Indexing MultiIndex with duplicates return Series instead of raw value

[LucasG0]

• 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, a copy-pastable example

single_index_no_duplicates = pd.Index([0, 1, 2])
single_index_with_duplicates = pd.Index([0, 1, 1])

multi_index_no_duplicates = pd.MultiIndex.from_tuples([(0, 0), (0, 1), (1, 0)])
multi_index_with_duplicates = pd.MultiIndex.from_tuples([(0, 0), (1, 0), (1, 0)])

>>> pd.Series(range(3), single_index_no_duplicates).loc[0]
0
>>> pd.Series(range(3), single_index_with_duplicates).loc[0]
0
>>> pd.Series(range(3), multi_index_no_duplicates).loc[(0,0)]
0
>>> pd.Series(range(3), multi_index_with_duplicates).loc[(0,0)]
0  0    0
dtype: int64

Problem description

When indexing a unique index value with .loc on a MultiIndex containing duplicated values,
the return type is a Series while it is a raw value for single Index.

Expected Output

I think we should expect a raw value to keep consistent with Index behavior.

>>> pd.Series(range(3), multi_index_with_duplicates).loc[(0,0)]
0

Output of pd.show_versions()

INSTALLED VERSIONS

commit : 2cb9652
python : 3.8.5.final.0
python-bits : 64
OS : Linux
OS-release : 5.8.0-55-generic
Version : #62~20.04.1-Ubuntu SMP Wed Jun 2 08:55:04 UTC 2021
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : fr_FR.UTF-8
LOCALE : fr_FR.UTF-8

pandas : 1.2.4
numpy : 1.20.3
pytz : 2019.3
dateutil : 2.7.3
pip : 20.0.2
setuptools : 45.2.0
Cython : None
pytest : 6.2.4
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : 1.1
pymysql : None
psycopg2 : None
jinja2 : None
IPython : None
pandas_datareader: None
bs4 : None
bottleneck : None
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pyxlsb : None
s3fs : None
scipy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
numba : None

[attack68]

slightly subjective what should happen here.

see #39775 for many similar cases, this is possibly a duplicate.

[Dr-Irv]

I think this is a bit different than #39775, which is more related to what happens if there are missing keys.

Here, the issue is that if the MultiIndex contains duplicates, and you ask for a key that is not duplicated in the index, you get a Series. If there are no duplicate keys in the MultiIndex, you get a single value. I ran into this in some code I was writing, where I knew there might be duplicates, or not, after some merging, and then the type of the result was inconsistent, dependent on whether there were duplicate keys in the index.

Here's another example:

import pandas as pd

df = pd.DataFrame(
    [["a", 1, 10], ["a", 1, 20], ["b", 2, 30]], columns=["ab", "ot", "val"]
).set_index(["ab", "ot"])
print(df)

s2 = df["val"].loc[("b", 2)]

print("result")
print(s2)
print()
print("s2 type", type(s2))
print()

df2 = pd.DataFrame(
    [["a", 1, 10], ["c", 1, 20], ["b", 2, 30]], columns=["ab", "ot", "val"]
).set_index(["ab", "ot"])
print(df2)

s3 = df2["val"].loc[("b", 2)]

print("result")
print(s3)
print()
print("s3 type", type(s3))

Here's the output:

       val
ab ot
a  1    10
   1    20
b  2    30
result
ab  ot
b   2     30
Name: val, dtype: int64

s2 type <class 'pandas.core.series.Series'>

       val
ab ot
a  1    10
c  1    20
b  2    30
result
30

s3 type <class 'numpy.int64'>

In both cases, I am doing .loc using the key ("b", 2). In the first case, other indices are duplicated, and I get a Series as the result. In the second case, no indices are duplicated, and I get a single value as the result.

So now I have to check the type of the result to determine what computation to do next. This is very unfriendly!

[CharlesD38]

I can confirm this bug still exists for 2d DataFrame.

Copy Paste Example Below:

import pandas as pd

df = pd.DataFrame(
    index = pd.MultiIndex.from_tuples(
        [
            ('a', 'b'),
            ('a', 'b'),
            ('a', 'c'),
            ('b', 'c'),
            ('b', 'd')
        ]
    ),
    columns = ['Tens', 'Hundreds']
)
df['Tens'] = [10,20,30,40, 50]
df['Hundreds'] = [100,200,300,400, 500]

print(df)
print('\n\n')
print('Returning Series Instead of Value')
print(df.loc[('b', 'c'), 'Tens'])


df = pd.DataFrame(
    index = pd.MultiIndex.from_tuples(
        [
            ('a', 'b'),
            ('a', 'd'),
            ('a', 'c'),
            ('b', 'c'),
            ('b', 'd')
        ]
    ),
    columns = ['Tens', 'Hundreds']
)
df['Tens'] = [10,20,30,40, 50]
df['Hundreds'] = [100,200,300,400, 500]

print()
print(df)
print('\n\n')
print('Returning Series Instead of Value')
print(df.loc[('a', 'c'), 'Tens'])

Output:

     Tens  Hundreds
a b    10       100
  b    20       200
  c    30       300
b c    40       400
  d    50       500



Returning Series Instead of Value
b  c    40
Name: Tens, dtype: int64
     Tens  Hundreds
a b    10       100
  d    20       200
  c    30       300
b c    40       400
  d    50       500



Returning Value Instead of Series
30