API: Meta-issue for making consistent API's to refer to column names and index names

I would like to propose that any pandas API that allows specification of a column name also works when specifying an index level name. Today, this works in some places, but not all. Here is a list of places where things work, and where things don’t work and could be improved. (The list is most likely incomplete, so additions are welcome). References to existing issues are given when they already exist (and if I knew about them):

Hoping to include this in the roadmap #27478

Things that work:

• DataFrame.query() allows queries to use column names and index level names
• DataFrame.merge() and pd.merge() allow both column names and index level names to be specified as key fields to use in the merge (but see below)
• DataFrame.groupby() allows both column names and index level names to be mixed in the groupby list

Things that could be improved:

• Allow .loc() and .getitem() to specify index level names wherever it allows column names.
• Modify DataFrame.rename() to allow renaming of index levels with a dict argument. (#20421)
• Allow index names to be specified like columns in pd.Grouper() (#19542)
• When merging on a subset of MultiIndex levels, preserve the levels not included in the merge. (#13371 is somewhat related)
• Have .itertuples() return a named tuple that includes index names (#27407)
• Allow .assign to refer to columns that correspond to level names (although the .loc and .getitem() suggestion above might handle this

[max-sixty]

Yes, thanks @Dr-Irv . That narrows the gap, though not all the way: being able to specify the indexes without the context of the current indexes would be more declarative and clearer, particularly to a reader of the code.

[sangrey]

As someone who is a heavy user of Multiindexes, I couldn’t agree more that this would be very useful to have. I find working in a fairly complex production codebase that whether a particular “column” is a best thought of as an index or a column in its own right is constantly switching back and forth, and so my code ends up being littered with reset_index() calls and set_index() calls.