Pandas ignores index when assigning different dtypes. · Issue #8387 · pandas-dev/pandas (original) (raw)

Consider the following example:

import pandas

foo = pandas.Series( [ 0, 1, 2, 0 ] )

w = foo > 0 bar = foo[ w ].map( str ) print bar

foo[w] = bar print foo

foo[w] = bar print foo

gives the following output

1    1
2    2
dtype: object

0      0
1    NaN
2      1
3      0
dtype: object

0    0
1    1
2    2
3    0
dtype: object

It looks as if in the first assignemend foo[w] = bar the matching index of both sides is ignored.

Is this a bug or documented behavior?