Aug 2019

3 commits
Remove EState.es_range_table_array.
12 Aug 2019Tom Lanemaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 100% surelanded as postedfrom patchset #132

The four-file diff exactly matches #132, removing es_range_table_array and implementing exec_rt_fetch through list_nth; #136 confirms the push.

Rationalize use of list_concat + list_copy combinations.
12 Aug 2019Tom Lanemaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #1071 of 1 reviewer saw the final version

Matches #107 except that estimate_path_cost_size in contrib/postgres_fdw/postgres_fdw.c uses list_concat instead of list_concat_copy. David Rowley identified the unnecessary copy in #112, and Tom confirmed incorporating that fix in #135.

  • David Rowleysaw the final versionLLM analysis, 100% sureReviewed the patch, identified the unnecessary FDW copy in #112, and accepted the version incorporating that correction in #130; the commit also explicitly credits his review.
Cosmetic improvements in setup of planner's per-RTE arrays.
9 Aug 2019Tom Lanemaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 100% surepart of a posted patchsetsignificant changesfrom patchset #129

Takes the setup_append_rel_array merger and comment improvements from #129, touching five files while retaining simple_rte_array. The proposed removal of that array and its caller conversions were abandoned because testing showed no gain and possible regressions.

Jul 2019

8 commits
Make better use of the new List implementation in a couple of places
22 Jul 2019David Rowleymaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 100% surelanded as postedfrom patchset #1051 of 1 reviewer saw the final version

The nodeAppend.c and nodeMergeAppend.c changes match #105; the only difference is wrapping one comment onto a single line.

  • Tom Lanesaw the final versiontrailerInspected the loops and requested dense-bitmap measurements in #106, then assessed the results and supported proceeding in #109.
Fix sepgsql test results for commit d97b714a2.
17 Jul 2019Tom 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

No attachment contains these contrib/sepgsql/expected/misc.out changes; the commit records a buildfarm-driven adjustment to aggregate log ordering after d97b714a21, the behavioral change discussed in #100.

Avoid using lcons and list_delete_first where it's easy to do so.
17 Jul 2019Tom Lanemaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 100% surelanded as postedfrom patchset #1002 of 2 reviewers saw the final version

Combines both patches from #100, including list_delete_last, the lcons conversions and the aggregate expected-output change; differences in diff hunk placement do not change the resulting edits.

  • David Rowleysaw the final versionLLM analysis, 100% sureReviewed the second patch function by function in #101, examining observable ordering and statistics-selection effects, and is explicitly credited for review in the commit.
  • Daniel Gustafssonsaw the final versionLLM analysis, 100% sureReviewed the first patch's ordering and list_delete_last edge cases in #102, and is explicitly credited for review in the commit.
Remove lappend_cell...() family of List functions.
16 Jul 2019Tom Lanemaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 100% surewritten from the discussion, no matching patch

Implements Tom's proposal in #91 to rewrite the remaining GEQO lappend_cell caller using list_insert_nth and remove the lappend_cell family. Those removals are not present in any attached patchset.

Clean up some ad-hoc code for sorting and de-duplicating Lists.
16 Jul 2019Tom Lanemaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #91

Preserves the sorting and deduplication implementation in #91, with calls and comments updated from list_qsort to the newly chosen list_sort name.

Redesign the API for list sorting (list_qsort becomes list_sort).
16 Jul 2019Tom Lanemaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #901 of 1 reviewer saw the final version

Implements #90's in-place sorting and comparator changes, renaming list_qsort and its comparator typedef to list_sort and improving documentation. The naming change was proposed in #92 and confirmed after landing in #98.

  • David Rowleysaw the final versionLLM analysis, 95% sureProposed renaming the sorting function while its API was changing in #92, a concrete amendment Tom adopted in #95.
Remove dead code.
16 Jul 2019Tom Lanemaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 100% surewritten from the discussion, no matching patch

Removes the now-empty memory-context switching sequence in async.c's pg_listening_channels, identified by Jesper Pedersen in #75. The cleanup was deliberately deferred in #76 and is absent from the attachments.

Represent Lists as expansible arrays, not chains of cons-cells.
15 Jul 2019Tom Lanemaster13beta1REL_19_STABLE13beta1REL_18_STABLE13beta1REL_17_STABLE13beta1REL_16_STABLE13beta1REL_15_STABLE13beta1REL_14_STABLE13beta1REL_13_STABLE13beta1
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #832 of 4 reviewers saw the final version

Lands #83 with an additional rebased lnext call, documentation improvements and local API cleanup changing the lappend_cell family to return void. Reconstructing the posted list.c and pg_list.h against their git blobs confirms these differences; #89 announces the push.

  • David Rowleysaw the final versionLLM analysis, 100% sureBenchmarked and reviewed successive implementations, supplied the detailed review in #81, and explicitly checked the final posted revision in #86; the commit credits his review.
  • Jesper Pedersensaw an earlier versionLLM analysis, 99% sureReported check-world success with and without DEBUG_LIST_MEMORY_USAGE in #75, before the deletion-loop redesign in #78; the commit credits his review.
  • Andres Freundsaw an earlier versionLLM analysis, 99% sureProposed the adopted inline initial allocation in #25 and supplied measured benchmark and profiling results for #49 in #55, before later iterator redesigns.
  • Daniel Gustafssonsaw the final versionLLM analysis, 99% sureConfirmed in #88 that the final posted patch still passed tests after adapting the recently introduced lnext callsite.

Feb 2019

1 commit
Standardize some more loops that chase down parallel lists.
28 Feb 2019Tom Lanemaster12beta1REL_19_STABLE12beta1REL_18_STABLE12beta1REL_17_STABLE12beta1REL_16_STABLE12beta1REL_15_STABLE12beta1REL_14_STABLE12beta1REL_13_STABLE12beta1+1 more
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #301 of 1 reviewer saw the final version

Exactly matches the first, nine-file parallel-loop cleanup in #30; the separate last-cell-predicate patch was dropped, as Tom explains in #45.

  • David Rowleysaw the final versionLLM analysis, 100% sureReviewed the loop cleanup in #34, specifically examining get_tablefunc's assignments before its ordinality-column continue, and is explicitly credited for review in the commit.

Credits across 12 commits

Author
Tom LaneLLM8
David RowleyLLM1
Reviewer
Tom Lane1
Committer
Tom Lane11
David Rowley1

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