Mailman 3 January 2004 - Cplusplus-sig

RE: [C++-sig] Extending & embedding Python?
by James Gunn Jan. 10, 2004

Jan. 10, 2004

*** Before acting on this e-mail or opening any attachment you are advised to read the disclaimer at the end of this e-mail *** Hi Stefan, > From: Stefan Seefeld [mailto:seefeld@sympatico.ca] > > James Gunn wrote: > > > I almost had it working yesterday, however, when my script > had finished > > running > > the state had not been updated...I think this is because > instead of Python > > referencing the same C++ object it had created a new one. > Is there a way > > around > > this? I've looked through the docs and the documentation > but I'm a bit > > lost/stuck. > > > It's hard to make any guess without some more detail (such as code > snippets). If you are not sure whether the objects are passed by value > (i.e. copied) or by reference, you may want to have a close look at > boost.python's 'call policies' (for example > > http://www.boost.org/libs/python/doc/tutorial/doc/call_policies.html). > > What you have in mind should certainly work. Thanks for the quick response. I'll try and get something together to show exactly what I'm doing. Cheers, James ********************************************************************** Please note: This e-mail and its attachments contain only the opinions of the sender and do not necessarily reflect the policy(s) of the communisis group in general. Employees of the communisis group are required not to make any defamatory statements and not to infringe or authorise any infringement of copyright or any other legal right by e-mail. Any such communication is therefore outside the scope of employment of the individual concerned. The communisis group will not accept any liability in respect of such a communication. Confidentiality: This e-mail and any attachments, together with their contents, are confidential unless otherwise explicitly stated in writing by the sender of this e-mail and are for the intended recipient only. If they have come to you in error you must not take any action in respect of them, which includes but is not limited to reproducing, sending or storing them, other than to notifying the sender immediately of the mistake, and deleting the e-mail, any attachments and any reproductions made by replying to it. Viruses: This e-mail and any attachments have been scanned for viruses but we cannot guarantee that they are virus free. The recipient should check this e-mail and any attachments for viruses. The communisis group accepts no responsibility for any damage caused by any virus transmitted by this e-mail or any of its attachments. In the event of any unauthorised copying or forwarding, the recipient will be required to indemnify the communisis group against any claim for loss or damage caused by any viruses or otherwise. ********************************************************************** ________________________________________________________________________ This email has been scanned for all viruses by the MessageLabs Email Security System. For more information on a proactive email security service working around the clock, around the globe, visit http://www.messagelabs.com ________________________________________________________________________

1 0

Extending & embedding Python?
by James Gunn Jan. 10, 2004

Jan. 10, 2004

*** Before acting on this e-mail or opening any attachment you are advised to read the disclaimer at the end of this e-mail *** Hi all, I've been trying to use boost::python to extend and embed Python into my app. The problem is that I have a C++ object that I want to instantiate and then pass into Python so that my scripts can update it. When the script has finished I need to get access to the current state of my C++ object. I almost had it working yesterday, however, when my script had finished running the state had not been updated...I think this is because instead of Python referencing the same C++ object it had created a new one. Is there a way around this? I've looked through the docs and the documentation but I'm a bit lost/stuck. You may wonder why I need to do this...basically I need an Observer/Observable setup. I need to be able to get progress information back from the Python script. Any help or pointers to docs will be appreciated. Thanks in advance. James Gunn Software Developer Communisis Chorleys Ltd Computer Bureau Manston Lane Crossgates Leeds LS15 8AH Tel: 0113 305 3001 Email: James.Gunn(a)communisis-dm.co.uk ********************************************************************** Please note: This e-mail and its attachments contain only the opinions of the sender and do not necessarily reflect the policy(s) of the communisis group in general. Employees of the communisis group are required not to make any defamatory statements and not to infringe or authorise any infringement of copyright or any other legal right by e-mail. Any such communication is therefore outside the scope of employment of the individual concerned. The communisis group will not accept any liability in respect of such a communication. Confidentiality: This e-mail and any attachments, together with their contents, are confidential unless otherwise explicitly stated in writing by the sender of this e-mail and are for the intended recipient only. If they have come to you in error you must not take any action in respect of them, which includes but is not limited to reproducing, sending or storing them, other than to notifying the sender immediately of the mistake, and deleting the e-mail, any attachments and any reproductions made by replying to it. Viruses: This e-mail and any attachments have been scanned for viruses but we cannot guarantee that they are virus free. The recipient should check this e-mail and any attachments for viruses. The communisis group accepts no responsibility for any damage caused by any virus transmitted by this e-mail or any of its attachments. In the event of any unauthorised copying or forwarding, the recipient will be required to indemnify the communisis group against any claim for loss or damage caused by any viruses or otherwise. ********************************************************************** ________________________________________________________________________ This email has been scanned for all viruses by the MessageLabs Email Security System. For more information on a proactive email security service working around the clock, around the globe, visit http://www.messagelabs.com ________________________________________________________________________

2 1

(no subject)
by Gilles Orazi Jan. 9, 2004

Jan. 9, 2004

2 2

boost/python/detail/python22_fixed.h missing from Boost RC_1_31_0 installation
by Patrick Hartling Jan. 9, 2004

Jan. 9, 2004

I just made an installation of the Boost RC_1_31_0 branch, and I noticed that the file boost/python/detail/python22_fixed.h does not get installed. I followed the documented process for making an installation: bjam install --prefix=/foo/bar This was done from the top-level Boost source tree using sources checked out from about three hours ago (including a fresh bjam 3.1.9 compiled from the same sources). Sorry I don't have a patch to submit, but I don't know what to modify to fix this. -Patrick -- Patrick L. Hartling | Research Assistant, VRAC patrick(a)vrac.iastate.edu | 2274 Howe Hall Room 2624 http://www.137.org/patrick/ | http://www.vrac.iastate.edu/ PGP: http://wwwkeys.gpg.cz:11371/pks/lookup?op=get&search=0xEBF86398

2 3

Function with an argument passed by reference
by Gilles Orazi Jan. 6, 2004

Jan. 6, 2004

2 1

extending with C++ and typeinfo question
by John Jan. 5, 2004

Jan. 5, 2004

Hi, (I'm not sure this worked the first time. My apologies if it did.) I am attempting to implement a scripting language for my application by embedding the python interpreter and exposing an API to python. So far so good. My application is heavily event driven, I have developed a publish/subscribe style event manager to which classes subscribe to published events. Frex: class i_subscriber { void on_event(event* e, event_type* et) = 0; }; class x : public i_subscriber { x() { SUBSCRIBE(this, event_startup) } void on_event(event* e, event_type* et) { if (IS_EVENT(event_startup, et)) { .... } } } when a event_startup event is generated, the event manager scoots through all the subscribers for that particular event and calls the on_event handler which determines the event type and handles it appropriately. The system depends heavily on rtti to work. Now what I want to do is to expose this system to my Python scripts. I'd like to have a Python class subscribe to an event, and when that event is published, for the script to be notified and be able to handle it. Frex: class y(i_subscriber): # the subscriber interface has been exposed by boost.python def __init__(self): subscribe(self, event_startup) # subscribe is an exposed API def __del__(self): unsubscribe(self) # unsubscribe from all events def on_event(evt, evt_type): # handle the event here, but how do I determine what the # type is? y_inst = y() These are the areas I am having trouble getting my head around at the moment. The main application needs to register the Python class and know that it is_a subscriber. This means I need to transmit type information for both the subscriber and the event through the interpreter. When an event is published it's not known if the subscriber is c++ or python. If I call on_event on the subscriber interface for a python class is there some marshalling that I need to do? How do I transmit the type of a published event through the interpreter so that my on_event handler knows what event is being processed? I've been looking through the Python and Boost.Python documentation and so far I haven't seen what I need to do to get this working. I'm probably just missing something simple. Any pointers would be helpful. Of course I might be blowing against the wind here. That would be good to know too. :-) Has anyone here attempted to do this before? Are there any other web sites out there where this may have been attempted? Cheers John

2 1

Advanced (?) compiling, freezing, translating Python
by luszczek1＠netscape.net Jan. 5, 2004

Jan. 5, 2004

Hi, here's my problem: - I know it will take me 10 times as long to write a code in C than in Python - I need the speed of the C version - my program cannot depend/require on Python interpreter/libraries/modules Is there a way to write Python program that almost looks like (say) C and then transoform it into C. Transformation should be intelligent - I don't want dynamic objects with dictionaries of methods. I'm willing to provide type descriptions so when I say: myList = [0] * 100 I get in C: int *myList; myList = calloc(100, sizeof(int)); I'm aware of projects like Pyrex, Psyco, and Pypy but their scope is too broad: they are about Python in its entirety and dynamism. There was also discussion some time back about adding static typing to Python. This is not what I'm thinking about either. I gave C as an example, but what I meant was translating Python into another language with some additional guidance (command-line optons or "header files"). I believe that Python is rich enough to express most of semantics of most of languages. For example, C++ templates are included in Python by default: In Python: def sum(aList): v = 0 for e in aList: v += e In C++: template <class T> sum(Vector<T> aList) {...} In case anybody would care, even primitive data types from C/C++ like float, short, etc. can be emulated in Python with array module. I don't see equivalence to C++ reference, value, and pointer semantics, but most folks can live without it. Does anything like this exist? Would anybody find it useful? With compiler module now in standard Python, it seems like an easy job. Or am I missing something? Peter __________________________________________________________________ New! Unlimited Access from the Netscape Internet Service. Beta test the new Netscape Internet Service for only $1.00 per month until 3/1/04. Sign up today at http://isp.netscape.com/register Act now to get a personalized email address! Netscape. Just the Net You Need.

5 5

Re: Advanced (?) compiling, freezing, translating Python
by luszczek1＠netscape.net Jan. 5, 2004

Jan. 5, 2004

>Unless he's programming in some tiny embedded environment, I think his >specifications are pretty bogus in the first place. The language he I would be glad to provide more detail on my suggestion. I don't which part of it is "bogus", though. I'm not suggesting another Python spin-off. We've seen enough already. What I'm proposing has to do (but is not limited to) programming style: - if you only use functions I guess you would like your work translated into C - if you only use single inheritance you probably would like to translate it to Java - if you want mulitple inheritance you go with C++ translation - if you change objects at runtime by adding methods and attributes you should probably stick with Python. >proposes would be neat to have, but I'd rather see effort go into PyPy >or something; which could theoretically do what he asks someday from >what I understand of their bootstrapping efforts. First sentence from the PyPy web page: The PyPy project aims at producing a simple runtime-system for the Python language. I was quite specific about runtime: I don't want it. Depending on your programming style you might not need it. Peter __________________________________________________________________ New! Unlimited Access from the Netscape Internet Service. Beta test the new Netscape Internet Service for only $1.00 per month until 3/1/04. Sign up today at http://isp.netscape.com/register Act now to get a personalized email address! Netscape. Just the Net You Need.

1 0

Re: Advanced (?) compiling, freezing, translating Python
by luszczek1＠netscape.net Jan. 5, 2004

Jan. 5, 2004

Nicodemus <nicodemus(a)esss.com.br> wrote: >luszczek1(a)netscape.net wrote: > >>I'm aware of projects like Pyrex, Psyco, and Pypy but their scope is >>too broad: they are about Python in its entirety and dynamism. > >Have you really looked into Pyrex? Because it seems to have what you >need: Although it can generate code that takes advantage of Python's >dynamism, type declaration is possible, allowing it to generate direct C >code... but I have never used, so I may be wrong. Yes I have. I think the project is great. However, from my (admittedly twisted) persepective it has a few problems: 1. Pyrex code doesn't work in Python interpreter. This is a big drawback to me. I like Python and want to stick with it, I don't want to learn another language. 2. Pyrex brands itself as a tool for extensions. What if I want the entire program to be translated into C/C++/whatever? Probably it's possible, but it's not the main goal. 3. The generated code depends on Python runtime (the PyObject and friends) There exist similar projects that approach the problem from different perspective: - Cython - drop braces from your C code - Perthon - translate Python to Perl that is human readable It's not quite what I want. Peter __________________________________________________________________ New! Unlimited Access from the Netscape Internet Service. Beta test the new Netscape Internet Service for only $1.00 per month until 3/1/04. Sign up today at http://isp.netscape.com/register Act now to get a personalized email address! Netscape. Just the Net You Need.

1 0

object.slice() infinite recursion bug
by Jonathan Brandmeyer Jan. 5, 2004

Jan. 5, 2004

This simple test demonstrates a bug in Boost.Python that results in infinite recursion (and a subsequent crash) whenever you make a slice using slice_nil for the upper and lower bounds. I originally discovered it with the array class, but it also exists in the object class, so that is what I'm putting up here. This crash is exhibited with 1.30 and CVS HEAD using both Python 2.2 and Python 2.3 with GCC 3.3.2. -Jonathan Brandmeyer ------------------------------------recurse.cpp------ #include <boost/python.hpp> void test_slice_recursion( boost::python::object arr) { // Haven't really found the right combination of using declarations // to use _ directly, yet. using namespace boost::python; arr.slice(_,_) = 2.0; } BOOST_PYTHON_MODULE(recurse) { using boost::python::def; def( "recurse", &test_slice_recursion); } --------------------------------------recurse.py--- from recurse import recurse # Should throw an exception for a list, but I get # "segmentation fault" recurse( [1, 2, 3])

2 1