[renovate]

This PR contains the following updates:

Package	Change	Age	Confidence
@effect/tsgo (source)	0.62.0 → 0.63.0

---

Release Notes

Effect-TS/tsgo (@​effect/tsgo)

v0.63.0

Compare Source

Minor Changes

• b961835: Add failCauseToConstructor, a style suggestion for Effect v3 and v4 that rewrites Effect.failCause / Effect.failCauseSync wrapping a single Cause.fail / Cause.die into the direct constructor:

  Effect.failCause(Cause.die(new Error("boom")))   // -> Effect.die(new Error("boom"))
  pipe(error, Cause.fail, Effect.failCause)        // -> Effect.fail(error)
  Effect.failCauseSync(() => Cause.fail(error))    // -> Effect.failSync(() => error)

  Piped spellings are detected through piping flows, and later steps are kept. On v3, a lazy Cause.die becomes Effect.dieSync; v4 has no dieSync, so it is not reported there. Forwarded, stored, composed or transformed causes, explicit failCause<E> type arguments, as assertions and annotated thunks are skipped. Lazy forms written as a pipe(...) subject or with a piped body are reported without a quick fix, as are lazy values whose error type would change once returned from a failSync thunk (such as cond ? "A" : "B", which Cause.fail widens to string) and named-import calls where no value Effect module binding is in scope (type-only imports do not count).

• cbf958c: Add manualMetricTapToTrack, an Effect v4 style suggestion with a quickfix that rewrites Effect.tap / tapError / tapDefect calls which only call Metric.update into Effect.trackSuccesses / trackErrors / trackDefects:

  handleRequest.pipe(
    Effect.tap((size) => Metric.update(responseSize, size)),
    Effect.tapError(() => Metric.update(requestErrors, 1)),
    Effect.tap(Metric.update(rpcSuccesses, 1))
  )
  // becomes
  handleRequest.pipe(
    Effect.trackSuccesses(responseSize),
    Effect.trackErrors(requestErrors, () => 1),
    Effect.trackSuccesses(Metric.withConstantInput(rpcSuccesses, 1))
  )

  Inside callbacks, only metrics that are stable at the call site (earlier const bindings, declare const, enum members) are moved, because track* evaluates the metric once while a tap callback re-evaluates it on every run. A computed input stays inside a mapper so it is still evaluated on every run.

• e75ca43: Add mapMapErrorToMapBoth style diagnostic (default off, Effect v3 and v4) with a quickfix. Adjacent Effect.map + Effect.mapError steps (either order) anywhere in one piping flow are reported: in a pipe/.pipe/Effect.fn pipeline, across chained .pipe(...).pipe(...) calls, and in nested data-first, curried or mixed data-first/pipeable calls. The quickfix merges them into Effect.mapBoth and removes the other step, collapsing a .pipe(...) left without steps to its subject; on v4, Effect.mapEager + Effect.mapErrorEager suggests Effect.mapBothEager. The quickfix is withheld when it cannot keep behavior: a step callee that is not a plain or dotted name (such as getEffect().map), explicit type arguments on any call between and including the two steps (a curried application, a containing or empty pipe/.pipe), no binding in scope that reaches mapBoth, a rewrite that would newly start an expression statement with ( or another token that continues the previous line, or a pair split across calls where merging would move a callback that is not a function literal, or the work of an eager first step, across other code.

  // before
  fetchCount.pipe(
    Effect.map((n) => n > 0),
    Effect.mapError((message) => new TaskError({ message }))
  )
  // also reported: fetchCount.pipe(Effect.map((n) => n > 0)).pipe(Effect.mapError(...))
  // after
  fetchCount.pipe(
    Effect.mapBoth({
      onSuccess: (n) => n > 0,
      onFailure: (message) => new TaskError({ message })
    })
  )

Patch Changes

• a2751c7: catchAllToMapError and catchToOrElseSucceed now name the catch variant the callee resolves to in their message. Previously a named import, renamed import or local alias of Effect.catchAll was reported as Effect.catch:

  import { catchAll } from "effect/Effect"
  
  pipe(program, catchAll((error) => fail(error.length)))
  // before: ... more directly than `Effect.catch` followed by `Effect.fail`
  // after:  ... more directly than `Effect.catchAll` followed by `Effect.fail`

• a2751c7: Quick fixes, refactors and completions now build references to Effect modules (Effect, Layer, Schema, Data, Context, ...) from the file's imports instead of reusing the receiver of the call being rewritten, and support namespace imports of the package barrel.

  import * as X from "effect"
  
  const program = source.pipe(X.Effect.map(() => undefined))
  // effectMapVoid quick fix:
  const program = source.pipe(X.Effect.asVoid)

  When the call site reaches Effect through something other than an import, the fix now uses the import:

  import { Effect } from "effect"
  const E = Effect
  const program = source.pipe(E.map(() => undefined))
  // quick fix now produces Effect.asVoid instead of E.asVoid

• 5ebba91: Perf CI: base and head are now measured on the same runner, and only when a maintainer opts in with the perf label. Builds still run in parallel, one job per channel and side, each with that commit's own tooling. A separate measure job per channel downloads both binaries and runs 9 rounds of stock, base and head compiles in a rotating order. The report's main column is now head vs base directly, flagged by a paired Wilcoxon signed-rank test (p < 0.01), with base and PR overhead over stock tsgo kept as context. repoctl perf ci takes --base-bin/--base-commit/--base-version, --head-bin/--head-commit/--head-version and --rounds instead of --side/--commit/--version/--patched-bin/--pairs, and writes version 2 cells:

  repoctl perf ci --channel latest \
    --base-bin ./tsgo-base --base-commit "$BASE_SHA" --base-version 7.0.2 \
    --head-bin ./tsgo-head --head-commit "$HEAD_SHA" --head-version 7.0.2 \
    --workload-dir /tmp/effect-workload --output perf/latest.json

• 611fae3: Rust port bookkeeping: each origin marker now carries the hash of the Go declaration it was last synced to, so porting/manifest.tsv and porting/ported.tsv, which caused constant merge conflicts, are gone. The repoctl tsrs port-* commands compute the manifest in memory, port-check no longer fails on a stale marker line number, and port-ack writes the hash (and current line) into the markers:

  // go: internal/typeparser/effect_type.go:45 TypeParser.EffectType 8a9c52f8a845e081

• fd5ec0a: preferSchemaOverJson's Effect v4 hint names Schema.fromJsonString(Schema.Unknown) for unknown shapes instead of Schema.UnknownFromJsonString, which effect marks @internal and leaves out of its published types:

  Use `Schema.fromJsonString(Schema.Unknown)` for unknown shapes, `Schema.fromJsonString(schema)` for known ones, or `Schema.toCodecJson(schema)` when working with JSON values instead of strings.
  

• 1b8a25b: Update the TypeScript next tag to typescript@next, which ships TypeScript commit 97aedd3b75e8e18a3cf2786ffbdb2e6c88aea23f, and the Rust port to tsrs@0.20.1-ts7.1.0-beta. Symbol tables are read through per-provider helpers, since TypeScript next replaces the map-based SymbolTable with an insertion-ordered one.

---

Configuration

📅 Schedule: (UTC)

• Branch creation
  • At any time (no schedule defined)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.

[coderabbitai]

Caution

Review failed

The pull request is closed.

Finishing Touches

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}