argument* to a perl subroutine (not as a "return
> value"). I tried a few things with typemaps (see
> below), however, unable to figureout how to accomplish
> this.
>
> My situation is like this.
>
> - One of the members in a C structure is a char array
> typedef char NAME[100];
> typedef struct { .... NAME name; ...} MyStruct;
>
> - To pass the structure to Perl, I convert the entire C
> structure & push into the perl stack (right before
> invoking the perl subroutine). Using the
> SWIG_NewPointerObj is very conveient as I dont have to
> bother with the individual data members in the structure.
>
> XPUSHs(SWIG_NewPointerObj(SWIG_as_voidptr(p_data),
> SWIGTYPE_p_tMyStruct, SWIG_SHADOW));
>
> - The perl-sub uses a swig accessor function to retrieve
> the name field.
>
> ($arg) = @_;
> $t_rname = myModule::tMyStruct::swig_name_get($arg);
>
> When I print this, perl prints a
> "p_unsigned_char=SCALAR(0x8428ad0)".
>
> I'm not sure how to now influence/change NewPointerObj so
> it can "typemap" this field so, Perl will automatically
> interpret it as a null terminated string?
>
> I suspected it may not work, but tried adding an "out"
> typemap like this, but that doesnt work : as I understand,
> this typemap only modified the behavior of the accessor
> methods returning this type (NAME), but not how the
> structure is converted into the Perl object.
>
> %typemap(out) (NAME) {
> $result = newSVpv($1, strlen($1));
> }
>
> result..
>
> SV = NULL(0x0) at 0x837a210
> REFCNT = 1
> FLAGS = ()
>
>
> Is there a way to accomplish this with typemaps or some
> other SWIG feature?
>
> Thanks
>
> %typemap(out) (NAME) {
> $result = newSVpv($1, strlen($1));
> }
>
>
>
>
>

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Python C++ extension (multi threaded environment)

2009-12-06T18:02:25Z

In the example below, the example of int is working fine. The callback is called properly in Python and the int value is displayed. All good.

But when I pass the object of type RFAMessageWrapper, even type(msg) raises an exception. I am not able to send user defined c++ object to Python. Not sure what is wrong here.

Also the usage of

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

doesn't raise any exception. But I wonder if it succeeds in acquiring the GIL.

Please advice.

Regards, Alok

From: Jadhav, Alok
Sent: Saturday, December 05, 2009 12:32 PM
To: Brian Cole
Cc: swig-user@...
Subject: Re: [Swig-user] Python C++ extension (multi threaded environment)

I will try to use that but before using the SWIG macros I should be able to build the arguments list from the C objects.

arglist = Py_BuildValue("(O)",msg2); // Build argument list

Above line raises an Access violation exception.

msg2 is non nul C class object.

where as

int blah =1;

arglist = Py_BuildValue("i",blah); // Build argument list

this line doesn't raise any exception.

So it means it has to do with creating a PyObject from C object. (Works fine for basic datatypes.)

What is the issue in this case? How can this be solved?

Regards, Alok

This is a shot in the dark, but have you tried using SWIG's macros for acquiring and releasing the GIL?

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

-Brian

On Dec 3, 2009, at 11:57 PM, Jadhav, Alok wrote:

Guys,

I created a new wrapper class RFAMesageWrapper which is defined in SWIG
interface file.

Now the new code looks like

RFAMessageWrapper msg2(msg);
  arglist = Py_BuildValue("(O)",msg2);             // Build argument
list

Where as msg2 is identified by Python. Still this function raises
access violation exception. So my thinking that it was because
RFAMessage is not identified as RFAMessage was incorrect.

Now back to square one. . What is wrong here? Why does this line of code
raises exception?

Still couldn't identify why state = PyGILState_Ensure(); blocks the
thread.

Regards,
Alok

-----Original Message-----
From: Jadhav, Alok
Sent: Friday, December 04, 2009 3:05 PM
To: swig-user@...
Subject: [Swig-user] Python C++ extension (multi threaded environment)

Hello,

I have made some progress in using C library to call from Python. In
next step I have to register a callback function so that C can call
ProcessMessage() function from which is written in python. My code is
not working properly and I have couple of questions regarding this.

Backgroun
----------

Python creates RFAClient object. This happens in Python main thread. Now
Python calls a function to start postoffice. Postoffice spawns a new C++
thread and it runs in a while loop listening to some data. Now when a
message arrives on this thread it should call Python function to process
this data. Looks simple .. But multi threaded environment, C++ and
python all together makes it a bit messy :)

Issues
------

My callback registed in C looks like this

static void PythonCallBack(RFAMessage msg, void *clientdata) { PyObject
*func, *arglist;

PyGILState_STATE state;
state = PyGILState_Ensure();

  func = (PyObject *) clientdata;               // Get Python function
  arglist = Py_BuildValue("(O)",msg);             // Build argument
list
  PyEval_CallObject(func,arglist);     // Call Python
  Py_DECREF(arglist);                           // Trash arglist

PyGILState_Release(state);
}

Here 2 lines are giving me exceptions.. (Remember this is called from
postoffice thread.. which is not the main Python thread. .. I have
called
PyEval_InitThreads() function; at the start of main thread)

-> arglist = Py_BuildValue("(O)",msg);
   - Raises Access Violation error. I have no idea why this is
occuring.
My reasoning : msg is of type RFAMessage. RFAMessage is not part of my
swig interface file, but it is part of the library used by my C++
wrapper project. Can RAFMessage object be converted to PyObject if it is
not part of SWIG interface file?
What are your suggestions on this?

-> state = PyGILState_Ensure();
    The thread just gets blocked on this call. It doesn't do anything
after this. After searching on internet I couldn't find anything which
can describe why postOffice thread may get blocked here. Any reasons?
solutions?

Please advice. I have spent more than 2 days trying to tackle this
single issue.

Regards,
Alok
--
View this message in context:
http://old.nabble.com/Python-C%2B%2B-extension-%28multi-threaded-environ
ment%29-tp26635775p26635775.html
Sent from the swig-user mailing list archive at Nabble.com.

------------------------------------------------------------------------
------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

===============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
===============================================================================

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Python C++ extension (multi threaded environment)

2009-12-04T19:31:55Z

I will try to use that but before using the SWIG macros I should be able to build the arguments list from the C objects.

arglist = Py_BuildValue("(O)",msg2); // Build argument list

Above line raises an Access violation exception.

msg2 is non nul C class object.

where as

int blah =1;

arglist = Py_BuildValue("i",blah); // Build argument list

this line doesn't raise any exception.

So it means it has to do with creating a PyObject from C object. (Works fine for basic datatypes.)

What is the issue in this case? How can this be solved?

Regards, Alok

This is a shot in the dark, but have you tried using SWIG's macros for acquiring and releasing the GIL?

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

-Brian

On Dec 3, 2009, at 11:57 PM, Jadhav, Alok wrote:

Guys,

I created a new wrapper class RFAMesageWrapper which is defined in SWIG
interface file.

Now the new code looks like

RFAMessageWrapper msg2(msg);
  arglist = Py_BuildValue("(O)",msg2);             // Build argument
list

Where as msg2 is identified by Python. Still this function raises
access violation exception. So my thinking that it was because
RFAMessage is not identified as RFAMessage was incorrect.

Now back to square one. . What is wrong here? Why does this line of code
raises exception?

Still couldn't identify why state = PyGILState_Ensure(); blocks the
thread.

Regards,
Alok

-----Original Message-----
From: Jadhav, Alok
Sent: Friday, December 04, 2009 3:05 PM
To: swig-user@...
Subject: [Swig-user] Python C++ extension (multi threaded environment)

Hello,

I have made some progress in using C library to call from Python. In
next step I have to register a callback function so that C can call
ProcessMessage() function from which is written in python. My code is
not working properly and I have couple of questions regarding this.

Backgroun
----------

Python creates RFAClient object. This happens in Python main thread. Now
Python calls a function to start postoffice. Postoffice spawns a new C++
thread and it runs in a while loop listening to some data. Now when a
message arrives on this thread it should call Python function to process
this data. Looks simple .. But multi threaded environment, C++ and
python all together makes it a bit messy :)

Issues
------

My callback registed in C looks like this

static void PythonCallBack(RFAMessage msg, void *clientdata) { PyObject
*func, *arglist;

PyGILState_STATE state;
state = PyGILState_Ensure();

  func = (PyObject *) clientdata;               // Get Python function
  arglist = Py_BuildValue("(O)",msg);             // Build argument
list
  PyEval_CallObject(func,arglist);     // Call Python
  Py_DECREF(arglist);                           // Trash arglist

PyGILState_Release(state);
}

Here 2 lines are giving me exceptions.. (Remember this is called from
postoffice thread.. which is not the main Python thread. .. I have
called
PyEval_InitThreads() function; at the start of main thread)

-> arglist = Py_BuildValue("(O)",msg);
   - Raises Access Violation error. I have no idea why this is
occuring.
My reasoning : msg is of type RFAMessage. RFAMessage is not part of my
swig interface file, but it is part of the library used by my C++
wrapper project. Can RAFMessage object be converted to PyObject if it is
not part of SWIG interface file?
What are your suggestions on this?

-> state = PyGILState_Ensure();
    The thread just gets blocked on this call. It doesn't do anything
after this. After searching on internet I couldn't find anything which
can describe why postOffice thread may get blocked here. Any reasons?
solutions?

Please advice. I have spent more than 2 days trying to tackle this
single issue.

Regards,
Alok
--
View this message in context:
http://old.nabble.com/Python-C%2B%2B-extension-%28multi-threaded-environ
ment%29-tp26635775p26635775.html
Sent from the swig-user mailing list archive at Nabble.com.

------------------------------------------------------------------------
------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

===============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
===============================================================================

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Python C++ extension (multi threaded environment)

2009-12-04T10:51:43Z

This is a shot in the dark, but have you tried using SWIG's macros for acquiring and releasing the GIL?

SWIG_PYTHON_THREAD_BEGIN_BLOCK;

SWIG_PYTHON_THREAD_END_BLOCK;

-Brian

On Dec 3, 2009, at 11:57 PM, Jadhav, Alok wrote:

Guys,

I created a new wrapper class RFAMesageWrapper which is defined in SWIG
interface file.

Now the new code looks like

RFAMessageWrapper msg2(msg);
  arglist = Py_BuildValue("(O)",msg2);             // Build argument
list

Where as msg2 is identified by Python. Still this function raises
access violation exception. So my thinking that it was because
RFAMessage is not identified as RFAMessage was incorrect.

Now back to square one. . What is wrong here? Why does this line of code
raises exception?

Still couldn't identify why state = PyGILState_Ensure(); blocks the
thread.

Regards,
Alok

-----Original Message-----
From: Jadhav, Alok
Sent: Friday, December 04, 2009 3:05 PM
To: swig-user@...
Subject: [Swig-user] Python C++ extension (multi threaded environment)

Hello,

I have made some progress in using C library to call from Python. In
next step I have to register a callback function so that C can call
ProcessMessage() function from which is written in python. My code is
not working properly and I have couple of questions regarding this.

Backgroun
----------

Python creates RFAClient object. This happens in Python main thread. Now
Python calls a function to start postoffice. Postoffice spawns a new C++
thread and it runs in a while loop listening to some data. Now when a
message arrives on this thread it should call Python function to process
this data. Looks simple .. But multi threaded environment, C++ and
python all together makes it a bit messy :)

Issues
------

My callback registed in C looks like this

static void PythonCallBack(RFAMessage msg, void *clientdata) { PyObject
*func, *arglist;

PyGILState_STATE state;
state = PyGILState_Ensure();

  func = (PyObject *) clientdata;               // Get Python function
  arglist = Py_BuildValue("(O)",msg);             // Build argument
list
  PyEval_CallObject(func,arglist);     // Call Python
  Py_DECREF(arglist);                           // Trash arglist

PyGILState_Release(state);
}

Here 2 lines are giving me exceptions.. (Remember this is called from
postoffice thread.. which is not the main Python thread. .. I have
called
PyEval_InitThreads() function; at the start of main thread)

-> arglist = Py_BuildValue("(O)",msg);
   - Raises Access Violation error. I have no idea why this is
occuring.
My reasoning : msg is of type RFAMessage. RFAMessage is not part of my
swig interface file, but it is part of the library used by my C++
wrapper project. Can RAFMessage object be converted to PyObject if it is
not part of SWIG interface file?
What are your suggestions on this?

-> state = PyGILState_Ensure();
    The thread just gets blocked on this call. It doesn't do anything
after this. After searching on internet I couldn't find anything which
can describe why postOffice thread may get blocked here. Any reasons?
solutions?

Please advice. I have spent more than 2 days trying to tackle this
single issue.

Regards,
Alok
--
View this message in context:
http://old.nabble.com/Python-C%2B%2B-extension-%28multi-threaded-environ
ment%29-tp26635775p26635775.html
Sent from the swig-user mailing list archive at Nabble.com.

------------------------------------------------------------------------
------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

===============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
===============================================================================

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

C->Perl : converting C array into perl string.

2009-12-04T05:18:14Z

Hello,

I am trying to map C array into perl a string and pass it as an *argument* to a perl subroutine (not as a "return value"). I tried a few things with typemaps (see below), however, unable to figureout how to accomplish this.

My situation is like this.

- One of the members in a C structure is a char array
typedef char NAME[100];
typedef struct { .... NAME name; ...} MyStruct;

- To pass the structure to Perl, I convert the entire C structure & push into the perl stack (right before invoking the perl subroutine). Using the SWIG_NewPointerObj is very conveient as I dont have to bother with the individual data members in the structure.

XPUSHs(SWIG_NewPointerObj(SWIG_as_voidptr(p_data), SWIGTYPE_p_tMyStruct, SWIG_SHADOW));

- The perl-sub uses a swig accessor function to retrieve the name field.

($arg) = @_;
$t_rname = myModule::tMyStruct::swig_name_get($arg);

When I print this, perl prints a "p_unsigned_char=SCALAR(0x8428ad0)".

I'm not sure how to now influence/change NewPointerObj so it can "typemap" this field so, Perl will automatically interpret it as a null terminated string?

I suspected it may not work, but tried adding an "out" typemap like this, but that doesnt work : as I understand, this typemap only modified the behavior of the accessor methods returning this type (NAME), but not how the structure is converted into the Perl object.

%typemap(out) (NAME) {
$result = newSVpv($1, strlen($1));
}

result..

SV = NULL(0x0) at 0x837a210
REFCNT = 1
FLAGS = ()

Is there a way to accomplish this with typemaps or some other SWIG feature?

Thanks

%typemap(out) (NAME) {
$result = newSVpv($1, strlen($1));
}

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Problems with Java typemaps to replace an input object parameter with a different type

2009-12-04T02:56:14Z

Hi all,
In case anyone is interested and followed this thread.

William suggested renaming the imported classes like this:
>> and the following:
>> %rename(BFoo) B::Foo;
>> %typemap(jstype) B::Foo & "B.Foo"
>> %typemap(javain) B::Foo & "B.Foo.getCPtr($javainput)"
>>
>> generates:
>> public void f(B.Foo x) {
>> exampleJNI.Problematic_f__SWIG_1(swigCPtr, this,
B.Foo.getCPtr(x), x);
>> }
>>
>> which is what I think you want.

This doesn't completely work because BFoo is not the Java fully
qualified name and then AJNI.java doesn't compile (A was the module
applying the typemap). The solution is simply doing:
%rename("B.Foo") B::Foo;
only in the modules that use the module/namespace B and no typemap is
required at all, everything just works.
SWIG will complain B.Foo is not a valid identifier and B::Foo can't be
wrapped. But indeed that's what we want because B::Foo is to be wrapped
in a different module.

Regards,
Juan

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Problems with shared_ptr and Python

2009-12-04T02:08:22Z

Hi William

Yes, this is indeed working now. Thank you very much for debugging my script.

Best regards

Andreas

---- William S Fulton <wsf@...> schrieb:

> You have three problems with your interface file. Here is the corrected
> version which works:
>
> %include "boost_shared_ptr.i"
> SWIG_SHARED_PTR(Foo, Foo)
>
> %inline %{
> #include <boost/shared_ptr.hpp>
> struct Foo {};
> void test1(boost::shared_ptr<Foo> f) {}
> void test2(boost::shared_ptr<Foo> f) {}
> %}
>
> William
>
> andreas.held@... wrote:
> > Hi William
> >
> > This is my interface file boost_example.i:
> > %module example
> >
> > // For plain classes, do not use for derived classes
> > %include "boost_shared_ptr.i"
> > MY_SWIG_SHARED_PTR(Foo, Foo)
> >
> > struct Foo {};
> > void test1(shared_ptr<Foo> f) {}
> > void test2(boost::shared_ptr<Foo> f) {}
> >
> > -----------------------------------------------------------
> >
> > My command line:
> >
> > $ swig -c++ -python boost_example.i
> > boost_example.i(6): Error: Syntax error in input(1).
> >
> > I do get the same results whether I run the command from a MSYS shell or from the Windows Power Shell, obviously, both on Windows (XP to be accurate).
> >
> > Finally, the version I am using:
> > $ swig -version
> >
> > SWIG Version 1.3.40
> >
> > Compiled with i586-mingw32msvc-g++ [i686-pc-linux-gnu]
> > Please see http://www.swig.org for reporting bugs and further information
> >
> >
> > Best regards and thanks for your help.
> >
> > Andreas
> > ---- William S Fulton <wsf@...> schrieb:
> >> andreas.held@... wrote:
> >>> Hi William
> >>>
> >>> Thank you for your quick reply. Sorry for my mistake in the command line. As you pointed out, the correct command line should be:
> >>> swig -c++ -python example.i
> >>>
> >>> Still, even with that command line I do get the same error. I now stronly suspect that something is wrong with my boost installation.
> >> It is impossible for SWIG to be affected by your boost installation
> >> unless you tell it to via an include path (the -I option). However,
> >> there might be a problem when your c++ compiler compiles the generated
> >> code, as that will then use your boost installation.
> >>
> >>> Anyway, I have now found a way around my problem by using
> >>> %template
> >>>
> >> I don't know what you are doing wrong, but the SWIG_SHARED_PTR macro
> >> invokes the %template directive for you. I think you need to post your
> >> non-working example again.
> >>
> >> William
> >
> >
>

Re: Python C++ extension (multi threaded environment)

2009-12-03T22:57:50Z

Guys,

I created a new wrapper class RFAMesageWrapper which is defined in SWIG
interface file.

Now the new code looks like

RFAMessageWrapper msg2(msg);
arglist = Py_BuildValue("(O)",msg2); // Build argument
list

Where as msg2 is identified by Python. Still this function raises
access violation exception. So my thinking that it was because
RFAMessage is not identified as RFAMessage was incorrect.

Now back to square one. . What is wrong here? Why does this line of code
raises exception?

Still couldn't identify why state = PyGILState_Ensure(); blocks the
thread.

Regards,
Alok

-----Original Message-----
From: Jadhav, Alok
Sent: Friday, December 04, 2009 3:05 PM
To: swig-user@...
Subject: [Swig-user] Python C++ extension (multi threaded environment)

Hello,

I have made some progress in using C library to call from Python. In
next step I have to register a callback function so that C can call
ProcessMessage() function from which is written in python. My code is
not working properly and I have couple of questions regarding this.

Backgroun
----------

Python creates RFAClient object. This happens in Python main thread. Now
Python calls a function to start postoffice. Postoffice spawns a new C++
thread and it runs in a while loop listening to some data. Now when a
message arrives on this thread it should call Python function to process
this data. Looks simple .. But multi threaded environment, C++ and
python all together makes it a bit messy :)

Issues
------

My callback registed in C looks like this

static void PythonCallBack(RFAMessage msg, void *clientdata) { PyObject
*func, *arglist;

PyGILState_STATE state;
state = PyGILState_Ensure();

func = (PyObject *) clientdata; // Get Python function
arglist = Py_BuildValue("(O)",msg); // Build argument
list
PyEval_CallObject(func,arglist); // Call Python
Py_DECREF(arglist); // Trash arglist

PyGILState_Release(state);
}

Here 2 lines are giving me exceptions.. (Remember this is called from
postoffice thread.. which is not the main Python thread. .. I have
called
PyEval_InitThreads() function; at the start of main thread)

-> arglist = Py_BuildValue("(O)",msg);
- Raises Access Violation error. I have no idea why this is
occuring.
My reasoning : msg is of type RFAMessage. RFAMessage is not part of my
swig interface file, but it is part of the library used by my C++
wrapper project. Can RAFMessage object be converted to PyObject if it is
not part of SWIG interface file?
What are your suggestions on this?

-> state = PyGILState_Ensure();
The thread just gets blocked on this call. It doesn't do anything
after this. After searching on internet I couldn't find anything which
can describe why postOffice thread may get blocked here. Any reasons?
solutions?

Please advice. I have spent more than 2 days trying to tackle this
single issue.

Regards,
Alok
--
View this message in context:
http://old.nabble.com/Python-C%2B%2B-extension-%28multi-threaded-environ
ment%29-tp26635775p26635775.html
Sent from the swig-user mailing list archive at Nabble.com.

------------------------------------------------------------------------
------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

===============================================================================
Please access the attached hyperlink for an important electronic communications disclaimer:
http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
===============================================================================

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Python C++ extension (multi threaded environment)

2009-12-03T22:05:04Z

Hello,

I have made some progress in using C library to call from Python. In next step I have to register a callback function so that C can call ProcessMessage() function from which is written in python. My code is not working properly and I have couple of questions regarding this.

Backgroun
----------

Python creates RFAClient object. This happens in Python main thread. Now Python calls a function to start postoffice. Postoffice spawns a new C++ thread and it runs in a while loop listening to some data. Now when a message arrives on this thread it should call Python function to process this data. Looks simple .. But multi threaded environment, C++ and python all together makes it a bit messy :)

Issues
------

My callback registed in C looks like this

static void PythonCallBack(RFAMessage msg, void *clientdata)
{
PyObject *func, *arglist;

PyGILState_STATE state;
state = PyGILState_Ensure();

func = (PyObject *) clientdata; // Get Python function
arglist = Py_BuildValue("(O)",msg); // Build argument list
PyEval_CallObject(func,arglist); // Call Python
Py_DECREF(arglist); // Trash arglist

PyGILState_Release(state);
}

Here 2 lines are giving me exceptions.. (Remember this is called from postoffice thread.. which is not the main Python thread. .. I have called PyEval_InitThreads() function; at the start of main thread)

-> arglist = Py_BuildValue("(O)",msg);
- Raises Access Violation error. I have no idea why this is occuring.
My reasoning : msg is of type RFAMessage. RFAMessage is not part of my swig interface file, but it is part of the library used by my C++ wrapper project. Can RAFMessage object be converted to PyObject if it is not part of SWIG interface file?
What are your suggestions on this?

-> state = PyGILState_Ensure();
The thread just gets blocked on this call. It doesn't do anything after this. After searching on internet I couldn't find anything which can describe why postOffice thread may get blocked here. Any reasons? solutions?

Please advice. I have spent more than 2 days trying to tackle this single issue.

Regards,
Alok

%newobject and perl5?

2009-12-03T10:40:44Z

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

Hi,
After reading the swig documentation on the %newobject directive (
http://www.swig.org/Doc1.3/Library.html ), I tried using it. However
when I run valgrind on the perl interpreter using this new module, it
tells me that there is a definite memory leak with the return values
from the functions that I have declared as %newobject.

I'm wondering if there is some detail about usage that I've missed?

In my .i file I have the following declarations:

%newobject bro_util_port;
void *bro_util_port(const char *);

%newobject bro_util_subnet;
void *bro_util_subnet( const char *);

And here's the code being wrapped in the .c file:

void *bro_util_port(const char *port){

BroPort *portout;
char proto[80];

portout = (BroPort *)malloc(sizeof(BroPort));
if (sscanf(port,"%u/%s", &(portout->port_num), proto)) {
if (strcmp(proto, "ip") == 0) {
portout->port_proto = IPPROTO_IP;
} else if (strcmp(proto, "icmp") == 0) {
portout->port_proto = IPPROTO_ICMP;
} else if (strcmp(proto, "tcp") == 0) {
portout->port_proto = IPPROTO_TCP;
} else if (strcmp(proto, "udp") == 0) {
portout->port_proto = IPPROTO_UDP;
} else {
free(portout);
fprintf(stderr, "Unknown port protocol: %s\n", proto);
return(NULL);
}
return( (void *)portout);
} else {
fprintf(stderr, "Could not parse port string %s into
####/proto\n", port);
return( NULL);
}
}

void *bro_util_subnet(char *subnet ){

BroSubnet *subnetout;
char *mask;

if (mask = strchr( subnet, '/')) {
subnetout = (BroSubnet *)malloc(sizeof( BroSubnet));

sscanf( ++mask, "%u", &subnetout->sn_width);
*--mask = (char)0;
subnetout->sn_net = bro_util_inaddr(subnet);
return((void *)subnetout);

}
fprintf(stderr, "Could not parse subnet %s into address/width\n",
subnet);
return( NULL);
}

I realize there are some input validation that the code needs in
order to be genuinely tight and correct, but for right now, I'd just
like to figure out if SWIG can handle the automatic free()'ing of the
memory that I've malloc'd.

Thanks for any help!
Steve
-----BEGIN PGP SIGNATURE-----
Version: GnuPG/MacGPG2 v2.0.12 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAksYBiwACgkQcVd2YI1BWAikCACfR95WoZY5HtADRgBEbCbb+lpH
YkwAnjKZOXb/5Pt6FvnKkZJOUy9OKdyt
=YGGB
-----END PGP SIGNATURE-----

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Problems with Java typemaps to replace an input object parameter with a different type

2009-12-03T08:24:22Z

Hello William,

Thanks for your answer

> I think might be missing something, but the type of the premature
> garbage collection prevention parameter is always long
Using the default typemaps, isn't the pgcp the same type as the input
parameter (Foo and B::Foo in this cases) and 'long' the actual parameter
passed to the underlying JNI function? Or am I misundertanding the
reference manual?

> and the following:
> %rename(BFoo) B::Foo;
> %typemap(jstype) B::Foo & "B.Foo"
> %typemap(javain) B::Foo & "B.Foo.getCPtr($javainput)"
>
> generates:
> public void f(B.Foo x) {
> exampleJNI.Problematic_f__SWIG_1(swigCPtr, this, B.Foo.getCPtr(x), x);
> }
>
> which is what I think you want.

I'm still having a small compilation problem but the %rename was the
piece I was missing, it has to be guarded by %ifdef clauses depending on
the module swig is processing but it basically works. Thanks for
pointing it out! I think I'll be able to work out the rest on my own.

> PS don't forget you can used named typemaps by changing the typemap
> types from B::Foo & to B::Foo &x to target just the x variable.
Sure, I've read it in the docs. However in this case I'm looking for a
global typemap if possible.

Best regards,
Juan

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

How do I generate throws clauses on Java SwigDirector__ functions

2009-12-03T08:00:28Z

Hi,
I have been happily using Swig to generate C# and Java on a clollection of classes and need to make callbacks on one of them. How do get Swig to generate a throws clause on SwigDirector_<Module>_<functionName> signatures in the JNI file?
For example if I have the following interface:

%module (directors="1") example
%{
#include "example.h"
%}
%javaexception("Exception") f3 "$action"
%feature("director") A;
%include "example.h"

And h file:

namespace X
{
class A {
public:
virtual void f3()=0;
};
};

I can see in the exampleJNI.java file:

public final static native void A_f3(long jarg1, A jarg1_) throws Exception;

public static void SwigDirector_A_f3(A self) {
self.f3();
}

How do I get Swig to generate "throws Exception" clause as per:
public static void SwigDirector_A_f3(A self) throws Exception {
self.f3();
}

Thanks

Re: newbie boost shared_ptr

2009-12-02T10:33:54Z

Marcel Breuers wrote:

> Marcel Breuers wrote:
> >> Dear readers,
> >>
> >> I recently started using SWIG to make C++ code available under
> Matlab by
> > generating a Java wrapper that can be called in Matlab.
> >> For basic functionality this works as expected, but boost shared
> >> pointers are giving me some trouble.
> >>
> >> For me it is not clear how to dereference the smart pointer to obtain
> >> access to the underlying pointer or object.
> >> Also the smart pointer seems to loose it ability to free memory that is
> >> no longer referenced.
> >>
> >> I was under the impression that boost_shared_ptr.i was going to solve
> >> all the memory management issues related the smart pointers and that
> >> this would also create methods to dereference a smart pointer.
> >>
> >> Any help and links to examples are appreciated.
> >>
> >> Thanks!
> >>
> >> -----------------------------------
> >>
> >> Below are samples from my code:
> >>
> >>
> >> I created a dummy C++ class Angle that contains a method to create
> a new
> >> Angle object:
> >>
> >> Angle::PointerAngle Angle::createAngleSmrt()
> >> {
> >> PointerAngle smartPtr( new Angle );
> >> return smartPtr;
> >> }
> >>
> > Angle must be a nested class. SWIG doesn't support nested classes, but
> > the next version has improved support.
> >
> >> In the header file I define:
> >> typedef boost::shared_ptr<Angle> PointerAngle;
> >>
> >> In the interface file I include:
> >>
> >> %{
> >> #include "Angle.h"
> >> %}
> >>
> >> %include "boost_shared_ptr.i"
> >> %template(SmartPtrAngle) boost::shared_ptr< Angle >;
> >>
> >> SWIG_SHARED_PTR(SmartPtrAngle,SmartPtrAngle);
> >>
> > SmartPointerAngle isn't a C type, so you have that completely wrong.
> > Probably you should have
> > SWIG_SHARED_PTR(SmartPtrAngle, Angle);
> > and skip the %template statement.
> >
> >> Angle::PointerAngle createAngleSmrt();
> >>
> > You don't show all the code, so I've not idea what PointerAngle is nor
> > Angle nor inheritance hierarchies. Please see the CHANGES file for an
> > example of SWIG_SHARED_PTR.
> >
> > William
>
> =============================
>
> Thanks for your response.
>
> I have now included the full code (except for some trivial methods).
>
> Could you please explain to me when to use:
>
> %template(SmartPtrAngle) boost::shared_ptr< Angle >;
>
> and/or
>
> SWIG_SHARED_PTR(SmartPtrAngle,SmartPtrAngle);
>
>
> You wrote
> > SWIG doesn't support nested classes, but the next version has
> improved support
>
> Does this mean I have to introduce a C++ helper class that will only be
> used to create my working horse class?
>
> Thanks in advance.
>
>

The following are the essential elements for your interface file:

%{
#include "Angle.h"
%}

%include "boost_shared_ptr.i"
//%template(SmartPtrInt) boost::shared_ptr< int >;
//%template(SmartPtrAngle) boost::shared_ptr< Angle >;
SWIG_SHARED_PTR( Angle , Angle);

// make it easy to load native code in Matlab
%pragma(java) modulecode=%{
static {
try {
System.loadLibrary("Angle");
} catch (UnsatisfiedLinkError e) {
System.err.println("Native code library failed to load. \n" + e);
System.exit(1);
}
}
%}

%include "Angle.h"

Note that you don't use %template when using SWIG_SHARED_PTR... actually
this macro has a %template in it. Your createArray() method will never
work as you are putting an array into shared_ptr... a big no no, read
the shared_ptr docs. I've ignored shared_ptr<int> because of this and
also I don't know how one would want to handle a shared_ptr of a single
primitive type from Java. SWIG_SHARED_PTR is for non-primitive types as
it requires a proxy class.

William

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Callbacks to .NET (tcl proc ivoke .NET delegate)

2009-12-01T15:02:27Z

Hi,

I would like to get the TCL stdout redirected to .NET GUI application.

Would it be sufficient to register a .NET callback to tcl_CreateChannelHandler API?

Any comments on this topic are welcome.

Thanks,

Ronen

newbie boost shared_ptr

2009-12-01T14:10:21Z

Marcel Breuers wrote:

> > Dear readers,
> >
> > I recently started using SWIG to make C++ code available under Matlab by
> generating a Java wrapper that can be called in Matlab.
> > For basic functionality this works as expected, but boost shared
> > pointers are giving me some trouble.
> >
> > For me it is not clear how to dereference the smart pointer to obtain
> > access to the underlying pointer or object.
> > Also the smart pointer seems to loose it ability to free memory that is
> > no longer referenced.
> >
> > I was under the impression that boost_shared_ptr.i was going to solve
> > all the memory management issues related the smart pointers and that
> > this would also create methods to dereference a smart pointer.
> >
> > Any help and links to examples are appreciated.
> >
> > Thanks!
> >
> > -----------------------------------
> >
> > Below are samples from my code:
> >
> >
> > I created a dummy C++ class Angle that contains a method to create a new
> > Angle object:
> >
> > Angle::PointerAngle Angle::createAngleSmrt()
> > {
> > PointerAngle smartPtr( new Angle );
> > return smartPtr;
> > }
> >
> Angle must be a nested class. SWIG doesn't support nested classes, but
> the next version has improved support.
>
> > In the header file I define:
> > typedef boost::shared_ptr<Angle> PointerAngle;
> >
> > In the interface file I include:
> >
> > %{
> > #include "Angle.h"
> > %}
> >
> > %include "boost_shared_ptr.i"
> > %template(SmartPtrAngle) boost::shared_ptr< Angle >;
> >
> > SWIG_SHARED_PTR(SmartPtrAngle,SmartPtrAngle);
> >
> SmartPointerAngle isn't a C type, so you have that completely wrong.
> Probably you should have
> SWIG_SHARED_PTR(SmartPtrAngle, Angle);
> and skip the %template statement.
>
> > Angle::PointerAngle createAngleSmrt();
> >
>
> You don't show all the code, so I've not idea what PointerAngle is nor
> Angle nor inheritance hierarchies. Please see the CHANGES file for an
> example of SWIG_SHARED_PTR.
>
> William

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

Thanks for your response.

I have now included the full code (except for some trivial methods).

Could you please explain to me when to use:

%template(SmartPtrAngle) boost::shared_ptr< Angle >;

and/or

SWIG_SHARED_PTR(SmartPtrAngle,SmartPtrAngle);

You wrote
> SWIG doesn't support nested classes, but the next version has improved support

Does this mean I have to introduce a C++ helper class that will only be
used to create my working horse class?

Thanks in advance.

My code (I have removed some trivial functions):
------------------------------------------------
%module AngleSwig

%{
#include "Angle.h"
%}

%include "boost_shared_ptr.i"
%template(SmartPtrInt) boost::shared_ptr< int >;
%template(SmartPtrAngle) boost::shared_ptr< Angle >;

SWIG_SHARED_PTR( SmartPtrAngle , SmartPtrAngle);

// make it easy to load native code in Matlab
%pragma(java) modulecode=%{
static {
try {
System.loadLibrary("Angle");
} catch (UnsatisfiedLinkError e) {
System.err.println("Native code library failed to load. \n" + e);
System.exit(1);
}
}
%}

class Angle
{
public:

enum Type { Radians, Degrees };

Angle();
explicit Angle(double i_angle, const Type& i_type = Degrees);
Angle(const Angle& i_otherAngle);
virtual ~Angle();

double radians() const;
double degrees() const;

double sin() const;
double cos() const;

static double degreesToRadians(double i_degrees);
static double radiansToDegrees(double i_radians);

void reset();

static const double PI;

Angle::PointerInt createArray();
Angle::PointerAngle createAngleSmrt();
Angle* createAngle();

private:

double m_radians;
double m_degrees;
double m_sin;
double m_cos;

void normalize();
};

-------------
HEADER_FILE:
-------------

#ifndef DDHK_ANGLE_H
#define DDHK_ANGLE_H

// STL Includes
#include <iostream>
#include <vector>

#include <boost/shared_ptr.hpp> //mb!!

class Angle
{
public:

// Type definitions
typedef std::vector<Angle> VectorType;

typedef boost::shared_ptr<int> PointerInt;
typedef boost::shared_ptr<Angle> PointerAngle;

enum Type { Radians, Degrees };

int *arrayTest;

// == I want to test how SWIG handles the next three methods ==
PointerInt createArray();
PointerAngle createAngleSmrt();
Angle* createAngle();

Angle();
explicit Angle(double i_angle, const Type& i_type = Degrees);
Angle(const Angle& i_otherAngle);
virtual ~Angle();

double radians() const;
double degrees() const;

static double degreesToRadians(double i_degrees);
static double radiansToDegrees(double i_radians);

static const double PI;

private:

double m_radians;
double m_degrees;
double m_sin;
double m_cos;

};

#endif // !DDHK_ANGLE_H

-----
CPP
-----

// Class declaration.
#include "Angle.h"

// STL Includes
#define _USE_MATH_DEFINES
#include <math.h>

// Pi constant
const double Angle::PI = M_PI;

Angle::Angle()
: m_radians(0.)
{
normalize();
}

Angle::PointerInt Angle::createArray()
{
arrayTest = new int[50000];
PointerInt smartPtr( arrayTest );
return smartPtr;
}

Angle::PointerAngle Angle::createAngleSmrt()
{
PointerAngle smartPtr( new Angle );
return smartPtr;
}

Angle* Angle::createAngle()
{
return new Angle;
}

Angle::Angle(double i_angle, const Type& i_type)
{
switch (i_type)
{
default:
case Degrees:
{
m_radians = degreesToRadians(i_angle);
break;
}
case Radians:
{
m_radians = i_angle;
break;
}
}

normalize();
}

Angle::Angle(const Angle& i_otherAngle)
{
m_radians = i_otherAngle.m_radians;
normalize();
}

Angle::~Angle()
{
// empty
}

double Angle::radians() const
{
return m_radians;
}

double Angle::degrees() const
{
return m_degrees;
}

double Angle::sin() const
{
return m_sin;
}

double Angle::degreesToRadians(double i_degrees)
{
return (i_degrees / 180.) * PI;
}

double Angle::radiansToDegrees(double i_radians)
{
return (i_radians / PI) * 180.;
}

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Getting started

2009-12-01T11:16:25Z

David Piepgrass wrote:

>>> 1. if the system header #includes other headers, you have to %import
>> those manually, and
>>> 2. SWIG gets the mistaken impression that structures defined in the
>> system headers, such as RECT, are wrapped in another SWIG module. This
>> only causes trouble if you actually need SWIG to wrap one of those
>> structures.
>>> Hey William, if you're reading this, don't you figure there should be
>> a %systeminclude command that would avoid these problems?
>>
>> How will it know which types to not import given a mix of struct,
>> primitive types and typedefs using structs, typedefs and struct
>> declared
>> in one line statment etc? There is an easy solution though, just ignore
>> any types which you won't be wrapping before it is imported using
>> %ignore, that is %ignore types that you want to treat as an opaque type
>> without a proxy class.
>
> Ahh yes, but what if you actually want SWIG to produce a wrapper for a system structure like RECT? Then you would have to use %include, and %ignore all structures that happen to be in the same system header file, other than the one you want. Or can you somehow use %ignore and still get SWIG to produce a wrapper?
>

Well yes, either you need to %import or %include a header file and
%ignore types selectively. Otherwise selectively copy types from the
header into an interface file.

> I was thinking a recursive "%systemimport" would be really handy for system header files, but then again, in the real world SWIG might get tripped up by compiler extensions like __declspec. Oh well.
>
The -importall option does this and extensions like __declspec are easy
to deal with. Usually system headers contain all sorts of junk which is
not usually needed for wrappers though.

William

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Getting started

2009-11-30T13:05:19Z

> > 1. if the system header #includes other headers, you have to %import
> those manually, and
> > 2. SWIG gets the mistaken impression that structures defined in the
> system headers, such as RECT, are wrapped in another SWIG module. This
> only causes trouble if you actually need SWIG to wrap one of those
> structures.
> >
> > Hey William, if you're reading this, don't you figure there should be
> a %systeminclude command that would avoid these problems?
>
> How will it know which types to not import given a mix of struct,
> primitive types and typedefs using structs, typedefs and struct
> declared
> in one line statment etc? There is an easy solution though, just ignore
> any types which you won't be wrapping before it is imported using
> %ignore, that is %ignore types that you want to treat as an opaque type
> without a proxy class.

Ahh yes, but what if you actually want SWIG to produce a wrapper for a system structure like RECT? Then you would have to use %include, and %ignore all structures that happen to be in the same system header file, other than the one you want. Or can you somehow use %ignore and still get SWIG to produce a wrapper?

I was thinking a recursive "%systemimport" would be really handy for system header files, but then again, in the real world SWIG might get tripped up by compiler extensions like __declspec. Oh well.

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: C structural "inheritance"?

2009-11-30T12:57:12Z

Sean wrote:

> William S Fulton wrote:
>>> Add the following to your interface file:
>>>
>>> %types(Derived = Base);
>>>
>>> This will modify the runtime type system to accept a Base * where a Derived *
>> is expected with a simple cast from Derived * to Base *. Hopefully that is all
>> you need.
>>
>> Er, correction... accept a Derived * where a Base * is expected. Sorry for
>> confusion.
>
>
> Thanks! "%types()" seems like what I am looking for, but it only does half the
> job. It does the C-level conversions, but doesn't set up the language-level
> inheritance relationships. For example in the generated Python wrapper file, you
> still get this:
>
> class Derived(_object):
> ...
>
> where you should have this:
>
> class Derived(Base):
> ...
>

This isn't possible at the moment. Well, nearly. If you use -py3, and
the abc feature:

%feature("python:abc") Derived "Base";

it will do as you want. Quite why it is only a py3 feature, I don't know.

We ought to have a general way to modify base classes for the scripting
languages as there is a way to achieve this with the statically typed
target languages.

> I'm not sure how the current (1.3.35, anyways) behavior is useful -- is it a bug?
> Should this use of %types be documented somewhere?
There is a bit in Typemaps.html and the CHANGES file.
William

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Problems with Java typemaps to replace an input object parameter with a different type

2009-11-30T12:56:42Z

Juan Hernando wrote:

> Dear all,
>
> I'm trying to wrap different C++ namespaces into different Java packages
> with SWIG 1.3.36 and so far I've been able to work out all the details
> except one.
>
> Suppose I have two namespaces A and B that map to Java packages A and B.
> In both namespaces there is an homonym class Foo which we don't want to
> rename (actually renaming would solve much of the problems but that's
> inconvenient from the user perspective).
> In A there is a class like this:
> class Problematic
> {
> void f(Foo &x);
> void f(B::Foo &x);
> };
> As the documentation says this type of code causes name clashes in the
> target language because namespaces are flattened. I could rename one of
> the overloads but then the C++ code and Java code wouldn't be the same
> and I want to avoid that.
>
> So, the solution seems to be providing a typemap for B::Foo &, but I'm
> not able to get it right.
>
> First I define a typemap for jstype
> %typemap(jstype) B::Foo & "B.Foo"
> so the Java code looks like this:
> ...
> public void f(B.Foo x) {
> ModuleAJNI.f__SWIG_1(swigCPtr, this, B.getCPtr(x), x);
> }
> ...
> The Java interface is exactly what I want but it doesn't work for 2
> reasons. First getCPtr is called on the wrong class, so I have to
> provide a typemap for javain. And then, premature garbage collection
> parameter doesn't match the type in ModuleAJNI.f__SWIG_1 because I
> haven't changed jstype. I've tried many possible combinations but
> without success. The key points are that there is no way to override the
> type of the permature garbage collection parameter (and I don't want to
> get rid of it) and that the JNI type is unique so I can't do tricks with
> the argument list without affecting other points of the C++ code. The
> closer I got was working fine but since I changed the jni type it
> affected wrapping of Foo & as a return type.
>
> Passing the object directly instead of the pointer seems to be a
> possible solution but that involves direct access to fields in the C++
> side, which is cumbersome and not less efficient that getCPtr.
> I'd appreciate very much any idea about how to solve this problem neatly?
>

I think might be missing something, but the type of the premature
garbage collection prevention parameter is always long and the following:

%rename(BFoo) B::Foo;
%typemap(jstype) B::Foo & "B.Foo"
%typemap(javain) B::Foo & "B.Foo.getCPtr($javainput)"

generates:

public void f(B.Foo x) {
exampleJNI.Problematic_f__SWIG_1(swigCPtr, this, B.Foo.getCPtr(x), x);
}

which is what I think you want.

William

PS don't forget you can used named typemaps by changing the typemap
types from B::Foo & to B::Foo &x to target just the x variable.

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: SWIGTYPE_p_p_char : how to use it ?

2009-11-30T12:56:22Z

Otmane Lahlou wrote:

> Hi List,
>
> The C type char ** is wrapped in SWIGTYPE_p_p_char when i wrapped my
> c++ class to Java.
> This class has a method with a c++ argument.
>
> How can I deal with the SWIGTYPE_p_p_char ?? How can I initialize it,
> access to its data??
>
> Thank you for help

Read about "Type Wrapper" classes and pointer to pointers in the Java
chapter. Depending on what the char ** does, write some typemaps or a
helper function to deal with this. The typemaps in Lib/java/various.i
might be useful.

William

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Swig + C# + multiple modules in different C# namespaces

2009-11-30T12:56:16Z

Herve Blanc wrote:

> I have tried to use advice
> @ http://old.nabble.com/Swig-%2B-C--%2B-multiple-modules-in-different-C--namespaces-td25286199.html,
> without success.
> I am cutting/pasting below a simple test case below to be precise. From
> the information provided earlier I would have expected the
> derivedclass.cs created by swig to include the
> "using CsharpNamespace.Base;" but it seems my typemap statement in
> derivedmodule.i is not taken into account.
> There must be something obvious another pair of eyes familiar with SWIG
> should catch easily I hope?
> Thanks in advance for any help,
> Herve.
>
> === baseclass.h ===
> namespace BaseNamespace {
> class baseclass {
> public:
> int foo();
> };
> }
>
> === derivedclass.h ===
> #include "base.h"
> namespace DerivedNamespace {
> class derivedclass : public BaseNamespace::baseclass {
> public:
> int bar();
> };
> }
>
> === basemodule.i ===
> %module BaseModule
>
> %{
> #include "baseclass.h"
> %}
>
> %include "baseclass.h"
>
> === derivedmodule.i ===
> %module DerivedModule
>
> %{
> #include "derivedclass.h"
> %}
>
> %typemap(csimports) DerivedModule
> %{
> using CsharpNamespace.Base;
> %}

DerivedModule does not name a type. Change to:

%typemap(csimports) DerivedNamespace::derivedclass
%{
using CsharpNamespace.Base;
using System;
using System.Runtime.InteropServices;
%}

William

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Getting started

2009-11-30T12:56:04Z

David Piepgrass wrote:

> Including header files is often not recommended because they tend to contain things SWIG doesn't understand: everything from special compiler-specific keywords like __declspec to mundane aliases like DWORD and PVOID that are defined in system headers (note that SWIG doesn't follow #include, because you don't want it to follow, say, <stdio.h> and create wrappers for fopen() and printf().)
>
> When I use SWIG, I do %include my header files directly. That's because I have control over my header files, and any time SWIG doesn't understand something I can always
>
> - wrap it in a "#ifndef SWIG" block so that the C++ compiler sees it but not SWIG.
> - add a typedef or #define that helps SWIG understand something, or an %ignore to make SWIG ignore it
>
> One other tip about including: you can include system headers with %import instead of %include when you want to pull in standard #defines and typedefs but ignore functions. For example I use (on Windows)
>
> %import "WinDef.h"
>
> The only problems are that
>
> 1. if the system header #includes other headers, you have to %import those manually, and
> 2. SWIG gets the mistaken impression that structures defined in the system headers, such as RECT, are wrapped in another SWIG module. This only causes trouble if you actually need SWIG to wrap one of those structures.
>
> Hey William, if you're reading this, don't you figure there should be a %systeminclude command that would avoid these problems?

How will it know which types to not import given a mix of struct,
primitive types and typedefs using structs, typedefs and struct declared
in one line statment etc? There is an easy solution though, just ignore
any types which you won't be wrapping before it is imported using
%ignore, that is %ignore types that you want to treat as an opaque type
without a proxy class.

William

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Temporarily disable std_vector.i?

2009-11-30T12:48:35Z

In article <4B0C61AC.7080207@...>,
William S Fulton <wsf@...> wrote:

> Russell Owen wrote:
> > Thank you very much for fixing this bug. It will be a huge help to our
> > project. (I can confirm that SWIG 1.3.40 fixes the problem for me; I was
> > at 1.3.36).
> >
> > Could you please expand a bit on your statement: "Yes, it is better to
> > use %template with a class type (Bar in this case) once SWIG has
> > parsed."? I'm not sure what you are recommending.
> >
> I'm trying to say that the order of your statements is important. SWIG
> must parse Bar before you use Bar in %template. This way SWIG knows all
> the type information before you use the types.

I see. Any idea what the risk is of violating this rule by defining the
objects after the %template statement?

Our real code is complex enough that it will be messy to split the
definitions into multiple headers and get the dependency order right.

> > I see that the warnings about an overlaoded method is still printed, but
> > I can't see anything to warn about anymore. Should I report it as a
> > minor bug?
> >
> The warning is there because std::vector of any type is given the same
> precedence in the overloading ranking mechanism. Actually I don't know
> if it is a problem or not! Try it out and see if you can get both the
> overloaded constructors to work. Normally you get this warning when a
> multiple C++ types have just one equivalent in the target language, eg
> Foo*, Foo&, const Foo*, const Foo& etc.

The original example (at start of this thread) demonstrates that all
versions of the C++ constructor are available. So I think the warning is
unnecessary and have filed a bug report accordingly.

-- Russell

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: Warning(303): %extend defined for an undeclared class

2009-11-30T12:47:24Z

Bill Spotz wrote:

> Hello,
>
> Can somebody explain to me what triggers a Warning(303)? I have
> successfully used %extend many times before %include-ing the
> previously undeclared class header file. So I am not even sure what
> the warning is trying to tell me.
>
> I am getting this warning now within a fairly complicated module, and
> it is preventing the wrapping of a class.
>
> I have played with directive ordering and namespace handling, all to
> no avail.
>
> I have tried to construct a simple example, but I cannot yet recreate
> the problem in a simple setting.
>
> FWIW, I am trying to wrap a class with a single constructor. I want
> to %ignore the constructor and substitute a new one with %extend. The
> class is nested in two namespaces. I have tried everything I can
> think of.
>

The following triggers it because Bah doesn't exist:

%extend Bah
{
void foo() {}
};
%inline %{
namespace Space {
struct Bah {};
}
%}

If you use instead:

%extend Space::Bah

all is well. You must have a simple namespace or typo problem somewhere.
Try using the "-debug-module 4" option to swig, which will then display
the parse tree for everything. You can check spellings, namespace that
SWIG is actually using. The svn version of SWIG also has the
-debug-csymbols option which displays all the C symbols and namespaces
they are in, which might be helpful.

William

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Re: newbie boost shared_ptr

2009-11-30T12:47:19Z

Marcel Breuers wrote:

> Dear readers,
>
> I recently started using SWIG to make C++ code available under Matlab by
> generating a Java wrapper that can be called in Matlab.
> For basic functionality this works as expected, but boost shared
> pointers are giving me some trouble.
>
> For me it is not clear how to dereference the smart pointer to obtain
> access to the underlying pointer or object.
> Also the smart pointer seems to loose it ability to free memory that is
> no longer referenced.
>
> I was under the impression that boost_shared_ptr.i was going to solve
> all the memory management issues related the smart pointers and that
> this would also create methods to dereference a smart pointer.
>
> Any help and links to examples are appreciated.
>
> Thanks!
>
> -----------------------------------
>
> Below are samples from my code:
>
>
> I created a dummy C++ class Angle that contains a method to create a new
> Angle object:
>
> Angle::PointerAngle Angle::createAngleSmrt()
> {
> PointerAngle smartPtr( new Angle );
> return smartPtr;
> }
>

Angle must be a nested class. SWIG doesn't support nested classes, but
the next version has improved support.

> In the header file I define:
> typedef boost::shared_ptr<Angle> PointerAngle;
>
> In the interface file I include:
>
> %{
> #include "Angle.h"
> %}
>
> %include "boost_shared_ptr.i"
> %template(SmartPtrAngle) boost::shared_ptr< Angle >;
>
> SWIG_SHARED_PTR(SmartPtrAngle,SmartPtrAngle);
>

SmartPointerAngle isn't a C type, so you have that completely wrong.
Probably you should have
SWIG_SHARED_PTR(SmartPtrAngle, Angle);
and skip the %template statement.

> Angle::PointerAngle createAngleSmrt();
>

You don't show all the code, so I've not idea what PointerAngle is nor
Angle nor inheritance hierarchies. Please see the CHANGES file for an
example of SWIG_SHARED_PTR.

William

------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing.
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Casting parameters to director methods

2009-11-30T09:42:37Z

I'm trying to wrap a few classes for Python. A simplified version of
them would be the following:

class Message
{
};

class Child
{
public:
void sendMessage(const Message &msg) = 0;
};

class Parent
{
public:
void sendMessage(const Message &msg)
{
// For each contained child
// call child.sendMessage(msg)
// End for
}
};

I want to extend the Message and Child classes in Python and have them
converted to correct Python classes when back in Python. I.e. I want
PyChild.sendMessage to be able to call msg.pyMethod if the msg parameter
is of type PyMessage and PyMessage has a method called pyMethod.

I have tried to use just directors but that doesn't result in the
correct casting when in Python (they come as plain Message classes, not
PyMessage), I'm guessing it's because the parameter to
Parent.sendMessage has a different memory address than the ultimate
parameter that is passed to PyChild.sendMessage, because I have seen
SWIG correctly cast to a Python extended type but then each Python
object was pointing to the same address.

Could someone please help me find the correct way to do this or suggest
an alternative interface.

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

newbie boost shared_ptr

2009-11-29T03:53:34Z

Dear readers,

I recently started using SWIG to make C++ code available under Matlab by
generating a Java wrapper that can be called in Matlab.
For basic functionality this works as expected, but boost shared
pointers are giving me some trouble.

For me it is not clear how to dereference the smart pointer to obtain
access to the underlying pointer or object.
Also the smart pointer seems to loose it ability to free memory that is
no longer referenced.

I was under the impression that boost_shared_ptr.i was going to solve
all the memory management issues related the smart pointers and that
this would also create methods to dereference a smart pointer.

Any help and links to examples are appreciated.

Thanks!

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

Below are samples from my code:

I created a dummy C++ class Angle that contains a method to create a new
Angle object:

Angle::PointerAngle Angle::createAngleSmrt()
{
PointerAngle smartPtr( new Angle );
return smartPtr;
}

In the header file I define:
typedef boost::shared_ptr<Angle> PointerAngle;

In the interface file I include:

%{
#include "Angle.h"
%}

%include "boost_shared_ptr.i"
%template(SmartPtrAngle) boost::shared_ptr< Angle >;

SWIG_SHARED_PTR(SmartPtrAngle,SmartPtrAngle);

Angle::PointerAngle createAngleSmrt();

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Swig-user mailing list
Swig-user@...
https://lists.sourceforge.net/lists/listinfo/swig-user

Nabble - swig-user

Passing structure pointings from a library to the other

Re: Python C++ extension (multithreadedenvironment)(Somethign useful ? )

Re: Python C++ extension (multi threadedenvironment)(Somethign useful ? )

Re: Python C++ extension (multi threaded environment)(Somethign useful ? )

Converting c++ object to Python object

Re: Python C++ extension (multi threaded environment) (Somethign useful ? )

Re: Python C++ extension (multi threaded environment) (Somethign useful ? )

Re: Python C++ extension (multi threaded environment)

Re: C->Perl : converting C array into perl string.

Re: Python C++ extension (multi threaded environment)

Re: Python C++ extension (multi threaded environment)

Re: Python C++ extension (multi threaded environment)

C->Perl : converting C array into perl string.

Re: Problems with Java typemaps to replace an input object parameter with a different type

Re: Problems with shared_ptr and Python

Re: Python C++ extension (multi threaded environment)

Python C++ extension (multi threaded environment)

%newobject and perl5?

Re: Problems with Java typemaps to replace an input object parameter with a different type

How do I generate throws clauses on Java SwigDirector_*_* functions

Re: newbie boost shared_ptr

Callbacks to .NET (tcl proc ivoke .NET delegate)

newbie boost shared_ptr

Re: Getting started

Re: Getting started

Re: C structural "inheritance"?

Re: Problems with Java typemaps to replace an input object parameter with a different type

Re: SWIGTYPE_p_p_char : how to use it ?

Re: Swig + C# + multiple modules in different C# namespaces

Re: Getting started

Re: Temporarily disable std_vector.i?

Re: Warning(303): %extend defined for an undeclared class

Re: newbie boost shared_ptr

Casting parameters to director methods

newbie boost shared_ptr

How do I generate throws clauses on Java SwigDirector__ functions