Inspired by this hacker news post which links to ModernCSV and terminal based visidata.

Problem: dealing with inconsistently formatted/not properly normalized data from sources you can't control. Eyeballing tedious and error prone. Surfacing problems is hard. Fixing is intensely manual.

What about a tool for dealing with "dirty" data, that displayed results in table. For each column, you could write a validator - that could include data from other columns. Regex? Lua? And for each column, you could write a transformer, that could likewise include source or transformed data from other columns. Add columns so you could could parse one column into many.

Everything would be live as you typed it.

Could be web-based; if so could use Wasm for validator/transformer parser.

Maybe you couldn't even edit individual cells.