BUG: Fix df.loc slice support

[Licht-T]

• closes BUG: loc.__setitem__ over MultiIndex with slice key and step != 1 does strange things #8856
• tests added / passed
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

[jreback]

whatsnew 0.22

[jreback]

make separate test

[jreback]

can u add some comments here

[jreback]

is this a duplicate of the other test?

[toobaz]

(.iloc vs. .loc)

[jreback]

can u add several cases using slices

also if other testing of slice u can pull to this section

[Licht-T]

@jreback Oops, sorry! I overlooked another test.
The MultiIndex loc slicing work at the first level.
https://github.com/pandas-dev/pandas/blob/master/pandas/tests/test_multilevel.py#L1829

Input:

s = pd.Series(np.arange(8), pd.MultiIndex.from_product(([0, 1, 2, 3], list('ab'))))
s.loc[::2] = 200
s

Output:

0  a    200
   b    200
1  a      2
   b      3
2  a    200
   b    200
3  a      6
   b      7

But, this returns the crazy result.

Input:

index = pd.MultiIndex(levels=[['foo', 'bar', 'baz', 'qux'], ['one', 'two','three']],
                      labels=[[0, 0, 0, 1, 1, 2, 2, 3, 3, 3], [0, 1, 2, 0, 1, 1, 2, 0, 1, 2]],
                      names=['first', 'second'])
frame = pd.DataFrame(np.random.randn(10, 3), index=index, columns=pd.Index(['A', 'B', 'C'], name='exp'))
frame.loc['foo':'bar':4]

Output:

exp                  A         B         C
first second
foo   one     0.870785  1.900457 -0.907938
bar   two    -0.290714 -0.211697 -0.957994

[jreback]

@toobaz if you'd have a look

[Licht-T]

I think slicing by string is not good since slice seems not supporting to handle string.

[toobaz]

@Licht-T : I'm confused. Unless I'm missing anything, your changes to code are limited to pd.MultiIndex.get_loc, but in up-to-date master,

• none of the problematic examples in BUG: loc.__setitem__ over MultiIndex with slice key and step != 1 does strange things #8856 hits pd.MultiIndex.get_loc
• none of the tests you add addresses any of the problematic examples in BUG: loc.__setitem__ over MultiIndex with slice key and step != 1 does strange things #8856
• none of the tests you add hits pd.MultiIndex.get_loc

... ?

[jreback]

closing as stale. ping if you want to update.