Module extract 

Read-only reference extraction for linting. Walks a document body and reports every wikilink, markdown link/image, heading, fenced code block and directive with (approximate, see below) 1-based source positions — without touching the render/build path.

Positions: comrak populates sourcepos on AST nodes during parsing, so lines come straight from the AST. Two approximations apply:

• Wikilink col is the start column of the folded Text/HtmlInline run that contains the [[…]], advanced by the character offset of the link within that run. Markdown escapes or entity references consumed by the parser can make this drift a little within the line; the line is exact.
• Lines inside a block directive’s body are the directive’s opening line plus the 1-based line within inner_md — exact for the common case, but nested rewrites deeper down inherit the same additive scheme, so treat inner positions as best-effort.

Directive extraction runs first (via directivepass::extract), so everything the render pipeline would treat as directive content is reported under its directive, and sentinel-substituted lines are re-padded so the content after a block directive keeps its original line numbers.

Structs

DirectiveRef

One directive occurrence (block or leaf).

DocRefs

Every reference found in one document body.

FenceRef

One fenced code block (indented code blocks are not reported).

HeadingRef

One heading, any level h1–h6.

MdLinkRef

One markdown link or image, url as written.

WikilinkRef

One [[target#anchor|label]] wikilink occurrence.

Functions

extract_refs

Extract all references from a markdown body (frontmatter already stripped). Read-only: shares parsing helpers with the render path but never mutates render behavior. Directive bodies are recursed into; their references are reported with lines offset by the directive’s opening line.