[ParamTanna]

Description

LinkInlineProcessor.getLink scans forward from the ( of a link target to pair the brackets. The treeprocessor calls it for every [text]( in the text, and when the parentheses never close each call scans to the end, so conversion time grows with the square of the number of unclosed (. Timings for markdown.markdown(s * n) on CPython 3.12:

'[a]('    x4000  (16000 bytes): 2.27s -> 0.015s
'[a](x '  x4000  (24000 bytes): 3.56s -> 0.018s
'[a](<u>' x4000  (28000 bytes): 4.11s -> 0.064s

This is the getLink companion to #1664 (which fixed the same problem in getText and noted that getLink was left out because its quote/title handling is more involved).

The quadratic inputs all turn out to be the ones with no quote before the matching ) — there the scan is just a plain bracket match. Those are now resolved from a table of closing positions built in a single pass per text (_findParenEnds, same monotonic-stack idea as #1664's getText), guarded by a per-text "next quote" table so the fast path only fires when no quote can start a title. Inputs where a quote may begin a title fall back to the original character-by-character loop unchanged; I measured those and they were already linear, so nothing is lost.

To be sure the fast path is behaviour-preserving, I compared the new getLink against the old loop on the raw (href, title, index, handled) tuple at every index of:

• all 1,111,111 strings up to length 6 over the ten characters [ ] ( ) ' " < > a and space, and
• 2,000,000 random strings of length 1–60 over a wider alphabet.

The results were identical in every case. The existing test suite passes unchanged (1166 tests), plus two new regression tests. test_many_unclosed_link_parens takes several seconds without the fix and finishes instantly with it.

AI Assistance Disclosure

• No AI tools were used in preparing this PR.

• AI tools were used and I have disclosed which ones, and have fully reviewed and verified their output.

  AI tools were used to assist with the following:

  • Code
  • Code Comments
  • Documentation
  • Changelog Entry
  • Commit Message
  • The Description Above
  • Other: the regression tests, the equivalence/fuzz comparison harness, and the timing script used to find the slow inputs

  Tool used: Claude Code (Claude Opus 4.8). I reviewed and verified all of it before submitting, including running the equivalence comparison and the full test suite myself.

Checklist

• This PR follows the contribution guidelines.
• The code follows the Code Style Guide.
• The commit message follows the Commit Message Style Guide.
• I have added or updated relevant docs, including release notes if applicable which follow the Documentation Style Guide.
• I have added or updated relevant tests.
• I have not requested, and will not request, an automated AI review for this PR.