model-writtenrun 3576 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Peter Eisentraut proposed adding alias, join_using_alias, funcordinality, and lateral to RangeTblEntry jumbling, asking which omissions were intentional and noting that WITH ORDINALITY changes query results #1.
- Alvaro Herrera raised a related unresolved problem: queries resolving identical table names through different search paths receive distinct query IDs but can have indistinguishable stored query text #2.
- Julien Rouhaud argued that aliases should remain excluded because they do not change query behavior and stored query text cannot reliably identify every spelling; he initially supported adding both funcordinality and lateral #3.
- Tom Lane accepted funcordinality but opposed lateral because meaningful lateral references already affect the hash, while an otherwise redundant LATERAL keyword should not distinguish queries #4.
- Julien described the search-path display ambiguity as a longstanding problem without a solution in this thread #5, and Michael Paquier endorsed the review arguments while suggesting tests for alias-related behavior #6.
- Peter committed only the unchanged funcordinality addition from the original patchset #1 as 8b29a119fd, then confirmed that the other three fields were left alone #7; no tests or solution to the search-path display problem landed with this change.