groupby on single col df results in ValueError on Master, works on 0.18.1

[alan-wong]

On 0.18.1 the following works without error

df = pd.DataFrame({'a':list('abssbab')})
df.groupby('a').count()

this results in an empty df:

Out[255]:
Empty DataFrame
Columns: []
Index: [a, b, s]

On Master it results in an error see related: #13393

[jorisvandenbossche]

Hmm, it now does work on master:

In [1]: df = pd.DataFrame({'a':list('abssbab')})

In [2]: df.groupby('a').count()
Out[2]:
Empty DataFrame
Columns: []
Index: [a, b, s]

In [3]: pd.__version__
Out[3]: '0.18.1+156.g3944a36'

I fetched an rebuilt in the meantime, so maybe my pandas master was not built correctly when I tested that. Sorry about that.

Anyway, the empty frame is the correct result I think, so we could add a test for this. Do you want to do a PR for this?

[alan-wong]

Wish I could but I can't compile pandas on my dev machine due to compiler issues

[jorisvandenbossche]

@alan-wong If you want tro try to solve those issues and have questions about that, just ask! (but better on https://gitter.im/pydata/pandas, that's a better medium for this instead in an issue)

[alan-wong]

@jorisvandenbossche thanks for the link I was completely unaware of that channel I'll join now

[sinhrks]

Will add tests and close this.