[tae898]

Documentation Validation fails on main (the latest runs and every open PR, e.g. #511) with three broken cross-references and two anchor names outside the lowercase-with-hyphens convention. This fixes all five; python docs-validator.py passes locally.

• how-to/connectivity/postgres.adoc: <<Column type mapping,...>> pointed at a section title, which the validator does not resolve; the section gets [[postgres-column-type-mapping]] and the reference uses it.
• reference/settings.adoc: <<_a_statement_that_fails_halfway>> relied on the generated section id; the section in concepts/transactions.adoc gets [[statement-fails-halfway]].
• reference/sql/chapter.adoc: <<vector-neighbours,...>>, the anchor is spelled vector-neighbors.
• reference/sql/sql-functions.adoc: [[cchShortestPath]] and [[duanSSSP]] become [[cch-shortest-path]] and [[duan-sssp]], like the other function anchors ([[both-v]]); the three references to the first are updated (graph-olap.adoc, path-finding.adoc, sql-functions.adoc). Links from outside the docs to #cchShortestPath or #duanSSSP would need the new names.

🤖 Generated with Claude Code

[mergify]

This pull request does not currently match the merge queue conditions, so it cannot be queued from here. The box comes back if it matches again.

[tae898]

With this change python docs-validator.py passes in CI too ("Documentation passed all validation checks"). The job is still red at its next step, the HTML check of the generated site, which main never reached because the validator failed first: 21 errors, all in the generated index.html.

count	error
15	No "code" element in scope but a "code" end tag seen.
2	Stray end tag "mark".
1	The string following "&" was interpreted as a character reference. ("&" probably should have been escaped as "&".)
1	Bad value "file://exports/" for attribute "href" on element "a": Illegal character in path segment: "<" is not allowed.
1	Named character reference was not terminated by a semicolon. (Or "&" should have been escaped as "&".)
1	Stray end tag "sub".

They come from markup inside the sources (a #, ~ or & read as formatting or an entity inside monospace text, an <file-name> placeholder in a link) and need a site build to map back to their lines, so I left them for a separate PR to keep this one to the validator.

[tae898]

CI is green now. The "Validate HTML output" step (html5validator on the single-page build of src/main/asciidoc/index.adoc) failed with 21 errors that main has too; main never got that far because the validator step failed first. The cause was in the sources: Asciidoctor applies strong, emphasis, mark and subscript before monospace, so a *, _, # or ~ in two different backtick spans paired up across them (for example `wal.group.*` and `commit.lock.*` rendered as <code>wal.group.<strong></code> and <code>commit.lock.</strong></code>). The same markup also rendered the live pages wrong (bold running across code spans). The affected spans, in 23 files, are now literal monospace (`+...+`, the form sql-projections.adoc already used), plus an anchor renamed and two spans in redis.adoc that main gained since this PR was opened. No check was loosened or skipped. Ready for review whenever you have a moment; happy to adjust.

[lvca]

@claude review this

[lvca]

Super