DataFrame.rename renames wrong elements

[fonnesbeck]

When specifying a dictionary map for renaming the index of a DataFrame, the renaming ends up including column values as well, even when they are not of the same type. For example, in the screen shot below, I have demographic information on race coded as integers 1-7 that I am remapping to string representations, and boolean values for the column headers specifying case status. When I specify a dictionary of mappings from integers to strings as the index argument, it treats the True value in the column header as a 1 and substitutes it.

Even if I specify value for columns it does not change the behavior.

[jreback]

This reproduces.....

In [17]: df = DataFrame(np.arange(15).reshape(5,3),columns=[False,True,2])

In [18]: df
Out[18]: 
   False  True   2
0      0     1   2
1      3     4   5
2      6     7   8
3      9    10  11
4     12    13  14

In [19]: mapper = { 0 : 'foo', 1 : 'bar', 2 : 'bah' }

In [20]: df.rename(columns=mapper)
Out[20]: 
   foo  bar  bah
0    0    1    2
1    3    4    5
2    6    7    8
3    9   10   11
4   12   13   14

Here's the problem

In [21]: True in mapper
Out[21]: True

In [22]: False in mapper
Out[22]: True

This is trying each element of the columns (in the example) and doing:

if x in mapper:
     return mapper[x]
else:
     return x

Not sure of an easy way around this, maybe just use the strings True and False in your columns?

[jtratner]

but OP is passing the mapping to index not columns...right? Otherwise I totally get it because True==1 and False==0

[jreback]

@jtratner its before an unstack so that's correct....

[DriesSchaumont]

take