BUG: DataFrame.__setitem__ raising ValueError when setting multiple values to dup columns

[phofl]

• closes DataFrame.__setitem__ fails with multiple values and duplicate column names #15695
• tests added / passed
• Ensure all linting tests pass, see here for how to run them
• whatsnew entry

Setting with a scalar key which is duplicated in columns should be the same as using a listlike key

[phofl]

Hacky, will try to improve

[phofl]

Hm this is a hacky edge case:

df = pd.DataFrame([[1, 2, 3], [4, 5, 6]], columns=["a", "b", "a"])

df["a"] = df.index

This is tested in a unit test test_column_dups_operations

If this should work, we can't really implement

df["a"] = (1, 2)

because here is not clear what to do, interpret (1,2) as a column, which should result in (this is tested with the case above)

   a  b  a
0  1  2  1
1  2  5  2

or interpret as a row (as described in the op) which would result in

   a  b  a
0  1  2  2
1  1  5  2

The loc case interprets this as

df.loc[:,"a"] = df.index
   a  b  a
0  0  2  1
1  0  5  1

which results in an Error if len(index) != number of "a" in columns

[jbrockmendel]

not necessarily for this PR, but it might be worth implementing a Index.count method (mirroring list.count)?

[phofl]

Yeah this would be helpful probably, something like that would have helped in #39278

[jreback]

thanks @phofl

[toobaz]

Doesn't this slow down every setting operation with a single key? I think it's fine as a temporary measure, but really we shouldn't be doing multiple lookups for a single operation.

[toobaz]

Doesn't this slow down every setting operation with a single key?

Well no, I guess only with a list of keys. Still, we could check if columns is unique (which is O(1), right?) before the test involving get_indexer_for?

[phofl]

That's probably a good idea, yes