harmonize the testing namespace with tm.TestCase

[jankatins]

Currently these two have different places to look for the relevant equality tests: TestCase has assert_numpy_array_equivalent and assert_numpy_array_equal but the testing module has assert_frame_equal and assert_series_equal.

It would be nice if all substantial code is available form the testing module and TestCase implements a small wrapper around these methods. That way someone outside pandas can import tm and us ethe methods without subclassing TestCase.

[jreback]

yep, some of these are added, but the inherting from TestCase is pretty recent so the entire conversion hasn't been done yet. (converting the usage of tm.assert_series_equal to self.assert_series_equal for example)

[josef-pkt]

I haven't looked at this yet, I just started to look at the failing tests in statsmodels.

I don't think inheriting from TestCase makes sense for statsmodels. What we need is the equivalent of the numpy.testing assert_xxx functions.
For example, up to now we could use numpy.testing.assert_equal to compare anything. Now it fails with, for example, comparing two pandas index or dates. We will have to go through the failing test cases and find what the compatible version of assert_equal is.

[jankatins]

@josef-pkt

import pandas.util.testing as tm
# for two series
tm.assert_series_equal(res, exp)
#for two frames
tm.assert_frame_equal(res, exp)

[jorisvandenbossche]

Closed by #8923