Implemented numpy.partition via cffi and introselect functions

Merged
#7 · Created  · Last updated

Merged pull request

Merged in numpy-partition (pull request #7)

b2a4ef4·Author: ·Closed by: ·2016-03-06

Description

I have started implementing numpy.partition for numpypy. I tried to follow the recommendations of Matti Picus https://gist.github.com/mattip/ab34268b049b859554ad . It works only for 4 types so far (improvements to be continued).

I used apply_along_axis to deal with axis parameter. apply_along_axis relies on tolist() that seems has a bug. So I created pypy patch in PR #407 https://bitbucket.org/pypy/pypy/pull-requests/407/ for better results.

The PR help to reduce the number of failed numpy tests:

Pypy 82592-56afe29ed2bf, numpy 8990e6279c9452c935f8e3adfaf7479adcbd35ec:

FAILED (SKIP=41, errors=673, failures=118)

pypy 82592-56afe29ed2bf, numpypy with this PR

FAILED (SKIP=41, errors=668, failures=118)

pypy with patch https://bitbucket.org/pypy/pypy/pull-requests/407/, numpypy with this PR

FAILED (SKIP=41, errors=638, failures=126)

0 attachments

0 comments

Loading commits...