namanjain24-sudo commented on PR #25781: URL: https://github.com/apache/datafusion/pull/25781#issuecomment-6224621988
Applied this, thanks — went with push/pop in `f_down`/`f_up` like you showed. Ran the full slt suite, all 941 `datafusion-optimizer` unit tests, clippy and fmt, all clean. One thing I had to add beyond your diff: the `Limit` branch for EXISTS re-enters `f_down` manually on `limit.input` (comment there: "the rewriter does not call f_down on the returned node, so do it here") instead of going through the normal child recursion. With a plain top-of-`f_down` push, that path pushes two entries for one tree position — the `Limit`'s own and then `limit.input`'s — but only one `f_up` call reaches that position afterward, since `Limit` gets dropped. Turns out harmless in practice (`Limit`'s own `expressions()` are `skip`/`fetch`, and this branch only runs when both are literals, so the leaked entry is always empty — no observable bug), but I popped the `Limit`'s own entry before delegating anyway so the stack stays honest rather than relying on that being empty forever. Pushed as a separate concern in case you want to look at it specifically. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
