4 Feb 2006 09:00
Re: xapian python bindings
Olly Betts <olly <at> survex.com>
2006-02-04 08:00:46 GMT
2006-02-04 08:00:46 GMT
I'm cc:-ing my reply to xapian-devel, since others may have an opinion. In particular, James has probably done most work on the python bindings. On Fri, Jan 27, 2006 at 03:56:23PM -0800, Jason Toffaletti wrote: > I'm curious if boost::python has been considered for creating more > pythonic bindings to Xapian. I've used it to wrap several C++ libraries > and found the resulting python bindings to be much nicer than SWIG. First off, I'm a relative novice at python, so you'll need to explain what you mean by "more pythonic" if you want me to understand why boost::python might be better. Note that the SVN sources have a number of improvements for python (and other languages) over 0.9.2 - snapshots available here if you want to take a look: http://www.oligarchy.co.uk/xapian/trunk/ I'd not heard of boost::python before. The disadvantages compared to SWIG would appear to be: * We already have working SWIG bindings for python. * James and I have a reasonable understanding of how to wrap stuff using SWIG. * boost::python would require its own interface description, which is yet another place to need to add a new method. We already need to add it to xapian.i for SWIG, to the Java JNI bindings in several places, and to the appropriate .xs file for Perl, in addition to changing the C++ code and headers. SWIG's xapian.i is trivial to update in many cases - just copying the method prototype over generally works.(Continue reading)
RSS Feed