model-writtenrun 2420 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Alexander Pyhalov proposed postgres_fdw semi-join pushdown in #1, representing semi-joins as EXISTS conditions propagated to an enclosing WHERE clause and rejecting target lists that reference inaccessible inner-relation columns.
- Ashutosh Bapat questioned whether a date expression was safe to ship in #2; Pyhalov explained in #3 that the expression used an immutable date conversion and an already-typed constant, so no change was needed.
- Yuuki Fujii tested and stepped through the rebased patch, asking about target-list restrictions, anti-join support, naming, and SQL spacing in #6; Pyhalov expanded the explanations and fixed spacing in #7, while leaving anti-join support unexplored.
- Tomas Vondra's review in #9 led to clearer additional_conds and hidden_subquery_rels names, improved comments, and simpler control flow in #10.
- After a gap of roughly nine months, Alexander Korotkov resumed substantive review in #11, proposing semi-join-specific target checking, clearer documentation, list-based condition handling, and tests for unsupported cases.
- Pyhalov implemented the target-check and test improvements in #12, explaining that inner-column equivalences and previously rejected join orders still limited pushdown; Korotkov accepted limiting the feature's scope rather than making the patch substantially more complicated.
- Korotkov converted additional-condition handling from StringInfo buffers to lists of strings in #13; Pyhalov accepted the implementation and documented its propagation rules in #14, after which Korotkov supplied final editorial cleanup in #15.
- The feature landed as 824dbea3e4 from #15, with only the semijoin_target_ok() comment correction requested by Pyhalov in #16.
- Pavel Luzanov reported missing optimization and an apparent crash in #17; Korotkov explained unsupported anti-joins and costing behavior in #19, and Luzanov withdrew the crash report after rebuilding successfully in #20.
- Robins Tharakan reopened the correctness discussion with bug #18852 in #21, identifying an unexpected error after the semi-join feature landed.
- Pyhalov diagnosed restrictions escaping subqueries beneath LEFT/RIGHT joins and supplied hidden_subquery_rels guards in #23; after Korotkov questioned broader regression risk in #24, Pyhalov explained in #25 why joins without underlying semi-joins were unaffected.
- Korotkov proposed narrower direct JOIN_SEMI checks and a more informative regression test in #26, questioning whether the original guards unnecessarily restricted higher joins.
- Pyhalov supplied a nested-join counterexample in #27 and a complete reproducer patch in #29, explaining that subquery creation and restriction extraction must use matching criteria even when the immediate child is not itself a semi-join.
- The bugfix landed as 023fb51275, combining the revised tests from #26 with the original broader guards from #23; the extra failing example in #29 was omitted, and the fix was backpatched to PostgreSQL 17.
- Tender Wang requested attention to a separate target-check simplification in #30; the thread ends with Korotkov confirming the bugfix in #31 and leaving future work to prevent lower semi-joins from unnecessarily limiting optimization higher in the join tree.