chore: release packages

[btravers]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

unthrown@5.1.0

Minor Changes

• 5ead919: Close a boundary hole where an async function's rejection escaped
  qualification, stop toBeErrTagged counting a message as payload, and name the
  value in NonExhaustiveError.

  fromThrowable / fromSafeThrowable now reject an async fn. These wrap a
  synchronous function, so they only ever see a synchronous throw: an async
  fn rejects long after the boundary has returned, and its rejection could never
  reach qualify. It used to produce Ok(<Promise>) — un-triaged — and the
  rejection then floated as an unhandled rejection, which terminates the process on
  Node by default:

  const f = fromThrowable(
    async () => {
      throw new Error("boom");
    },
    (cause, defect) => defect(cause)
  );
  f(); // before: Ok(<Promise>) + an unhandled rejection
  // now:    Defect(TypeError: … `fn` returned a thenable …)

  The orphaned rejection is adopted and silenced, so it cannot float. Reach for
  fromPromise / fromSafePromise for async work.

  This is caught at runtime rather than by the type system — the one thenable ban
  in the library that is not a compile error. Putting NotThenable on fn's return
  also makes generic functions unassignable, so fromSafeThrowable(structuredClone)
  would stop compiling with T collapsed to unknown; the phantom rest-tuple guard
  fromPromise uses fares worse. The success type is therefore slightly over-stated
  (Result<Promise<T>, E> is spellable but never inhabited) — the mirror of
  recoverErrCases's never under-stating the error channel.

  toBeErrTagged's exact-payload form now ignores every reserved key. The
  documented way to set a TaggedError's message is a subclass field —
  override message = "…" — which lands as an own enumerable property, so it
  leaked into the payload and failed an exact assertion on the very pattern the
  library prescribes:

  class HttpError extends TaggedError("HttpError")<{ status: number }> {
    override message = `http ${this.status}`;
  }
  expect(Err(new HttpError({ status: 500 }))).toBeErrTagged("HttpError", {
    status: 500,
  });
  // before: failed — the payload was seen as { status: 500, message: "http 500" }
  // now:    passes

  The matcher now skips _tag, name, message and stack — exactly the keys
  TaggedErrorInstance omits and the constructor types ?: never, so none of them
  can legitimately be payload. Assertions using expect.objectContaining are
  unaffected (they were already tolerant of the extra key).

  NonExhaustiveError now names the value it could not match. JSON.stringify
  returns undefined — it does not throw — for a function, a symbol, or
  undefined, so the String(input) fallback never fired and the message read
  "no pattern matched the value undefined" for exactly the rogue inputs this error
  exists to describe. It now falls back correctly (… the value function rogueFn() {}).

@unthrown/oxlint@5.1.0

Minor Changes

• 829b9d3: Add the no-unused-matcher rule (Add an @unthrown/oxlint rule: no-unused-matcher (…Cases callbacks that ignore the injected matcher) #171) — and enable it in the recommended
  preset.

  A …Cases callback (the five error combinators, and match's errCases
  handler) that never uses the matcher it was handed sources its exhaustiveness
  from a builder bound to some other value. The type checker cannot see it —
  the ExhaustiveMatch constraint is structural, and noUnusedParameters never
  fires because the parameter is not unused, it is simply never declared — and
  the runtime runs the borrowed builder against whatever value it closed over:
  the wrong branch is chosen silently (a plausible wrong value, with the Err
  channel typing as fully handled), or nothing matches and the modeled error
  becomes a Defect.

  The rule reports a callback whose matcher parameter is absent (() => …) or
  never read, and — separately, to catch a trivial void matcher reference
  fronting for a foreign builder — any second unthrown / ts-pattern match(...)
  built in the callback's own body. Branch handlers (nested functions) stay free
  to match their payload. There is no escape hatch: a …Cases callback that
  does not use its matcher is never what you meant.

  Enabling the rule in recommended can surface new lint errors in an existing
  codebase; each one is a real bug of the shape above.

@unthrown/prisma@0.2.0

Minor Changes

• ab7194f: Make E domain outcomes only: infrastructure failures become defects, and fix
  an unsound error channel on tryCreate / tryUpsert.

  E now carries only what you would actually branch on. Three P-codes
  describe a domain outcome and stay modeled — UniqueConstraintViolation (P2002,
  409), ForeignKeyViolation (P2003, 400), RecordNotFound (P2025/P2018, 404).
  Everything infrastructural is now a defect: a dropped connection, a pool
  timeout, a deadlock, an unmapped P-code, a malformed query, a client that could
  not start, an engine panic.

  Nobody writes domain logic for a severed TCP connection — you log it and return a
  500, which is exactly what match's defect arm already does. Modelling those
  failures only forced every call site to carry an arm that duplicated its own
  defect arm:

  // before — the same handling, written twice, at every call site:
  errCases: (matcher) => matcher
    .with(P.tag("UniqueConstraintViolation"), (e) => resp.conflict(e.fields))
    .with(P.tag("DriverError"), (e) => resp.serverError(e)),
  defect: (cause) => resp.serverError(cause),

  A defect is not a crash: it flows through the pipeline untouched and is folded at
  the edge like any other unmodeled failure.

  Consequences:

  • The DriverError class is removed. Nothing routes to it any more; the
    original Prisma error reaches your defect arm unwrapped, with its code,
    meta and stack intact.
  • A read has no modeled failure: tryFindMany is
    AsyncResult<User[], never>. Absence is still null.
  • Pagination gains InvalidCursor, the one carve-out. A cursor is an opaque
    string from a client, so a parseCursor that rejects it — or a query Prisma
    refuses to validate — is anticipated input you answer with a 400. A throw out
    of getCursor, which reads rows you just fetched, is a bug and stays a defect.
  • Retrying a deadlock (P2034) or pool timeout (P2024) now goes through
    recoverDefect rather than a tag match — one place in a codebase, versus an
    arm at every call site.

  tryCreate and tryUpsert now carry RecordNotFound. A nested connect
  pointing at a record that does not exist raises P2025, even though neither
  operation has a row of its own to miss:

  db.post.tryCreate({ data: { title, author: { connect: { id: 999 } } } });
  // Err(RecordNotFound) — P2025

  Their unions previously excluded it, so the runtime produced an error the type
  said was impossible. An exhaustive mapErrCases over the declared union then had
  no arm for the value that arrived, the matcher threw NonExhaustiveError, and
  the throw-to-defect net turned a modeled database failure into a Defect. P2018
  — the same failure from the to-many side of a nested write — now maps to
  RecordNotFound too. The batch mutations (tryCreateMany / tryUpdateMany and
  their *AndReturn twins) accept no nested writes, so they remain free of it.

  withCursor's after and before are now mutually exclusive. Passing both
  type-checked and silently ignored after.

  Breaking, shipped as a minor: qualifyPrismaError takes the injected defect
  helper as a second argument (it is a qualify, so passing it to a boundary is
  unchanged — only direct invocation needs updating), the DriverError class is
  gone, and every error union changed. Each of those is a compile error at the call
  site rather than a silent behaviour change — which is the point.

Patch Changes

• Updated dependencies [5ead919]
  • unthrown@5.1.0

@unthrown/vitest@5.1.0

Minor Changes

• 5ead919: Close a boundary hole where an async function's rejection escaped
  qualification, stop toBeErrTagged counting a message as payload, and name the
  value in NonExhaustiveError.

  fromThrowable / fromSafeThrowable now reject an async fn. These wrap a
  synchronous function, so they only ever see a synchronous throw: an async
  fn rejects long after the boundary has returned, and its rejection could never
  reach qualify. It used to produce Ok(<Promise>) — un-triaged — and the
  rejection then floated as an unhandled rejection, which terminates the process on
  Node by default:

  const f = fromThrowable(
    async () => {
      throw new Error("boom");
    },
    (cause, defect) => defect(cause)
  );
  f(); // before: Ok(<Promise>) + an unhandled rejection
  // now:    Defect(TypeError: … `fn` returned a thenable …)

  The orphaned rejection is adopted and silenced, so it cannot float. Reach for
  fromPromise / fromSafePromise for async work.

  This is caught at runtime rather than by the type system — the one thenable ban
  in the library that is not a compile error. Putting NotThenable on fn's return
  also makes generic functions unassignable, so fromSafeThrowable(structuredClone)
  would stop compiling with T collapsed to unknown; the phantom rest-tuple guard
  fromPromise uses fares worse. The success type is therefore slightly over-stated
  (Result<Promise<T>, E> is spellable but never inhabited) — the mirror of
  recoverErrCases's never under-stating the error channel.

  toBeErrTagged's exact-payload form now ignores every reserved key. The
  documented way to set a TaggedError's message is a subclass field —
  override message = "…" — which lands as an own enumerable property, so it
  leaked into the payload and failed an exact assertion on the very pattern the
  library prescribes:

  class HttpError extends TaggedError("HttpError")<{ status: number }> {
    override message = `http ${this.status}`;
  }
  expect(Err(new HttpError({ status: 500 }))).toBeErrTagged("HttpError", {
    status: 500,
  });
  // before: failed — the payload was seen as { status: 500, message: "http 500" }
  // now:    passes

  The matcher now skips _tag, name, message and stack — exactly the keys
  TaggedErrorInstance omits and the constructor types ?: never, so none of them
  can legitimately be payload. Assertions using expect.objectContaining are
  unaffected (they were already tolerant of the extra key).

  NonExhaustiveError now names the value it could not match. JSON.stringify
  returns undefined — it does not throw — for a function, a symbol, or
  undefined, so the String(input) fallback never fired and the message read
  "no pattern matched the value undefined" for exactly the rogue inputs this error
  exists to describe. It now falls back correctly (… the value function rogueFn() {}).

@unthrown/boxed@5.1.0

Patch Changes

• Updated dependencies [5ead919]
  • unthrown@5.1.0

@unthrown/effect@5.1.0

Patch Changes

• Updated dependencies [5ead919]
  • unthrown@5.1.0

@unthrown/neverthrow@5.1.0

Patch Changes

• Updated dependencies [5ead919]
  • unthrown@5.1.0

@unthrown/standard-schema@5.1.0

Patch Changes

• Updated dependencies [5ead919]
  • unthrown@5.1.0

[Copilot]

Pull request overview

This is an automated Changesets release PR to publish the @unthrown/* fixed version group at 5.1.0, including a minor release of @unthrown/oxlint adding the new no-unused-matcher rule (enabled in the recommended preset).

Changes:

• Bump package versions from 5.0.0 → 5.1.0 across the fixed release group.
• Update package changelogs for 5.1.0 (notably @unthrown/oxlint with the new rule notes).
• Remove the applied changeset file after generating changelogs.

Reviewed changes

Copilot reviewed 15 out of 15 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
packages/vitest/package.json	Version bump to 5.1.0.
packages/vitest/CHANGELOG.md	Add 5.1.0 section header.
packages/standard-schema/package.json	Version bump to 5.1.0.
packages/standard-schema/CHANGELOG.md	Add 5.1.0 patch entry referencing unthrown@5.1.0.
packages/oxlint/package.json	Version bump to 5.1.0.
packages/oxlint/CHANGELOG.md	Document the no-unused-matcher rule release notes for 5.1.0.
packages/neverthrow/package.json	Version bump to 5.1.0.
packages/neverthrow/CHANGELOG.md	Add 5.1.0 patch entry referencing unthrown@5.1.0.
packages/effect/package.json	Version bump to 5.1.0.
packages/effect/CHANGELOG.md	Add 5.1.0 patch entry referencing unthrown@5.1.0.
packages/core/package.json	Version bump to 5.1.0.
packages/core/CHANGELOG.md	Add 5.1.0 section header.
packages/boxed/package.json	Version bump to 5.1.0.
packages/boxed/CHANGELOG.md	Add 5.1.0 patch entry referencing unthrown@5.1.0.
.changeset/seven-rules-no-unused-matcher.md	Remove the consumed changeset after release generation.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.