How to fill true values of a dataframe with column names?

First replace booelan to int and then use mask or where with inverting mask by ~:

df = df.astype(int).mask(df, df.columns.to_series(), axis=1)
print (df)
   A  B  C  D
0  0  B  C  0
1  0  0  C  0
2  A  B  0  0

df = df.astype(int).where(~df, df.columns.to_series(), axis=1)
print (df)
   A  B  C  D
0  0  B  C  0
1  0  0  C  0
2  A  B  0  0

Thank you John Galt for improvement in new versions of pandas 0.21.x:

df = df.astype(int).mask(df, df.columns, axis=1)

numpy solution:

a = np.tile(df.columns, [len(df.index),1])
print (a)
[['A' 'B' 'C' 'D']
 ['A' 'B' 'C' 'D']
 ['A' 'B' 'C' 'D']]

df = pd.DataFrame(np.where(df.astype(int), a, 0), columns=df.columns, index = df.index)
print (df)
   A  B  C  D
0  0  B  C  0
1  0  0  C  0
2  A  B  0  0

pandas 1.5.2

df = df.mask(df.astype(bool), df.columns.to_series(), axis=1)

not astype(int) but astype(bool) or astype('bool')

Otherwise ValueError: Boolean array expected for the condition, not uint8

can't remove .to_series()

Otherwise ValueError: other must be the same shape as self when an ndarray