Documentation is part of the code
This site is part of Verbora's public interface, and the repository treats it that way. This page is what that means for you as a reader.
What is verified before a page ships
| Check | What it guarantees |
|---|---|
check-snippets.py | Every Rust snippet here is extracted into a real crate, compiled against the actual verbora-* crates and run. A wrong assert_eq! on this site fails the build exactly like a wrong test. |
vitepress build | Every page renders and every sidebar entry has a page behind it. |
check-links.py | Every internal link and #anchor resolves — including inside raw HTML — and no page is stranded outside the navigation. |
All three run together:
cd site && npm run checkRustdoc is held to the same standard from the Rust side: public examples run as doctests under cargo test --workspace.
What you can rely on when reading
Every API named here exists. Where something does not exist — a parallel API, a scratch buffer, a stemmer for a given language — the page says so plainly instead of describing a plausible design.
Every published number names its benchmark, the machine it ran on and the command that repeats it. Guidance derived from the shape of the code rather than from a measurement is labelled as such, and never presented as a measurement.
Wherever there is more than one API for the same problem, the page says what each variant is for, what it costs, and which to reach for by default. See Choosing the right API and Allocation behaviour.
Found something wrong?
A documentation error is a bug, and reporting one is as useful as reporting a crash — an example that no longer works costs a reader more time than no example at all. Issues and corrections go to the repository.