Skip to content

ENH: Introduce type-safe constructors for Timestamp and Timedelta. #58475

Open
@randolf-scholz

Description

@randolf-scholz

Feature Type

  • Adding new functionality to pandas

  • Changing existing functionality in pandas

  • Removing existing functionality in pandas

Problem Description

The default constructors pd.Timestamp.__new__ and pd.Timedelta.__new__ can return NaT, which is a different type. This can lead to silent type errors, depending on the type-checker used. Consider the following example:

import numpy as np
from pandas import Timedelta, Timestamp

t: Timestamp = Timestamp("2024-04-29T18:00:00")
t2: Timestamp = Timestamp(np.datetime64("nat"))  # actually NaTType!
dt: Timedelta = Timedelta(1, "h")
dt2: Timedelta = Timedelta(np.timedelta64("nat"))  # actually NaTType!

Type-checking results:

  • mypy --strict: No errors (w/ and w/o pandas-stubs)
  • pyright with useLibraryCodeForTypes = true: 4 errors (only formally correct result)
  • pyright with useLibraryCodeForTypes = false: no errors

Feature Description

Introduce new constructors timestamp and timedelta (in analogy to how pyarrow does constructors), which are guaranteed to return pd.Timestamp and pd.Timedelta types, or raise an exception in the case when NaT is encountered.

Alternative Solutions

Split pd.NaT into two different types, Timestamp("NaT") and Timedelta("NaT") (as is the case in numpy), which are instances of the respective types. (#24983)

Additional Context

Metadata

Metadata

Assignees

No one assigned

    Labels

    EnhancementNeeds TriageIssue that has not been reviewed by a pandas team member

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions