Excel goes in-house
.xlsx, .xlsb and .xls are now read by our own parser - about 4,400 lines of new reader code - and calamine is gone, taking five transitive crates with it.
What that buys:
- All three containers share one cell model and one number-format renderer, so dates, currencies and rounding come out the same whichever way a workbook was saved.
- Workbook bugs are now fixable in this repo instead of waiting on an upstream release.
- Six fewer crates to build and audit.
Smarter Markdown escaping
Backslashes now only appear where emphasis could actually form. a *b 2 * 3 comes out as written, not a \*b 2 \* 3.
Fixes
|inside a code span no longer splits table rows- docx no longer fuses words together at run boundaries
- epubs with a repeated spine part no longer duplicate the chapter
Full Changelog: https://github.com/firecrawl/anydoc/compare/v0.1.9...v0.2.0
