Re: [openbabel-scripting] Accessing SSSR or Internal Coordinates from Python

[Geoffrey H. <ge...@ge...>]

On Feb 24, 2006, at 4:19 AM, Noel O'Boyle wrote:

> I'd like to help with this if I can - I think this issue is going to
> occur more than once. Is there some particularly good documentation
> available somewhere on wrapping with SWIG, that can make this all very
> clear to me?


Well, SWIG has a pretty decent manual. I've been living pretty  
bleeding-edge for SWIG, since the C++ support is continually  
improving and I've found periodic bugs in the SWIG releases (e.g.,  
1.3.28 doesn't work, but the current CVS does).

So I download the SWIG CVS: (See the SWIG directory)
http://sourceforge.net/cvs/?group_id=1645

Documentation for SWIG is in SWIG/Doc/Manual/index.html
C++ documentation is in SWIG/Doc/Manual/SWIGPlus.html

There's also a pretty responsive mailing list:
http://lists.sourceforge.net/lists/listinfo/swig-user

I've found a few catches, which is why this is somewhat slow-going:
* Sometimes I'll add in a new header to openbabel.i, and everything  
dies. More recently, I've been carefully adding in #ifndef SWIG /  
#endif  lines in the header to tell SWIG to ignore certain parts of  
the code. This seems to help, since some C++ syntax apparently  
confuses SWIG or simply cannot be wrapped correctly.

* It's not always clear to me what needs to be explicitly added to  
openbabel.i (e.g., this vector issue) and what SWIG figures out for  
itself. In retrospect, the SWIG documentation now makes perfect sense  
-- but I think the scripting language wrappers will be evolving for  
the foreseeable future.

* The current modules are growing pretty large. It's probably going  
to be a good idea to start splitting openbabel.i into multiple pieces  
so that compiling openbabel_python.cpp and openbabel_perl.cpp don't  
take so long. The current file on my disk is ~1.3MB of C++ code.

And of course, I'd be glad to give pointers/tips/feedback to anyone  
willing to help out with the wrappers.
-Geoff