model-writtenrun 9629 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Pavel Stehule proposed accepting explicit VARIADIC arrays in LEAST and GREATEST in #1; when Vik Fearing suggested making them ordinary functions in #2, Andrew Gierth and Pavel explained that their common-type coercion behavior exceeded ordinary polymorphic functions' capabilities in #3 and #4.
- Tom Lane objected to both the array syntax and the separate MinMaxExpr argument representation, preferring dedicated array functions in #5, while Pavel defended consistency with ordinary variadic calls in #6.
- After a gap, Chapman Flack found build failures caused by the changed function-call API, pointer warnings, and missing tests and documentation in #7; Pavel refreshed the patch in #8, and Chapman subsequently verified its behavior, including empty and null arrays, in #9.
- Pavel added documentation and regression tests in #10, Chapman supplied grammatical corrections in #14, and his testing report in #15 marked that version ready for a committer.
- Tom's review in #11 identified an ExprEvalStep size-constraint violation and criticized storing arguments in different fields; Pavel responded in #16 by using new operation constants and the existing args list, then consolidated the executor's comparison loops in #17.
- Chapman supplied another documentation correction in #19 and found function-name selection, error-reporting, and repeated comparison-test issues in #20; Pavel's revisions in #21, #23, and #25 addressed these through better diagnostics, precomputed flags, restored switch statements, and corrected prose.
- During that review, Chapman argued in #22 that documentation could explain why these constructs only resemble functions, leaving array operations to ordinary functions; Pavel continued to favor the more uniform user interface in #23.
- Chapman confirmed that the final feature patch passed installcheck-world and addressed his suggestions in #26, but David Steele reported opposition from Tom and Andres and proposed rejection in #27.
- Andrew Dunstan supported rejecting the feature while documenting the limitation in #28, and Chapman proposed placing a note covering COALESCE, GREATEST, and LEAST at the top of the conditional-expressions section in #29.
- Andrew formally rejected the implementation and posted the independent documentation patch in #31; Tom reviewed its wording in #34, and Andrew committed the posted patch unchanged as 5e74a42785.
- Andrew acknowledged Tom's wording suggestion in #35, and David G. Johnston offered alternative prose after the commit in #36; the supplied commit still contains the original wording, the documentation change landed, and the VARIADIC feature was rejected without replacement array functions being submitted.