replace() involving booleans affects integers and floats equal to 0 and 1

[krassowski]

Code Sample

Series([0, 1, 2, True]).replace({0: 'a', True: 'x', False: 'y'}) 

Output:

	0
0	y
1	x
2	2
3	x

Problem description

I expected zero to be replaced with 'a', and 1 to remain untouched.

• documentation does not clarify whether replace() checks for equality or for identity.
• both inexperienced and experienced users may fall in the trap of the code as shown above; if the current behaviour is desired, would it be reasonable to add a warning when the user's attempts can lead to certainly unintended results (as above)?

Expected Output

	0
0	a
1	1
2	2
3	x

Due research

Stack Overflow:

• How to map True and False to 'Yes' and 'No' in a pandas data frame for columns of dtype bool only? (6k views, 6/8 upvotes) works around this problem by only using booleans in columns with dtype == bool
• I wonder when replace 0 on a Series(or dataframe), will False also be replaced? demonstrates that replacing zeros will not touch booleans if the dtype == bool.
• I verified that my problem exists with dtype in {int, float} (i.e. after removing True from Series)

Output of pd.show_versions()

INSTALLED VERSIONS

commit : None
python : 3.8.0.final.0
python-bits : 64
OS : Linux
OS-release : 5.3.0-26-generic
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_GB.UTF-8
LOCALE : en_GB.UTF-8

pandas : 1.0.0
numpy : 1.18.1
pytz : 2019.3
dateutil : 2.8.0
pip : 19.3.1
setuptools : 41.2.0
Cython : None
pytest : 5.2.1
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 2.10.3
IPython : 7.8.0
pandas_datareader: None
bs4 : None
bottleneck : None
fastparquet : None
gcsfs : None
lxml.etree : None
matplotlib : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pytables : None
pytest : 5.2.1
pyxlsb : None
s3fs : None
scipy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
xlsxwriter : None
numba : None

[krassowski]

It seems that I have made this to complicated an example...

First,

{1: 1, True: True} == {1: True}

So my dict is never truly used as I wish it would be. However, if I remove the zero key which is undistinguishable from False, i.e.:

Series([0, 1, 2, True]).replace({True: 'x', False: 'y'})

I still get a result different from what I would have expected.

[MarcoGorelli]

Thanks for the report. I tried your last example and got

0    y
1    x
2    2
3    x
dtype: object

Since False == 0 and True == 1, this is what I would've expected

[krassowski]

Would it make sense to add "replacement compares equality, not identity" somewhere in the documentation then?

[MarcoGorelli]

IMO it wouldn't hurt to mention it, perhaps with an example. Interesting in submitting a pull request?