Breaking change multiplication bool and int

[Tux1]

Code Sample, a copy-pastable example if possible

import pandas as pd
a = pd.DataFrame([[False, True], [False, False]])
print(a.dtypes)
b = a * 1
print(b.dtypes)

Problem description

With 0.20.3, I get b as DataFrame int32.
With 0.23.3, I get b as DataFrame object.

Expected Output

With 0.20.3

a.dtypes =
0 bool
1 bool
dtype: object

b.dtypes =
0 int32
1 int32
dtype: object

With 0.23.3

a.dtypes =
0 bool
1 bool
dtype: object

b.dtypes =
0 object
1 object
dtype: object

Versions

pandas 0.20.3 vs pandas 0.23.3

[jbrockmendel]

What do you get if you multiply a series instead of a dataframe?

[Tux1]

Don't understand your question
EDIT: I amended my example adding .dtypes. maybe it is clearer for you now

[jorisvandenbossche]

A Series works infers the dtype correctly:

In [71]: a[0]
Out[71]: 
0    False
1    False
Name: 0, dtype: bool

In [72]: a[0] * 1
Out[72]: 
0    0
1    0
Name: 0, dtype: int64

[Tux1]

Any idea about the root cause ? If you give me some tips, I can have a look

[jbrockmendel]

No idea what changed it, but it'll be fixed by the follow-up after #22074.