BUG: Let DataFrame.quantile() handle datetime

[TomAugspurger]

Closes #6965

[TomAugspurger]

@jreback I'm having trouble with a bit of this.

With a DataFrame

In [3]: df = DataFrame({'a': pd.to_datetime(['2010', '2011']), 'b': [0, 5]})

In [5]: df.dtypes
Out[5]: 
a    datetime64[ns]
b             int64
dtype: object

In [6]: df.quantile(.5, numeric_only=False)

For the implementation, I've got a function f that I apply to each column essentially via df.apply.
When f gets the first column a the dtype is object, and I'm not able to convert from object back to datetime64.

ipdb> arr
0    2010-01-01 00:00:00
1    2011-01-01 00:00:00
Name: a, dtype: object
ipdb> _values_from_object(arr)
array([Timestamp('2010-01-01 00:00:00'), Timestamp('2011-01-01 00:00:00')], dtype=object)
ipdb> _values_from_object(arr).view('i8')
*** TypeError: Cannot change data-type for object array.
ipdb> arr.convert_objects(convert_dates=True)
*** TypeError: Cannot change data-type for object array.

What's to best way to get from these timestamps to viewing as i8 so that the quantiles can be computed?

[jreback]

just iterate over with iteritems() and concat the results

apply has a lot of heuristics that u don't need

[TomAugspurger]

Thanks. That's a lot better.

[jreback]

I don't think you actually need the if (if its empty it just won't iterate over anything)...but no biggie

[TomAugspurger]

I left it out originally, but I hit an error somewhere, I can't remember exactly what.

[jreback]

no prob...this looks fine otherwise