API: Add DataFrame.droplevel

[TomAugspurger]

As a way to make dropping an index level possible in an method chain.

>>> df = pd.DataFrame(np.arange(12).reshape(4, 3), columns=['a', 'b', 'c']).set_index(['a', 'b'])
>>> df.droplevel(0)

roughly equivalent to

df2 = df.copy()
df2.index = df.index.droplevel(0)
df2

cc @twiecki

[ericmjl]

@woohoo! I'd love for this to come into pandas, so that I don't have to worry about implementing it in pyjanitor! (Issue on my own package, which I'm more than willing to close if this happens: pyjanitor-devs/pyjanitor#8)

[ericmjl]

Also, a hacky implementation to hopefully kickstart things: https://github.com/ericmjl/pyjanitor/blob/index-manipulation/janitor/functions.py#L324

and

https://github.com/ericmjl/pyjanitor/blob/index-manipulation/janitor/dataframe.py#L57

[twiecki]

Thanks @TomAugspurger! This would be a great addition.

One could make the same case for all other Index methods that return an index besides rename_axis and droplevel(), too. I wonder if it's worth thinking about a more general solution. E.g. something along the lines of a forward of df.index_*() -> df.index = df.index.*().

[jschendel]

Not against this, but will it have any different functionality than reset_index with drop=True?

Seems like that covers what you'd want to do, unless I'm misinterpreting what this is supposed to do:

In [2]: df = pd.DataFrame(np.arange(16).reshape(4, 4), columns=list('abcd')).set_index(list('abc'))

In [3]: df
Out[3]:
           d
a  b  c
0  1  2    3
4  5  6    7
8  9  10  11
12 13 14  15

In [4]: df.reset_index(level=1, drop=True)
Out[4]:
        d
a  c
0  2    3
4  6    7
8  10  11
12 14  15

In [5]: df.reset_index(level='c', drop=True)
Out[5]:
        d
a  b
0  1    3
4  5    7
8  9   11
12 13  15

In [6]: df.reset_index(level=['a', 'c'], drop=True)
Out[6]:
     d
b
1    3
5    7
9   11
13  15

[TomAugspurger]

Ah, good point. I suspect that covers all the needs. This would be good to document in reset_index and in MultiIndex.droplevel.

…

On Wed, Mar 14, 2018 at 11:47 AM, jschendel ***@***.***> wrote: Not against this, but will it have any different functionality than reset_index with drop=True? Seems like that covers a lot of what you'd want to do: In [2]: df = pd.DataFrame(np.arange(16).reshape(4, 4), columns=list('abcd')).set_index(list('abc')) In [3]: df Out[3]: d a b c0 1 2 34 5 6 78 9 10 1112 13 14 15 In [4]: df.reset_index(level=1, drop=True) Out[4]: d a c0 2 34 6 78 10 1112 14 15 In [5]: df.reset_index(level='c', drop=True) Out[5]: d a b0 1 34 5 78 9 1112 13 15 In [6]: df.reset_index(level=['a', 'c'], drop=True) Out[6]: d b1 35 79 1113 15 — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#20342 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABQHIg2mu7QcbSgsWAfFPM-vKhZmLflmks5teUm2gaJpZM4SqWfj> .

[tliu30]

Don't think it totally covers everything...sometimes find a need to drop a column-index level, and i don't think reset_index supports columns (for good reason)

[TomAugspurger]

@jschendel had a good point that this will overlap with reset_index with drop=True. The advantages of a dedicated droplevel being

• droplevel is just for dropping a level, whereas reset_index can drop or insert
• droplevel can drop on either axis.

I don't have a strong opinion on whether we should add it, or better document reset_index.