Support numpy 1.8

[jtratner]

Numpy 1.8 was released recently - https://github.com/numpy/numpy/blob/master/doc/release/1.8.0-notes.rst

We should probably support it and add/change a Travis build for it. Additionally, might be interesting to benchmark numpy's nanmean, nanvar and nanstd against pandas versions...

numpy also notes that they are planning to make more datetime changes in 1.9 - whee!

[jreback]

@cpcloud were u going to setup a wheel for this one? (1.8)?

[jreback]

https://travis-ci.org/jreback/pandas/builds/13386292

works....takes 8 minutes longer (total)...w/o wheels

[cancan101]

What is a wheel?

[cancan101]

Also how closely does the pydata team collaborate with the numpy team on matter such as datetime support?

[jreback]

@cpcloud basically made binary installables for long-to-compile (e.g. numpy/scipy) that work on the travis boxes. so you just download the image and done, no need to get a version and compile, much faster.

2nd question - no collaboration, just if we find bugs they are pushed upstream (though mainly they are worked around , e.g. 1.6.2 basically broken).

[jtratner]

And numpy tests pandas around releases. Is there anything we really wish
numpy did with datetimes?

[jreback]

datetime support is fine (well maybe if they add proper tz support but who knows)