Problem accessing .loc with tuple for MultiIndex Series

[lionel42]

Describe the bug
When accessing with tuples a MultiIndex Serie, I get an Invalid index type error

The example code is from the offical pandas user guides:

https://pandas.pydata.org/docs/user_guide/advanced.html#advanced-indexing-with-hierarchical-index

It seems the type is simply not in the expected types, but it should by, as this is an offical way of accessing multi-index timeseries.

To Reproduce
Minimal runnable pandas example that is not properly checked by the stubs:

 import pandas as pd
s = pd.Series(

    [1, 2, 3, 4, 5, 6],

    index=pd.MultiIndex.from_product([["A", "B"], ["c", "d", "e"]]),

)

s.loc[[("A", "c"), ("B", "d")]]

using mypy

receiving:

 error: Invalid index type "list[tuple[str, str]]" for "_LocIndexerSeries[int]"; expected type "Series[bool] | ndarray[Any, dtype[bool_]] | list[bool] | Index[Any] | Sequence[float] | list[str] | slice | tuple[Index[Any] | Series[bool] | ndarray[Any, dtype[bool_]] | list[bool] | str | bytes | date | datetime | timedelta | datetime64 | timedelta64 | bool | int | float | Timestamp | Timedelta | complex | list[Any] | slice | tuple[str | bytes | date | datetime | timedelta | datetime64 | timedelta64 | bool | int | float | Timestamp | Timedelta | complex, ...], ...] | Callable[..., Any]"  [index]

• OS: Windows
• OS Version 10
• python version 3.12
• version of type checker 1.9.0
• version of installed pandas-stubs 2.2.1.240316

[Dr-Irv]

Thanks for the report.

I think that if we add list[_IndexSliceTuple] to:

pandas-stubs/pandas-stubs/core/series.pyi

Lines 194 to 201 in 0342d90

	idx: (
	MaskType
	| Index
	| Sequence[float]
	| list[str]
	| slice
	| _IndexSliceTuple
	| Callable

that will fix this. PR with tests welcome.

[lionel42]

I updated a test case for the issue in the PR.

I tested the above suggested solution which fails with mypy .

It gives

tests\test_series.py:186: error: Invalid index type "list[tuple[str, str]]" for "_LocIndexerSeries[int]"; expected type "Series[bool] | ndarray[Any, dtype[bool_]] | list[bool] | Index[Any] | Sequence[float] | list[str] | slice | tuple[Index[Any] | Series[bool] | ndarray[Any, dtype[bool_]] | list[bool] | str | bytes | date | datetime | timedelta | datetime64 | timedelta64 | bool | int | float | Timestamp | Timedelta | complex | list[Any] | slice | tuple[str | bytes | date | datetime | timedelta | datetime64 | timedelta64 | bool | int | float | Timestamp | Timedelta | complex, ...], ...] | list[tuple[Index[Any] | Series[bool] | ndarray[Any, dtype[bool_]] | list[bool] | str | bytes | date | datetime | timedelta | datetime64 | timedelta64 | bool | int | float | Timestamp | Timedelta | complex | list[Any] | slice | tuple[str | bytes | date | datetime | timedelta | datetime64 | timedelta64 | bool | int | float | Timestamp | Timedelta | complex, ...], ...]] | Callable[..., Any]"  [index]

instead of

tests\test_series.py:186: error: Invalid index type "list[tuple[str, str]]" for "_LocIndexerSeries[int]"; expected type "Series[bool] | ndarray[Any, dtype[bool_]] | list[bool] | Index[Any] | Sequence[float] | list[str] | slice | tuple[Index[Any] | Series[bool] | ndarray[Any, dtype[bool_]] | list[bool] | str | bytes | date | datetime | timedelta | datetime64 | timedelta64 | bool | int | float | Timestamp | Timedelta | complex | list[Any] | slice | tuple[str | bytes | date | datetime | timedelta | datetime64 | timedelta64 | bool | int | float | Timestamp | Timedelta | complex, ...], ...] | Callable[..., Any]"  [index]

I also tried adding an additional overload option:

    # Multiindex case
    @overload
    def __getitem__(
        self,
        idx: list[_IndexSliceTuple],
    ) -> Series[S1]: ...

This one worked with mypy (no error raised) but it failed for pyright

Poe => test

===========================================
Beginning: 'Run mypy on 'tests' (using the local stubs) and on the local stubs'
===========================================

Success: no issues found in 226 source files

===========================================
End: 'Run mypy on 'tests' (using the local stubs) and on the local stubs', runtime: 39.946 seconds.
===========================================


===========================================
Beginning: 'Run pyright on 'tests' (using the local stubs) and on the local stubs'
===========================================


test_series.py:186:23 - error: No overloads for "__getitem__" match the provided arguments (reportCallIssue)
test_series.py:186:23 - error: Argument of type "list[tuple[Literal['A'], Literal['c']] | tuple[Literal['B'], Literal['d']]]" cannot be assigned to parameter "idx" of type "list[_IndexSliceTuple[Unknown]]" in function "__getitem__"
    "list[tuple[Literal['A'], Literal['c']] | tuple[Literal['B'], Literal['d']]]" is incompatible with "list[_IndexSliceTuple[Unknown]]"
      Type parameter "_T@list" is invariant, but "tuple[Literal['A'], Literal['c']] | tuple[Literal['B'], Literal['d']]" is not the same as "_IndexSliceTuple[Unknown]"
      Consider switching from "list" to "Sequence" which is covariant (reportArgumentType)
test_series.py:186:23 - error: "assert_type" mismatch: expected "Series[int]" but received "Unknown" (reportAssertTypeFailure)
3 errors, 0 warnings, 0 informations

===========================================
Step: 'Run pyright on 'tests' (using the local stubs) and on the local stubs' failed!
===========================================

[Dr-Irv]

Instead of list[_IndexSliceTuple] that I suggested here: #899 (comment)

use Sequence[_IndexSliceTuple] . I tried it on the example above and it worked.