[SciPy-dev] Inclusion of cython code in scipy

[Nathan Bell]

On Wed, Apr 23, 2008 at 1:54 PM, Matthew Brett <matthew.brett at gmail.com> wrote:
>
>  But, it seems to me that no-one is going to use cython much if it's
>  not obviously useful.  It's the usual open-source idea of allowing
>  people to try tools if they fit

The usual open-source approach is fueled by people who enjoy
reinventing the wheel over, and over,...

>  I don't think cython needs to make an argument,

It does if it wants to replace existing code

>  nor do I think it commands blind allegiance.

I have a hard time reconciling this statement with the previous.

>  It's just going to be one of those tools that looks very promising,
>  and will show its worth if people find they like using it and move towards it. And if, in
>  practice, it creates good maintainable, readable code.  We'll know
>  it's good, if that happens.  We'll know it's not good, if it doesn't.
>  And if it is good, it is very likely to survive and flourish.

And if it's a bad choice, can we blame you for deflecting valid
criticism in favor of a wait-and-see approach?

The question is not so much whether Cython should be used, but rather
to what extent.  I have expressed concern that writing hundreds of
thousands of lines of Cython code (as SAGE has done) is potentially
dangerous.  On the other hand using Cython as a wrapper generator
carries little risk and might represent an improvement over SWIG for
instance.

It's irresponsible to continue promoting an unproven technology like
Cython without first addressing these concerns.  Furthermore, it's
profoundly ignorant and extremely offensive to argue for replacing
existing implementations without first making a solid case of the
deficiencies of the status quo.  This sort of fanboyism causes people
like me, who help maintain an existing part of SciPy, to treat Cython
with skepticism.

-- 
Nathan Bell wnbell at gmail.com
http://graphics.cs.uiuc.edu/~wnbell/