test(spec): the third data/ file group's test titles state each cited decision in words instead of a tracker number (stage 17)

[objectstack-fleet]

Part of #20749
Clause-②: no

Stage 17 of this card: the next area of class (e), the test strings shipped under packages/spec/src, as ruled in 5902360492 on #20513. This stage takes the third name-ordered file group directly under packages/spec/src/data/: the 20 test files from filter-comparand-type.test.ts to filter-view-operator-parity.test.ts. They carried 95 messages and 99 tracker ids, citing 36 records. Every one of those ids now either states what its record decided, in words (form D), or is dropped where the title already says it. Text only: no assertion, identifier, test count or code comment changes.

Census at the base (2799155678)

Instruments: census10.cjs (md5 9d08602ab972b4b8643c90d64d40fa41), census.cjs (md5 6e42a45a926d375013c32d62f16a296e), census-wide.cjs (md5 c98410a19529c439adb0afbfb00026a2) and dirtable.cjs (md5 dda605c54745b4a60cc14c9a686e4eff). They are byte-identical to the copies stages 10 to 16 used. A literal counts as a test title when its folded message is argument 0 of a describe / it / test call, .each / .skip / .only chains included. Everything else is an "other" string.

Both instruments read 1140 messages / 1207 ids in 245 files, the seat's reading at 2799155678. That is stage 16's head reading (1137 / 1204) plus the three ids that PRs #21819 and #21825 added in new ui/ test titles, which ride the ui/ stages.

directory	files	messages / ids	titles	other
ui/	84	396 / 419	378 / 401	18 / 18
data/ (this PR: the third 20 files)	55	280 / 299	265 / 282	15 / 17
api/	40	189 / 201	181 / 193	8 / 8
system/	34	154 / 165	128 / 138	26 / 27
(files directly in src/)	30	118 / 120	117 / 119	1 / 1
ai/	1	2 / 2	0	2 / 2
contracts/	1	1 / 1	0	1 / 1
total	245	1140 / 1207	1069 / 1133	71 / 74

The group reads 95 messages / 99 ids in 20 files, the seat's figures, file for file:

file (under data/)	messages / ids	titles	other
filter-comparand-type.test.ts	9 / 10	9 / 10	0
filter-cross-field-comparison-class.test.ts	4 / 4	4 / 4	0
filter-dotted-head.test.ts	2 / 2	2 / 2	0
filter-empty-operator.test.ts	5 / 5	5 / 5	0
filter-equality-array-schema-door.test.ts	5 / 5	5 / 5	0
filter-field-reference-lowering.test.ts	1 / 1	1 / 1	0
filter-icontains-parse-door.test.ts	3 / 3	3 / 3	0
filter-like-code-point.test.ts	3 / 3	3 / 3	0
filter-like-nul-pattern.test.ts	1 / 1	1 / 1	0
filter-like-wire-lowering.test.ts	3 / 3	3 / 3	0
filter-lowering.test.ts	6 / 6	1 / 1	5 / 5
filter-ne-array-schema-door.test.ts	5 / 6	5 / 6	0
filter-number-comparand-declared-type.test.ts	17 / 17	17 / 17	0
filter-operator-vocabulary.test.ts	2 / 4	1 / 1	1 / 3
filter-preset-comparand.test.ts	2 / 2	2 / 2	0
filter-save-door-face-parity.test.ts	14 / 14	7 / 7	7 / 7
filter-text-comparand.test.ts	3 / 3	3 / 3	0
filter-text-conformance.test.ts	2 / 2	1 / 1	1 / 1
filter-text-operator-declared-type.test.ts	6 / 6	6 / 6	0
filter-view-operator-parity.test.ts	2 / 2	2 / 2	0
20 files	95 / 99	81 / 83	14 / 16

Two more test files sit in the same name range and carry no id: filter-subtree-provenance.test.ts and filter-verdict.test.ts. They are not touched.

• Controls. Lit, a title: data/filter.test.ts, outside the group, reads 18 / 19 at the head as at the base. Lit, "other" strings: the twin expect messages in system/message-queue-retirement.test.ts (:87, :127) still read at the head. Dark: data/filter-operator-vocabulary.test.ts reads 0 / 0 at the head while 11 of its comment lines still carry a number. Planted in a scratch copy of the head data/filter-like-nul-pattern.test.ts: an id put into a title reads 1 / 1, and an id put into a comment reads 0.
• A wider pattern (any # plus digits) reads the same totals as the gate pattern in all 20 files at the base, and 0 in all 20 at the head.
• At the head: 1045 messages / 1108 ids in 225 files. The 20 files read 0 / 0, and no other file moved.

How the area was chosen

data/ has no subdirectory to split by, so its stages take name-ordered file groups near the ~100-id bound. Stage 16's re-cut named this group at 99, and this census reads 99, so the rule needed no re-cut.

Named for the next stages (re-cut from the head census, 1045 / 1108; data/ 185 / 200 left, the same cut stages 15 and 16 named):

• data/ in two more stages, name-ordered:
  1. filter.test.ts to object.test.ts: 17 files, 103 messages / 114 ids. object.test.ts alone carries 42, so no cut lands nearer the bound;
  2. query-transport.test.ts to validation.test.ts (11 files, 34 / 34) with data/driver/ (7 files, 48 / 52): 86 ids.
• ui/ 419, about four stages. api/ 201, two. system/ 165, two. The files directly in src/, 120, one.
• The three docblock needles (ai/build-progress.test.ts:236, :237, contracts/approval-service.test.ts:274), one stage with their docblocks.

What each id became

20 literals (23 ids) now state a decision in words. 3 literals (4 ids) get their subject back in words where the number stood in for it. 72 literals (72 ids) drop a number the title already explains.

Every cited record was read with its comments through REST: 35 answer 200. #14426 answers 404, and its decision was read from what landed: commit 40a44b91b and its CHANGELOG entry (the undefined comparand refusal names the null predicate by its full spellings, so following it never lands in another refusal). The title already says that, so the number is dropped.

record(s)	literal (under data/)	now reads
#7872	filter-comparand-type.test.ts:41	"the accepted set — one set, defined at the shared door for every driver". The ruling put the comparand-type set at the shared compile face, the measured superset, instead of per-driver policy.
#5869	filter-comparand-type.test.ts:255	"does not judge a non-array list-operator comparand — that SHAPE is refused 400 by the engine’s shape gate". The record's 500 became an enveloped 400 at the engine's list-shape gate.
#20347	filter-cross-field-comparison-class.test.ts:44	"the class table — one classification, read by every judge". One comparison classification, lifted from the SQL driver, shared by the lint arms and the driver.
#8371	filter-dotted-head.test.ts:24	"classifyDottedFilterHead — the FILTER axis dotted verdict, by the type of the head segment". Option 2: a type-directed verdict on the head segment, with the structured head left unjudged.
#19757	filter-empty-operator.test.ts:148	"the empty list stays refused in the equality slot — the ruled refusal of an array there is untouched". Ruling 乙: an array in the equality slot is refused at the shared face, the empty array included.
#7597	filter-field-reference-lowering.test.ts:56	"equality triples with a { $field } comparand — lowered to $eq, never a bare field spec".
#20041	filter-like-nul-pattern.test.ts:25	"hasNulInLikePattern — one predicate, so every door refuses a pattern holding U+0000". One spec predicate beside the dangling-escape one, asked at every door.
#7536	filter-like-wire-lowering.test.ts:51	"the wire lowering of like / ilike — its own operator, never folded onto $contains". The caller binds the wildcards.
#20116, #19886	filter-ne-array-schema-door.test.ts:243	"§5 — the stored-filter carrier walk refuses $ne: [...] on save too". The save door refuses what the query faces refuse. #19886 is dropped: the title already says the slot refuses an array.
#20336 (2)	filter-number-comparand-declared-type.test.ts:59, :174	"the numeric grammar — a JSON number literal naming a finite double" and "numberComparandDoorVerdict — a numeric string narrows to its number, a non-numeric one is refused".
#20510	filter-number-comparand-declared-type.test.ts:308	"aggregated and boundByDriver — the refusal says only what is true at its position". The refusal names an aggregated column as such, and names the PostgreSQL server error only where the driver binds.
#5701, #3948, #6520	filter-operator-vocabulary.test.ts:82, :89	One declared expect message: "(measured when $icontains was staged — a dropped predicate WIDENS, which on an RLS read scope is an unauthorized read)" and "Admitting $icontains with an arm on every JS face, in one PR, is the worked example of the clearing direction."
#4706	filter-operator-vocabulary.test.ts:121	"covers the operators retired when $icontains replaced $regex". Option B: $regex retired, $icontains added.
#8793	filter-preset-comparand.test.ts:31	"FilterConditionSchema — a bare preset name in an ordering comparand is refused at publish".
#19889 (2)	filter-save-door-face-parity.test.ts:308, :309	Two declared it.each labels: "an implicit list (an equality-slot refusal, unchanged)" and "a $eq list (an equality-slot refusal, unchanged)". Ruling A: the save door refuses an array in the equality slot.
#18113	filter-text-comparand.test.ts:51	"isRefusedTextComparand — the one predicate every face reads, driven through every FILTER_TEXT_CASES case". Ruling D: the predicate lives with the contract's owner.
#4706	filter-text-conformance.test.ts:162	A declared expect message: "a case-folding $contains returns both — the answer ruled out when $contains was declared case-sensitive". The record's second question, answered A.
#15661	filter-text-operator-declared-type.test.ts:184	"textOperatorDoorVerdict — a text operator over a type that never stores a string is refused". C-deny.

Subject restored (4 ids): #5222 and #7597 at filter-comparand-type.test.ts:225 ("leaves a FieldReference alone at every position — cross-field push-down, scalar comparison and the $eq lowering own its fate"), #8296 at filter-dotted-head.test.ts:40 ("a virtual head is refused: the unmaterializable-field verdict finally reaching the dotted spelling") and #19514 at filter-save-door-face-parity.test.ts:743 ("… and the text-comparand arm stays silent").

Dropped only (72 ids): #5146 (2), #5234, #5240, #5298 (3), #5701, #5869, #7536 (2), #7872 (2), #8220, #8793, #8934, #14079, #14426, #15661 (5), #18113 (2), #19514 (3), #19886 (4), #19889 (5), #20116 (7), #20143 (3), #20311 (3), #20336 (5), #20347 (3), #20446, #20502 (8), #20510, #21448 (4).

• Each of these titles already states the decision it pins: for example "$ne a value: a row with no value satisfies it" for 非否定路径上的 $ne / $nin / $notContains:driver-sql 排除 NULL 行,driver-memory / formula 返回它们(#5146 只裁定了 $not) #5298, "§1 — the enumeration: the save door refuses exactly what the query faces refuse" for spec: FilterConditionSchema (the save-time door) admits a non-boolean $null / $exists flag, so a stored dataset or widget filter carrying one saves clean and is refused 400 on every query face #20116, or "matchesLikePattern — _ is one code point" for spec + SQLite drivers: $like _ matches one code point on SQLite but one UTF-16 unit in formula / driver-memory — $like: '_' returns a stored emoji over REST; the spec never says which "character" is meant #20143.
• The §1 … §7 markers stay. They number each file's own sections, which its comments (// §1 …) and header name; only the card number in front of them goes.
• Files that carried a record as a file-wide tag (filter-cross-field-comparison-class, filter-number-comparand-declared-type, filter-text-operator-declared-type, filter-like-wire-lowering, filter-text-comparand): the verdict's describe now states the decision (above). The others name the contract part they pin (the judged fields, the refusal message, the fixture, the case table), and their child titles state what each part does. The files' header comments still name the records; comments are not touched here.
• Declared labels and messages: the five filter-lowering.test.ts NULL_ROWS names feed it(row.name) and the seven filter-save-door-face-parity.test.ts labels feed a %s title; vitest prints both whole, as the head run's names show.

Readers

• Test-name filters: none. No tracked script, workflow or config passes -t / --testNamePattern.
• Snapshots: none. No __snapshots__ directory is tracked under packages/spec.
• Projects: none of the 20 files is listed in packages/spec/vitest.repo-tests.json; all 20 run in the local project.
• By substring: every old literal, plus a window around each id (260 needles), was searched across the tracked tree outside its own file. No gate, doc, filter, snapshot or scripts/check-*.mjs self-test reads one. The 7 needle hits fall on 5 lines: a code comment in filter-comparand-type.ts:392, and sibling titles in objectql (engine-number-comparand-declared-type-door.test.ts:447, :459, :484) and rest (data-number-comparand-door.test.ts:269). None reads a spec test title.
• The files by name: scripts/check-driver-conformance.mjs names filter-text-conformance.ts, the source module, not the test file's titles. The other references to these file names are code comments, plus one design-doc line (docs/design/predicate-compilation-convergence.md:377) that points at filter-lowering.test.ts's INTRODUCED constant, not at a title.

Text-only proof

Stage 10's scratch tool (textonly10.cjs, md5 d5e4801dbb4329ab1984da91e92fc47c) compares base and head file by file on three legs:

1. Skeleton: the full AST, with string pieces masked. It must be identical.
2. Comments: every comment, byte-equal.
3. Strings: each changed string leaf must sit in a test-call title position or on a declared line, must carry a tracker id before, and must carry no # plus digits after. The declared lines are filter-lowering.test.ts:81, :83, :85, :93 and :95, filter-operator-vocabulary.test.ts:82 and :89, filter-save-door-face-parity.test.ts:308, :309, :311 to :314 and :743, and filter-text-conformance.test.ts:162.

• Result: 20 of 20 files SAME on all three legs, with the per-file counts predicted in writing before the run.
• Totals: 96 changed string leaves in 95 literals, 81 titles and 15 declared. The diff's + and - lines are exactly the 96 planned lines, and every file keeps its line count.
• Controls (10 of 10 as predicted, on scratch copies, each anchor hit once): identifier rename DIFF; numeric literal DIFF; comment edit COMMENT DIFF; a non-title string given an id VIOLATION; a rewritten title given a new id VIOLATION; a title that was id-free at base edited VIOLATION; one title reverted to base SAME; a declared string given a new id VIOLATION; an undeclared expect message changed VIOLATION; a title re-split into a + chain DIFF.

Test counts: the 20 files were run at the base, in a separate base worktree, and at the head, with --project local --project repo. Both sides read 671 tests, all passed, with the same count and status sequence per file in 20 of 20. 543 full test names change, and each equals the base name with the planned replacements applied (0 mismatches). No full name repeats on either side.

main since the base

Re-fetched just before this PR opened, origin/main was three commits past the base (5b2d189e28). None of them touches packages/spec or any of the 20 files, so main was not merged.

Changeset: skip-changeset

Measured, not assumed:

• npm pack --dry-run of @objectstack/spec lists 2068 files. 0 of the 20 touched files are in it, and no *.test.ts at all. The controls src/data/filter.zod.ts and dist/index.mjs are in it.
• In the built dist/, a new phrase and an old literal each read in 0 files. The control Unrecognized key reads in 42.

So this PR publishes nothing, and no changeset is added.

Verification (at 2ceeffd7e7)

• pnpm turbo run build over all packages: 71 / 71.
• @objectstack/spec:
  • vitest run --project local: 615 files, 18384 passed, 1 todo.
  • typecheck exit 0, including check:test-typecheck (52 files / 246 errors / 135 pinned signatures held). Its program holds all 20 touched files, counted with tsc --listFilesOnly -p tsconfig.test.json.
  • check:generated: all 15 generated artifacts up to date.
• Gates: dispatch-gates --commands derived 79 families, the same set as stages 13 to 16, and all 79 exit 0. --ran reconciles: 79 derived, 79 run, 0 NOT-MEASURED, 0 UNRUN, every family with its exit code recorded.
  • The five roster families whose rosters sit under a touched directory were also run, and each exits 0: check:meta-url-spelling, check:spec-changes, check:authz-resolver, check:error-code-casing and check:filter-alias-parity.
• ESLint, a proven narrowing: --no-inline-config over the 20 files reads 0 errors and 0 warnings. The population comes from ESLint's own config: 20 configured, 0 ignored. No file sets parserOptions.project or projectService, so no untouched file's verdict can move.
• check-governed-merges --test: NOT governed, 192 changed lines.

Acceptance notes

• No needle in this group. Every id was a title or a declared string; no expected value of an assertion over a source docblock was found. The three known needles are untouched.
• Same-id test titles in this card's later stages go with those stages: 6 lines in packages/spec/src. Two are in the next data/ group (data/filter.test.ts:576 and :679), and four are in ui/ (component-object-grid-default-filters.pin.test.ts:56, component.test.ts:3518, dataset-filter-nested-relation-list.test.ts:220, view-filter-rule-value-shape.test.ts:249).
• Same-id test titles in other packages are their lanes' test-string shares. A search of describe / it / test lines outside packages/spec finds 187 lines in 16 packages and one example app citing ids this PR handled: driver-sql 32 (14 files), service-analytics 32 (18), formula 24 (9), lint 21 (8), objectql 21 (11), driver-memory 13 (10), driver-turso 11 (10), plugin-security 9 (5), driver-mongodb 8 (5), driver-sqlite-wasm 6 (5), rest 3 (2), cli 2, metadata-protocol 2, and one each in core, runtime and examples/app-todo.
• Code comments still carry ids in these files and their sources, for example the // [#21448] line above filter-save-door-face-parity.test.ts:311 and the header of filter-operator-vocabulary.test.ts. Comments are not this card's share, and none is touched here.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

Nothing in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 0 changed package(s)), so this run has no opinion about the docs.

What this run could not see

• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 0 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 5b2d189e28d5563cbcaa84ac912219017692ea97 → packageMentionDocs.