model-writtenrun 8988 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
Feb - Mar 2018 · 2 entries
- Nikolay Shaplov proposed replacing StdRdOptions with separate structures for each relation kind and renaming option-access macros in #1, seeking smaller allocations and a consistent foundation for broader options refactoring.
- Andres Freund questioned the memory-saving justification in #2; Nikolay explained the intended reusable options framework in #3, and Andres urged keeping related work together for review in #4.
Sep - Oct 2018 · 1 entry
- After several months, Nikolay rebased the patch and incorporated the new B-tree cleanup option in #6, then refreshed it again in #8.
Nov 2018 - Jan 2019 · 3 entries
- Dmitry Dolgov reported an MSVC build failure in #9, and Nikolay removed the empty partitioned-options array and confirmed a successful Windows build in #10.
- Álvaro Herrera opposed unnecessary public API renaming because extensions could depend on the existing names in #11 and #13; Nikolay restored those names and added assertions in #16, exposing a parallel_workers relation-kind issue that he also raised separately.
- Álvaro rejected new IsHeapRelation and IsViewRelation APIs in #17, leading Nikolay to use direct relkind checks and relax the parallel_workers handling in #18.
Mar 2019 · 1 entry
- Kyotaro Horiguchi and Aya Iwata questioned the new heap-versus-TOAST decisions at callers in #20 and #21; Nikolay maintained his separate-representation approach and rebased it in #22.
Jul - Aug 2019 · 2 entries
- Dent John encountered newly added vacuum options while trying the patch in #27; Nikolay incorporated them in #30, which Dent then tested successfully while criticizing duplicated option descriptions in #31.
- Tomas Vondra proposed a shared parsing wrapper and objected to leaking heap/TOAST distinctions into callers in #32, concerns Michael Paquier reinforced when returning the patch with feedback in #34.
Sep - Nov 2019 · 12 entries
- Nikolay narrowed the work to index AMs in #35 and corrected the leftover B-tree field in #36; he explained in #37 that partitioned-table and view-assertion work had separate threads and the heap/TOAST split could be postponed or omitted.
- Dent examined the remaining dependency on central option definitions in #38, while Amit Langote insisted that repeated parsing and allocation code should first become a shared function in #42.
- Amit supplied the independent BuildRelOptions() patch in #44, allowing the index refactor to build on a simpler API rather than duplicating more parsing code.
- Michael's review led Amit to privatize the old helpers, add an assertion, rename the new function to build_reloptions() and improve its documentation across #46 and #48.
- Testing dummy_index_am exposed the structure-size assertion as unreliable, so Michael replaced it with an option-count check in #52; Amit corrected the return-format documentation in #55.
- Michael committed the shared helper and caller conversions from #55 as 3534fa2233, but deliberately retained the public low-level helpers for possible extension uses, explaining that decision in #57.
- Amit questioned the remaining public macros in #58, Álvaro explained their historical purpose in #61, and Michael proposed removing them and privatizing the old helpers in #63 after finding no external uses.
- Nikolay resumed the index refactor in #66, adopting the newly committed build_reloptions() and retaining the narrower index-only scope.
- Michael committed the cleanup from #63 as 50d22de932, retaining GET_STRING_RELOPTION() as a useful accessor and explaining in #67 why he left further movement of type definitions out.
- Michael requested index-kind assertions in #68, explained how relam identifies the AM in #70, and supplied #71 with AM-specific assertions and page-space helper macros, also covering GIN and BRIN.
- Nikolay accepted the final revision while questioning assertions for extension AMs in #73; Michael explained why those checks were unnecessary for custom AMs in #75, and Amit caught the stale StdRdOptions comment in #74.
- Michael committed #71 with the comment correction, formatting and typedef-list additions as 4cb658af70 and confirmed completion in #77; the shared helper, API cleanup and index refactor had landed, while the broader heap/TOAST separation remained deferred.