[Biopython-dev] Re: Work towards getting KDTree compiling

[Brad Chapman]

Michiel and Thomas;

> The latest version works with Microsoft Visual Studio and the mingw32 
> compiler. Good job!

Sweet. Thanks for all your work helping to debug this. That's good
enough for me for now -- making a release with what we've got -- and
we'll deal with these next problems for the upcoming release. 

> > o) The source in Bio/PDB/mmCIF/lex.yy.c uses some Unix-specific
> > commands from  unistd.h. Microsoft's compiler barfs on these, the
> > mingw32 compiler does not. It  may be a good idea anyway to check if
> > the mingw32-compiled version actually works.
> 
> lex.yy.c is a lex-generated file. I presume it will look different when
> generated on a windows system. lex.yy.c is part of the distribution
> because (a) not everyone has lex installed and (b) it's not clear to me
> how to fire up lex from distutils. I have no objections if you want to
> comment this out in setup.py.

What kind of work would be involved with generating this lex file on
Windows? I know next to nothing about lex, but if we could get
different copies that work on Windows and Unix then we could again
build up the extension in the setup.py file so it would compile on
with all compilers.

> > o) Recently I found out that the official Numerical Python version is
> > now  numarray. The Numeric module is now unsupported (see
> > http://www.pfdubois.com/numpy). We'll need to decide if Biopython is
> > going over  to numarray, or whether to stick with Numeric for now. Some
> > modifications will  be needed in the Python and C code if we start
> > using numarray.
> 
> Good point. We should indeed move to numarray, I think, though we might
> wait some months until numarray is more widely used.
> Cheers & thanks for the C++ related work,

I'd agree that a move to numarray makes the most sense. We need to
move in the direction development is going. I'd like to make this
something we can plan to do soon-ish so it can be tested and ready
for the next release. Right now, it looks like the following code
would need to be migrated:

-> Affy -- Harry
-> KDTree -- Thomas
-> PDB -- Thomas
-> SVDSuperimposer -- Thomas
-> Cluster -- Michiel
-> Stasitics/lowess -- Michiel
-> LogisticRegression -- Jeff?
-> MarkovModel -- Jeff?
-> NaiveBayes -- Jeff?
-> distance -- Jeff?
-> kNN -- Jeff?

What I really don't want is to make people download both Numeric and
numarray to use Biopython, so I'd hope to make a coordinated switch
between releases. If we decide to do this, we can split up the tasks
and have a go at it.

Let me know what people think about this and we can get it
coordinated. I'd like to wait at least a week before starting to do
it, so we can make sure I didn't mess anything up with the new
release that will require a quick fix-it release :-).

Brad