model-writtenrun 9848 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Tatsuo Ishii proposed clarifying that pgbench’s prepared mode uses named prepared statements, updating both the manual and a source comment because extended mode also uses prepared statements #1.
- Fabien COELHO tested the patch successfully and investigated the protocol distinction #2; Tatsuo explained that prepared mode avoids repeated parse messages and parse analysis on subsequent executions #3.
- Robert Haas questioned whether the existing wording needed changing #4, while Fabien advocated explicitly describing reuse #5 and Tatsuo preferred “reused” over “named” as the more useful distinction #6.
- Peter Eisentraut favored putting detail in the reference page and later requested a fuller explanation of the query modes #7 #10; Fabien clarified that the patch already targeted the manual and that the modes exercise different libpq interfaces #8 #11.
- Tatsuo supplied benchmark results and replaced the original terminology changes with a manual-only paragraph explaining how reuse of parse-analysis results makes prepared mode faster, leaving the existing mode descriptions and source comment unchanged #12.
- Alvaro Herrera approved the revision #13, and Tatsuo committed it unchanged as 2472ea0a53, confirming completion in the thread’s final message #15.