BUG: DataFrame.stack with sort=True and unsorted MultiIndex levels

[rhshadrach]

Ref: #53282 (comment)

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)
#      0  1
# 0 1  0  2
#   0  1  3

# Expected
#      0  1
# 0 0  1  3
#   1  0  2

When the MultiIndex levels are not sorted in ascending order, DataFrame.stack with sort=True does not sort the values in the result.

[rhshadrach]

This was a misunderstanding of what sort=True is supposed to do. When sort=True, the levels of the MultiIndex get sorted, but not the values. I plan to clarify the docs here.

[jorisvandenbossche]

To be honest, I don't even really understand your last sentence ;)
What do you mean with the "levels get sorted"? Do you mean the labels of one level (eg df_stacked.index.levels[-1]), or the integer codes of the levels (df_stacked.index.codes[-1]) while keeping the labels the same?

The above example might also not be the best, because I don't see any difference between sort=True vs sort=False