PKG/DEP: Make importable without pytest

[TomAugspurger]

Closes #16065

Edit: removed the top-level imports of pytest in pandas.util.testing

[TomAugspurger]

Actually, thinking more about this I probably prefer removing the line from panads.init and have people import it separately

import pandas.testing as tm

[chris-b1]

Since pytest isn't actually needed for the api functions (assert_frame_equal, etc), maybe those could be defined in pandas.testing, then imported back into pandas.util.testing?

edit: actually this is lot less trivial than I was originally thinking, so maybe not a good idea

[chris-b1]

Looking at testing.py again - almost every use pytest already has function level imports, so may be best to bite the bullet and do it for the few remaining cases?

[TomAugspurger]

so may be best to bite the bullet and do it for the few remaining cases?

OK, let me give this a shot.

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@2f9c854). Click here to learn what that means.
The diff coverage is 66.66%.

@@            Coverage Diff            @@
##             master   #16088   +/-   ##
=========================================
  Coverage          ?   90.84%           
=========================================
  Files             ?      159           
  Lines             ?    50780           
  Branches          ?        0           
=========================================
  Hits              ?    46132           
  Misses            ?     4648           
  Partials          ?        0

Flag	Coverage Δ
#multiple	88.62% <66.66%> (?)
#single	40.36% <50%> (?)

Impacted Files	Coverage Δ
pandas/util/testing.py	79.5% <66.66%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2f9c854...ed1ca67. Read the comment docs.

[TomAugspurger]

Ok, not too bad at all. Hopefully this passes.

And here it is working without pytest, which is nice

In [1]: import pandas as pd

In [2]: pd.testing.assert_series_equal(pd.Series([1]), pd.Series([2]))
---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
<ipython-input-2-46d7e85da706> in <module>()
----> 1 pd.testing.assert_series_equal(pd.Series([1]), pd.Series([2]))

/Users/taugspurger/Envs/pandas-dev/lib/python3.6/site-packages/pandas/pandas/util/testing.py in assert_series_equal(left, right, check_dtype, check_index_type, check_series_type, check_less_precise, check_names, check_exact, check_datetimelike_compat, check_categorical, obj)
   1310                                        check_less_precise=check_less_precise,
   1311                                        check_dtype=check_dtype,
-> 1312                                        obj='{0}'.format(obj))
   1313
   1314     # metadata comparison

as/pandas/util/testing.pyx in pandas.util.libtesting.assert_almost_equal (pandas/util/testing.c:4156)()

as/pandas/util/testing.pyx in pandas.util.libtesting.assert_almost_equal (pandas/util/testing.c:3274)()

/Users/taugspurger/Envs/pandas-dev/lib/python3.6/site-packages/pandas/pandas/util/testing.py in raise_assert_detail(obj, message, left, right, diff)
   1137         msg = msg + "\n[diff]: {diff}".format(diff=diff)
   1138
-> 1139     raise AssertionError(msg)
   1140
   1141

AssertionError: Series are different

Series values are different (100.0 %)
[left]:  [1]
[right]: [2]

[jreback]

@TomAugspurger btw, make sure you can locally conda build (or pip), install into a fresh env and do a
python -c "import pandas; pandas.test()" works (we tests this on travis, but this is a check).

also after you tag, but before pushing, make sure the same works for the __version__

[jreback]

thanks @TomAugspurger