GH1217 Modify items/iterros/itertuples methods to return Iterator

[loicdiridollou]

• Closes Return type for DataFrame/Series iteration methods should be Iterator, not Iterable #1217
• Tests added: Please use assert_type() to assert the type of any return value

[Dr-Irv]

Can we add a test of this form for iterrows() and itertuples():

for t1, t2 in df.iterrows():
    check(assert_type(t1, Hashable), Hashable)
    check(assert_type(t2, pd.Series), pd.Series)

[Dr-Irv]

We shouldn't be using _PandasNamedTuple here in the call to check because what is happening internally is that pandas is creating a new class of type pandas.core.foo.Pandas, and if you use the name="Foobar" argument, the class is of the type pandas.core.foo.Foobar

But it's a dynamic class, so I think something like this would be appropriate:

for t1 in df.itertuples():
    assert_type(t1, _PandasNamedTuple)
    assert(t1.__class__.__name__ == "Pandas")
    assert(isinstance(t1.Index, int))
    assert(isinstance(t1.col1, int))  
    assert(isinstance(t1.col2, int))
    for k in [0, 1, 2]:
        assert(isinstance(t1[k], int))

And you could add a similar test for having name="Foobar" as the argument, and only the line that checks the name of the class would need to be changed.

[Dr-Irv]

This loop and the one below should be for k in [0,1,2] since the tuple has 3 components.

[loicdiridollou]

Ahh yes correct I thought it was iterating over the number of elements.

[Dr-Irv]

thanks @loicdiridollou

[tilsche]

Thank you for the quick fix, merge and release.