Groupby.cumsum()'s skipna option does not work

[qiaoqiaozhu]

Code Sample, a copy-pastable example if possible

df = pd.concat([pd.DataFrame({'x':np.random.randn(3), 'gp': 'a'}), pd.DataFrame({'x':np.random.randn(3), 'gp': 'b'})])
df.groupby('gp')['x'].cumsum(skipna=False)

Problem description

DataFrame.cumsum(skipna=False) works.
But DataFrame.groupby['var'].cumsum(skipna=False) does not work.

as seen from the above codes. Error message says numpy operations are not valid with groupby. This is inconsistent behavior, and is in conflict with docs http://pandas.pydata.org/pandas-docs/version/0.17.0/generated/pandas.core.groupby.DataFrameGroupBy.cumsum.html

[jreback]

pls show_versions() - you are pointing to very old docs

[qiaoqiaozhu]

I'm using pandas 0.22.0 build py36h0a44026_0

[jreback]

yeah we are being a little strict on this and immeditately rejecting this to prevent users from doing np.cumsm(df.groupby(...)) which is an invalid construct. So skipna just needed to be added to the signature and passed thru (for all for the cum* methods). want to do a PR?

cc @gfyoung

[shangyian]

If no one else is working on this, I'll take a stab at a PR

[gfyoung]

Go for it!