potiuk commented on PR #71550: URL: https://github.com/apache/airflow/pull/71550#issuecomment-5341812423
Thanks for actually running it — that's the useful kind of review. I think the snippet may be the wrong buffer though. I diffed it against the file as it stands before this PR and it comes out byte-identical apart from a trailing newline — no `ThreadPoolExecutor`, no `iter_pypi_data`, no request timeout, and no `OSError`/`JSONDecodeError` handling. Applying it would revert the change rather than reorder it, so I've left it aside for now. On the ordering itself: `executor.map` yields in submission order and the consumer zips it against the package list in constraints-file order, so the table should still come out in file order. I couldn't reproduce unsorted output here. Your prompt did surface something real, though. Chasing it I realised that if results ever *did* come back out of order, the table wouldn't visibly reshuffle — rows are labelled from the package list, so each row would silently show another package's versions. My first attempt at a test for this passed even with ordering deliberately broken. I've rewritten it to assert the package-to-version pairing and confirmed it fails under `as_completed` and passes as written, so that property is now pinned down. If you did see genuinely wrong output, could you share the command and a snippet? That would be a different bug and worth chasing separately. --- Drafted-by: Claude Code (Opus 5); reviewed by @potiuk before posting -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
