Commits on Jun 11, 2026
-
Complete type hints, refresh docstrings and README, rebuild docs" -m …
…"Fully annotate the public API now that py.typed ships: case() -> bool (normalized to always return a bool), __enter__/__exit__, result, and closed_range, using 3.9-safe modern annotation syntax. Rewrite the docstrings that render as the API reference, fixing the context-manager description, default() ordering caveat, case() return contract, fall-through result semantics, and closed_range wording. README: version + docs badges with a link to the hosted docs, uv install option, corrected feature list with s.result, new fall-through/results and 'Why not just match?' sections, f-strings throughout. Tooling: ruff target-version py39 to match requires-python; add PyPI keywords and mark the project production/stable. Rebuild the docs site. Closes #13
-
Fix closed_range step overshoot and result sentinel comparison
closed_range() built its inclusive bound as stop + step, so any step greater than 1 overshot: closed_range(1, 6, 2) produced 1, 3, 5, 7 and values outside the stated range could match a case. Build the range with stop + 1 instead, and reject step < 1 with a clear ValueError. First tests with a non-default step. .result compared its no-result sentinel with == instead of is, so a matched case returning an always-equal object (numpy arrays are the real-world case) raised "No result has been computed". Compare by identity, with an assertIs regression test. Rebuild the docs for the updated closed_range reference; every page also picks up the just-added Umami analytics snippet. Ignore uv.lock. Closes #14
-
Add behavior tests pinning documented guards and semantics
Seven new tests (29 total) for documented-but-untested behavior: the closed_range start/stop guard, adjacent closed ranges raising duplicate-case, exception propagation from the block body and from case actions, the default-before-case ordering quirk, non-callable case actions, and None as a legitimate result. Takes switchlang to 100% line coverage.
