Re: [openbabel-scripting] openbabel and swig

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

On Jun 13, 2005, at 1:43 PM, S Joshua Swamidass wrote:

> Let me point out I'm only familiar wiht python, not perl at all. So
> everything i say here has that slant.

That's fine. I think right now, there's more interest in Python  
wrappers than Perl.

> One sticky issue in particular is the '>>' and '<<' methods of
> ... Molecule input/output in openbable relies heavily on this
> feaure.

Not anymore. These are removed completely in the 2.0 development  
code. There are probably ways to bring them back, but honestly, it's  
nice to separate the conversion code completely from an OBMol.

Even in earlier versions, the << and >> operators were not  
extensively used by internal code.

> will allow open bable to read a molecule from any
> file-emulating-object of the scripting target language, not just real
> files (very important for python).

It's already there. If you pass me a C++ iostream, then I'll hand it  
off to the new OBConversion framework and it'll go to work. I would  
assume that should handle most cases for Python, though I'll admit I  
haven't extensively looked at the new SWIG wrappers.

> The other major problem i was stuck on was compiling the wrapper swig
> produced. I never got to the bottom of which i couldn't. I can go back
> and look again. In my (limited) experience, this is a sign swig
> misunderstood some more complex uses of c++ syntax.

No, it was a problem with the declarations for C replacement  
functions missing on some platforms. SWIG saw them in the OpenBabel  
namespace and had problems. First I updated the development version  
to use std:: so that SWIG would correctly ignore them. Now, they're  
out of mol.h entirely.

Please take a look at the latest CVS snapshot on the openbabel-new- 
framework branch or the CVS development snapshot. There's even a SWIG  
openbabel.i file there. :-)

At the very least, I can vouch that the wrapper compiles correctly.

Cheers,
-Geoff