# @systemfsoftware/effect-gherkin-spec ## 7.0.1 ### Patch Changes - Updated dependencies: - @systemfsoftware/vitest@2.0.0 ## 7.0.0 ### Major Changes - Require `effect` `^4` (stable 4.0.0). Release candidates are no longer supported: imports follow 4.0.0's module layout (`effect/http`, `effect/rpc`, `effect/http-api`, `effect/cluster`, `effect/process`, `effect/socket`, `effect/persistence`, `effect/reactivity`, `effect/encoding`, `effect/Arbitrary`), which does not exist in any 4.0.0 RC. ### Patch Changes - Updated dependencies: - @systemfsoftware/effect-spec-runtime@1.0.0 - @systemfsoftware/vitest@1.0.0 ## 6.0.1 ### Patch Changes - Updated dependencies: - @systemfsoftware/vitest@0.3.0 ## 6.0.0 ### Major Changes - The lawful test runner is now a peer dependency under its own name, `@systemfsoftware/vitest`, instead of the `@effect/vitest` alias that pointed at it. Install `@systemfsoftware/vitest` and import `it`, `layer`, `expect`, and the rest from `@systemfsoftware/vitest`. `@systemfsoftware/effect-gherkin-spec` re-exports `it` and `layer` from `@systemfsoftware/vitest`. - Live cases now declare a required reason, and every other scenario runs on the simulation kernel under the profile's seeded schedules. - `Feature('...').liveClock()` → `Feature('...').live('the reason it waits on real I/O')`, at every builder stage. - `FeatureLayerOptions` is gone: `withLayer(shared, opts)` → `withLayer(shared)`. - `makeFeature` takes bindings with only `it`: `makeFeature({ it, layer })` → `makeFeature({ it })`. - One scenario or outline stays live through its own options: `scenario('...', { live: 'why it needs the real clock' }, pipeline)`, where `live` is a reason string, not a boolean. The package re-exports only `it` and `layer` from `@effect/vitest`; import anything else from `@effect/vitest` directly. ### Minor Changes - `StepError` now has a message naming the keyword, the resolved step text and a summary of the cause, for example `Given "a user logs in" failed: Unauthorized {"user":"bob"}`, and its stack leads with the spec line that wrote the step. Steps are recorded as spans, so a failing scenario reports which steps passed and which failed. - `Then`, `And` and `But` bodies take `(state, expect)` and return exactly one check (or an Effect of one); `expect` is no longer imported from `@effect/vitest`. Only `Given` and `When` open a new observed state, so a `Then` followed by an `And` or `But` on the same state is refused: merge the facts into one `Then` over a record. `Then.soft`, `Then.poll` and `When.poll`, with their `And`/`But` forms, are removed, as are `checkSoftFailures`, `owned` and `recordAssertion`. A background pipeline may carry the check service. ### Patch Changes - Update peer and runtime dependency on `effect` and companion packages to `4.0.0-rc.117`. - Updated dependencies: - @systemfsoftware/vitest@0.2.0 ## 5.0.0 ### Minor Changes - `Then`, `And` and `But` bodies take `(state, expect)` and return exactly one check (or an Effect of one); `expect` is no longer imported from `@effect/vitest`. Only `Given` and `When` open a new observed state, so a `Then` followed by an `And` or `But` on the same state is refused: merge the facts into one `Then` over a record. `Then.soft`, `Then.poll` and `When.poll`, with their `And`/`But` forms, are removed, as are `checkSoftFailures`, `owned` and `recordAssertion`. A background pipeline may carry the check service. ### Patch Changes - Updated dependencies: - @systemfsoftware/vitest@0.1.0 ## 4.4.0 ### Minor Changes - Requires `@systemfsoftware/vitest` installed under the `@effect/vitest` name (`"@effect/vitest": "npm:@systemfsoftware/vitest"`) in place of the upstream package. A scenario whose steps all pass counts as asserted, a failing `expect` inside a step fails that step, and a feature's layer is built once and shared by its scenarios. ### Patch Changes - Updated dependencies: - @systemfsoftware/effect-spec-runtime@0.2.0 ## 4.3.1 ### Patch Changes - Update peer and runtime dependency on `effect` and companion packages to `4.0.0-rc.117`. ## 4.3.0 ### Minor Changes - Feature runtime, outline expansion, do-notation, case-layer and pairwise helpers that took their subject first can now also be called data-last. Existing calls keep working. ### Patch Changes - Updated dependencies: - @systemfsoftware/effect-spec-runtime@0.1.0 ## 4.2.3 ### Patch Changes - `VitestTaskRef` is now the task reference owned by `@systemfsoftware/effect-spec-runtime`, so its context key changed from `@systemfsoftware/effect-gherkin-spec/VitestTask` to `@systemfsoftware/effect-spec-runtime/VitestTask`. Code that imports `VitestTaskRef` from this package is unaffected; code that re-declared a reference under the old key no longer receives the running test's task. Every exported type, the stage vocabulary, and every other runtime behaviour are unchanged. ## 4.2.2 ### Patch Changes - Unconstrained type holes are defaulted type parameters (``) instead of a type argument `unknown`. Calls that omit those arguments are unchanged. ## 4.2.1 ### Patch Changes - Bump the stryker catalog to pull in the latest patch releases of the stryker-js ecosystem. ## 4.2.0 ### Minor Changes - Add compile-time BDD stage validation, Then.soft composite assertions, polling step combinators, Effect equality tester integration, and Vitest task reporting ## 4.1.0 ### Minor Changes - These packages now peer-depend on Vitest 5. Install `vitest` ^5; Vitest 4 is unsupported. ## 4.0.2 ### Patch Changes - Releases the workspace so its published versions track the shared dependency graph this change moves. - The `@systemfsoftware/source` export condition is gone. It resolved to a package's TypeScript sources for editors and in-repo typechecks; each package now exports only its built entry. If your tsconfig sets `customConditions: ["@systemfsoftware/source"]`, or a bundler config names that condition, remove it — resolution falls back to the built entry, which is what every consumer already got. ## 4.0.1 ### Patch Changes - Peer Effect requirement advances to 4.0.0-rc.112. No API changes. ## 4.0.0 ### Major Changes - AtomRuntime.fn now infers precise success and error types from the function you pass, in both direct and curried forms. Values that are not functions are rejected at compile time; previously they failed when the resulting atom was first read. withScope now requires scope bindings to be effects with no service requirements. Bindings that require services no longer compile — provide those services with withLayer and read them inside steps instead. onToolResult now fails with PlatformError instead of an untyped error. ### Patch Changes - Re-published against the oxc-based instrumenter: workspace dependency ranges move to the new instrumenter major; no package's own behavior changes in this release. ## 3.0.0 ### Major Changes - Scenario titles are now enforced as prose at the type level: `scenario` and `scenarioOutline` reject any title that either starts with "Should" (the `Should_[Behavior]_When_[Condition]` unit-test naming convention) or lacks an ASCII space between words (every concatenated-token shape — PascalCase, snake_case, CamelCase — that reads as a unit-test name rather than prose). A rejected title fails to type-check with the rule in the diagnostic. Titles that are not string literals are unaffected. ## 2.0.2 ### Patch Changes - Peer Effect requirement advances to 4.0.0-rc.112. No API changes. ## 2.0.1 ### Patch Changes - The peer requirements for `effect` and for the Effect test-runner integration now accept any compatible `4.0.0-rc` release, instead of demanding one exact release candidate. Installing alongside a newer release candidate no longer reports an unmet peer dependency or resolves a second copy of `effect` into the dependency tree. ## 2.0.0 ### Major Changes - Collapse projection tags into the ports they projected A `*ExecutorDeps` tag whose service type was indexed off other ports recorded which members one operation happened to reach for. Exporting it turned internal composition into a surface commitment: a consumer had to discover and provide an aggregator for each operation, where the port it came from already served. The tags rode the `R` channel of exported signatures, so a consumer met them only at their own call site. Each was removed the way its evidence directed — no tag was renamed, and none was replaced: - `AttwCliExecutorDeps`, `CheckPackageExecutorDeps` — deleted. Zero consumers, no Live layer; callers already required the real capabilities directly. - `HookDispatcherExecutorDeps` — was a 1:1 alias of `Scope.Scope`. Requiring `Scope` names the same service. - `InjectInstructionsExecutorDeps` — packed `FileSystem`, `Path` and `TomlLoader` with no logic. Consumers require the three ports. - `EffectVitestDeps` → `EffectVitestBindings`. Never a `Context.Tag`: a plain type alias already passed as an ordinary parameter, so only the name misdescribed it. A stale api report that no `api-extractor` regenerates was deleted with it. A consumer providing a port instead of an aggregator provides no more than before, since a service `{ a, b }` is assignable to a requirement `{ a }`. ### Minor Changes - cut over to effect v4 (4.0.0-rc.108): public surface derives from effect types; peers flip effect ^3→^4 ### Patch Changes - Array types are spelled one way. `Array` and `ReadonlyArray` in emitted declarations become `T[]` and `readonly T[]`, which the type checker cannot tell apart: no exported type changes, only how it is written. - New version is published through npm trusted publishing, so it carries a provenance attestation you can verify. - Fix two `test:types` scripts that could not pass. `tstyche` exits 1 with "No test files were selected using current configuration" when nothing matches `testFileMatch`, so a watcher pointed at files that do not exist is a script guaranteed to fail the moment anyone runs it. Measured directly: `effect-gherkin-spec`'s `tstyche.json` matched `src/**/*.tst.ts` and the package holds no such file, so its `test:types` script is removed rather than left as a gate attached to nothing. `stryker-plugins` now pins `TSTYCHE_TYPESCRIPT_MODULE` so its own `test:types` resolves the TypeScript module it needs. Both packages publish, and both scripts ship in the published `package.json`, so a consumer inspecting or running them sees the change — a real patch rather than an empty intent. ## 1.0.0 ### Major Changes - Collapse projection tags into the ports they projected A `*ExecutorDeps` tag whose service type was indexed off other ports recorded which members one operation happened to reach for. Exporting it turned internal composition into a surface commitment: a consumer had to discover and provide an aggregator for each operation, where the port it came from already served. The tags rode the `R` channel of exported signatures, so a consumer met them only at their own call site. Each was removed the way its evidence directed — no tag was renamed, and none was replaced: - `AttwCliExecutorDeps`, `CheckPackageExecutorDeps` — deleted. Zero consumers, no Live layer; callers already required the real capabilities directly. - `HookDispatcherExecutorDeps` — was a 1:1 alias of `Scope.Scope`. Requiring `Scope` names the same service. - `InjectInstructionsExecutorDeps` — packed `FileSystem`, `Path` and `TomlLoader` with no logic. Consumers require the three ports. - `EffectVitestDeps` → `EffectVitestBindings`. Never a `Context.Tag`: a plain type alias already passed as an ordinary parameter, so only the name misdescribed it. A stale api report that no `api-extractor` regenerates was deleted with it. A consumer providing a port instead of an aggregator provides no more than before, since a service `{ a, b }` is assignable to a requirement `{ a }`. ### Minor Changes - cut over to effect v4 (4.0.0-rc.108): public surface derives from effect types; peers flip effect ^3→^4 ### Patch Changes - Array types are spelled one way. `Array` and `ReadonlyArray` in emitted declarations become `T[]` and `readonly T[]`, which the type checker cannot tell apart: no exported type changes, only how it is written. - New version is published through npm trusted publishing, so it carries a provenance attestation you can verify. - Fix two `test:types` scripts that could not pass. `tstyche` exits 1 with "No test files were selected using current configuration" when nothing matches `testFileMatch`, so a watcher pointed at files that do not exist is a script guaranteed to fail the moment anyone runs it. Measured directly: `effect-gherkin-spec`'s `tstyche.json` matched `src/**/*.tst.ts` and the package holds no such file, so its `test:types` script is removed rather than left as a gate attached to nothing. `stryker-plugins` now pins `TSTYCHE_TYPESCRIPT_MODULE` so its own `test:types` resolves the TypeScript module it needs. Both packages publish, and both scripts ship in the published `package.json`, so a consumer inspecting or running them sees the change — a real patch rather than an empty intent.