@hyperfixi/types-browser 4.6.0 → 4.8.0

package/CHANGELOG.md

@@ -7,6 +7,124 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [4.8.0] - 2026-10-16
+
+### Added
+
+- **`@lokascript/intent` / `@lokascript/framework`: a role's word list, `RoleSpec.words`** (domains
+  rethink Phase 5). Per language, with `'*'` for every language; the core engine (`engine: 'core'`)
+  reads such a slot only from one of its words, compared lower-cased, so a word the grammar does not
+  know is left unread and the input refused, instead of read as the value. It is for a domain's fixed
+  words: behaviorspec's `when clicks zzqx` read actor `clicks`, action `zzqx`, and compiled to a
+  comment. The schema-pattern generator carries the pattern language's list on the role token
+  (`RolePatternToken.words`); a grammar that lists no words generates the same patterns. The classic
+  engine and semantic ignore it.
+- **`@lokascript/framework/testing/gate`: `GateFixture.commands`**, commands the gate writes besides
+  the generated ones, where one role's value depends on another's (behaviorspec's
+  `[given subject:#box value:visible]` beside `[given subject:viewport value:1024x768]`).
+- **`@lokascript/framework`: `UnreadInputError.hint`.** Where the last word a parse read as a value is
+  followed by another bare word, the refusal adds `if "buy milk" is one value, quote it` (todo's
+  `add buy milk to groceries`, whose item is quoted).
+
+### Changed
+
+- **`@lokascript/framework/testing/gate`: leftover input read into a role is `pass`.** An unknown word
+  after a command used to be `silent` wherever the parse accepted it. It is `pass` when the parse is
+  the text's own with the word in one role (an open-text role extended by it, sql's
+  `where active zzqx`, or a role the text left empty, todo's `list zzqx`) and the word reaches the
+  compiled code; `refused` when the input stops compiling; `silent` otherwise (read as something
+  else, or the word goes nowhere). `readsWord` is exported. On the domains (domains 4.1.0, this
+  framework packed): 74 leftover cases accepted → read, 0 worse, 0 parses moved.
+- **`@lokascript/framework`: the stock reader's optional slot does not read a marker the rest of the
+  pattern owes**, through its optional groups, as a required slot already does not read the marker
+  next to it (behaviorspec's `when user types into #search`, whose typed text is optional and has no
+  marker). No domain parse moves.
+
+## [4.7.0] - 2026-10-16
+
+### Added
+
+- **`@lokascript/framework`: the matcher's skeleton, `MatchSkeleton`** (domains rethink Phase 4.3), with
+  its seam (`ValueReader`, `MatchView`, `RoleSlot`) and the `ConfidenceModel` / `ConfidenceContext` /
+  `ConfidenceBreakdown` interfaces, all exported. It is the skeleton semantic's hyperscript matcher runs:
+  `@lokascript/semantic`'s `PatternMatcher` now extends it and passes `HyperscriptValueReader`, and its
+  confidence types are framework's narrowed to hyperscript's names. Framework's pattern types take the
+  fields semantic's carried alone (`LanguagePattern.parseOnly`, `LiteralPatternToken.renderSuppress`,
+  `RolePatternToken.valueShape`, `GroupPatternToken.renderRequired`, `ExtractionRule.valueIsExpression`,
+  `PatternMatchResult.roleStarts` / `eventSource`). `createMultilingualDSL` still uses its own matcher.
+  Every hyperscript parse is unchanged (27,194 checked); semantic's browser bundles grew ~150 bytes
+  gzipped.
+- **`@lokascript/framework`: `createSchemaPatternGenerator`** (domains rethink Phase 4.4), the pattern
+  generator semantic's hyperscript parser runs, moved from `@lokascript/semantic`: the main pattern per
+  command and language and its variants (simple, omit-role, widened-type, required-role, SOV verb-first,
+  value-prefix-literal), with the grammar's hook `positionOrder` (`'ascending'`: hyperscript's 1-based
+  positions; `'descending'`: a domain's, as framework's `generatePattern` reads them). Also exported:
+  `withoutVariantOnlyRoles`, `resolveMarkerForRole`, `schemaMarkerAlternatives`,
+  `legacyMarkerAlternatives`. semantic's `generatePattern` / `generatePatternVariants` / … keep their
+  names and signatures over it; its fused event-handler patterns stay in semantic. Every generated
+  pattern in all 24 languages is byte-identical (5,899), and every parse unchanged (27,194).
+- **`@lokascript/framework`: rendering through patterns, and the translation-loss checks** (domains
+  rethink Phase 4.5), moved from `@lokascript/semantic`. `selectRenderPattern` scores a language's
+  patterns for a node (role coverage, values a pattern pins) and `renderPattern` walks the chosen one;
+  a grammar's preferences and its value writing are `PatternRenderHooks`. `findTranslationLoss` runs
+  the fail-loud checks in order (unread input, forms nothing writes, a read-back that does not read or
+  differs, verbatim values) with a grammar's `TranslationLossRules`; `LossyTranslationError`,
+  `TranslationLoss` and `unconsumedSpans` moved with it (semantic re-exports them, so `instanceof`
+  holds). semantic's renderer and `translate` run them with hyperscript's hooks and rules; its
+  collision rule (`verified-render`) stays in semantic. Every translation of every command-shape case
+  in all 24 languages is unchanged, rendered (27,240) and with refusals on (704 refused, the same
+  kinds and losses).
+- **`@lokascript/framework/testing/gate`: the round-trip gate kit and the grammar linter** (domains
+  rethink Phase 4.6), moved from lokascript-domains: `runRoundTrip` renders every command of a
+  grammar in each of its languages and reads it back (`pass` / `refused` / `silent`, plus each render
+  with an unknown word after it, which must be refused); `compare` / `listingOf` / `countsOf` keep a
+  shrink-only baseline and `snapshotOf` / `snapshotChanges` a parse snapshot; `lintDomain` runs the
+  domain-toolkit rules R1–R10, its schema-structure rule on the validator the input passes
+  (`validateSchema`; framework cannot import semantic). No test runner needed. On the 8 domains, with
+  their own fixtures, it reads 3,488 of 3,488 cases as the domains' gate does, with no drift from
+  their baseline.
+- **`@lokascript/framework`: the stock value reader and confidence model for domain grammars** (domains
+  rethink Phase 4.7a). `createStockReader()` (`StockValueReader`) is a `ValueReader` for the matcher's
+  skeleton: a greedy slot reads up to the next marker its pattern owes (`readGreedy`,
+  `greedyStopMarkers`), any other slot one value (`readRoleValue`: a selector, a quoted string, a number,
+  a boolean, a duration, a URL, an identifier, or a dotted or called form, checked against the slot's
+  types), a required slot never takes the marker its pattern wants next, and an optional one may be
+  skipped for it. `stockConfidenceModel` scores as the old matcher did. Both are the old matcher's code,
+  moved: `createMultilingualDSL`'s matcher calls them, and the domain gate on this build moves no case
+  and no parse. `createSchemaPatternGenerator` also takes a domain schema's `greedy`, its
+  post-verb `sovSlot` and its `markerPosition` (hyperscript's patterns: unchanged, 5,899).
+- **`@lokascript/framework`: `engine: 'core'`, a domain DSL on the core** (domains rethink Phase 4.7b):
+  a `DSLConfig` option. `createMultilingualDSL({ ...config, engine: 'core' })` returns the same
+  `MultilingualDSL`, read by the core semantic's hyperscript parser runs: patterns from
+  `createSchemaPatternGenerator` (positions descending), matched by `MatchSkeleton` with the stock
+  reader and confidence model, and rendered through those patterns where the domain's own renderer
+  declines. Parsing's whole-input rule, the refusals, translation's read-back, compile, explicit syntax
+  and extensions are shared with the default, `engine: 'classic'`, which is unchanged; any other engine
+  name throws. The stock reader also gives a required slot back the value an optional, marker-less slot
+  just before it read (the old matcher's backtrack). Measured on the 8 domains, each on the core
+  (the domain gate on a packed build): 0 cases worse, 30 better (behaviorspec 306 → 336
+  round trips), no case silent that was not before.
+
+### Changed
+
+- **`@lokascript/intent`: one schema type for every grammar** (domains rethink Phase 4.1). `RoleSpec` and
+  `CommandSchema` now hold the fields hyperscript's schemas use as well as the domains' (`valueShape`,
+  `variantOnly`, `markerOptional`, `markerLegacy`, `valuePrefixLiteral`; `bareKeyword`, `structuralOnly`
+  and the `omitRoleVariants` / `widenTypeVariants` / `rolePrefixLiteralVariants` / `requiredRoleVariants`
+  lists), each documented with the code that reads it, and take type parameters for a grammar's action,
+  role, value and category names. The parameters default to today's types, so a domain schema reads as
+  before; framework's `docs/API_SURFACE.md` prints them spelled out. `@lokascript/semantic`'s `RoleSpec`
+  is this type with hyperscript's names, and its `CommandSchema` extends it with the fields only its AST
+  builder and error docs read (`ast`, `errorCodes`, `preconditions`, `recoveryHints`). One field keeps two
+  meanings for now: framework's generator sorts `svoPosition` / `sovPosition` descending (higher =
+  earlier), semantic's ascending; the field doc says so.
+- **`@lokascript/semantic`: the pattern matcher reads every value through a seam** (domains rethink
+  Phase 4.2). The matcher is now a grammar-neutral skeleton (`parser/match-skeleton.ts`: literals,
+  groups and their rollback, the speculative skip of an optional slot, spans, ranking, confidence) and
+  hyperscript's `HyperscriptValueReader`, which reads each role slot and answers the skeleton's grammar
+  hooks (`parser/value-reader.ts`). `PatternMatcher` keeps its constructor, methods and import path. Every
+  parse is unchanged (27,194 checked).
+
 ## [4.6.0] - 2026-10-15
 
 ### Added
@@ -1736,7 +1854,9 @@ _Synchronized version release. See git history for details._
 - npm access token stored in GitHub Secrets
 - 2FA recommended for npm organization
 
-[Unreleased]: https://github.com/codetalcott/hyperfixi/compare/v4.6.0...HEAD
+[Unreleased]: https://github.com/codetalcott/hyperfixi/compare/v4.8.0...HEAD
+[4.8.0]: https://github.com/codetalcott/hyperfixi/compare/v4.7.0...v4.8.0
+[4.7.0]: https://github.com/codetalcott/hyperfixi/compare/v4.6.0...v4.7.0
 [4.6.0]: https://github.com/codetalcott/hyperfixi/compare/v4.5.0...v4.6.0
 [4.5.0]: https://github.com/codetalcott/hyperfixi/compare/v4.4.0...v4.5.0
 [4.4.0]: https://github.com/codetalcott/hyperfixi/compare/v4.3.0...v4.4.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hyperfixi/types-browser",
-  "version": "4.6.0",
+  "version": "4.8.0",
   "description": "TypeScript definitions for the hyperfixi browser globals: window.hyperfixi / window._hyperscript (the engine), LokaScriptSemantic, LokaScriptI18n",
   "type": "module",
   "types": "dist/index.d.ts",
@@ -30,7 +30,7 @@
   "author": "LokaScript Contributors",
   "license": "MIT",
   "devDependencies": {
-    "@hyperfixi/engine": "^4.6.0",
+    "@hyperfixi/engine": "^4.8.0",
     "typescript": "^5.8.3"
   },
   "publishConfig": {