model-writtenrun 4055 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Peter Eisentraut proposed \gp to execute parameterized queries through the extended protocol, motivated by transparent column encryption and easier protocol regression testing in psql #1.
- Simon Riggs requested examples showing repeated execution on separate lines, while Peter distinguished that usage from future support for named prepared statements #5 #6 #7.
- Corey Huinker identified that treating every \gp argument as a parameter would obstruct output to files or pipes, prompting his proposal to separate parameter binding from execution #8 #14.
- Daniel Verite demonstrated that existing psql variables already worked as positional parameter values, while subsequent proposals explored named-variable interpolation and automatic protocol selection without reaching adoption #11 #15 #20 #22.
- Peter adopted a one-shot \bind command without an explicit end marker, preserving the existing parameter-handling machinery while allowing all execution-command variants and repeated execution of the previous query #23.
- Corey tested the revised patch successfully and requested explicit documentation of zero-parameter extended execution and the binding state's one-shot lifetime #25.
- A global extended-protocol setting remained outside the patch after Peter questioned its practical use case and Daniel identified no additional benefit beyond exercising the protocol #24 #26 #27 #28.
- Peter committed #23 as 5b66de3433, adding only Corey's documentation clarifications, and announced completion in #29.
- After landing, Corey checked quoting, semicolon handling, and repeated execution with \watch; Peter explained that these behaviors followed ordinary psql backslash-command syntax #30 #31.
- In September 2023, Tobias Bussmann reported that \bind failed with FETCH_COUNT and supplied a patch passing parameters when declaring the cursor, with regression tests; he also suggested bypassing cursor execution as an alternative #32.
- Alvaro Herrera questioned whether those tests covered suspected interactions with libpq pipeline support, leaving the FETCH_COUNT fix without a recorded resolution in the supplied thread #33.
- The thread ends with the original feature committed but follow-up questions unresolved, including Aleksander Alekseev's February 2024 request for a direct way to bind SQL NULL values #34.