Effect-TS/effect @effect/vitest@4.0.0-rc.113

on GitHub

Patch Changes

• #7254 a63dcbf Thanks @gcanti! - Add the experimental Schema-first effect/unstable/arbitrary/Arbitrary module for native generation without
  fast-check. Arbitrary.schema derives an opaque arbitrary from the decoded Schema Type, Arbitrary.sampleEffect
  provides interruptible sampling with typed exhaustion, and Arbitrary.checkEffect returns structured property results.
  The initial implementation supports bounded discards, shrinking, replay, and recursive and mutually recursive Schemas.
  SampleError and Exhausted include the effective seed so discarded runs remain reproducible even when the caller did
  not provide one. Arbitrary.isArbitrary identifies values through the module's nominal protocol. Numeric constraints
  retain NaN when it is accepted by their supported Order.Number bounds. Union derivation validates oneOf
  exclusivity and isolates lazy cross-member shrinking from unrelated random generation. Object derivation keeps
  optional-property selection constructive when candidate fields have different recursive costs.
  Struct, Record, JSON-object, and record-shaped Arbitrary.all outputs periodically use a null prototype as an edge
  case, preserving that prototype throughout shrinking and replay without perturbing structural PRNG choices. The change
  adds 0.01–0.03 KB gzip to representative Arbitrary fixtures and leaves production-only bundle sentinels unchanged.

  Add Arbitrary.map, Arbitrary.flatMap, Arbitrary.filter, Arbitrary.filterMap, and Arbitrary.all for composing
  derived Arbitraries without exposing a second catalog of primitive constructors. Filtering remains bounded and
  promotes valid shrink descendants through rejected nodes. maxShrinks bounds every inspected shrink candidate,
  including candidates rejected before property evaluation, while retaining the best shrunk input found when the
  budget is exhausted. flatMap provides deterministic dependent generation, source-first shrinking, post-source PRNG
  checkpoints, and one shared residual recursion budget. all combines tuples, iterables, and records with a shared
  budget, randomized internal generation order, stable output shape, and independent member shrinking. Arbitrary values
  implement Pipeable for composition with data-last combinators.

  Add the experimental Schema arbitraryConstraint and toCodecArbitrary annotations and their
  Schema.Annotations.ToArbitrary types. Declarations can provide a Schema Link optimized for generation, while filters
  can contribute native semantic constraints. The callback receives decoded type parameters and normalized constraints.
  The compiler owns efficient representations for common built-ins, including JSON, RegExp, URL, Date, byte arrays,
  ReadonlyMap, and ReadonlySet. Effect-specific HashMap, HashSet, Chunk, Graph, BigDecimal, and date-time declarations keep
  local generation Links, while declarations with productive canonical codecs require no arbitrary-specific annotation.
  Schema.isUniqueKey provides key-based Map uniqueness for explicit array representations.

  The same ownership policy applies to formatter and equivalence derivation: implementations for common declarations
  live in their compiler, while domain-specific and dynamically constructed declarations retain local annotations.
  Declarations whose intrinsic Equal implementation already matches their Schema equivalence need no annotation or
  compiler special case. This keeps unused common callbacks out of production Schema bundles.

  Against the previous layout, schema-toArbitrary decreases from 36.68 KB to 33.24 KB gzip and
  arbitrary-combinators decreases from 37.16 KB to 33.70 KB. schema-toFormatter increases from 18.92 KB to 19.49 KB
  and schema-toEquivalence increases from 19.05 KB to 19.39 KB because callers that explicitly derive these capabilities
  now retain the common declaration handlers. Generic production fixtures remain unchanged; an equivalence-specific
  production fixture using common declarations decreases from 20.75 KB to 20.48 KB, while declarations whose intrinsic
  equality is sufficient decrease from 23.42 KB to 23.34 KB. An Arbitrary-specific production fixture using common
  declarations decreases from 20.35 KB to 19.61 KB, while one using the locally annotated BigDecimal and date-time
  declarations increases from 18.34 KB to 23.01 KB.
  The complete 31-scenario native Arbitrary comparison reports no statistically classified runtime regression; the five
  moved BigDecimal and date-time scenarios remain within measurement noise.

  Add SchemaGetter.forbiddenEncoding, a reusable getter for the encode side of decode-only Schema transformations.

  Remove the fast-check bridge from the effect package, including Schema.toArbitrary and
  effect/testing/FastCheck. Replace the legacy Schema.Annotations.ToArbitrary callback contract with the native
  Schema-first types. The effect package no longer depends on fast-check.

  Migrate TestSchema.Asserts.verifyLosslessTransformation and TestSchema.Asserts.arbitrary().verifyGeneration to the
  native runner. Both methods now accept native check options directly, bound unsuccessful generation, and include the
  shrunk input and replay token in property failures.

  Use the Arbitrary runner for all @effect/vitest property tests. Property inputs may combine Schemas and Arbitraries,
  and are composed directly with Arbitrary.all; check options are available through arbitrary. Raw fast-check
  arbitraries and the fastCheck options object are no longer supported. As with the previous fast-check adapter, thrown
  exceptions, defects, and typed failures from a property are shrinkable falsifications; Effect interruption remains an
  interruption.

  Optimize constructive regular-expression generation by caching feasible lengths on the compiled pattern, computing
  sequence-suffix feasibility once, and precomputing character-class metadata. Seeded generation, shrinking, and replay
  remain unchanged.

  Optimize BigDecimal.Order and BigDecimal.Equivalence with a shared hybrid comparator. Ordinary scale differences
  use cached, bounded coefficient alignment, while large differences are compared without materializing their decimal
  zeroes. BigDecimal.make now rejects scales that are not safe integers.

  Before its removal, the materialized fast-check bridge fixture
  schema-toArbitrary-materialized-fast-check.ts measured 79.00 KB minified and gzipped.

  Representative runtime measurements against corresponding hand-written fast-check 4.9.0 arbitraries are shown below.
  Values are median latency on Node 24.12.0 and Apple M3; lower is better. Both implementations validate the
  same output domains, although their generation distributions are not identical. Native speedup is fast-check latency
  divided by Native latency, so higher is better.

  Scenario	fast-check	Native	Native speedup
  32 recursive samples	150 µs	103 µs	1.45x
  128 optional Struct samples	244 µs	86.0 µs	2.84x
  128 constrained strings	742 µs	49.7 µs	14.86x
  RegExp derivation and first sample	13.4 ms	30.8 µs	429.02x
  64 RegExp strings	595 µs	919 µs	0.64x
  RegExp failure and shrinking	168 µs	88.2 µs	1.91x
  128 bounded numbers	68.9 µs	21.8 µs	3.18x
  128 Uint8Array samples	98.3 µs	74.4 µs	1.32x
  128 BigDecimal samples	66.6 µs	56.3 µs	1.18x
  128 DateTime.Utc samples	71.2 µs	50.5 µs	1.42x
  128 named time zones	52.2 µs	27.9 µs	1.85x
  128 time zones	63.7 µs	33.8 µs	1.89x
  128 zoned date-times	130 µs	112 µs	1.16x
  32 samples through Schema filter	65.9 µs	49.4 µs	1.33x
  32 unique arrays	156 µs	132 µs	1.18x
  128 literal samples	40.0 µs	3.70 µs	10.78x
  128 mapped samples	59.0 µs	14.1 µs	4.21x
  128 samples through passing filter	58.9 µs	13.9 µs	4.23x
  32 samples through selective filter	66.1 µs	42.9 µs	1.54x
  128 filterMap samples	75.7 µs	31.5 µs	2.40x
  Filtered failure and shrinking	12.7 µs	7.71 µs	1.66x
  128 all tuple samples	43.5 µs	18.5 µs	2.35x
  128 all record samples	81.0 µs	30.4 µs	2.66x
  128 dependent flatMap samples	125 µs	67.2 µs	1.86x
  flatMap failure and shrinking	20.1 µs	6.71 µs	2.99x
  Replay flatMap shrink path	14.3 µs	6.57 µs	2.17x
  Passing property, 100 runs	42.3 µs	27.1 µs	1.56x
  TestSchema, 100 generations	44.5 µs	35.9 µs	1.24x
  First failure plus one shrink	8.77 µs	1.30 µs	6.75x
  Replay recorded failure	6.35 µs	1.19 µs	5.36x

  Cold recursive derivation is not included because the native fixture constructs and compiles a Schema, while the
  fast-check fixture constructs a hand-written arbitrary; it is not a like-for-like warm-generator comparison.

  Add a guide for the native module and a migration guide from the fast-check bridge published in effect@4.0.0-rc.109.

• #8154 3993340 Thanks @tim-smart! - Await asynchronous test and layer finalizers after an Effect test times out before starting the next test.

• #8104 e9915d5 Thanks @tim-smart! - Require Vitest >=5.0.0 <6.0.0 and Node.js ^22.12.0 || ^24.0.0 || >=26.0.0.

  Breaking changes

  • Replace .sequential and { sequential: true } with { concurrent: false }.
  • Use bench from the test context and await bench(name, fn).run(). The top-level benchmark API is removed.
  • Use Assertion<void, T> or Assertion<Promise<void>, T>. Define custom matchers through vitest.Matchers, not @vitest/expect.
  • Import reporter types from vitest/node and environment/snapshot APIs from vitest/runtime. Set outputFile when consuming JSON reports.
  • Await asynchronous assertions. Mock history now clears before each test.

  See the Vitest migration guide for removed types and other upstream changes.

• #8104 e9915d5 Thanks @tim-smart! - Add a concurrent option to named layer and it.layer suites. Omitted options and anonymous layers preserve inherited concurrency.

• Updated dependencies [49e3901, b945ded, d6422f4, 5a80204, 53511ef, 79ae49f, 0d083ba, be0f822, debe8fd, a8588f9, f17eb0a, f30cbfe, 78cc9c0, ccbdbd5, a63dcbf, b845b18, 381b794, 4ffcaf4, ba2fd82, 02be94c, 115d8c2, 1452635, 77f85fe, a3f2b31, 310f8d3, 291d616, 48dbbb2, b43bfd6, 1c89c78, 9bbe1a5, dd99ab0, 8f397ed, d7ae6b6, 8cf1203, 87654c5, f05ae0b, 436f5eb, d8ff960, 8766475, b64f406, 4697aaa, cec6c2d, 9956f0e, 5c7eed0, 183c2ea, 1e92dbd, 4eb0fa7, 534b8b9, b76a1cf, acc1e53, 4950a91, c020987, abe95d1, 027ceb9, 3d203b7, a29b8f4, ce4aa65, a8ea807, a3ebb7f, 473bd81, 53843f6, c5eca65, d6f9eba, 8d1e97a, b28ab48, 9960708, 8ac53b6, 84ad49a, 72cfa24, 95c2581, d5c7cd2, fe4fed1, d150a64, 05b1e80, 414dc90, 3f51acd, d68ff05, f3cf1e6, 6525771, 4ab4e83, f7f1d78, 47b358a, 45ffa72, 6232650, 4b73e1b, 284050c, c85fc0b, 7999b07, 7d455f5, d681c2e, 84d2a47, ed74b18, 2245997, d386979, fc9fedf, 7750dbe, fc91af6, 39b9738, 88093b5, d14c463, d473bd3, 84864bc, e2ae724, f921ed3, 1df933d, 829aff9, aa0aba3, a71140f, 0276a27, 10d2c98, c86c999, 975f758, 2a3a478, 1e6e206, fa6027b, 7616f73, fc668b6, d425c8c, ce120f4, 248201f, e80d397, f1a941d, 73bc3a1, b628bb1, 6e3ae7b, e891247, 53e6c73, 1579d6f, d3c6b73, 4a59c6a, 145d8e1, f74282c, 0a38623, 0a08ae0, fa6a56b, d60c5d4, 07ffd25, 9b517ad, 604b1c1, ccc2e02, 14d810a, a9d1ee3, a31adbe, 7245f87, 3a0828b, cd83544, 6550a07, 6f090d4, b505c0d, 186dd49, 9f37e58, 4446451, 58be972, 1320075, ba53b64, f7490d4, c8ea602, 62d82f4, a2c9e7c, 97dd022, f984ee8, f1b2910, a29e05a, 1aa1d8b, bb99734, 222e7ca, 797c9e3, b4d5398, c8349ed, 8426e5f, 26e0085, 1a86166, 0856631, 0af0985, bc582c9, 2c63f1e, 0d98213, ca6f0dc, d8cc9ed, 42fd969, 5641ad3, 629870d, d592c14, 10d2c98, 53909a9, 78a4269, 06c6307, 0847c41, 5a77084, 96f99b3, 7bb8781, 4907e9b, 2a30248, 8364ddd, ad67d8c, ec0c087, fa6a56b, a2c1ce6, 91e9af0, 7bd3f34, ef16581, 1693a87, df3fc47, e11be41, 2e39e8b, 11c5ee7, 1742d2f, 8efc70e, 9642776, 6680828, c34edcb, 7b2c5bd, 1a2ccee, db995df, af0ccdd, addeaea, 7704034, e72b12f, 310dd9c, 1c2afc1, 44f44ca, f43b9d6, 56e72b3, 50ef80e, fc3b718, ee336d8, d12f922, 46d8310, 59812fd, f9d0dec, 4372c79, 81485ef, bd393d6, 0c95c04]:

  • effect@4.0.0-rc.113