[SciPy-dev] Inclusion of cython code in scipy

Prabhu Ramachandran prabhu@aero.iitb.ac...
Fri Apr 25 11:55:14 CDT 2008


David Cournapeau wrote:
> So I don't see how a discussion about C++ and stl is relevant for 
> numpy/scipy. The original email was about including cython code: if the 
> choice is between no code and cython code, I know which side I am on.

The thread has meandered far from the original topic.  Sorry for helping 
it along.  I'll start another thread once I put together what I've been 
talking about.

Just to be clear.  I have nothing against Cython or Pyrex and even use 
Pyrex in some of my own code.  There are cases when it is an excellent 
option.  I don't think Nathan had said anything of the sort either. 
I'll elaborate the problem I see later on and start a new thread about it.

> Swig has problems too, BTW. For example, when I was working on 
> scipy.cluster last year, I could not regenerate the source file because 
> of SWIG version mismatches, and at the end, I just found easier to 
> rewrite the extension in pure C (it was really small).

Of course SWIG has its fair share of problems.  Not the least the fact 
that it changes so fast, so much -- at least in the past this was a problem.

cheers,
prabhu


More information about the Scipy-dev mailing list