Blocked-by: #6250 (the shared predicate must land first — PR #6541)
Filed by the #6250 developer agent per that issue's ruling: "OUT OF SCOPE — the two out-of-lane consumers. ... packages/drivers/driver-sql/** (domain:drivers) must NOT be edited. Migrating them onto the predicate is a follow-up per lane, filed once the predicate exists." The predicate now exists, so this is that filing.
What is left
packages/drivers/driver-sql/src/sql-driver.ts judges the same question with an inline regex, no name and no doc comment:
if (nullSafe.size > 0 && /unique constraint failed|duplicate entry|duplicate key value/i.test(msg)) {It reads only err.message — the code / errno channel is not consulted at all, so a driver that surfaces SQLSTATE 23505 or ER_DUP_ENTRY with unremarkable prose is invisible to it. That is the same blind spot #6250 measured on the REST side (a Postgres error carrying only code: '23505' came back 500), one layer down.
Why this is finding and not a defect
Nothing a user hits today, as far as this card measured: on the three dialects the repo ships, the message channel happens to carry the words. The defect is structural — a fourth vocabulary that can be taught a dialect the other three never learn. Grade it in triage rather than trusting the label.
The migration
@objectstack/driver-sql already depends on @objectstack/types. Swap for isUniqueViolationError(error) and delete the inline regex. Two things to check while there, because they are what makes this more than a mechanical swap:
- This site tests
msg, not the error object. The shared predicate accepts a bare string on the message channel, so a minimal swap works — but passing the error instead is what buys the code/errno channels, which is the point of migrating.
- The
nullSafe.size > 0 guard is the site's own business logic and must survive untouched; only the discriminator moves.
Note the sibling spelling in this package's own tests (/UNIQUE constraint failed|duplicate key value/, in sql-driver-schema.test.ts, sql-driver-unique-tenancy.test.ts, adr0120-three-posture-conformance.test.ts). Those are assertions on what a real driver emitted, not a discriminator, so they are arguably fine as-is — but they are a fifth spelling of the same vocabulary, and worth a decision in the same pass rather than left to the next reader.
Blocked-by: #6250 (the shared predicate must land first — PR #6541)
Filed by the #6250 developer agent per that issue's ruling: "OUT OF SCOPE — the two out-of-lane consumers. ...
packages/drivers/driver-sql/**(domain:drivers) must NOT be edited. Migrating them onto the predicate is a follow-up per lane, filed once the predicate exists." The predicate now exists, so this is that filing.What is left
packages/drivers/driver-sql/src/sql-driver.tsjudges the same question with an inline regex, no name and no doc comment:It reads only
err.message— thecode/errnochannel is not consulted at all, so a driver that surfaces SQLSTATE23505orER_DUP_ENTRYwith unremarkable prose is invisible to it. That is the same blind spot #6250 measured on the REST side (a Postgres error carrying onlycode: '23505'came back 500), one layer down.Why this is
findingand not a defectNothing a user hits today, as far as this card measured: on the three dialects the repo ships, the message channel happens to carry the words. The defect is structural — a fourth vocabulary that can be taught a dialect the other three never learn. Grade it in triage rather than trusting the label.
The migration
@objectstack/driver-sqlalready depends on@objectstack/types. Swap forisUniqueViolationError(error)and delete the inline regex. Two things to check while there, because they are what makes this more than a mechanical swap:msg, not the error object. The shared predicate accepts a bare string on the message channel, so a minimal swap works — but passing the error instead is what buys the code/errno channels, which is the point of migrating.nullSafe.size > 0guard is the site's own business logic and must survive untouched; only the discriminator moves.Note the sibling spelling in this package's own tests (
/UNIQUE constraint failed|duplicate key value/, insql-driver-schema.test.ts,sql-driver-unique-tenancy.test.ts,adr0120-three-posture-conformance.test.ts). Those are assertions on what a real driver emitted, not a discriminator, so they are arguably fine as-is — but they are a fifth spelling of the same vocabulary, and worth a decision in the same pass rather than left to the next reader.