ENH: Add sort keyword to stack #53282
Conversation
mroeschke
added
Enhancement
Reshaping
pandas/core/reshape/reshape.py
Outdated
| level_codes = sorted(set(mi_cols.codes[-1])) | ||
| level_codes = unique(mi_cols.codes[-1]) | ||
| if sort: | ||
| level_codes = sorted(level_codes) |
There was a problem hiding this comment.
this changes level_codes to a list? can/should we preserve it as ndarray?
There was a problem hiding this comment.
Good idea. Used np.sort here instead
|
if the default were False (mentioned in the #15105), would it be straightforward to reproduce the current behavior with something like |
|
Yeah that's a likely future behavior where we default to False and the deprecate this keyword, but we need to set it to true for backward compat |
|
Any other comments here @jbrockmendel? |
* ENH: Add sort keyword to stack * Removed commented * Use np.sort
| @@ -711,7 +714,7 @@ def _convert_level_number(level_num: int, columns: Index): | |||
| roll_columns = roll_columns.swaplevel(lev1, lev2) | |||
| this.columns = mi_cols = roll_columns | |||
|
|
|||
| if not mi_cols._is_lexsorted(): | |||
| if not mi_cols._is_lexsorted() and sort: | |||
There was a problem hiding this comment.
I'm wondering if this is behaving as intended. In the example below, I would think that the rows would be swapped (the stacked 2nd level of the index being 0 1 instead of 1 0)
levels = ((0, 1), (1, 0))
stack_lev = 1
columns = MultiIndex(levels=levels, codes=[[0, 0, 1, 1], [0, 1, 0, 1]])
df = DataFrame(columns=columns, data=[range(4)])
df_stacked = df.stack(stack_lev, sort=True)
print(df_stacked)
# 0 1
# 0 1 0 2
# 0 1 3
# Expected?
# 0 1
# 0 0 1 3
# 1 0 2
mi_cols._is_lexsorted() is checking if the codes are lexsorted (they are [[0, 0, 1, 1], [0, 1, 0, 1]] here) but not if the values are sorted (they are [[0, 1], [1, 0]] here).
Should sort be sorting the level values?
There was a problem hiding this comment.
Nice find. Yeah I agree and I think the level values should be sorted here
There was a problem hiding this comment.
Thanks, I plan on putting up a PR for this.
* ENH: Add sort keyword to stack * Removed commented * Use np.sort
sort=Falseoption to stack/unstack/pivot #15105 (Replace xxxx with the GitHub issue number)doc/source/whatsnew/vX.X.X.rstfile if fixing a bug or adding a new feature.