model-writtenrun 3095 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Aleksander Alekseev proposed exposing PostgreSQL's existing CRC32 implementation through text- and bytea-input SQL functions returning hexadecimal text, citing user demand and compatibility with external software in #1.
- Nathan Bossart proposed adding CRC32C in #2; after Aleksander initially favored keeping the scope smaller in #3, Nathan argued for reusing both existing implementations, dropping the text-input overload, returning bytea, and describing results as CRC values in #4, which Aleksander implemented in #5.
- Nathan's review in #6 removed unnecessary byte-at-a-time processing and suggested placing the wrappers in src/backend/utils/hash/pg_crc.c; Aleksander incorporated both changes in #7.
- Nathan reopened the return-type question in #8, and Aleksander illustrated the awkward bytea-to-bigint conversion in #9; Nathan consequently suggested reconsidering text in #10, leading to hexadecimal text results in #11.
- Peter Eisentraut advocated an exact numeric result in #12, Nathan weighed integer, bigint, and numeric in #13, and Tom Lane endorsed bigint to represent unsigned 32-bit values in #14; Nathan then posted the bigint-returning version, still credited to Aleksander, in #15.
- Aleksander approved the staged patch in #16, and Nathan committed it as 760162fedb with only a comment edit, catalog-version bump, and commit-message adjustments, announcing completion in #17; both bytea-input CRC functions landed with bigint results and no remaining patch work was identified.