test(objectql): conjoin $or/$and with sibling filters in six driver doubles

[huangyiirene]

Part of #7620

⚠️ Part of, not Fixes, and deliberately so. #7620 names 16 files across four packages. This PR carries the six in packages/objectql/src — the domain:engine-core lane, per the claim comment on the card. The remaining ten (6 plugin-sharing, 3 plugin-security, 1 runtime) belong to two other seats and are routed separately, so the card must survive this merge.

What

Six in-memory driver doubles built a WHERE matcher that returned early on $and/$or, discarding every sibling equality key in the same object:

if (Array.isArray(where.$and)) return where.$and.every((w) => matchesWhere(row, w));
if (Array.isArray(where.$or))  return where.$or.some((w) => matchesWhere(row, w));
for (const [k, v] of Object.entries(where)) { /* the siblings, never reached */ }

A real driver ANDs them. So a query shaped like SysMetadataRepository.listDrafts's —

{ state: 'draft', package_id: 'app.x', $or: [{ organization_id: ORG }, { organization_id: null }] }

— would have been answered on the $or alone, handing back rows matching neither state nor package_id. Not a stricter or looser edge case: a different query, with the suite staying green while testing it.

The fix folds $and/$or into the entries loop so they compose with their siblings, matching the corrected form protocol-revert-org-scope.test.ts already carries from #7619 (re-verified intact on current main).

file	matcher calls observed	$or/$and seen
publish-meta-response-conformance.test.ts	60	0
protocol-recorded-by-null.test.ts	44	0
protocol-save-meta-repo-path-real-engine.test.ts	21	0
protocol-registry-shadow.test.ts	7	0
save-meta-response-conformance.test.ts	0	0
plugin.authoring-channel.test.ts	0	0

Live vs dormant — measured, not assumed

The card names establishing this as step one. A probe was installed in each of the six matchers, logging every where it was handed, and the suites run:

132 matcher calls across the six suites, and not one $or or $and. All six are dormant.

The two zero-call files never invoke their matcher at all — those suites drive writes, not reads. The control that makes the silence evidence rather than a dead probe is the 132 plain-equality calls the same instrumentation did record.

Consequence: no existing test outcome changes, and none should. There was no correction-induced failure to report — the interesting result this card could have produced (a test that was passing against a query nobody wrote) did not materialise in this lane. The card's own note that sharing and security are the likeliest to be live is unaffected by this measurement and still worth the other seats' attention.

Dormant is still worth closing: nothing distinguished "this double is faithful here" from "this double quietly changed the fixture", and the next test to add an $or would have inherited a matcher that lies.

Capability preservation

Each matcher keeps exactly the operator surface it already had — $eq unwrapping, the undefined→null comparison normalisation, and the skip for any other $-prefixed key. A non-array $and/$or still falls through to that skip, exactly as before. Measured first: the six proved semantically identical, differing only in indentation (one is 2-space) and an intermediate rowVal variable, so there was no lowest-common-denominator to flatten anything to.

No shared helper — stated deliberately

A helper local to packages/objectql would genuinely have reduced six to one. Rejected anyway, for three reasons:

1. publish-meta-response-conformance.test.ts carries the repo's own written rationale against it: "a gate that imports its own substrate from another gate's file couples two tripwires that must be able to fail independently."
2. fix(metadata-protocol,metadata): revert reads the history row under the key the writer stored it with (#7559) #7619's reference correction is inline for the same reason, and it is the template this card says to copy.
3. An objectql-local helper cannot serve the ten remaining files in plugin-sharing / plugin-security / runtime — a cross-package one is out of scope by construction — so it would add a second convention rather than consolidate to one.

Verification

All run locally in a dedicated worktree at b54aaab.

gate	result
packages/objectql vitest (full)	185 files / 3274 tests passed, 0 failed
pnpm check:query-options-erasure	exit 0 — 67 non-test sites, none new; baseline key set verified against b54aaab, no files added
pnpm check:type-check-debt	exit 0 — 33 ledger entries re-measured, none above its recorded number
pnpm check:durability-log-level	exit 0
pnpm check:engine-double-contract	exit 0 — 155 pinned, unchanged
node scripts/check-engine-split-ratio.mjs	exit 0
eslint on the six	clean

Gate list derived with node scripts/pm/dispatch-gates.mjs <the six paths>, not enumerated by hand. Neither ratchet baseline was raised. Nothing here let me lower a ledger entry: packages/objectql is fully type-checked and carries no entry, and the surplus the debt gate reports sits in unrelated packages.

Not in this PR

• The other ten files of finding: 16 test doubles short-circuit $or in their WHERE matcher, dropping sibling filters — the suite stays green while testing a different query #7620 — other lanes, deliberately untouched.
• objectql test driver doubles are any-annotated at scale — the retired-capability tombstone has no tsc channel in 64 files #7264 (these same doubles are any-annotated at scale) — that is the type channel over the same fixtures; this is their runtime semantics. Different fix, different round.
• No product code, and no test assertion, changed.

---

Generated by Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
objectstack	Ignored		Aug 11, 2026 8:33pm

[github-actions]

📓 Docs Drift Check

No hand-written docs reference the 0 changed package(s). ✅