GroupBy.apply calling function twice

[cancan101]

def foo(x):
    print "hello"
_ = pd.DataFrame({'a':[1]}).groupby(["a"]).apply(foo)

hello
hello

[jreback]

dupe of #2936 and its mentioned here in the docs

[cancan101]

@jreback I am confused in reading those docs: what is the single run version of apply that can be used when I want to return a new dataframe for each grouping that is that concatenated to form one resulting dataframe?

e.g what is the simples way to apply this:

def foo(x):
    print 'xxx'
    return x

and have it called once per group and end up with the same dataset that went in coming out?

[jreback]

df.groupby(...).apply(foo)

not sure why it matters if its called twice. This is only to determine there is mutation or not.

[cancan101]

foo is a very expensive operation that I would prefer is run just once.

[jreback]

then loop your self

``pd.concat([ (g, foo(grp)) for g, grp in df.groupby(....) ])`

[cancan101]

Okay.
Should be:

pd.concat([ foo(grp) for _, grp in df.groupby(...) ])