[Biopython-dev] [GSoC] Gsoc 2014 aspirant

[Nigel Delaney]

As a quick $0.02 from a library user on this.  Back in 2006 when I first
started using biopython I was working with Bio.pairwise2, and learned that
it was too slow for the task at hand, so wound up switching to
passing/parsing files to an aligner on the command line to get around this,
as (on windows back then), I never got the compiled C code to work properly.
Although pypy may be fast enough and has been very impressive in my
experience (plus computers are much faster now), I think wrapping a library
that is cross platform and maintains its own binaries would be a great
option rather than implementing C-code (I think this might be what the GSoC
student did last year for BioPython by wrapping Python).

Mostly though I just wanted to second Peter's wariness about adding C-code
into the library.  I have found over the years that a lot of python
scientific tools that in theory should be cross platform (Stampy, IPython,
Matplotlib, Numpy, GATK, etc.) are really not and can be a huge timesuck of
dealing with installation issues as code moves between computers and
operating systems, usually due to some C code or OS specific behavior.
Since code in python works anywhere python is installed, I really appreciate
the extent that the library can be as much pure python as allowable or
strictly dependent on a particular downloadable binary for a specific
OS/Architecture/Scenario.