API: DataFrame.select_dtypes should accept scalar

[chris-b1]

In [164]: df = pd.DataFrame({'a': [1, 2, 3], 'b': ['a', 'b', 'c']})

In [165]: df.select_dtypes(include='object')
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-165-04044faa1a5a> in <module>()
----> 1 df.select_dtypes(include='object')

~\AppData\Local\Continuum\Anaconda3\lib\site-packages\pandas\core\frame.py in select_dtypes(self, include, exclude)
   2355         include, exclude = include or (), exclude or ()
   2356         if not (is_list_like(include) and is_list_like(exclude)):
-> 2357             raise TypeError('include and exclude must both be non-string'
   2358                             ' sequences')
   2359         selection = tuple(map(frozenset, (include, exclude)))

TypeError: include and exclude must both be non-string sequences

In [166]: df.select_dtypes(include=['object'])
Out[166]: 
   b
0  a
1  b
2  c

Problem description

Only a convenience thing, but basically anywhere else we take list-likes, we accept a single string and I think should do the same here.

pandas 0.20.2

[TomAugspurger]

+100 :) We should do the same for exclude

[Ffisegydd]

I was looking at picking this up as one of my first contributions. A quick question for clarity though.

select_dtypes allows strings ('category', 'datetimetz', etc) but also allows numpy.number. For example:

df.select_dtypes(include=['category', numpy.number])

is valid.

My question is: what's the expected behaviour for df.select_dtypes(include=np.number)? The original issue only mentions allowing strings but it seems silly to exclude np.number, as such I'll continue by assuming that's the way to go but would be good to get some clarity.

[TomAugspurger]

Correct. It should convert any scalar type to an iterable.

…

On Jul 8, 2017, at 07:56, Keiron Pizzey ***@***.***> wrote: I was looking at picking this up as one of my first contributions. A quick question for clarity though. select_dtypes allows strings ('category', 'datetimetz', etc) but also allows numpy.number. For example: df.select_dtypes(include=['category', numpy.number]) is valid. My question is: what's the expected behaviour for df.select_dtypes(include=np.number)? The original issue only mentions allowing strings but it seems silly to exclude np.number, as such I'll continue by assuming that's the way to go but would be good to get some clarity. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.