ENH: Add optional argument index to pd.melt to maintain index values #33659
Conversation
|
Does the CI always have this many issues? |
There was a problem hiding this comment.
Thanks @Rik-de-Kort almost there. generally lgtm. just a couple more comments.
pandas/core/reshape/melt.py
Outdated
| result = frame._constructor(mdata, columns=mcolumns) | ||
|
|
||
| if not ignore_index: | ||
| new_index = np.tile(frame.index, K) |
There was a problem hiding this comment.
both MI and Index already have a .repeat() method, I think we could add a .tile() method to make this easier. (or just use repeat)
There was a problem hiding this comment.
Index(["foo", "bar"]).repeat(2) yields Index(['foo', 'foo', 'bar', 'bar'], dtype='object'), where as np.tile(["foo", "bar"]) yields array(['foo', 'bar', 'foo', 'bar', 'foo', 'bar'], dtype=object). The latter corresponds to the layout used in melt so it's very not trivial to use repeat instead of tile.
I tried having a look at implementing tile on indices but then I would also have to do it for multiindices and document it and tests, and argument validation which I've never even looked at before and I think it's a big hassle that I will not undertake.
There was a problem hiding this comment.
we already have _tile_compat in pandas\core\reshape\util.py. This may allow futher simplification here.
There was a problem hiding this comment.
This converts to object dtype. Can you use pandas.core.reshape.util._tile_compat?
There was a problem hiding this comment.
You should be able to remove the next section as well, since you won't be converting to an ndarray.
There was a problem hiding this comment.
Thanks, that simplifies the code a lot! Build is failing but that's due to a worker crashing.
Co-authored-by: Simon Hawkins <simonjayhawkins@gmail.com>
pandas/core/reshape/melt.py
Outdated
| result = frame._constructor(mdata, columns=mcolumns) | ||
|
|
||
| if not ignore_index: | ||
| new_index = np.tile(frame.index, K) |
|
@jreback I though I gave you a ping, but I don't see it, so here it is! |
|
thanks @Rik-de-Kort |
|
You're welcome |
Finishing up a stale PR idea: #28859 and #17459
Has some tests and better code.
I think it's fair to duplicate the index values and not bend over backwards to maintain uniqueness like in previous iterations.
Apologies for the mess, it was a quick job and I didn't want to spend an hour fiddling with the commits.
Finally, I deleted some ignore type comments for mypy because the commits weren't going on my system. Is there some other fix for that? Other than that I think it's good to go.
black pandasgit diff upstream/master -u -- "*.py" | flake8 --diff