model-writtenrun 8463 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
May 2018 · 3 entries
- Simon Muller proposed allowing HEADER with COPY's text format in #1, adding output support, documentation, and a regression test.
- Andrew Dunstan questioned the need given tab-delimited CSV support in #5, while David G. Johnston and Robert Haas supported producing the exact format users needed in #7 and #15.
- Isaac Morland proposed checking imported headers against destination column names in #8; Tom Lane supported this safeguard while resisting broader column-remapping functionality in #12, and David G. Johnston suggested HEADER true, false, or match in #13.
Jul - Aug 2018 · 3 entries
- Simon corrected file_fdw expectations in #16, and Cynthia Shang supplied test relocation and COPY FROM coverage suggestions in #18 that Simon incorporated in #20.
- Daniel Verite and Cynthia Shang challenged the changed error code in #22 and #26, leading Simon to preserve ERRCODE_FEATURE_NOT_SUPPORTED in #27.
- Michael Paquier requested stronger round-trip and file_fdw tests, a corrected header_line comment, and further consideration of header matching in #32; with that feedback unanswered, he returned the patch in #33.
Feb - Mar 2020 · 1 entry
- Rémi Lapeyre revived the work in February 2020 with #34, combining text-header support with an enum-based HEADER MATCH mode that checked input field counts and names.
Jul - Oct 2020 · 6 entries
- Peter Eisentraut requested separate patches, clearer commit explanations, and fuller matching documentation in #41, leading to the split submissions #43 and #44, subsequently bundled in #46.
- Vignesh C found hardcoded paths, a declaration-order warning, and indentation problems in #47, which Rémi corrected in #48.
- Daniel Verite reproduced an error-stack crash caused by the PG_TRY/PG_CATCH parser and identified documentation issues in #49; Vignesh supplied #50, replacing exception-based parsing with direct option recognition.
- Rémi discovered that repeated HEADER options could escape redundancy checking in #51, and Michael extended the diagnosis to FREEZE and recommended explicit specified flags in #52.
- Rémi included a duplicate-option fix in #53, but Michael landed his independently prepared patch from #459104 in thread #43013 as 10c5291cc2, with broader option tests, and explained in #54 that he had missed Rémi's attachment.
- Rémi dropped the now-resolved duplicate-option patch in #55 and kept the remaining features current through the COPY source split in #58 and a later rebase in #60.
Mar - Apr 2021 · 1 entry
- Zhihong Yu's review in #61 and #63 led to clearer field-count diagnostics and wording corrections in #62 and #64, followed by adaptation to the reorganized regression tests in #65.
Dec 2021 - Feb 2022 · 4 entries
- Peter landed the text-header portion of #65 as 43f33dc018, with light wording and test changes, and requested a real column-name mismatch test for the remaining feature in #66.
- Rémi supplied the remaining matching patch and missing test in #67, arguing that HEADER MATCH avoided the contradictory combinations a separate matching option would introduce.
- Peter identified ineffective file_fdw tests and questioned parser complexity and null-header handling in #68 and #69; Daniel defended accurate tri-state error reporting and agreed that null header values should fail in #71.
- Peter committed #67 as 072132f04e, retaining HEADER MATCH while refining names, diagnostics, null handling, and tests, and announced completion in #73.
Jun 2022 · 4 entries
- Julien Rouhaud reported in #74 that matching used relation attribute numbers to index input fields, causing incorrect validation or unsafe access with reordered or dropped columns, and also questioned accepting MATCH for COPY TO.
- After Andrew Dunstan supported rejecting MATCH on output in #75, Julien supplied #76 with corrected field indexing, an output restriction, documentation, and regression tests.
- Michael expanded dropped-column coverage and revised the error code in #78; discussion involving Julien, Peter, and Daniel in #79, #80, and #81 ended with agreement to retain consistency with existing COPY errors in #82.
- Michael landed the lightly corrected #78 as ca7a0d1d36 and announced it in #87; Julien's thanks in #88 closed the thread with both features and the identified follow-up fixes committed.