Add method to sort *within* a single MultiIndex level

[wesm]

No description provided.

[cpcloud]

related

http://stackoverflow.com/questions/18189981/reordering-columns-rows-of-a-pivot-table/18191656#18191656

[hayd]

I think this is fixed by #6135

[dragoljub]

So did we implement sorting a column MultiIndex by specific subset of levels?
For example: df.sortlevel(axis=1, level=[2,1], sort_remaining=True) to sort only by level 2 then 1 ignoring all other levels?

Does column indexing work correctly if the column multindex is not fully lexsorted? I know this is a limitation of the row multiindex.

[jreback]

#8526 is going to be in 0.15.2 but you in General should always fully lexsort

[toasteez]

Is there a way to custom sore the level for example for a normal dataframe you can create a list in the custom order and use something like:

def sort_df(df, sort_list, column): df[column] = pd.Categorical(df[column], sort_list) df = df.sort_values(column) return df

[jorisvandenbossche]

@toasteez #3942 is an issue about adding a keys kwarg to be able to do something like this. WIth 0.19.0 (which preserves Categoricals in a MultiIndex), you will also be able to do something similar as your workaround for columns.