model-writtenrun 9867 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
Jan - Apr 2018 · 9 entries
- Daniel Verite proposed native psql CSV output in #1 so scripts, variables and metacommand results could use CSV without COPY's query-wrapping and single-line limitations.
- Pavel Stehule requested a convenient --csv option and challenged the inherited pipe separator in #3, while Daniel explained the difficulties of automatically changing shared formatting settings in #5.
- Fabien COELHO built and tested the initial patch, found missing quoting in headers and expanded output, and requested regression tests in #12; Daniel fixed those bugs in #13 and added tests with an independent fieldsep_csv setting in #23.
- David Fetter questioned a dedicated command-line shortcut and automatic header choices in #19 and #21, while Fabien's detailed review in #31 requested alphabetical format ordering, broader tests and reconsideration of shared separators.
- Daniel's revision in #46 reused fieldsep and recordsep, made --csv select comma-separated output, dropped the short -C option and added array, JSON and NULL test cases.
- Pavel demonstrated that changing argument order could silently replace an explicitly requested separator in #47, then supplied a competing implementation with format-specific defaults in #49.
- Pavel developed that alternative through #67, adding default restoration and then a reset command, while Fabien's review in #61 exposed hidden-state problems and Daniel's review in #66 found erroneous parameter reporting.
- The separator disagreement prevented agreement for PostgreSQL 11: Fabien sought a compromise in #70, Pavel rejected the shared-separator version in #73, and David G. Johnston advocated an independent CSV setting in #78.
- Peter Eisentraut proposed treating CSV as unaligned output with configurable quoting in #94, but Daniel explained its remaining setting interactions and favored returning to an independent CSV format in #100.
Aug - Nov 2018 · 10 entries
- After renewed discussion in June, Daniel resumed the independent fieldsep_csv branch in #107, leaving the format-default state machinery behind.
- Fabien's reviews in #108 and #111 led to documentation and help improvements, removal of the leftover -C binding, explicit platform-dependent line-ending documentation and separator validation in #110 and #112.
- After the patch was returned with feedback in #114, Daniel rebased it and added rejected-separator tests in #115; Fabien verified compilation, regression tests and documentation generation in #116.
- Michael Paquier extracted the alphabetical format ordering from #115 into add9182e59, announced the split in #117, and Daniel resent the remaining CSV work in #118.
- Michael's deeper review produced #123, restricting separators to one byte and improving defaults and tests, but he subsequently questioned duplicated backend functionality and differences from COPY's interface in #128.
- Daniel, David G. Johnston and Fabien defended the client-side use cases in #129, #130 and #132, explaining why native CSV formatting remained useful without reproducing every COPY option.
- Tom Lane took up the patch in #133, emphasizing that the existing recommendation to use unaligned output for data interchange was unsafe, and posted a rebased version with revised documentation and cosmetic cleanup.
- The final discussion retained configurable separators but rejected arbitrary separator strings and multibyte support; Tom selected the name csv_fieldsep in #142, and Daniel accepted the single-byte restriction in #143.
- Daniel identified COPY's special backslash-dot end marker in #143 and an expanded-output variant in #144; Tom found additional combinations and implemented conservative field quoting, explaining the correction in #145.
- Tom committed the CSV feature from #133 with the renamed option, simpler separator storage, revised tests and end-marker protection in aa2ba50c2c; the thread ended with Daniel's acknowledgment in #147, with both the ordering cleanup and CSV feature landed.