model-writtenrun 4886 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Bertrand Drouvot proposed generating relation statistics accessors with two macros, crediting an earlier idea from Andres and choosing to rename internal fields rather than SQL functions to avoid changes to views and tests #1.
- Nathan Bossart suggested hard-coding the common function prefix, then supplied shorter macro names and clearer ordering; Bertrand incorporated those changes and chose RELENTRY names to accommodate both tables and indexes, producing the final relation patch #2 #3 #4 #5.
- Nathan accepted the relation patch, and Michael Paquier suggested extending the approach to database statistics; Bertrand and Michael agreed to keep that extension separate so the preparatory relation work could proceed independently #6 #8 #9 #10.
- Michael committed Bertrand's final relation patch unchanged as 83a1a1b566, then posted his own database extension generating 21 integer and five floating-point accessors while preserving function names and renaming internal fields #5 #11.
- Nathan requested alphabetizing the database accessors, while Bertrand endorsed renaming their fields for consistency after considering an alternative that retained the n_ prefix through token concatenation #12 #13.
- Bharath Rupireddy asked about extending macro generation to background-writer and transaction statistics, but Michael expressed less enthusiasm because those cases offered less code reduction; no patch for that extension followed in this thread #15 #16.
- Michael committed the database patch with only accessor ordering changes as 8018ffbf58, retaining the field renames rather than adding more token concatenation #11 #17.
- Tom Lane reported that the landed macros triggered empty-declaration warnings on buildfarm member wrasse, verified that removing invocation semicolons did not confuse pgindent, and also recommended removing terminal backslashes that could accidentally absorb subsequent lines #18.
- Nathan supplied both corrections, Michael confirmed independently preparing the same fix, and the attachment landed unchanged as 66dcb09246, with both credited as authors #20 #21.
- The thread ended with Bertrand thanking Tom, Nathan, and Michael for the diagnosis and fix; both accessor conversions and their follow-up correction had landed, with no attached patch remaining pending #22.