BUG: losing Index/Series names master issue

[jreback]

cases where index names are lost in various operations, most of these involve an issue in concat

Index.name

• Panel.fillna, Panel.fillna() loses index names #3570
• groupby on empty, BUG: Group-by on an empty data object dtype loses the index name (cython aggregation is ok) #8093
• .append, Adding new rows drops index name #9857 (closed by BUG: fix for index name lost #9857 and also added the corresponding test in test_index.py #9876)
• groupby .size, BUG: GroupBy.size doesn't attach index name properly if grouped by TimeGrouper #9925
• timedelta ops, BUG: index name lost with timedelta ops #9926
• .intersection/.union, BUG: index name lost when indexing with another index #9943
• drop_duplicates, BUG: drop_duplicates drops name(s). #10116
• to_datetime/to_timedelta BUG: Index name is not preserved during some conversions #10875 (closed by BUG: Index name lost in conv #10875 #10945)
• Index.__new__ API/BUG: Index creation with named Index resets name #11193

Series.name

• rolling_* funcions BUG: #10565 Series.name lost in rolling_* funcions #10659
• dt accessor funcions BUG: Series.dt ops reset name #10712
• quantile ERR/DEPR: Fix quantile error message / remove percentile_width #10881

[ksanghai]

I am working on this issue now.. lets see how far I can get..

[jreback]

In [1]: Index(['a'],name='foo').append(Index(['b']))
Out[1]: Index([u'a', u'b'], dtype='object')

In [2]: Index(['a'],name='foo').append(Index(['b'])).name

[sinhrks]

How about making testing.assert_index_equal to check name attribute (maybe optionally) to detect this kind of problems?

[jorisvandenbossche]

+1 testing.assert_frame_equal does this already, so it seems logical that assert_index_equal can do this as well (and then we can also clean up assert_frame_equal a bit)

[jreback]

IIRC the checking names option was turned off as a default quite a while ago, because too many things were breaking; I suspect it would be much closer nowadays. so +1 here as well.

[sinhrks]

Added #9925 and #9926 to the list, which found by changing assert_index_equal to check index names.

[sergeny]

Independently stumbled upon the same thing: #9943

[Licht-T]

@jreback #8093 seems already fixed.
#8093 (comment)

[Dr-Irv]

@jreback #11193 is closed by #11193
#3570 was closed since it is for Panel (remark from you is "won't fix")