Make pivot values work with most iterables

[cscanlin]

Previously the values check for pivot tables was only looking for lists or tuples. This patch allows for any generators or other iterables (such as dict.keys() in python 3) to be passed as values. They will be converted to lists when they are identified as iterables, excluding strings. The performance hit from the extra list type assignment to iterables that are already lists or tuples should be minor.

[jreback]

what issue is this supposed to address?

need tests

[max-sixty]

use is_list_like?

[cscanlin]

Yes that's exactly what I was going for, thanks for the tip!

[cscanlin]

The problem is that right now the only objects that can be passed in as values to a pivot table are lists and tuples. I would like to expand this so that other iterable objects like generators and (in python 3) dictionary methods like .keys() can be passed in as well. My use case is something like this:

import pandas as pd
import numpy as np

df =pd.DataFrame({'A': ['foo', 'foo', 'foo', 'foo',
                             'bar', 'bar', 'bar', 'bar',
                             'foo', 'foo', 'foo'],
                       'B': np.random.randn(11),
                       'C': np.random.randn(11)})

print(df)

aggs = {'B':np.sum, 'C':np.mean}

pivot_output = pd.pivot_table(
    df,
    index=['A'],
    values=aggs.keys(),
    aggfunc=aggs,
)

print(pivot_output)

This works in Python 2, but returns an error in Python 3 because .keys() returns a dict_keys object instead of a list of keys. This forces you to explicity change it to a list before passing it through which feels clunky: values=list(aggs.keys()),

@MaximilianR pointed out that using is_list_like is probably preferable in the source, instead of the conditional I wrote. I am also wondering if its neccessary to expicitly convert other iterable objects to a list, or if they would work appropriately as is. I need to test more.

As far as unit tests go, I'm still learning and I'm not sure how the best way to write those for this would be, but I'll continue looking into it.

I may have jumped the gun on this PR a bit, so please let me know if theres a better way I should proceed.

[max-sixty]

In lieu of anyone more experienced answering your questions:

• If you can use one of the standard checks, that's great. is_list_like looks good.
• Unit tests are awesome. You can start with that use case if you want - just assert that the result is what you want. If you haven't run any nosetests before, have a look at the contributor docs. If you need help writing the tests, have a look at the existing tests, or post questions here.
• Coercing to a list might be necessary, but probably isn't. Write the tests; if you have problems see if coercing helps.

Congrats on being halfway to your first PR.

[jreback]

this needs tests

[jreback]

add the issue number as a comment here

[jreback]

lgtm. pls add a whatsnew note in enhancements section, use this PR number as the issue number

[jreback]

thanks!

[cscanlin]

Thanks @jreback and @MaximilianR for all your help and suggestions on my first PR!