MAINT: Numpy deprecated boolean unary '-' operator

[jseabold]

I'm sure there are more of these, but I just caught this one when I was raising on all warnings.

[jreback]

ahh..prob a fair number of these....

[jseabold]

Might be trivial to write a log only on this deprecation warning to wrap the test suite in. I have no idea though.

[jreback]

this is under numpy 1.9-dev right?

[jseabold]

Yeah 1.9.0.dev-3e61f86

[jreback]

we test with numpy master in every build: https://travis-ci.org/jreback/pandas/jobs/23688029

is their a way to make nose print out the warnings (i'll have to turn off the deprecation stuff) because we hide that (or maybe that's all I need to do?)

[jseabold]

Deprecation warnings are filtered in python by default. I'm not really sure how to show them using nosetest. This is the code to show them

import warnings
warnings.simplefilter('always', DeprecationWarning)

But I'm not sure how to get nosetest to respect that. Maybe you could put it in .pystartup. Or put it at the beginning of a pandas.test() function, but I don't think pandas test suite can be run that way.

[jseabold]

I had a filter to 'error', because I wanted to debug another warning I had, but my code raised on this first.

[jreback]

ok...easy enough...to run it and look for these

[jreback]

going to close in favor of #6960 their are like a ton (think I got them though)

[jreback]

@jseabold this turned into a project!

numpy really turned on the spigot with depr warnings! (of course the conversions that we were doing were a bit suspect anyhow)..

e.g. you can't compare np.datetime('NaT') with tslib.NaT (e.g. its a datetime64 cmp with an integer (but np.datetime('NaT').astype('int64') == tslib.NaT works just fine)