May 2020

2 commits
Rework EXPLAIN format for incremental sort
12 May 2020Tomas Vondramaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 99% surepart of a posted patchsetlight changesfrom patchset #3462 of 2 reviewers saw the final version

The EXPLAIN-format component of #346 matches explain.c and incremental_sort test changes, apart from a preserved blank line; the separate comment patch was committed independently. Its attachment header identifies Justin as author.

  • James Colemansaw the final versiontrailerIn #346, he combined Justin’s formatting patches, corrected indentation, standardized Pre-sorted naming and updated tests.
  • Tomas Vondrasaw the final versioncommittercommitted the patch
Fix typos and improve incremental sort comments
12 May 2020Tomas Vondramaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 99% surepart of a posted patchsetlight changesfrom patchset #3461 of 1 reviewer saw the final version

Takes the comment component of #346, omitting one ExecInitIncrementalSort comment hunk; the remaining explain.c, nodeIncrementalSort.c and tuplesort.c edits match. Justin’s patch header and the commit credit identify Justin and sender James as co-authors.

  • Tomas Vondrasaw the final versioncommittercommitted the patch

Apr 2020

7 commits
Stabilize incremental_sort tests
8 Apr 2020Tomas Vondramaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 99% suremodified before commitlight changesfrom patchset #3341 of 1 reviewer saw the final version

Matches the larger-table regression fix in #334, with each bare ANALYZE narrowed to ANALYZE t as James requested in #335.

  • James Colemansaw the final versiontrailerIn #335, he verified that mode-transition coverage remained intact and requested table-specific ANALYZE.
Minor improvements in Incremental Sort explain
7 Apr 2020Tomas Vondramaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 97% surepart of a posted patchsetlight changesfrom patchset #3201 of 1 reviewer saw the final version

Combines James’s group-invariant comments from #309 with the Maximum-to-Peak subset of Justin’s later #320, adding corresponding regression-output updates. The latest matching submission is Justin’s, but the commit explicitly credits James.

  • Tomas Vondrasaw the final versioncommittercommitted the patch
Consider Incremental Sort paths at additional places
7 Apr 2020Tomas Vondramaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 99% surepart of a posted patchsetlight changesfrom patchset #2861 of 1 reviewer saw the final version

Takes the additional-planner-path component of #286, retaining its allpaths.c, planner.c and shared equivalence-helper changes. Differences are removal of an XXX comment and updated parallel-query expected output; the other series components are excluded.

  • James Colemansaw the final versiontrailerHe reviewed the planner extension in #155 and incorporated the final pathkey-list optimization in #279, after which only light changes remained.
Fix representation of SORT_TYPE_STILL_IN_PROGRESS.
7 Apr 2020Tom Lanemaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 99% surewritten from the discussion, no matching patch

Tom rejected the NULL-check proposal in #309 and restored the zero-valued SORT_TYPE_STILL_IN_PROGRESS sentinel, explaining the shared-memory dependency in #312 and testing his fix in #316. The commit also replaces the sizeof(bits32) loop bound with NUM_TUPLESORTMETHODS.

Fix failures in incremental_sort due to number of workers
6 Apr 2020Tomas Vondramaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 99% surewritten from the discussion, no matching patch

Adds max_parallel_workers_per_gather = 2 to the SQL and expected output after Tom’s report #300, with Tomas announcing the fix in #301; no attached patch contains these two additions.

Fix show_incremental_sort_info with force_parallel_mode
6 Apr 2020Tomas Vondramaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 99% surewritten from the discussion, no matching patch

Implements Tomas’s diagnosis of the premature return in show_incremental_sort_info #294. Although #297 contains the same code plus a test, it was posted at 21:27 UTC, after this commit at 21:19 UTC, so it was not the source.

Implement Incremental Sort
6 Apr 2020Tomas Vondramaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1Also in: PoC: Partial sort
LLM analysis, 99% surepart of a posted patchsetlight changesfrom patchset #2861 of 9 reviewers saw the final version

Takes the core implementation from #286, dropping an unrelated pathnode.c comment hunk and adding parallel regression tests; startup-cost and additional-path components remain separate. This 2020 source supersedes the last 2016 submission listed for thread #30555, and the commit credits James Coleman and Alexander Korotkov.

  • Tomas Vondrasaw the final versiontrailerHe reviewed costing, memory lifetime, planner integration and final revisions throughout the thread, and committed the final version.
  • Andreas Karlssonsaw an earlier versiontrailerHe tested the series and identified a const warning in #221, then proposed unconstify in #223, before later rescan and instrumentation changes.
  • Marti Raudseppno review foundtrailerNo review activity by him appears in this thread; the commit also credits work associated with thread #30555.
  • Peter Geoghegansaw an earlier versiontrailerHe challenged benchmark assumptions in #21 and investigated comparator and memory effects in #123, before the hybrid executor rewrite.
  • Robert Haassaw an earlier versiontrailerHe reviewed naming and executor separation in #2 and raised skew and benchmark concerns in #26 and #31.
  • Thomas Munrosaw an earlier versiontrailerHe examined prefix comparisons and suffix-only sorting in #36 and #40, before substantial executor and planner changes.
  • Antonin Houskasaw an earlier versiontrailerHe reviewed costing, group counting and postgres_fdw tests in #35, #41, #44 and #50.
  • Andres Freundsaw an earlier versiontrailerHe proposed reproducible EXPLAIN output for the unstable tests in #200, before subsequent memory, rescan and instrumentation fixes.
  • Alexander Kuzmenkovsaw an earlier versiontrailerHe supplied cleanup and planner redesigns in #73 and #80, before Coleman’s hybrid executor rewrite.

Credits across 9 commits

Author
James Coleman3
Justin Pryzby2
Tomas Vondra1 LLM2
Alexander Korotkov1
Reviewer
Tomas Vondracommitter4
James Coleman3
Alexander Kuzmenkov1
Andreas Karlsson1
Andres Freund1
Antonin Houska1
Marti Raudsepp1
Peter Geoghegan1
Robert Haas1
Thomas Munro1
Committer
Tomas Vondra8
Tom Lane1

N LLM means N of the credits come from LLM analysis, not from commit trailers.