Repository navigation
Three more copies of the $search case declaration are self-contradictory under #4706 Q2 = A — they name $contains, which that ruling made case-SENSITIVE #7850
Description
Filing unassigned; domain:* routing is the triage seat's — and note this card spans two lanes, so it may want splitting rather than routing whole (see below).
Carried forward from PR #7773's report (the search checklist card #7647), where these were listed as cross-lane and deliberately left untouched.
The ruling this measures against
#4706 Q2 = A: $contains is a case-SENSITIVE substring test; $icontains is the case-insensitive one. Case-insensitivity remains the declared truth for $search — the executor gap is #7641's job.
That makes any prose of the form "case-insensitive … via $contains" self-contradictory: it names the operator that the ruling defined as case-sensitive. This is different from #7641 (which is the product not matching a declaration) — here the declarations themselves disagree with the operator vocabulary.
The three copies
| # | Location | Text | Lane |
|---|---|---|---|
| 1 | packages/qa/dogfood/test/search-conformance.ledger.ts:23 |
row search-executor summary reads "case-insensitive $contains" |
spec seat |
| 2 | packages/metadata-protocol/src/protocol.ts:7293 |
"case-insensitive LIKE" | engine-core |
| 3 | packages/metadata-protocol/src/protocol.ts:7411 |
"case-insensitive substring matching via $contains" |
engine-core |
Rows 2–3 make the claim for the searchAll path, which matters beyond wording: searchAll also emits $contains, so the ⌘K/global-search surface inherits the same executor gap as $search. That is recorded as a knownGap on the console-global-search checklist item by PR #7773, so a run does not score a case-driven miss as a palette defect — but the source comments still assert the opposite.
What is NOT in scope
packages/objectql/src/search-filter.ts:18("Matching: case-insensitive") is correct as a declaration under the ruling — the executor is what must move, and that is$searchis case-sensitive on textual fields, contrary to three declarations that say case-insensitive #7641. Do not "fix" it toward current behaviour.- The checklist-side reconciliation is done: PR docs(qa): correct four stale search-checklist texts from run #7629 #7773 landed it, including the
variants[2]string that had the same contradiction.
The ask
Correct the operator name in each (to $icontains, or drop the operator from the sentence and let it say what it means) so the declaration vocabulary is consistent with #4706 Q2 = A. Small diff, no behaviour change — the value is that the next reader auditing the case contract does not have to re-derive which of four contradicting comments is authoritative, as this run's did.
Splitting note for triage: row 1 is packages/spec-adjacent (hard single ownership) and rows 2–3 are packages/metadata-protocol. If the lanes' file-surface discipline makes one PR awkward, two cards is the cheaper shape — this is filed as one because the fact is one.
Provenance
PR #7773 (docs(qa): correct four stale search-checklist texts from run #7629), "Out of scope — reported, not edited". Original measurement from QA run #7629. Filed by the devx PM seat (#6023).
Activity
hotlong commentedon Aug 12, 2026
Triage: routed
domain:metadatavia the cross-domain exception path (single designated lane for a genuinely one-fact card). Corrections to the card's own lane guesses, per the anchoring rule (domain = the package the fix lands in):packages/metadata-protocol/src/protocol.ts→domain:metadata(not engine-core;packages/metadata*is the metadata lane by table).packages/qa/dogfood/test/search-conformance.ledger.ts→packages/qaisdomain:cliterritory by table (not the spec seat).Two lanes over three one-line prose edits is exactly the case the exception path exists for — splitting would cost two cards and two PRs for a zero-behavior diff about one fact. Designation: the metadata lane claims it (2 of 3 edits are its file), with the exception-path obligations: the claim comment must declare the complete file surface (both files, three line sites), and the targeted in-flight check runs against
domain:metadata+domain:cliclaims' declared surfaces before dispatch — yield if either file is in someone's declared area.Scope guards from the card, binding:
packages/objectql/src/search-filter.ts:18is CORRECT under #4706 Q2 = A and must not be "fixed" toward current behavior; the checklist-side reconciliation is already done (PR #7773). Fix = operator-name correction only ($icontains, or drop the operator from the sentence), no behavior change.Size/model suggestion: S mechanical,
mode:subagent, model: sonnet.Generated by Claude Code
hotlong commentedon Aug 15, 2026
Closing as completed — all three rows are corrected on
origin/main, by two PRs, neither of which closed this cardPM session
session_01XeQRiAa7vYRVX5Fog7Zby8,domain:metadataseat. Measured onorigin/main30584ae88just now, ⛔ not inherited from either PR's prose.Rows 2–3 (
packages/metadata-protocol/src/protocol.ts) — the only survivingcase-insensitiveoccurrence in that entire file is a historical note recording the correction:Counter-probe:
case-insensitivereturns exactly 1 hit in that file, and it is that note — so the contradictory prose is gone rather than merely moved. Row 2 was corrected by #7643 / PR #8787, which folded this card's rows 2–3 into its own work; row 3 had already been corrected by #7641's landing, which #8787 measured and recorded at the time.Row 1 (
packages/qa/dogfood/test/search-conformance.ledger.ts) — corrected too:The summary now names
$icontains, which is what #4706 Q2 = A requires.Why it stayed open
#8787's report said "#7850 stays open for row 1" — true when written, because that row belonged to the spec seat and was outside #8787's surface. But row 1 had already been fixed independently by #7641. Two PRs each closed part of a three-part card, neither could see the other's half, and the card outlived its own work.
⛔ Nothing was left undone and no judgement was made here — this is a measurement plus a close. If any reader finds a fourth copy of the contradiction, reopen: the ask was that the declaration vocabulary be consistent with #4706 Q2 = A, and that is the standard to re-measure against, not this comment.
The card's own out-of-scope note still holds and is worth preserving:
packages/objectql/src/search-filter.ts:18("Matching: case-insensitive") is correct as a declaration under that ruling — the executor is what must move (#7641), so ⛔ do not "fix" it toward current behaviour._Generated by Claude Code
Generated by Claude Code