PERF: follow up to ASV

[jreback]

xref #9715

• add docs on how to use / add tests (in contributing.md / contributing.rst)
• add/update vbenches to current
• whatsnew note (put in highlites that this is added)
• test on Windows (basicly works, but had to do some small adjustments)
• get it working with python 3
• clean up autogenerated tests
  • fix failing tests
  • remove redundancy (by putting things in classes and reusing setup, .

cc @qwhelan

[jorisvandenbossche]

Added:

• test on Windows (basicly works, but had to do some small adjustments):
  • eg needs libpython to build -> I don't know how it is possible to specify something windows only in the yaml file
• clean up autogenerated tests:
  • fix failing tests
  • remove redundancy (by putting things in classes and reusing setup, ..)

Maybe I can find some people for this on the EuroScipy sprint if it is not yet done before that.

[jreback]

@jorisvandenbossche awesome!

[jreback]

note that I disabled the 3.4 support (was failing anyhow) for now.

[qwhelan]

I've fixed the translation script a bit to move functions from inside setup() to the class-level, so the majority of failures outside eval.py should be fixed.

I also have the suite running under Python 3 - I'll have to look into the current results in the morning but it looks promising.

[TomAugspurger]

Great. @qwhelan I'm curious what cause the py3 failures? Was it a change in how imports are handled?

[qwhelan]

@TomAugspurger Partially, the changes were:

• from pandas_vb_common import * -> from .pandas_vb_common import *
• Replace xrange with range and add from pandas.compat import range in pandas_vb_common
• Replace string.uppercase with string.ascii_uppercase

There are probably a few others, but the vast majority seem to be working with the above changes.

[jorisvandenbossche]

@qwhelan could you push what you already did on this? Possibly there could be some people to sprint on this on Sunday, so then for that case best to start with your latest version!