model-writtenrun 7951 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- David G. Johnston opened the thread by proposing documentation for the rediscovered IS [NOT] OF syntax, including back branches through 9.5 #1.
- Tom Lane argued for removal because the implementation checked exact types during parse analysis rather than the SQL standard's subtype semantics, and noted that earlier documentation had already been commented out #2 #3.
- After Joe Conway supplied historical discussion, Tom explained that pg_typeof() now provided equivalent functionality and proposed removing the incompatible syntax; Joe and Bruce Momjian supported that direction #6 #7 #8 #9.
- Tom posted the removal patch, deleting parser and node support and commented-out documentation, adding a pg_typeof() pointer to the feature table, and converting regression tests to print their types directly #10.
- Joe replied with general approval but supplied no testing results or substantive patch analysis, so his response does not qualify for a suggested review under the stated criteria #11.
- Tom committed #10 unchanged as 926fa801ac and announced the push, explaining that the earlier removal proposal had lacked the replacement subsequently supplied by pg_typeof() #13.
- The thread closed with the removal landed and John Naylor deleting the obsolete TODO entry for documenting or improving IS OF #15.