Module code 

Code fences and inline code spans: finding them, and blanking them so a lexer never harvests out of them.

One lexer for every caller. This crate’s link scanners, a #tag lexer and any tool that rewrites markdown all ask “is this code”, and a second copy would be a second answer.

CodeRules picks between two rule sets.

CodeRules::CommonMark, the default, follows CommonMark:

• A fence opens on a line of up to three spaces and a run of three or more backticks or tildes. A backtick fence’s info string may not hold a backtick. The fence closes on a line holding only a run of the same character, at least as long, and whitespace. An unclosed fence runs to the end of the input.
• An inline span opens on a run of backticks and closes on the next run of exactly that many. A run with no close is literal text.
• A span stays inside its block. A blank line, a fence, an ATX heading, a setext underline, a thematic break, a block quote or a list item ends the block. A span opened on an ATX heading stays on that line.

CodeRules::SkillExport keeps the looser rules of some agent-skill renderers, for tools that must match them:

• A backtick fence’s info string may hold backticks.
• Any fence marker line of the same character and at least the same length closes a fence, whatever follows the run.
• A longer backtick run closes a shorter span.
• Only a blank line or a fence marker line ends a span’s block.

Both rule sets read one line at a time, so neither knows about container blocks. Indented code blocks are not code. A fence is found only when its marker starts within three spaces of the line start, so a fence nested in a list item or a block quote is missed. A list item or block quote marker ends a span’s block at any indent and with any start number, even with no text after it, so it may end a span that CommonMark would continue. Nested items are indented, and ending a span early keeps text visible.

CodeRules::CommonMark also treats a line starting with | after up to three spaces as a one-line block, table or not. A span cannot enter or leave that line, but a span in it may cross a cell. GFM reads a pipe line without a delimiter row as paragraph text, and splits a table row into cells before it reads spans.

Every marker is ASCII, so a byte scan never splits a UTF-8 character.

use markdown_link::code::{CodeRules, strip_code_with};

let text = "keep [[a]] `[[b]]`";
let blanked = strip_code_with(text, CodeRules::CommonMark);
assert_eq!(blanked, "keep [[a]]        ");

Structs

Fences

The fence state of a line-by-line scan: which fence, if any, is open.

Spans

Finds where inline code spans close, for a scan that reads its input once, left to right.

Enums

CodeRules

Which rules the lexer follows. See the module doc.

FenceLine

Where a line sits relative to fenced code, as Fences::line reports it.

Functions

strip_code

Blank out fenced and inline code so links and tags inside it are not harvested. strip_code_with under CodeRules::CommonMark.

strip_code_with

Blank out fenced and inline code under rules.