BUG: MultiIndex block assignment introduces NaNs in data

[fatvlady]

• 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.

---

Code Sample

import pandas as pd
import numpy as np

cols = pd.MultiIndex.from_product([[1], ['a', 'b']])
index = np.arange(2)
data = np.ones((len(index), len(cols)))
df = pd.DataFrame(data=data, index=index, columns=cols)
print(df)
df.loc[:, 1] = df.loc[:, 1]
print(df)

Problem description

When assigning to block of MultiIndex-ed dataframe (I tested with columnar multi-index), data becomes covered with NaNs instead. Same happens when doing right-assign arithmetic operations (*=, += etc.). It is not required for RHS to point to same data storage - any dataframe with corresponding index/columns gives same effect.

Expected Output

     1
     a    b
0  1.0  1.0
1  1.0  1.0
     1
     a    b
0  1.0  1.0
1  1.0  1.0

Output of pd.show_versions()

INSTALLED VERSIONS

commit : f2c8480
python : 3.8.2.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.18362
machine : AMD64
processor : Intel64 Family 6 Model 158 Stepping 10, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : English_United States.1251

pandas : 1.2.3
numpy : 1.18.1
pytz : 2020.1
dateutil : 2.8.1
pip : 21.0.1
setuptools : 41.2.0
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 2.11.1
IPython : 7.13.0
pandas_datareader: None
bs4 : 4.9.3
bottleneck : None
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : 3.2.0
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pyxlsb : None
s3fs : None
scipy : 1.4.1
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : 1.2.0
xlwt : None
numba : None

[alexprincel]

Confirmed this happens on master branch as well.

[phofl]

I think df.loc[:, 1] reduces the MuldiIndex column header in dimensionality, so we can't match this?

[alexprincel]

I am making very slow progress on understanding this but, interestingly, I just found out that indexing the dataframe instead of using loc does not have any of those issues (e.g. I can perform the operation in your code above with expected result, as well as, say, multiply by two a subset of multi-leveled columns).

See a working example :

import pandas as pd
import numpy as np

cols = pd.MultiIndex.from_product([[1], ['a', 'b']])
index = np.arange(2)
data = np.ones((len(index), len(cols)))
df = pd.DataFrame(data=data, index=index, columns=cols)
print(df)
df[1] = df[1]
print(df)

[bashtage]

I have a proof of concept

import pandas as pd
mi = pd.MultiIndex.from_product([["a","b","c"],[1,2,3],["z","y","x"]])
df = pd.DataFrame(index=mi,dtype=float)
mi2 = pd.MultiIndex.from_product([["a","b","c"],[1,2,3]])
s = pd.Series(index=mi2, dtype=float)
s.iloc[:]=3.14
df["new"] = s
df

On 1.2.5 and 1.3.0 we get

        new
a 1 z  3.14
    y  3.14
    x  3.14
  2 z  3.14
    y  3.14
    x  3.14
  3 z  3.14
    y  3.14
    x  3.14
   ...

On master we get

       new
a 1 z  NaN
    y  NaN
    x  NaN
  2 z  NaN
    y  NaN
    x  NaN
  3 z  NaN
    y  NaN
    x  NaN

This bug is causing failures on statsmodels when testing against pandas master.

[mesvam]

This is the same as underlying issue as #45751 right?

here's another example

df = pd.DataFrame(
    np.arange(3*4).reshape((4, 3)),
    index=pd.MultiIndex.from_product([['a', 'b'], ['c', 'd']]),
    columns=['A', 'B', 'C']
)
print(df)
df.loc['a', 'A'] = df.loc['b', 'B']
print(df)

     A   B   C
a c  0   1   2
  d  3   4   5
b c  6   7   8
  d  9  10  11
       A   B   C
a c  NaN   1   2
  d  NaN   4   5
b c  6.0   7   8
  d  9.0  10  11

very unintuitive behavior