Extract KaTeX sources before Pandoc and render placeholders in plugin#1585
Extract KaTeX sources before Pandoc and render placeholders in plugin#1585davidmerfield wants to merge 1 commit into
Conversation
|
Bugbot is not enabled for your account, so this pull request was not reviewed. Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c70a7dff10
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| // indentation in text! More discussion of this issue: | ||
| // https://github.com/jgm/pandoc/issues/1841 | ||
| debug("Pre-katex-source", text); | ||
| text = safely(extractMathSources, text); |
There was a problem hiding this comment.
When a markdown/txt entry contains $$...$$ and the blog has plugins.katex.enabled = false, this unconditional conversion still rewrites the source to .blot-katex-source HTML before the plugin stage decides to skip KaTeX. The build pipeline runs converters before Plugins.convert, and the existing disabled-plugin case in app/build/tests/plugins.js:125-135 expects the literal $$a+b$$ to remain; with this change the final HTML leaks <span class="blot-katex-source" ...> instead. Gate this extraction on the plugin being enabled or restore the original delimiters when the plugin is skipped.
Useful? React with 👍 / 👎.
| let position = 0; | ||
|
|
||
| while (position < input.length) { | ||
| const start = input.indexOf(delimiter, position); |
There was a problem hiding this comment.
For markdown inputs that document TeX literally, such as raw <code>$$a+b$$</code>/<pre>$$a+b$$</pre> or fenced/backtick code, this raw string scan replaces the delimiters before Pandoc has identified code regions. That bypasses the renderer's SKIP_TAGS behavior and the existing KaTeX fixtures (skip-code.txt, skip-pre.txt) will either render math inside code/pre or show placeholder HTML in code blocks instead of the literal $$...$$. The extractor needs to be Markdown/HTML-context aware or run only on eligible text nodes.
Useful? React with 👍 / 👎.
| const display = isDisplayMath(input, start, sourceStart, afterEnd); | ||
|
|
||
| output += input.slice(position, start); | ||
| output += placeholder(source, display); |
There was a problem hiding this comment.
When a markdown entry contains literal dollar pairs inside an HTML attribute, such as <img alt="price $$x$$"> or a link/image title, this replacement inserts a quoted <span class="blot-katex-source" ...> fragment into the attribute before Pandoc parses the HTML. The previous renderer only visited parsed text nodes, so attributes stayed literal; this now terminates the attribute value early and can leave malformed HTML or stray placeholder nodes in the output. Limit extraction to real text content rather than scanning attribute strings.
Useful? React with 👍 / 👎.
| const escaped = escapeHtml(source); | ||
|
|
||
| if (display) { | ||
| return '<pre class="blot-katex-source" data-display="true">' + escaped + "</pre>"; |
There was a problem hiding this comment.
For a normal standalone display equation such as $$\nE=mc^2\n$$, this emits a block-level <pre> before Pandoc runs, so Pandoc preserves it as raw HTML instead of producing the paragraph that the existing KaTeX display fixture expects (<p><span class="katex-display">...). When the KaTeX plugin later replaces the <pre>, the result is a top-level display span and surrounding Markdown paragraph/list structure can change. Use an inline-safe placeholder or rewrap display output to keep the converter's prior HTML shape.
Useful? React with 👍 / 👎.
1 participant
Motivation
$$...$$with Pandoc-safe HTML placeholders containing only the raw TeX (no$$), so the KaTeX plugin can render them reliably after conversion.$$...$$after Pandoc and keep the existing-tex_math_dollarsPandoc extension to prevent Pandoc from interfering with math handling.Description
extractMathSources(input)inapp/build/plugins/katex/source.jsthat finds balanced$$...$$, leaves unmatched$$untouched, stores only the TeX source (no delimiters), HTML-escapes&,<, and>, and emits inline placeholders<span class="blot-katex-source" data-display="false">...</span>or display placeholders<pre class="blot-katex-source" data-display="true">...</pre>depending on block shape.app/build/converters/markdown/convert.jsso extraction runs before the indentation pass and before writing topandoc.stdinviatext = safely(extractMathSources, text);.app/build/plugins/katex/index.jsto render preserved placeholders directly with a newrenderSourcePlaceholders($)step that replaces.blot-katex-sourceelements withkatex.renderToStringoutput based on the element body anddata-displayattribute, rather than restoring$$-delimited text.-tex_math_dollarsPandoc extension active so Pandoc will not produce its own math output.Testing
extractMathSourcesbehavior for inline, display, and unmatched cases; these assertions passed (node - <<'NODE' ... NODE).node -conapp/build/plugins/katex/source.js,app/build/converters/markdown/convert.js, andapp/build/plugins/katex/index.js, all of which succeeded../scripts/tests/invoke.sh app/build/plugins/katex) but it could not complete in this environment because Docker is not available (test runner requires Docker).Codex Task