Top 5 Hacker Discussions
Discussions this week span logical replication conflict logging, index prefetching, and row pattern recognition correctness. The conflict log table patch saw active review addressing duplicate logging, naming inconsistencies, and tuple descriptor safety. The REPACK CONCURRENTLY feature was confirmed deferred to PG20. A proposal to clean up stale publication entries for UNLOGGED tables was abandoned after reviewers questioned its utility. Row pattern recognition testing against Oracle and Trino exposed three matcher defects and two wrong answers in context absorption.
|
Adding REPACK [concurrently]
The discussion concerns the proposed `REPACK [CONCURRENTLY]` feature for PostgreSQL. Alvaro Herrera had previously suggested letting the patch slip from PG19, recommending it be r…
Read Discussion
|
|
Row pattern recognition
Henson Choi reports a systematic cross-check of PostgreSQL's Row Pattern Recognition (RPR) matcher against Oracle 23ai and Trino 471, using roughly 1,000 PATTERN statements. Three…
Read Discussion
|
|
index prefetching
Peter Geoghegan posted V31 of the index prefetching patch set, noting it addresses bit rot against the current master branch with only minor differences from the previous version.…
Read Discussion
|
|