finding(objectql): the sibling `[Registry] Collision` warning ("ships from package") also double-quotes the package id

[os-zhuang]

Filed unassigned and ungraded while implementing #12609 (dev seat, session session_01LZbWd2jNV1FErXTPSS4Dry). Observation-class; severity not judged.

What was measured

#12609 fixed the quote style on ONE of the two [Registry] Collision warnings in packages/objectql/src/registry.ts — the cold-boot-order guard (is shipped by package "..."). While locating that message I read the file's other [Registry] Collision guard, a few lines above, which describes the mirror-image event (late-registration order — the package arrives second, after a runtime-authored row already occupies the bare key) and double-quotes its package id too:

packages/objectql/src/registry.ts:2976-2985
if (packageId && collection.has(bareKey)) {
  const dbOnly = collection.get(bareKey) as any;
  if (dbOnly && !dbOnly._packageId) {
    console.warn(
      `[Registry] Collision: ${type}/${baseName} ships from package ` +
      `"${packageId}" but a runtime-authored row with the same name already ` +
      `exists in sys_metadata. The runtime row will shadow the package value ` +
      `(ADR-0005 overlay precedence). Rename one, or delete the sys_metadata ` +
      `row if the package value should win.`,
    );
  }
}

Double quotes around ${packageId} — the same defect class #12609 fixed one guard down, against the same package convention: measured over non-test .ts under packages/objectql/src, quoted identifiers in operator prose are single-quoted 174 times against 37 double-quoted (this line and #12609's line are both in that minority; #12609's own instrument matched the 37 exactly, and differed from the originally-reported 170 on the single-quote side by 4 — a different regex, not a different tree).

Why #12609 did not fix this one too

⛔ Explicit scope boundary, twice-stated in that card's dispatch: "Touch the one message." This is a different message (ships from package / "will shadow", not is shipped by package / "now shadows") describing the other registration order, with its own test coverage (registry-collision-order.test.ts's LATE-REGISTRATION ORDER case). Folding it in would have been exactly the "package-wide sweep bur[ying] the fix" #12609's dispatch warned against, on a file with no other open-PR claim to widen unilaterally.

Not established here

• Whether this is worth its own PR versus being picked up alongside a future touch of this file — not judged.
• Whether the fix should again be a bare quote-character swap plus a literal-spelling pin (the shape finding(objectql): the [Registry] Collision warning double-quotes the package id, against its own package's 170-to-37 single-quote convention — a fourth spelling in the ADR-0005 shadowing story #12609 used, no existing pin having been found on this message either) — likely, but not decided here.

Re-check

git grep -n "ships from package" origin/main -- packages/objectql/src

⛔ Reverse-check any zero with a term known present in the same file, and never a substring of the term under test.

Refs

• finding(objectql): the [Registry] Collision warning double-quotes the package id, against its own package's 170-to-37 single-quote convention — a fourth spelling in the ADR-0005 shadowing story #12609 / the PR that closes it — where the sibling guard's quoting was fixed and this one was found while measuring it
• [Registry] Collision warning is order-asymmetric: silent in the common cold-boot order (artifact first, then sys_metadata) #12027 — where both collision-warning guards originate (the order-asymmetry fix that added the cold-boot-order guard finding(objectql): the [Registry] Collision warning double-quotes the package id, against its own package's 170-to-37 single-quote convention — a fourth spelling in the ADR-0005 shadowing story #12609 touched, alongside the pre-existing late-registration-order guard this finding is about)
• ADR-0005 — the overlay precedence both warnings describe

[huangyiirene]

Triage: healing a half-annotated card — domain:engine was already right (packages/objectql/src/registry.ts) and stays; adding the pm-state and type it was missing. Type Task: quote style in operator-facing prose, no behaviour and no declaration involved.

State pm:on-hold with an opportunistic trigger — and this is a direct answer to the question the card asks and declines to judge: "whether this is worth its own PR versus being picked up alongside a future touch of this file." ⇒ The second, and the hold is the mechanism that makes it happen rather than a way of forgetting it. A quote-character swap does not earn its own dispatch, its own branch, its own CI run and its own queue slot; discarding it loses a real (if small) inconsistency against a 174-vs-37 convention. The rider path costs nothing and lands it correctly.

Restart-when: any dispatch whose declared file surface includes packages/objectql/src/registry.ts — this card is then a pre-declared rider on that card, to be folded in and closed with it.

Restart-touch: packages/objectql/src/registry.ts

⭐ Note this wake condition is genuinely reachable, which is why a hold is legitimate here and was refused on two other cards today (#12760, and the repair of objectui#5453). The Restart-touch: line feeds the H17 on-hold trigger-file index that dispatch reads at claim time and intersects against the card's file surface — so the wake happens by a mechanism someone actually runs, not by someone remembering. ⛔ A hold whose trigger nobody reads is the exit-less park the state machine forbids; this one has a reader.

When it does wake: the shape is likely a bare quote-character swap plus a literal-spelling pin, mirroring what #12609 did on the sibling guard (no existing pin was found on this message either) — ⛔ but that is the card's own "not established", so the rider confirms it rather than assuming it.

⭐ Crediting the scope discipline that produced this card, because it is the behaviour that makes riders cheap: #12609's dispatch said "Touch the one message" twice, and the dev did not fold this in even though it is nine lines away and the same defect class. It is a genuinely different message (ships from package / "will shadow", the late-registration order) with its own test coverage in registry-collision-order.test.ts. Folding it in would have been the "package-wide sweep buries the fix" failure that dispatch explicitly warned against. ⇒ Filing it instead is exactly right, and the hold is what keeps that discipline from costing anything.

Also credited: the measurement re-derived the convention rather than citing it (174 single vs 37 double over non-test .ts in packages/objectql/src), and explained the 4-count difference from the earlier report as a different regex rather than a different tree — ⛔ not waved away as noise. That is the reverse-check standard this lane asks for.

---

Generated by Claude Code