Hi,

I wrote another “interesting” filter base on the same idea, but in reverse: pantable/pantable2csv.py at master · ickc/pantable. It converts all tables in pandoc to a YAML-CodeBlock-styled CSV table defined in pantable.

Effectively, it adds a “CSV Writer”, where pantable is kind of “CSV Reader”.

I can kind of achieve idempotence here, but only at $P^3 = P^2$ , not $P^2 = P$ (it’s from pandoc though). Basically it captures all info from pandoc’s AST, so the conversion pantable2csv did should be “lossless”.

These are at least important to me because I can safely jump between the 2 formats (native pandoc table and csv table in code-block) without worrying too much. I could jump to csv for edit (more low-level, e.g. width control) and jump back to pandoc tables for better readability.