commit 21a92741e	a gate and a script for the plan-choice tests; no statement gets faster
commit 8aab48d7e	CHANGELOG line written after the commit; case point_scattered_subselect
commit da8761c87	CHANGELOG line written after the commit; case topn_text_tiebreak_global_order
commit ea24f1cdb	CHANGELOG line written after the commit; case order_layout_range
commit 4a4950427	CHANGELOG line written after the commit; case wbuf_layout_offset
commit e4337adbf	CHANGELOG line written after the commit; case topn_page_dirty_pages_20pct
commit f725d7dcb	a few blocks on a point read (9 / 13 / 17 to 9 / 11 / 15 for an equality, a short range and an IN list of 16 keys); no statement of the registry isolates the read of the directory header, and the time of no statement moves; test/sql/col_gorder_point.sql counts the blocks
commit 95bf906bf	the planner price of the keys of an IN list on an order array; the point_scattered_N cases plan the point directory in the builds before and after (60 keys: 2.7 million instructions and 561 against 563 blocks in both), so they do not show it; test/sql/col_gorder_point.sql covers the lists
commit cf7a0678d	a plan choice between the scan and the b-tree plan of one table; the b-tree variants of the registry are tables of their own, so no case shows it; test/sql/agg_unique_key_in_list.sql covers it
commit 672d457c2	the decode of a compact chunk costs 5.5 instructions an entry with AVX2 against 45 (a unit test and the tools/gquant microbenchmark); the statements that read compact chunks are order_layout_range and order_layout_offset_deep, whose time is not decode-bound
commit 570bd2526	the case wbuf_narrowest_condition_first shows no difference from PostgreSQL's plan at 300,000 rows (0.58 ms against 0.63 ms, 41 blocks each): the 1,203-block shape of the commit is not reproduced at this size; test/sql covers the rows
commit a3f3dfef0	a plan price (514 against 82 for k = ANY(300 keys) AND k = 7) held by test/sql; keys_in_few_256 runs the same plan as PostgreSQL at 300,000 rows (0.65 ms against 0.64 ms), so no case shows a gain
commit 466aa2f9f	a lateral top-1 over a key-cut grouped order array of 1,000,000 rows (24.2 to 17.8 million instructions, 2.6 to 2.0 ms) is a size no registry case runs at 300,000 rows; test/sql/col_gorder_group.sql checks the rows
commit 1148c0f74	2,000 hash partitions planned from a new connection, a statistics-sharing test; test/t/leaf_stats_shared.sh checks that every connection finds the records shared
commit 170199310	a planner price of the rows a lookup gets through a b-tree index; test/plan/bands.tsv, calibrate.tsv, planning.tsv and test/expected/agg_lookup_rows_read.out hold the plan choice of 140 entities of mid rank
commit f786fbc57	repairs the cost that 5ce93b74c added on every read of a kept batch set (116 to 43 blocks, 5,287,447 to 5,228,487 instructions on the narrow statement of a 15-page buffer); not a gain over a released version
commit c7468b07c	comment and option text only; no code path changes
commit ca46315a9	comment and option text only; the choice of the layout is not changed
commit 5ce93b74c	a repair of wrong rows after VACUUM in a development version that never shipped; its cost is measured: 69,711 instructions to stamp a buffer of 15 pages and 20,151 for each read that finds the set (0.5% of the page statement)
commit 9e574ccc6	CHANGELOG line written after the commit; case topn_offset_after_vacuum
commit 68f4de4d2	CHANGELOG line written after the commit; case distinct_on_latest_per_entity
commit f2178aebd	the cost of a probe that started over at each LATERAL row (80.8 to 55.9 thousand instructions by callgrind) is part of what lateral_top1_kept_source measures: 200 probes over 300,000 rows take 8.3 million instructions and 1.01 ms (a b-tree 6.8 million, 0.79 ms). The commit alone, built with and without it on the same data (200 probes, key_cut grouped order array, 300,000 rows): 21.1 to 17.8 million instructions, 2.4 to 2.1 ms in the same 3,214 blocks (a b-tree 5.5 million, 0.6 ms)
commit 96387b908	a test commit: a counter (sift._test_col_gdead_stats, third column) that pins that the cut of an OFFSET is taken, asserted by test/sql/agg_topn_offset_bulk.sql (30,000 to 40,000 entries cut for OFFSET 40,000, none with the cut off); it gives no speed of its own, and the OFFSET cases measure the statement
commit d446b8f67	a planner price held by test/sql/agg_join_filter_price.sql (failing first: the filtered join of 20 of 40,000 rows was the hash join, it stays the scan) and the plan matrix shape semi_expr_key; the commit's figures are costs and the time of the plan chosen, not a speed of a case
commit f23436160	held by test/sql/agg_topn_walk_refused.sql (failing first: 303 heap fetches for the latest row of an entity alone and per entity, 0 and 0 now, the rows equal the normal plan's); the LATERAL cases lateral_top1_text_filter and lateral_top1_kept_source run the statements
commit 6ff768eeb	a planner price held by test/sql/agg_par_cost.sql (the lookup, category and bucket classes, failing first: workers on the previous price) and agg_workers_guessed_keys.sql; the commit's times are the plan matrix's best of 25 runs
commit 079bb110e	held by test/sql/agg_exact_units_dirty.sql (failing first: the cost of a month's count with 5% of the pages dirty at the end is under twice the clean one, with 1 page in 20 dirty anywhere it is over twice); the plan matrix with the dirty dimension is the time proof
commit 478e9cb3b	a planner price held by test/sql/agg_topn_runs_filter.sql (failing first: the page of 5,000 of 300,000 rows with 10% passing is the scan and cheaper than the normal plan, and its price grows with the page); the commit's table is the cost against the time of the plan matrix
commit e7265c883	a planner price held by test/sql/agg_topn_walk_cost.sql (failing first, four booleans: the price of the walk follows the pass rate, 394 to 139 at 50%, 434 to 114 at 98%, the give-up boundary at 1%); col_topn_gorder now expects the array
commit eda053aae	CHANGELOG line written after the commit; case topn_offset_churn_text
commit 3d11cb1fb	a planner price of the probes of a repeated index scan held by test/sql/col_join_probe_price.sql (failing first: the arrival-order heap took the hash join, the same rows either way) and the plan matrix shape added with it
commit 74ccf0039	a planner price of the keys an index looks up in its order array held by test/sql/col_point_btree_price.sql (failing first: 500 keys took the sift_col index, the b-tree now, the same rows)
commit a33bd8e5f	a planner price of when the index paths of an aggregate over a few keys stay, held by test/sql/col_point_lookup_plan.sql (failing first: a column of 667 rows a key kept the index scan; 250,000 rows, the 216 rows of one day priced 76 against 187 for the aggregate scan and taking 1.4 ms against 0.55)
commit 1b3b58e06	repair of the baseline-ISA check (scripts/isa_check.sh): best level 30% fewer instructions per rank (116/356/676 to 83/251/475 at 4/16/32 words), the whole query under 0.2% (428.3 M to 427.6 M for 80 conjunctions); the scalar level counts by arithmetic, 151/487/935, more than the libgcc helper it replaces
commit 2d4d217b5	a planner price of the probes of a repeated index scan (the first unit looked up, the others walked, each heap row paid) held by test/sql/agg_topn_plan_keys.sql and col_list_equality.sql (failing first: both red on the tip; the nested loop of 124 rows a probe took 4 to 5 ms and 3,127 buffer hits where the top-N scan takes 1.4 ms and 50)
commit ab98e13a3	a planner price of the rows an index-only scan hands over (0.07 us plus 0.005 for each column of the index, measured 0.1 us for 3 columns, 0.22 for 28) held by test/sql/col_ios_full_price.sql (failing first: the hash join over a full index-only scan of 20,000 rows, 8 ms where the heap takes 0.8) and the join of 1,000 keys: 41 ms before, 7.8 ms after, 100,000 rows
commit 390c12b8b	CHANGELOG line written after the commit (the Added line of key_order); cases key_order_latest and key_order_offset
commit 43f9c7f63	a test-only floor (sift.table_max_segments 16 to 8, a setting that exists only in the testing build) so that tbl_writepath and tbl_bigtx reach the directory pressure with a quarter of the rows (71 s to 15 s and 11 s to 2 s of test time); the library a release ships is unchanged, held by test/sql/tbl_writepath.sql and tbl_bigtx.sql (failing with dir_pressure() always false)
commit 37eaaf50c	a test gate (sift.test_merge_gate) that holds a merge of the maintenance worker, replacing a sleep setting in the testing build; no statement gets faster, held by test/t/maint_lock.sh
commit 6f3edd674	which keys get the grouped order array by default is held by test/sql/col_gorder_autogroup_kinds.sql (a date of 365 values, a timestamp in heap order, an integer and a decimal, an entity of 20 rows and one of 2, a timestamp cut by key_cut, a table with nothing to group, a two-column index); the case lateral_latest_row_default runs the shape of the latest row of each entity
commit 11e5e77b6	a planner price that repairs two regression tests after the point directory became the default of an integer key: test/sql/col_point_dir.sql holds the plan of an equality on a boolean (the index scan priced by the row against the index-only scan) and test/sql/col_index_return_str.sql the plain index scan of the one-row lookups; the plan matrix has no flipped cell
commit 05d41606e	a planner price of the restart of a top-N probe of the order array (2.5 units, measured 2.4 to 2.7 us a probe at 100,000 to 1,000,000 rows) held by test/sql/agg_topn_probe_cost.sql (failing first: priced 88 to 106 units a probe) and the case lateral_top1_kept_source
commit 11e5e77b6	a planner price of the heap row of a plain index scan (0.1 us a row, measured 146 ns a row of a clustered table) held by test/sql/col_point_dir.sql (failing first: the lookup of an equality on a bool, 16,853 rows, took the index scan at 1.55 ms where the index-only scan takes 1.08)
commit 6f3edd674	CHANGELOG line written after the commit; cases group_latest_entity and group_count_entity
commit 0df9bb51c	a test-only setting (sift.test_direct_max) that shortens a regression test from 41.5 s to 0.7 s; no code path of a release build changes
commit 4c1f89e29	a test-only setting of the table load that lets the tests of the ring and the workers use smaller units; the time is the test run, not a statement
commit ad95c2c1	the read of the tombstone log by page, measured by bench/churn/heapfree_churn_reads.py (the count of one tenant under track_changes, recent 20% updated)
the documents that searches walk in the open write buffer of a BM25	needs the maintenance worker to act on the reads of a hundred searches of one buffer, a state the single-session case harness cannot hold still; test bm25_wbuf_read_promote asserts the rule (walked, due) and the lab numbers are in program-scale/bm25lab (mix.sh, records)
the write buffer of a BM25 `sift` index is opened in buffer format 6	a buffer of a given size in a state the case harness does not hold (a read mix over a buffer of 2,000 documents); test bm25_wbuf_wide_bloom asserts the pages skipped and the rows, lab figures in program-scale/bm25lab
`sift.workers` lists the `sift_table` tables the caller can read next 	not a speed claim: a setting, function, message or tool change; nothing to measure
`sift.help(slug)` answers the slug of a refusal of `sift.explain_why` 	not a speed claim: a setting, function, message or tool change; nothing to measure
`sift.merge`, `sift.promote` and `sift.verify` take a `sift_col` index	not a speed claim: a setting, function, message or tool change; nothing to measure
`CREATE INDEX` and `ALTER INDEX .. SET` of a `sift` index with the rel	not a speed claim: a setting, function, message or tool change; nothing to measure
`sift.agg_lookup_max_rows` (deprecated since 3.2; `sift.agg_lookup_max	not a speed claim: a setting, function, message or tool change; nothing to measure
the developer method `unit_rewrite` of `sift.eval_methods` is removed 	not a speed claim: a setting, function, message or tool change; nothing to measure
the name `latest_hint` of `sift.eval_methods` is gone with the code be	not a speed claim: a setting, function, message or tool change; nothing to measure
the name `snapshot_masks` of `sift.eval_methods` is gone: it was accep	not a speed claim: a setting, function, message or tool change; nothing to measure
the developer method `skip_trigger_lookup` of `sift.eval_methods` is g	not a speed claim: a setting, function, message or tool change; nothing to measure
4.2 changes ten things worth reading before the update (`docs/UPGRADIN	not a speed claim: a setting, function, message or tool change; nothing to measure
the planner chooses the aggregate scan more often for a grouped aggreg	not a speed claim: a setting, function, message or tool change; nothing to measure
the aggregate scan prices an expression that PostgreSQL evaluates (`ma	not a speed claim: a setting, function, message or tool change; nothing to measure
`DROP INDEX CONCURRENTLY` of a `sift_col` index that tracks its change	not a speed claim: a setting, function, message or tool change; nothing to measure
a `sift_col` index with `track_changes` whose log is not complete only	not a speed claim: a setting, function, message or tool change; nothing to measure
`sift.refusal_actions()` returns what to change for each slug of `sift	not a speed claim: a setting, function, message or tool change; nothing to measure
options of one `sift_table`, as functions, because a table access meth	not a speed claim: a setting, function, message or tool change; nothing to measure
`bench/steps/fresh_backend_statements.sh` no longer counts the exit of	not a speed claim: a setting, function, message or tool change; nothing to measure
a correlated lookup (a `LEFT JOIN LATERAL .. LIMIT 1` into a large map	not a speed claim: a setting, function, message or tool change; nothing to measure
a run of a correlated `LIMIT 1` child that the scan probes in its btre	not a speed claim: a setting, function, message or tool change; nothing to measure
commit 1620b5c4b	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 301b6785d	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 80be326aa	refactor: one implementation of code that was repeated; no statement gets faster
commit a7e38ea21	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 68b7c57ee	refactor: one implementation of code that was repeated; no statement gets faster
commit 6b06e4db1	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 68aad4143	internal change of the planner or the scan with no CHANGELOG claim of its own
commit a6d9bcfb3	refactor: one implementation of code that was repeated; no statement gets faster
commit 9c5539e19	refactor: one implementation of code that was repeated; no statement gets faster
commit 894a559c7	refactor: one implementation of code that was repeated; no statement gets faster
commit 7729dc951	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 711bd6521	refactor: one implementation of code that was repeated; no statement gets faster
commit 12b2e2c85	refactor: one implementation of code that was repeated; no statement gets faster
commit 2cc886d59	refactor: one implementation of code that was repeated; no statement gets faster
commit 58761cf22	refactor: one implementation of code that was repeated; no statement gets faster
commit f3283936f	refactor: one implementation of code that was repeated; no statement gets faster
commit e2a494fbe	refactor: one implementation of code that was repeated; no statement gets faster
commit 87c284eb8	refactor: one implementation of code that was repeated; no statement gets faster
commit bfc5cd48a	refactor: one implementation of code that was repeated; no statement gets faster
commit 1416eae4a	internal change of the planner or the scan with no CHANGELOG claim of its own
commit ee2de2875	refactor: one implementation of code that was repeated; no statement gets faster
commit 48e056320	refactor: one implementation of code that was repeated; no statement gets faster
commit 8848c1a35	refactor: one implementation of code that was repeated; no statement gets faster
commit e18193ea3	refactor: one implementation of code that was repeated; no statement gets faster
commit f48c4fd46	refactor: one implementation of code that was repeated; no statement gets faster
commit dc2838be2	refactor: one implementation of code that was repeated; no statement gets faster
commit e920d6d54	refactor: one implementation of code that was repeated; no statement gets faster
commit 124d63570	refactor: one implementation of code that was repeated; no statement gets faster
commit af9077f21	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 62387634f	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 400b004b5	refactor: one implementation of code that was repeated; no statement gets faster
commit 985047524	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 8afef71ef	refactor: one implementation of code that was repeated; no statement gets faster
commit b01c0dd3a	refactor: one implementation of code that was repeated; no statement gets faster
commit 3fa5768c4	refactor: one implementation of code that was repeated; no statement gets faster
commit 2447e0234	refactor: one implementation of code that was repeated; no statement gets faster
commit 89b92cc05	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 2e15b8c56	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 905165e4c	internal change of the planner or the scan with no CHANGELOG claim of its own
commit e1fe65bc9	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 825fcb214	internal change of the planner or the scan with no CHANGELOG claim of its own
commit e08362f34	internal change of the planner or the scan with no CHANGELOG claim of its own
commit b6542fe08	internal change of the planner or the scan with no CHANGELOG claim of its own
commit b9180636c	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 82c0351b4	internal change of the planner or the scan with no CHANGELOG claim of its own
commit f5b4c7adf	internal change of the planner or the scan with no CHANGELOG claim of its own
commit d14fd6f52	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 5e7f88a8a	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 416e693ff	internal change of the planner or the scan with no CHANGELOG claim of its own
commit dff1a6579	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 394505b7e	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 4148f11b3	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 33a513378	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 13b8cb552	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 13d064d70	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 19047dd69	internal change of the planner or the scan with no CHANGELOG claim of its own
commit b35e77c54	internal change of the planner or the scan with no CHANGELOG claim of its own
commit ddc30f6f1	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 95044278b	internal change of the planner or the scan with no CHANGELOG claim of its own
commit faee82fb3	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 56f247585	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 29dbf3670	internal change of the planner or the scan with no CHANGELOG claim of its own
commit bce6dfca2	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 3f5232cc8	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 8db5f0169	internal change of the planner or the scan with no CHANGELOG claim of its own
commit bc548c9c1	internal change of the planner or the scan with no CHANGELOG claim of its own
commit b5080144b	internal change of the planner or the scan with no CHANGELOG claim of its own
commit a1454c982	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 36b41e7cc	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 675c705ab	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 9c7cd0856	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 61e3079aa	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 103c54d6a	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 15ca8e326	internal change of the planner or the scan with no CHANGELOG claim of its own
commit bdd8b8143	internal change of the planner or the scan with no CHANGELOG claim of its own
commit bc16c518d	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 4555942f5	internal change of the planner or the scan with no CHANGELOG claim of its own
commit a1f54611d	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 4d9b0198e	internal change of the planner or the scan with no CHANGELOG claim of its own
commit fec4f7f80	internal change of the planner or the scan with no CHANGELOG claim of its own
commit c422f29a3	internal change of the planner or the scan with no CHANGELOG claim of its own
commit b51b237d5	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 1761d534b	refactor: one implementation of code that was repeated; no statement gets faster
commit 679a93b6a	internal change of the planner or the scan with no CHANGELOG claim of its own
commit ec1d19465	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 1778d598e	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 173b535f9	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 8bc3e4241	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 4e456915b	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 803fdf494	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 27d75617a	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 92b89c6f6	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 14353f264	internal change of the planner or the scan with no CHANGELOG claim of its own
commit b0b11245c	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 7eb8a57a5	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 61c53f152	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 98c4654c8	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 4d4fce7cc	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 3b4de9472	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 0993f0cf2	internal change of the planner or the scan with no CHANGELOG claim of its own
commit c21cb2b02	internal change of the planner or the scan with no CHANGELOG claim of its own
commit fa43b024c	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 3774679ee	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 7e6fac745	internal change of the planner or the scan with no CHANGELOG claim of its own
commit f2d575012	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 8098b99e3	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 46ae42100	internal change of the planner or the scan with no CHANGELOG claim of its own
commit ed110b147	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 8e4c1d1d2	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 8a3ee5348	tooling, tests or docs; no statement gets faster
commit cda120493	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 300ef4109	internal change of the planner or the scan with no CHANGELOG claim of its own
commit b7a217a8d	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 3abc4d56c	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 1afdd03d0	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 04fe4d1a2	internal change of the planner or the scan with no CHANGELOG claim of its own
commit c05e3afd6	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 490a75d10	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 10efee8f9	internal change of the planner or the scan with no CHANGELOG claim of its own
commit f087552a4	internal change of the planner or the scan with no CHANGELOG claim of its own
commit bf6a230da	internal change of the planner or the scan with no CHANGELOG claim of its own
commit c74d2d3da	refactor: one implementation of code that was repeated; no statement gets faster
commit 904bd2acd	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 5e27cd63b	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 03440d882	internal change of the planner or the scan with no CHANGELOG claim of its own
commit d687b03ca	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 0bd50da38	internal change of the planner or the scan with no CHANGELOG claim of its own
commit fb418469d	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 7e39a564b	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 30c601dfc	internal change of the planner or the scan with no CHANGELOG claim of its own
commit f27748eaf	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 3e1ed81c9	internal change of the planner or the scan with no CHANGELOG claim of its own
commit c0302aec5	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 6103c5d53	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 1b51fb4ea	internal change of the planner or the scan with no CHANGELOG claim of its own
commit ce19dbb2d	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 7ff873340	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 8545fdeb8	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 32784171b	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 8809bf7ab	internal change of the planner or the scan with no CHANGELOG claim of its own
commit 4e565ae7c	internal change of the planner or the scan with no CHANGELOG claim of its own
commit d568b2374	a comment of the update script; no statement changes
commit b3219f120	includes of the core headers; no code runs differently
commit 9961efd4f	includes of the table headers; no code runs differently
commit 38eac31fc	includes of the column sources; no code runs differently
commit d0b95627d	includes of the table sources; no code runs differently
commit d5b8b4ba4	includes of the core sources; no code runs differently
commit b59e66ff2	includes of the aggregate sources; no code runs differently
commit e68ea604c	a ceiling on the rows of a segment of a key_cut build, not a speed: test/sql/col_key_cut_segment_cap.sql builds 3.2 million rows at the real 128-segment directory under an array limit and holds the bound; no statement of the registry reads the build arrays
commit 10ce9e0e6	a plan choice between the walk of the order array and the units for a list page under a second test (2.68 to 1.81 and 3.63 to 1.96 million instructions on 25,000 rows to a group); the registry has no list page under a second filter on a key_order group, test/sql/col_topn_walk_price.sql pins the choice on 1,000,000 rows
commit 783b4d09f	a ceiling on the memory of a key_cut build, not a speed: test/sql/col_key_cut_segment_cap.sql builds 3.2 million rows, a third of them one value, under an array limit at the real 128-segment directory, and test/sql/col_key_cut_hot_value.sql holds the rule; no statement of the registry reads the build arrays
commit ae68778a0	which step of a rewrite chooses the options of a sift_col index again from the data, not a speed: test/sql/col_key_cut_redecide.sql (worker off) and col_key_cut_redecide_async.sql (worker on) hold the choice; no statement of the registry rewrites a churned index
group the second order of the order arrays (`order_array_group = 'auto'`) also by an integer column that a non-unique b-tree	a choice of the build held by test/sql/col_gorder_entity_group.sql (failing first: no second order for any b-tree shape); measured on scratch tables of 150,000 to 5,000,000 rows, docs/INDEX-RECIPES.md; no registry case builds a table with a b-tree on a measure and a timestamp