DataFrame division returns infinity for 0 / 0

[keszybz]

xref #9286

0 / 0 does not have a value, and should return NaN, not infinity.

>>> x = pd.DataFrame([[0, 0], [1, 0], [-1, 0]])
>>> x[0] / x[1]
0    inf                              <------------- here
1    inf
2   -inf
dtype: float64

numpy does this correctly:

>>> y = np.array([[0, 0], [1, 0], [-1, 0]])
>>> y[:, 0] / y[:, 1]
array([ nan,  inf, -inf])

[jreback]

show your numpy/pandas versions.

numpy used to get this completely wrong for integer division (it would return 0,0,0) in 1.8.1

not sure if they fixed it in 1.9

[keszybz]

>>> pd.__version__
'0.14.1'
>>> np.__version__
'1.9.0'
>>> sys.version
'2.7.5 (default, Jun 25 2014, 10:19:55) \n[GCC 4.8.2 20131212 (Red Hat 4.8.2-7)]'

[jreback]

I think I can buy adding this

want to add a test here: https://github.com/pydata/pandas/blob/master/pandas/tests/test_series.py#L2491
(prob need to restrict this testing to >= 1.9) and test separately for < 1.9

and fix (which is a teeny bit more complicated) (see core/common/fill_zeros)

cc @seth-p

agree?

[seth-p]

I agree that 0/0 should be NaN, not inf.

[jreback]

cc @jaimefrio
(sorry mispelled before). I think 1.10 in numpy is now in sync AFAIK (< 1.10 not so much though)