[Docusaurus] html-math-method: webtex config ignored - math rendered as tex_math_dollars

[cderv]

The html-math-method: webtex configuration in _extension.yml is not being applied. Math equations are output as $$..$$ (tex_math_dollars) instead of webtex image URLs.

Expected behavior

Per documentation:

"By default, Quarto renders LaTeX math within Docusaurus projects using WebTeX"

Expected output:

![](https://latex.codecogs.com/svg.latex?e%20%3D%20mc%5E2)

Actual behavior

Output:

$$e = mc^2$$

Root cause

PR #7996 changed the docusaurus architecture:

Before:

writer: docusaurus_writer.lua  # Custom writer honored all options

After:

quarto-custom-format: docusaurus
filters:
  - at: post-finalize
    path: docusaurus_citeproc.lua

The docusaurus_citeproc.lua filter calls pandoc.write() with a hardcoded format string and only passes columns in options (added in PR #8365):

local cli_opts = { columns = PANDOC_WRITER_OPTIONS.columns }
local result_str = pandoc.write(result, 'markdown_strict+...+tex_math_dollars+...', cli_opts)

The html-math-method and other writer options from _extension.yml are not forwarded.

Evidence

The test snapshot at tests/docs/smoke-all/2024/01/18/docusaurus/equations.mdx.snapshot (created Jan 2024) contains $$ syntax, not webtex URLs - indicating this was already broken when the test was created.

Related

• PR restore citation behavior to Docusaurus projects #7996 - Architectural change that removed custom writer
• PR pass writer options correctly for docusaurus-md #8365 - Partial fix for columns only
• Issue [Docusaurus] Table having many columns output from knitr with df-print: kable seems broken (Quarto 1.4) #8332 - Table column wrapping (led to PR pass writer options correctly for docusaurus-md #8365)

Files

• src/resources/extensions/quarto/docusaurus/_extension.yml (line 31: html-math-method: webtex)
• src/resources/extensions/quarto/docusaurus/docusaurus_citeproc.lua (line 13-14: pandoc.write call)

---

This was uncovered while trying to understand snapshot change at #13249

And probably math is not the only problem we missed.