model-writtenrun 9499 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
Jun - Aug 2018 · 3 entries
- Pavel Stehule opened the thread because \dt+ reported zero for partitioned tables despite their populated partitions, and Ashutosh Bapat proposed explicitly identifying their type as partitioned table #1 #2.
- Pavel and Amit Langote agreed to pursue a separate command that aggregates partition sizes through a query while leaving ordinary \dt size behavior unchanged #7 #8.
- Pavel's first \dP prototype used a recursive query, and Mathias Brossard tested it and supplied a revision restricting support to declarative partitioning on PostgreSQL 10 or newer #11 #14 #16.
Oct - Dec 2018 · 6 entries
- After pg_partition_tree() became available, Pavel adopted it for PostgreSQL 12 while retaining recursive queries for PostgreSQL 10 and 11 #18 #19.
- Amit requested partitioned-index support, and Pavel introduced \dPi and \dPt alongside \dP, initially distinguishing index sizes, table sizes, and combined table-plus-index sizes #22 #28 #30.
- The naming improvement proceeded separately: Michael Paquier supplied explicit partitioned labels for relation and privilege listings, and Alvaro Herrera committed those changes with expanded object titles and regression updates as d56e0fde82 #35 #37.
- Michael's review prompted schema qualification, simpler dispatch, and regression coverage; Alvaro then caught developer-specific expected output, which Pavel replaced with a dedicated test role and narrower patterns #38 #41 #42 #43.
- Michael argued that multilevel listings needed parent information and patterned \dP should find indexes, while Pavel worried about counting the same storage repeatedly; Pavel's next revision showed roots by default and included indexes when a pattern was supplied #44 #45 #50.
- Pavel added the n modifier to include nested partitioned objects and their parents, initially reporting only directly attached partition sizes so displayed sizes could be added without overlap #52.
Feb - Apr 2019 · 10 entries
- Amit's multilevel tests showed why direct-only sizes could mislead, and discussion with Pavel and Alvaro led to separate direct-partition and total-partition size columns #54 #57 #58 #60.
- Amit supplied a documentation and code-cleanup delta that Pavel merged, then Justin Pryzby tested the result and proposed further wording and size-label changes that Amit incorporated into a full revision #61 #62 #63 #65.
- When Peter Eisentraut questioned the feature's scope and authorship, Pavel and Amit clarified that Pavel remained the principal author and that the separate command avoided changing ordinary describe commands to aggregate partition trees #70 #71 #72.
- Alvaro corrected translation-sensitive sentence construction, a wrong translated-column index, and translation state leaking between calls; his accidental Spanish translation edits were identified and later removed #74 #75 #76 #77 #78 #88.
- Kyotaro Horiguchi supplied deltas to include leaf partitions, but withdrew the proposal after Amit explained that the command intentionally listed partitioned objects; Alvaro nevertheless raised broader questions about moving partition detail out of ordinary \d output #79 #80 #81 #82.
- Alvaro retained the + requirement to avoid unconditional size-calculation costs, refactored modifiers to work in any order, and sought consistent table/index selection; Justin caught incomplete implementation and broken \dP+ dispatch, leading to a revision where \dP consistently lists both kinds #86 #88 #89 #92 #94.
- Before announcing the main feature's landing, Alvaro made patterns include matching nested objects and parent names, improved completion and ordering, and retained empty-table output when nothing matches #96 #98.
- Immediately afterward, Justin identified an unnecessary pg_class join following the switch to regclass casts and incorrect documentation about index sizes, then supplied a cleanup patch #99 #100.
- Amit's testing clarified that pg_table_size already includes TOAST storage, and Alvaro accepted that the new command matched ordinary describe behavior; moving leaf partitions out of \d remained a future proposal #102 #103 #104.
- Justin resubmitted the cleanups with a Table-column rename, and Alvaro confirmed all three landed as 421a2c4832, closing the submitted fixes while leaving broader leaf-listing and additional size-display ideas for future work #105 #106.