Clean up DataFrame.setitem behavior for duplicate columns #39403
Conversation
jreback
added
Indexing
pandas/core/indexers.py
Outdated
| @@ -376,6 +376,25 @@ def unpack_1tuple(tup): | |||
| return tup | |||
|
|
|||
|
|
|||
| def check_key_length(columns, key, value): | |||
| """Checks if a key used as indexer has the same length as the columns it is | |||
pandas/core/indexers.py
Outdated
|
|
||
| Parameters | ||
| ---------- | ||
| columns: The columns of the DataFrame to index. |
|
If you're interested, there might be something worth salvaging from _setitem_array here. Avoids going through iloc so that we dont have to worry about inplace vs not-inplace there (xref #38896). If/when I get #39302 in shape to be a real PR, something like that branch will have to be a part of it, but I'm optimistic you'll get around to it before I do. |
|
That looks good. I will try to incorporate this after this is merged. |
pandas/core/indexers.py
Outdated
| @@ -376,6 +376,27 @@ def unpack_1tuple(tup): | |||
| return tup | |||
|
|
|||
|
|
|||
| def check_key_length(columns, key, value): | |||
There was a problem hiding this comment.
can you type as much as you can here
There was a problem hiding this comment.
We would have to type wiht Index and DataFrame causing circular imports. Is there a way around this?
There was a problem hiding this comment.
do the import inside a if TYPE_CHECKING:
There was a problem hiding this comment.
Sorry, should have said that I have already tried this. If using TYPE_CHECKING and typing the signature (not the return value) this raises
NameError: name 'DataFrame' is not defined
There was a problem hiding this comment.
Did not know about the future import. Thx for the help
pandas/core/indexers.py
Outdated
|
|
||
| Parameters | ||
| ---------- | ||
| columns: Index The columns of the DataFrame to index. |
There was a problem hiding this comment.
space between "columns" and colon, same on the next two lines
|
thanks @phofl was this technically a bug fix? or just an incorrect test? |
|
Bug fix yes, but is covered by the whatsnew from #39341, follow up from that one |
cc @jbrockmendel This cleans the edgy behavior up for duplicate columns.
Previous test was wrong.