BUG / CoW: also return new object in case of null slice for both rows and columsn (.(i)loc[:, :])

[jorisvandenbossche]

Related to #49450 (also ensuring we always return new objects from indexing operations, but for a different case).

Currently, there is the specific corner case of df.(i)loc[:, :] that just returns df (this doesn't happen for df.(i)loc[:] or df[:], where only a single dimension gets indexed)

Given the current inconsistency with df.loc[:, :] vs df.loc[:], we could also consider changing this for 2.0 for the default behaviour as well, and not just for when CoW is enabled?

• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/vX.X.X.rst file if fixing a bug or adding a new feature.

xref #48998

[mroeschke]

Given the current inconsistency with df.loc[:, :] vs df.loc[:], we could also consider changing this for 2.0 for the default behaviour as well, and not just for when CoW is enabled?

Yeah it would be nice if there was a consistency in the returned object by whatever method if the entire object is sliced

[jorisvandenbossche]

So shall I change this for 2.0 also in the default mode? (so as small breaking change) @jbrockmendel thoughts?

[jbrockmendel]

So shall I change this for 2.0 also in the default mode? (so as small breaking change)

I like this idea.

[jorisvandenbossche]

OK, then I updated this PR to have this new behaviour for all cases (removed the check for CoW enabled), and added a whatsnew note mentioning it as a small API change.

[jbrockmendel]

pandas/tests/indexing/test_loc.py::TestLocBaseIndependent::test_identity_slice_returns_new_object failing

[jorisvandenbossche]

Ah, that was the edit I made to the test when assuming this is only changed for CoW, now it is always a new object. Should be fixed now.