BUG: DataFrame.replace is broken in 0.12-dev · Issue #4115 · pandas-dev/pandas (original) (raw)

the DataFrame.replace seems to be broken in 0.12 development version

this code:

df = pd.DataFrame({'Type':['Q','T','Q','Q','T'], 'tmp':2})
print df
print df.replace({'Type': {'Q': 0, 'T': 1}})

Return correct result in 0.11.0:

  Type  tmp
0    Q    2
1    T    2
2    Q    2
3    Q    2
4    T    2
5    T    2
  Type  tmp
0    0    2
1    1    2
2    0    2
3    0    2
4    1    2
5    1    2

and in 0.12.0.dev-795004d it does not work:

  Type  tmp
0    Q    2
1    T    2
2    Q    2
3    Q    2
4    T    2
5    T    2
   Type  tmp
0     0    2
1     1    2
2     0    2
3     1    2
4     0    2
5     1    2

Series.replace it works well, so the workaround for this bug is the following:

df['Type'] = df['Type'].replace({'Q': 0, 'T': 1})