Large documentation tables can make the current HTML-to-Markdown conversion path CPU-bound. We reproduced this with WezTerm's nerdfonts.html?q= page: the page contains a single table with roughly 10,752 rows and 21,504 cells. Plain Turndown and non-GFM conversion complete quickly, but @joplin/turndown-plugin-gfm@1.0.67 does not finish within two minutes for that table.
This issue tracks a broader table-handling improvement separate from the immediate hang mitigation.
Possible direction:
- Add a DOM preprocessing/cleanup step specifically for tables before markdown conversion.
- Remove unnecessary styling/formatting wrappers and layout-only divs while retaining inline semantic formatting where useful.
- Decide per table whether Markdown or minimized HTML is the better representation.
- Prefer minimized HTML tables when preserving table structure, background colors, column/row styling, code examples, or other complex cell content matters.
- Prefer Markdown tables for smaller/simple tabular content where GFM output is compact and reliable.
- Consider row-based splitting or chunking for very large tables so full content can be retained without forcing a single huge GFM table conversion.
The immediate production fix should remain smaller: detect oversized tables and avoid the pathological Joplin GFM conversion path.
Large documentation tables can make the current HTML-to-Markdown conversion path CPU-bound. We reproduced this with WezTerm's
nerdfonts.html?q=page: the page contains a single table with roughly 10,752 rows and 21,504 cells. Plain Turndown and non-GFM conversion complete quickly, but@joplin/turndown-plugin-gfm@1.0.67does not finish within two minutes for that table.This issue tracks a broader table-handling improvement separate from the immediate hang mitigation.
Possible direction:
The immediate production fix should remain smaller: detect oversized tables and avoid the pathological Joplin GFM conversion path.