BUG: breaking change in df.replace() from 1.0.5 to 1.1.0

[agnesbao]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

Note: Please read this guide detailing how to provide the necessary information for us to reproduce your bug.

Code Sample, a copy-pastable example

# Your code here
df = pd.DataFrame({"age": range(1,100)})
df["age"] = pd.cut(df["age"], [0, 50, 65, 99], right=False)
df.replace(to_replace={
    "age":{
        pd.Interval(0, 50, closed='left'): "Ages < 50",
        pd.Interval(50, 65, closed='left'): "Ages 50-64",
        pd.Interval(65, 99, closed='left'): "Ages 65+"
    }
})

Problem description

The above sample code works in 1.0.5 but error out in 1.1.0 and above with error msg:

TypeError: Cannot compare types 'ndarray(dtype=object)' and 'Interval'

It only happens with pd.Interval dtype, nested dict of string to string mapping works fine.

Expected Output

Output of pd.show_versions()

INSTALLED VERSIONS

commit : e3bcf8d
python : 3.7.6.final.0
python-bits : 64
OS : Darwin
OS-release : 19.6.0
Version : Darwin Kernel Version 19.6.0: Thu Jun 18 20:49:00 PDT 2020; root:xnu-6153.141.1~1/RELEASE_X86_64
machine : x86_64
processor : i386
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 1.2.0.dev0+165.ge3bcf8d19
numpy : 1.19.1
pytz : 2020.1
dateutil : 2.8.1
pip : 20.2.2
setuptools : 49.6.0.post20200814
Cython : 0.29.21
pytest : 6.0.1
hypothesis : 5.29.0
sphinx : 2.4.0
blosc : None
feather : None
xlsxwriter : 1.3.3
lxml.etree : 4.5.2
html5lib : 1.1
pymysql : None
psycopg2 : 2.8.4 (dt dec pq3 ext lo64)
jinja2 : 2.11.2
IPython : 7.17.0
pandas_datareader: None
bs4 : 4.9.1
bottleneck : 1.3.2
fsspec : 0.8.0
fastparquet : 0.4.1
gcsfs : None
matplotlib : 3.3.1
numexpr : 2.7.1
odfpy : None
openpyxl : 3.0.5
pandas_gbq : None
pyarrow : 0.15.1
pytables : None
pyxlsb : None
s3fs : 0.2.0
scipy : 1.5.2
sqlalchemy : 1.3.19
tables : 3.6.1
tabulate : 0.8.7
xarray : 0.16.0
xlrd : 1.2.0
xlwt : 1.3.0
numba : 0.50.1

[asishm]

#32107

c6e3e15 is the first bad commit
commit c6e3e15
Author: Daniel Saxton 2658661+dsaxton@users.noreply.github.com
Date: Sat Apr 25 18:58:51 2020 -0500

BUG: Allow addition of Timedelta to Timestamp interval (#32107)

[dsaxton]

Hmm, this is apparently due to

pandas/pandas/_libs/interval.pyx

Line 302 in a1f6056

__array_priority__ = 1000

which was added to allow np.timedelta64 + pd.Interval of timestamps in the linked PR: #32107 (comment). Not sure exactly what's happening to be honest but will take a look.

cc @jbrockmendel if any thoughts

[dsaxton]

So evidently __array_priority__ = 1000 has the not so nice side effect of no longer broadcasting things like equality comparisons (which is causing the bug in the OP):

In [1]: import numpy as np
   ...: import pandas as pd
   ...:
   ...: arr = np.array([pd.Interval(0, 1), pd.Interval(1, 2)])
   ...: arr == pd.Interval(0, 1)
   ...:
Out[1]: False

Probably this behavior is worse than not being able to add np.timedelta64 + pd.Interval(pd.Timestamp, pd.Timestamp).

[simonjayhawkins]

fixed in #35938