[Numpy-discussion] [OFFTOPIC] creating/working NumPy-ndarrays in C++

[Hänel Nikolaus Valentin]

* Chris Barker <chris.barker at noaa.gov> [2012-04-09]:
> 2012/4/9 Hänel Nikolaus Valentin <valentin.haenel at epfl.ch>:
> 
> http://www.eos.ubc.ca/research/clouds/software/pythonlibs/num_util/num_util_release2/Readme.html
> >>
> >> that looks like it hasn't been updated since 2006 -- I"d say that
> >> makes it a non-starter
> >
> > Yeah, thats what I thought... Until I found it in several production
> > codes...
> 
> are they maintaining it?

Well, no... thats «legacy code» that was handed down to me, more or
less.

> >> 4?
> >
> > http://www.boost.org/doc/libs/1_49_0/libs/python/doc/v2/numeric.html
> > (old)
> >
> > https://github.com/ndarray/Boost.NumPy
> > (new)
> >
> > http://code.google.com/p/numpy-boost/
>  (also pretty old -- I see this:)
> 
> - Numpy (numpy.scipy.org) (Tested versions: 1.1.1, though >= 1.0 should work)
> - Python (www.python.org) (Tested versions: 2.5.2, though >= 2.3 should work)
> 
> both pretty old versions.
> 
> http://www.eos.ubc.ca/research/clouds/software/pythonlibs/num_util/num_util_release2/Readme.html
> 
> 
> also pretty old.
> 
> So I'd go with the actively maintained on -- or Cython -- what I can
> tell you is that Cython is being very widely used in the
> numerical/scientific computing community -- but I haven't seen a lot
> of Boost users. Maybe they use different mailing lists, and dont go to
> SciPy or Pycon...

Yeah, I would choose cython... if I had a choice...

I have had boost.python mentioned a single time throughout the last four
editions of EuroScipy2012

> I'm not sure you made your use case clear -- are you writing C++
> specifically for calling form Python? or are you working on a C++ lib
> that will be used in C++ apps as well as Python apps?

Currently just curious about the different tools available to
fascilitate interoperability between numpy and boost.python.

V-