DataFrame.add with fill_value bug

[wesm]

Very, very surprised this bug isn't caught by the test suite.

df1 = pandas.DataFrame({'a':{True:1,False:5},'b':{True:7,False:8}})
df2 = pandas.DataFrame({'a':{True:1,False:5},'c':{True:7,False:8}})

In [15]: df1.add(df2, fill_value=0)
Out[15]: 
        a   b   c
False  10 NaN NaN
True    2 NaN NaN

[changhiskhan]

the joining turns it into a mixed type frame. The test that checks combineAdd was already float. The test that checks mixed type did not have differing columns.

[wesm]

crapola =/ figured it was something like that related to dtypes

[changhiskhan]

via 5c945e1