pandas has its own typing since 2.0

[davetapley]

Environment data

• Language Server version: v2023.11.10
• OS and version: Codespace mcr.microsoft.com/devcontainers/universal:2
• Python version (& distribution if applicable, e.g. Anaconda): ghcr.io/devcontainers/features/python:1 on 3.11

Code Snippet

from datetime import datetime
from pandas import Timestamp

Timestamp(datetime.now().isoformat()).as_unit('us')

Repro Steps

1. pip install pandas==2.1.0
2. paste above snippet

Expected behavior

Type check ok because as_unit

Actual behavior

Cannot access member "as_unit" for type "Timestamp"
  Member "as_unit" is unknownPylance[reportGeneralTypeIssues](https://github.com/microsoft/pyright/blob/main/docs/configuration.md#reportGeneralTypeIssues)

because as_unit not in the pandas-stubs.

because:

• "Timestamp" has no attribute "as_unit" [attr-defined] pandas-dev/pandas-stubs#797

But this raises the question: Do we need to use pandas-stubs now 2.0 has its own types?

[davetapley]

Related issue I found in my research:

• Pylance uses its own type stubs over actually installed libraries #1197
• Stubs not found for installed typed library #3084
• Switch pandas stubs to the ones maintained by pandas team python-type-stubs#191

[debonte]

This seems like a question for the folks at https://github.com/pandas-dev/pandas-stubs. If pandas has sufficient inline types to eliminate the need for stubs, then they should ship with a py.typed file. The presence of that file would cause type checkers to favor the inline types over any available stubs.

[davetapley]

@debonte oh, I naively assumed it was py.typed, but apparently not 😑

It looks like as of Feb they were recommending pandas-stubs, I'll ask there...

• Typing Stubs and PEP 561 compatibility pandas-dev/pandas#28142 (comment)