model-writtenrun 8321 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Josef Šimánek introduced COPY progress reporting in #1, following Pavel Stěhule's suggestion, with processed-row counters, ftell-based file-byte accounting and a new pg_stat_progress_copy view.
- Michael Paquier advised prioritizing useful progress points and documentation over difficult deterministic tests in #2, while Fujii Masao questioned STDIN/STDOUT behavior in #3 and Bharath Rupireddy proposed a uint64 byte counter to avoid ftell limitations in #4.
- Josef adopted the counter approach and posted #9, extending byte reporting to STDIN/STDOUT and adding documentation while retaining the existing progress framework.
- Fujii questioned the command-option columns and noticed that file_fdw operations also appeared in the view in #10; Josef confirmed that internal COPY users were included in #12, and Fujii requested documentation of that behavior while opposing the extra columns in #22.
- Vignesh C questioned counting buffered input before its rows were processed in #11, then reported debugger observations confirming that distinction in #17; Bharath separately requested performance comparisons and suggested reducing helper-call overhead in #21.
- Tomas Vondra requested total input-file bytes in #14, and Josef identified existing fstat information in #15; Tomas favored using that total while deferring expensive row counting and speculative query estimates in #16.
- Tomas rejected formatting the underlying counters with pg_size_pretty because it would obstruct calculations in #18, and Vignesh accepted client-side formatting in #24; Fujii also proposed supplying estimated input size through a COPY option for client-side imports in #23, but no implementation was attached here.
- Daniel Gustafsson resent Josef's unchanged revision in #25 to resolve automated testing confusion caused by duplicate representations of the same changes, without taking over authorship.
- The patch subsequently needed a rebase, as Fujii noted in #31, and Michael returned it with feedback after inactivity in #33; Josef ended the thread by expressing continued interest and promising to revisit it in #34, with the outstanding revisions still unsubmitted.
- The work later continued in thread #43478, whose January patchset #465920 is identified as the source of 8a4f618e7a; that landing added total-byte reporting and simplified the view, so it belongs to the later thread despite crediting this earlier discussion.