Revision as of 11:19, 3 June 2014

This page is a list of notes for building a numpy for IronPython, including Mono on Mac, Windows, Linux, etc.

Options

1. Write a version of numpy in pure Python
   1. Contacted Tom Hochberg for small bit of Numeric in Python
2. Write a version of numpy in C#
   1. Look to Jython's Numeric (could use IKVM to use directly)
      1. https://bitbucket.org/zornslemon/jnumeric-ra/overview - rewrite
      2. http://jyni.org/ - talk directly to C
3. Use a bridge to talk to Python and numpy (would require marshaling back and forth)
   1. http://rpyc.readthedocs.org/en/latest/ - rpyc
   2. http://codespeak.net/execnet/example/hybridpython.html - execnet
4. Wrap existing numpy
   1. IronClad - works by reimplementing the Python C API in C#
      1. https://code.google.com/p/ironclad/
   2. PyPy - written in RPython?
      1. https://bitbucket.org/pypy/numpy
      2. http://buildbot.pypy.org/numpy-status/latest.html - status
   3. Cython -
      1. http://blog.enthought.com/python/scipy-for-net/#.U42k3x92nfE - Blog on Enthought work
      2. https://www.enthought.com/repo/.iron/ - Enthought packages
      3. https://github.com/numpy/numpy-refactor - NumPy for .NET
      4. https://github.com/jasonmccampbell/scipy-refactor - SciPy for .NET
      5. https://bitbucket.org/jasonmccampbell/cython-for-ironpython - Cython for .NET
      6. https://github.com/jasonmccampbell/fwrap - Wrapping Fortran extensions

Notes

• http://legacy.python.org/dev/peps/pep-0465/ - Python3 will have a matrix multiplication operator
• http://technicaldiscovery.blogspot.com/2011/10/thoughts-on-porting-numpy-to-pypy.html