Having more than one table of data in a single spreadsheet does happen “in the wild”. Calthorpe often send me Excel spreadsheets like this. I’m not certain but I don’t think the CSV format can handle more than one table per file, so a real-world editor would need to accomodate more complicated formats.

A bonus of this is that the more complicated formats would, I assume, contain the structural metadata required to identify individual tables. They might go some way towards identifying table headers, too.

The file processor could apply some heuristic tests to try and deduce the structure of the table and could be quite successful in many cases. I think things like the tide table example, are always going to be tough to sort out, but this could ease the pain somewhat.

Providing previews and adjustments at the converting stage would be complicated to implement but should be quite possible.