fix(service-analytics)!: the read scope and the draft preview compare a temporal comparand in the column storage form (ADR-0053 D-A1 / D-A2)

[objectstack-fleet]

Fixes #21505
Clause-②: yes (narrowing)

ADR-0053 D-A1 binds every surface that puts a filter comparand into raw SQL to the driver's own temporal coercion, and D-A2 makes the comparand coercion and its column companion a pair. The analytics read scope bound a temporal comparand as written, and the draft preview compared temporal values as text. On a declared datetime column both faces then selected different rows from engine.find for the same filter, in both directions. Triage ruling 5964364198 (one coercion, the engine door's, no second copy) and claim revision 5964626028 set the shape built here.

What changed

The read scope (read-scope-sql.ts). ReadScopeCompileOptions gains two optional members, coerceTemporalFilterValue(field, value) and coerceTemporalFilterColumn(field, columnSql). They mirror StrategyContext.coerceTemporalFilterValue / coerceTemporalFilterColumn, bound to the object the scope reads, and reach the driver's temporalFilterValue / temporalFilterColumnSql. After the shared lowering, every value comparison (implicit equality, $eq, $ne, the four orderings, $in, $nin, $between) binds its comparand through the first and reads its column through the second. Null tests, $empty and the text arms read the column as stored, as the native where face does. An absent member is identity. The order is ADR-0053 D-E3's by construction: the lowering widens a bare day first, then the arm converts the bound.

The two callers. NativeSQLStrategy.applyReadScope binds the context's pair to the OBJECT, never the join alias. The ObjectQLStrategy read-scope echo binds it to its table. One call site each.

The draft preview (preview-evaluator.ts). It has no driver, so its counterpart of the engine door is the rule that door applies. @objectstack/core's temporalStorageForm, for the kind temporalComparandKind gives the declared type, is applied to both sides of each value comparison: the comparand, and the drafted row's value, as driver-memory reads them. It covers the where and the window. A column the host names no type for stays as written. The declared type is the reader #21417 already hands the evaluator, so analytics-service.ts is untouched.

No new export (the package index is unchanged), no @objectstack/spec edit, and no new dependency edge.

Changeset: @objectstack/service-analytics minor, BREAKING, with an ADR-0087 not-required (no-migration-prescription) disposition. The answer moves in both directions onto the engine's rows: on some filters fewer rows than before, on others more.

Measured before the change (at d2f452b; classes only, the cells live in the pins)

• Read scope vs engine.find, over 16 cells (12 datetime, 4 date): 7 of 12 datetime cells differed on SQLite, and 9 of 12 on PostgreSQL 16.14 under a non-UTC server (America/New_York). Both dialects included the admitting direction. The date cells differed on 0 of 4.
• Through the plugin's own composition with a host getReadScope, the native face differed on the same cells; the ObjectQL face (the engine path) differed on none.
• The draft preview differed on 7 of 12 datetime cells and 0 of 4 date cells. A window end written shorter than the stored instant left out the row at that instant.
• Core's temporalStorageForm equals the driver's coercion on SQLite and PostgreSQL. On MySQL the driver adds its own physical spelling, which is why the read scope takes the driver's pair rather than the core rule.

Pins

• read-scope-temporal-coercion.test.ts (new):
  • compiler shape: the hook sees the lowered comparand; every value comparison takes both halves; null, $empty and text arms are untouched; absent members equal identity members;
  • the 16 cells on SQLite and on live PostgreSQL, with a non-UTC server asserted: the read scope compiled with the driver pair equals engine.find;
  • end to end, the native face with a host getReadScope equals the engine on all 16;
  • the ObjectQL echo prints the coerced comparand;
  • the column half, on an uncertified SQLite datetime column (an external object, which the driver never backfills), against the driver's own find;
  • the draft preview through queryDataset previewDrafts: the 16 cells and two window ends written shorter than the stored instant equal engine.find, over drafted rows in the canonical spelling and over the same instants respelled.
• analytics-faces-one-lowering.test.ts: the read-scope (F9) test's PostgreSQL skip is gone; it passes the driver pair and runs on both databases.

Ablations (predicted first; mutated with scripts/ablation-replace.mjs, each restore verified blob == HEAD and git diff HEAD empty)

• Read-scope comparand member bypassed: 10 red, exactly the predicted set. The first failing cell on each database is one of the card's cells.
• Read-scope column member bypassed: 3 red, exactly the predicted set (two compiler-shape pins and the uncertified-column pin). Every certified-column cell stayed green.
• Preview comparand side bypassed: 2 red (the preview pin in both database blocks). The first failure is on the canonical drafted rows.
• Preview row side bypassed: 2 red. The first failure is on the respelled drafted rows; the canonical rows pass first.

The read-scope pair was ablated at 508b0bea9d and again at this head, with identical results.

Tests and gates (at f8113c0)

• pnpm --filter @objectstack/service-analytics test, with live PostgreSQL set: 175 files passed, 4403 tests passed, 2 skipped.
• pnpm --filter @objectstack/service-analytics typecheck: green.
• Downstream on the rebuilt dist, analytics files: rest 22 files / 291 passed, runtime 7 / 70, dogfood 8 / 70.
• dispatch-gates --commands: 64 families derived, 64 run, every one exit 0. The --ran reconciliation reads 0 NOT-MEASURED (derived from recorded exit codes).
• ESLint narrowed to the 6 changed code files: 0 errors, 0 warnings. The config enables no type-aware linting, so no untouched file's verdict can move. The full pnpm lint is CI's.

Acceptance notes

• CI's Temporal Conformance job runs this package's suite under a skewed process zone but sets no OS_TEST_POSTGRES_URL for it, so the PostgreSQL cells here (and #5930 step 4 (domain:services): the analytics faces delete their hand-copied filter meaning — the read scope (F9), the where tree and its compilers (F10), the draft preview (F11) — each naming its typed column reader #21417's in the same file) are named skips in CI. Each asserts a non-UTC server, so wiring the URL there would not pass vacuously. Noted, not filed.
• origin/main took #5930 step 4 (domain:services): the analytics faces delete their hand-copied filter meaning — the read scope (F9), the where tree and its compilers (F10), the draft preview (F11) — each naming its typed column reader #21417 as a squash (81e69ca), so this branch's commit list still shows #5930 step 4 (domain:services): the analytics faces delete their hand-copied filter meaning — the read scope (F9), the where tree and its compilers (F10), the draft preview (F11) — each naming its typed column reader #21417's pre-squash commits beneath this card's. The file diff is this card's 7 files only.
• A host that calls compileScopedFilterToSql directly keeps the as-written bind until it passes the pair from its driver: absent is identity.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/service-analytics, touching 17 documentable anchor(s).

⛔ 2 release-owned page(s) name something this change touched. These are read-only:

• content/docs/releases/v14.mdx (via generateSql (symbol, a method of class ObjectQLStrategy))
• content/docs/releases/v17/17-5.mdx (via generateSql (symbol, a method of class ObjectQLStrategy))

content/docs/releases/ is RELEASE-OWNED (AGENTS.md "Documentation Guardrails"): release
notes are written centrally at release time, and a code PR that edits them is the exact PR
that guardrail exists to stop. They are still audited — read-only. If one of them is actually
wrong, file an issue or open a dedicated docs-only PR; do not edit it here.

What this run could not see

• the SDK route bridge reached 54 of 206 client-bound route-ledger rows — the other 152 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 152: 0 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 55 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 97 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• 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 — 10 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 10454b3afa94d49e6e424cc16fbbff3a898f3ad8 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 2911ef0ed4ce171d71964f2a02e6b462d61db368 — the merge of head f8113c0a093dbd9ccc77c9346dabf65ba53d7690 into base 10454b3afa94d49e6e424cc16fbbff3a898f3ad8, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 2911ef0ed4ce171d71964f2a02e6b462d61db368 && git checkout 2911ef0ed4ce171d71964f2a02e6b462d61db368
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 10454b3afa94d49e6e424cc16fbbff3a898f3ad8 f8113c0a093dbd9ccc77c9346dabf65ba53d7690 && git checkout -B drift-repro 10454b3afa94d49e6e424cc16fbbff3a898f3ad8 && git merge --no-ff f8113c0a093dbd9ccc77c9346dabf65ba53d7690

node scripts/docs-audit/affected-docs.mjs --json 10454b3afa94d49e6e424cc16fbbff3a898f3ad8

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 10454b3afa94d49e6e424cc16fbbff3a898f3ad8 → pass the list as
args.docs, on the commit named under Which tree this was computed on.