dcdpr / jp / 25955687064

67%

Build Type

push

github

Committed by web-flow

Commit Message

chore(tools): Fix rustdoc page parsing in web fetch tool (#643)

The HTML section extractor had two failure modes when used against
rustdoc-generated pages (docs.rs or local `cargo doc` output).

First, the section listing was polluted with page-chrome IDs
(`rustdoc_body_wrapper`, `rustdoc-toc`, `rustdoc-modnav`,
`main-content`). These were picked up by the ancestor-id fallback in
`resolve_heading_id` because rustdoc wraps the whole page in elements
carrying those IDs. A new `RUSTDOC_SCAFFOLDING_IDS` filter drops them
from the listing while still allowing explicit `sections=[...]` requests
to target `main-content` on non-rustdoc pages.

Second, method and variant documentation was never returned. Rustdoc
places the signature heading inside `<section id="method.X">` and puts
the prose in a sibling `<div class="docblock">` outside the section —
optionally further wrapped in a `<details>` toggle. The extractor was
walking the section's own children, finding nothing, and returning an
empty string. Two new helpers, `extract_rustdoc_section` and
`preview_from_section_docblock`, resolve the sibling docblock in both
the flat and toggle layouts. `extract_section_html_from_doc` now
short-circuits into this path for any `<section>` target, and the
preview walk gains a matching fallback so listing previews agree with
extracted bodies.

Signed-off-by: Jean Mertz <git@jeanmertz.com>

Coverage Stats

58 of 61 new or added lines in 1 file covered. (95.08%)

1 existing line in 1 file now uncovered.

26310 of 39405 relevant lines covered (66.77%)

192.27 hits per line