Thanks for sharing this! The context-based approach is interesting. Maintaining state from past successful parses to resolve ambiguity is essentially what we do too, though we batch it (scan first 50 values) rather than doing it incrementally. The incremental approach has the advantage of adapting mid-column if formats shift, which we've seen in datasets that were manually concatenated from different sources.

Will take a look at the repo.

That's the ideal, but unfortunately not always an option when you're on the receiving end. We're building a data cleaning tool, so the whole point is dealing with messy user-uploaded CSVs where we don't control the export format. If we could mandate ISO 8601 everywhere, life would be much simpler. But the reality is people copy-paste from Excel, export from legacy systems, or hand-edit CSVs, and we need to handle what shows up.