fix: Markdown link parsing

[gibson042]

Fixes #2959

Improves alignment with CommonMark, although does not include support for unescaped nesting.

Can be reviewed commit-by-commit.

[Gerrit0]

Thank you!

[Gerrit0]

This still isn't quite right... I was looking at reducing the number of test cases, picking key things to check as it slowed down the tests by about 15% on my machine to generate a ludicrous number of checks (3.5 -> 4 seconds) and found these cases, which are incorrectly captured as links:

[

](./no)

[

`code`](./no)

[

text](./no)

[gibson042]

This still isn't quite right... I was looking at reducing the number of test cases, picking key things to check as it slowed down the tests by about 15% on my machine to generate a ludicrous number of checks (3.5 -> 4 seconds) and found these cases, which are incorrectly captured as links:

[

](./no)

[

`code`](./no)

[

text](./no)

Can you elaborate? Handling of those inputs looks correct to me:

code

$ npx tsx -e '
  import { lexCommentString } from "./src/lib/converter/comments/rawLexer.ts";
  import { lexBlockComment } from "./src/lib/converter/comments/blockLexer.ts";
  import { parseCommentString, parseComment } from "./src/lib/converter/comments/parser.ts";
  import { FileRegistry } from "./src/lib/models/FileRegistry.ts";
  import { TestLogger } from "./src/test/TestLogger.ts";
  import { MinimalSourceFile } from "#utils";

  const inputs = ["[\n\n](./no)", "[\n\n`code`](./no)", "[\n\ntext](./no)"];
  inputs.push(inputs.join("\n\n"));

  const makeParse = (lex, parse) => {
    const config = {
      blockTags: new Set("@param @remarks @module @inheritDoc @defaultValue".split(" ")),
      inlineTags: new Set(["@link"]),
      modifierTags: new Set("@public @private @protected @readonly @enum @event @packageDocumentation".split(" ")),
      jsDocCompatibility: { defaultTag: true, exampleTag: true, ignoreUnescapedBraces: false, inheritDocTag: false },
      suppressCommentWarningsInDeclarationFiles: false,
      useTsLinkResolution: false,
      commentStyle: "jsdoc",
    };
    return text => {
      const files = new FileRegistry();
      const logger = new TestLogger();
      const content = lex(text);
      const sourceFile = new MinimalSourceFile(text, "/dev/zero");
      const result = parse(content, config, sourceFile, logger, files);
      logger.expectNoOtherMessages();
      return result;
    };
  };
  const parseRaw = makeParse(lexCommentString, parseCommentString);
  const parseBlockComment = makeParse(lexBlockComment, parseComment);
  const embedInComment = input => {
    const lines = input.split("\n");
    const embedded = `/**\n${lines.map(line => " * " + line).join("\n")}\n */`;
    return embedded;
  };
  for (const rawInput of inputs) {
    console.log("\n\n==== raw input");
    console.log(rawInput);
    console.log("==== raw parse");
    console.log(parseRaw(rawInput));
    const commentInput = embedInComment(rawInput);
    console.log("\n==== comment input");
    console.log(commentInput);
    console.log("==== comment parse");
    console.log(parseBlockComment(commentInput));
  }
'

result

==== raw input
[

](./no)
==== raw parse
{
  content: [ { kind: 'text', text: '[\n\n](./no)' } ],
  frontmatter: {}
}

==== comment input
/**
 * [
 * 
 * ](./no)
 */
==== comment parse
Comment {
  summary: [ { kind: 'text', text: '[\n\n](./no)' } ],
  blockTags: [],
  modifierTags: Set(0) {},
  label: undefined
}


==== raw input
[

`code`](./no)
==== raw parse
{
  content: [
    { kind: 'text', text: '[\n\n' },
    { kind: 'code', text: '`code`' },
    { kind: 'text', text: '](./no)' }
  ],
  frontmatter: {}
}

==== comment input
/**
 * [
 * 
 * `code`](./no)
 */
==== comment parse
Comment {
  summary: [
    { kind: 'text', text: '[\n\n' },
    { kind: 'code', text: '`code`' },
    { kind: 'text', text: '](./no)' }
  ],
  blockTags: [],
  modifierTags: Set(0) {},
  label: undefined
}


==== raw input
[

text](./no)
==== raw parse
{
  content: [ { kind: 'text', text: '[\n\ntext](./no)' } ],
  frontmatter: {}
}

==== comment input
/**
 * [
 * 
 * text](./no)
 */
==== comment parse
Comment {
  summary: [ { kind: 'text', text: '[\n\ntext](./no)' } ],
  blockTags: [],
  modifierTags: Set(0) {},
  label: undefined
}


==== raw input
[

](./no)

[

`code`](./no)

[

text](./no)
==== raw parse
{
  content: [
    { kind: 'text', text: '[\n\n](./no)\n\n[\n\n' },
    { kind: 'code', text: '`code`' },
    { kind: 'text', text: '](./no)\n\n[\n\ntext](./no)' }
  ],
  frontmatter: {}
}

==== comment input
/**
 * [
 * 
 * ](./no)
 * 
 * [
 * 
 * `code`](./no)
 * 
 * [
 * 
 * text](./no)
 */
==== comment parse
Comment {
  summary: [
    { kind: 'text', text: '[\n\n](./no)\n\n[\n\n' },
    { kind: 'code', text: '`code`' },
    { kind: 'text', text: '](./no)\n\n[\n\ntext](./no)' }
  ],
  blockTags: [],
  modifierTags: Set(0) {},
  label: undefined
}

[Gerrit0]

Sorry about that! I had apparently added a test case with just one newline, but didn't have different links, so I didn't realize it was that link that cased it.