ENH: Index StringMethods should return MultiIndex when result dimension is more than one

[mortada]

Currently the Index.str methods only support returning Index results. However, analogous to the expansion from Series.str.* -> DataFrame, we should be able to do Index.str.* -> MultiIndex in certain cases, as discussed in #9870 (comment)

Some of the string methods that can support this are: str.get_dummies, str.extract, str.split

• split (DEPR: Deprecate str.split return_type #10085)
• rsplit (ENH: added rsplit to StringMethods #10303)
• partition / rpartition (ENH: Add StringMethods.partition and rpartition #9773)
• extract (ENH: str.extractall for several matches #11386, only supports one group )
• extractall (not implemented in Index), ENH/BUG: str.extractall doesn't support index #13156
• get_dummies (ENH: Add Index.str.get_dummies #12842)

Here are the related PRs: #9667, #9843, #9870, #9985

[sinhrks]

I should have comment here... #9773 is going to have a general logic for expand. Affected methods (in future) are summarized in #9870.

[sinhrks]

I understand current status is:
-> Checklist is moved to the top.

[sinhrks]

Index.str.extractall raises AttributeError. Just fix and close.

idx = pd.Index(["a1a2", "b1", "c1"], ["A", "B", "C"])
idx.str.extractall("[ab](?P<digit>\d)")
# AttributeError: 'Index' object has no attribute 'iteritems'

CC: @tdhock

[jreback]

thanks @sinhrks