I caught this bug while updating from version 0.18.0 to 0.18.1. The kwarg usecols no longer accepts unicode column labels.

Example below:

from io import StringIO import pandas as pd s = u'''AAA,BBB,CCC,DDD 0.056674973,8,True,a 2.613230982,2,False,b 3.568935038,7,False,a ''' buff = StringIO(s) print pd.read_csv(buff, usecols=[u'AAA', u'BBB']) >> ValueError: The elements of 'usecols' must either be all strings or all integers

I note that 0.18.1 introduced the requirement that usecols be all string or all ints. This makes sense but it looks like the implementation also throws away unicode strings.