bug in bool type series logical AND operation

[stsouko]

Code Sample, a copy-pastable example if possible

s1=pd.Series([True,False,True,True])
s2=pd.Series([True,True,False])

s1.index=pd.MultiIndex.from_tuples( [(0, 2), (1, 1), (1, 2), (2, 1)],names=['st', 'at'])
s2.index=pd.Index([0,1,2], name='st')

ds1=pd.DataFrame(s1)
ds2=pd.DataFrame(s2)

s3=pd.Series([True,False,True])
s4=pd.Series([True,True,False])

Expected Output

st  at
0   2      True
1   1     False
1   2      True
2   1     False
dtype: bool

For series, * and & give different result:

In: s1 & s2
Out: 
st  at
0   2     False
1   1     False
1   2     False
2   1     False
dtype: bool

In: s1 * s2
Out: 
st  at
0   2      True
1   1     False
1   2      True
2   1     False
dtype: bool

BUT for dataframes they give the same:

In: ds1 * ds2
Out: 
           0
st at       
0  2    True
1  1   False
1  2    True
2  1   False

In: ds1 & ds2
Out: 
           0
st at       
0  2    True
1  1   False
1  2    True
2  1   False

and last for series with single index, also both s3 & s4 or s3 * s4 give:

0     True
1    False
2    False
dtype: bool

output of pd.show_versions()

pandas: 0.18.1

[jorisvandenbossche]

It seems there is going something wrong with the alignment.

Also, if you switch the position, you get an error:

In [43]: s2 & s1
...
ValueError: Buffer dtype mismatch, expected 'Python object' but got 'long long'

[jreback]

this is a duplicate of #1134

you need to join before you do this - it's not s well defined operation as specified

[jorisvandenbossche]

I think it is a slightly different issue

Consider those serieses:

In [84]: s1 = pd.Series([True,False,True])

In [85]: s2 = pd.Series([True,True,False])

In [86]: s2b = pd.Series([True,True,False], index=[1,2,3])

The issue in #1134 is that the index is ignored, so using s2 or s2b gives the same:

In [87]: s1 == s2
Out[87]:
0     True
1    False
2    False
dtype: bool

In [88]: s1 == s2b
Out[88]:
0     True
1    False
2    False
dtype: bool

But in this case, s2 and s2b do not give the same:

In [89]: s1 & s2
Out[89]:
0     True
1    False
2    False
dtype: bool

In [90]: s1 & s2b
Out[90]:
0    False
1    False
2     True
dtype: bool

In the case of non-identical index, the result is also not the same as when the index is ignored:

In [91]: s1.values & s2b.values
Out[91]: array([ True, False, False], dtype=bool)