Add typing & type checking

[Zethson]

There's lots of untyped code. Let's also add the checker to the pre-commit setup. ty or mypy

[lauraluebbert]

Reopening because there's still substantive work left here.

What landed in c6afd210:

• Type annotations on 175 functions across all 28 source files in gget/ (from __future__ import annotations everywhere, behavior unchanged).
• mypy wired into .pre-commit-config.yaml and [tool.mypy] configured in pyproject.toml, with pandas-stubs / types-requests / types-tqdm / types-beautifulsoup4.
• Permissive baseline (disallow_untyped_defs = false, etc.) so existing errors don't block CI on day one.

Current baseline: Found 613 errors in 24 files (checked 30 source files). By category:

Count	Code	Pattern
157	[index]	pandas DataFrame/Series subscripts mypy can't narrow
135	[union-attr]	DataFrame | dict return types — calling a method on the union where only one branch has it; biggest single fix would be @typing.overload on the json=True/False pattern that's used across many modules
68	[attr-defined]	dynamically-built objects (JSON responses, etc.)
58	[call-overload]	pandas/numpy overload mismatches
56	[assignment]	width mismatches
53	[arg-type]	mostly str | None passed where str expected
~85	other	string ops, operators, return-value, etc.
21	[var-annotated]	empty dict = {} / list = [] literals needing explicit annotation — fastest wins

Suggested follow-up plan:

1. Quick wins: knock out the 21 [var-annotated] errors (mechanical, no judgement calls).
2. Add @typing.overload for the json=True / json=False return-type narrowing pattern. Should collapse ~135 [union-attr] errors in a single design change.
3. Tighten one module at a time when contributors touch it; add per-module mypy overrides (disable_error_code = []) as each module hits zero.
4. Once every module is clean, flip disallow_untyped_defs = true so new untyped code is rejected at CI.

[Qiaoli-Li-Res]

Hi! Would a small incremental PR fixing the remaining mypy errors in gget_pdb.py be useful? I’d keep it limited to type annotations for the URL fallback list, with no runtime behavior changes.

[lauraluebbert]

Update: mypy now runs in pre-commit via a baseline (new code is gated; existing errors grandfathered)

Adopting mypy wholesale was blocked by the ~550 pre-existing errors: a plain mypy hook exits non-zero, and since pre-commit.ci runs --all-files, it failed on every PR regardless of what the PR touched — which trains everyone to ignore the check.

Fixed on dev (e0e7e18f) by switching to mypy-baseline:

• .mypy-baseline.txt records the current 553 errors (line numbers normalized, so they survive unrelated line shifts).
• The pre-commit hook now runs mypy gget | mypy-baseline filter, which only fails on NEW errors not in the baseline. Existing errors are grandfathered.
• Works on pre-commit.ci (the baseline filters by error identity, not by changed files, so --all-files is fine).
• Config lives in [tool.mypy-baseline] in pyproject.toml; mypy-baseline was added to the test dependency group.

Verified: a clean tree passes (exit 0); a freshly introduced type error fails the hook (exit 1).

What this means going forward

• New/changed code must be type-clean — no new mypy errors can merge.
• The baseline can only shrink: when you fix some of the grandfathered errors, refresh it with mypy gget | mypy-baseline sync and the count goes down. mypy-baseline warns if a baselined error was resolved, nudging toward zero.
• This issue stays open as the tracker for paying the baseline down to 0 (the per-module cleanup plan above still applies; it's now a ratchet rather than a wall).

Current baseline: 553 errors across 24 files.