TYP: enable reportUnusedImport

[twoertwein]

pyright's reportUnusedImport checks py and pyi files for unused imports. If an import is explicitly marked as public (re-exported using as or in __all__) it is "used" (flake8 seems to use the same definition).

Adding unused imports to __all__ becomes messy when the file does not yet have __all__: need to list also all public symbols (constants, classes, functions).

[github-actions]

This pull request is stale because it has been open for thirty days with no activity. Please update and respond to this comment if you're still interested in working on this.

[mroeschke]

This generally looked okay if you want to merge main.

[mroeschke]

So just to confirm:

1. Nothing changed from the "public" API?
2. In terms of code flow, we will need to ensure all "unused" imports are defined in __all__?

[twoertwein]

__all__ as two functions: 1) it affects from xyz import * (imports only functions/classes inside __all__) and 2) in a py.typed library __all__ defines the public interface of a module (unless the module is private, starts with _).

1. Nothing changed from the "public" API?

Technically, users will import fewer modules if they use from xyz import * which could be a public API change.

2. In terms of code flow, we will need to ensure all "unused" imports are defined in `__all__`?

flake8/pyright will let you know about unused imports :) They can then either be removed, be ignored (if they have some side effects), or added to __all__ if they are meant to be public.

[mroeschke]

Technically, users will import fewer modules if they use from xyz import * which could be a public API change.

Could you do a quick before/after check when from pandas import * is run? I imagine it should hopefully not change in a negative way, and maybe can be noted as a benefit that less stuff is imported :)

[twoertwein]

Seems there is actually no differences between main and this PR when imported everything from a file:

import importlib
from pathlib import Path
import pickle

import ipdb

imported = {}
for py in Path("pandas").glob("**/*.py"):
    print(py)
    py = str(py).replace("/", ".").removesuffix(".py")
    mod = importlib.import_module("py")
    imported[py] = [x for x in dir(mod) if not x.startswith("_")]
with open("before.pickle", mode="wb") as file:
    pickle.dump(imported, file)

# ran the above first on this PR and save as after.pickle
# and then the entire script on main

after = {}
with open("after.pickle", mode="rb") as file:
    after = pickle.load(file)

for py, imports_b in imported.items():
    imports_a = set(after[py])
    imports_b = set(imports_b)
    remove = imports_b.difference(imports_a)
    added = imports_a.difference(imports_b)

    if len(remove) > 0 or len(added) > 0:
        ipdb.set_trace()  # never ended up here

[mroeschke]

Thanks for double checking @twoertwein!