How should DataFrame.isin() handle DataFrame input?

[cpcloud]

n = 5
df = DataFrame({'a': randint(3, size=n),
                'b': randn(n)})
df2 = DataFrame({'a': randint(4, size=n),
                 'b': df.b + (rand(n) > 0.5).astype(float)})
df.isin(df2)

yields

	a	b
0	False	True
1	False	True
2	False	False
3	False	False
4	False	True

I might expect it to go column by column calling isin on the intersection of the columns. I'm not sure if this was ever defined when DataFrame.isin() was implemented.

[TomAugspurger]

We did isin accepting DataFrames in a second PR. I'll see if I can find the discussion.

[TomAugspurger]

Here's the PR / discussion: #5199

[jorisvandenbossche]

It is also clearly stated in the docstring how it is handled: If values is a DataFrame, then both the index and column labels must match. http://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.isin.html#pandas.DataFrame.isin

[hayd]

see new issue on this #7258.

If you want to do it column-wise, pass a dict:

df.isin(dict(df2))

[jreback]

closing this in favor of #7258