Dataframe diff unexpected result

[kdebrab]

Pandas version 0.17.1:

In [1]: import pandas as pd
   ...: data = pd.DataFrame({'d':[1,2,3],'e':[1.5,2,3],'f':[2,2,3]},index=list('abc'))
   ...: data.diff(axis=1)
Out[1]: 
    d   e  f
a NaN NaN  1
b NaN NaN  0
c NaN NaN  0

Hint: the following gives expected result

In [2]: data.astype(float).diff(axis=1)
Out[2]: 
    d    e    f
a NaN  0.5  0.5
b NaN  0.0  0.0
c NaN  0.0  0.0

[TomAugspurger]

This seems similar to #10930

This comment suggests that the fix there won't work for mixed dtypes (but you only have ints)

[jreback]

this is a dupe of #10539