v0.17.0

Added

• Persistent color alpha channel: Added alpha property to ColorSymbol (stored outside value channels, defaults to null). Alpha is now preserved across all color format conversions (RGB → HSL → Hex, etc.). All color types (RGB, HSL, Hex) now support alpha via .alpha property. CSS color formatting includes alpha as fourth parameter when set and < 1 (e.g., rgb(255, 0, 0, 0.5)).

Breaking Changes

• Color alpha property: Replace rgbaColor.value.a with rgbaColor.alpha
• Color alpha access: Use color.alpha property on any color type instead of RGBA-specific handling
• ColorSymbol constructor: Constructor signature changed from (value, subType?, config?, alpha?) to (value, subType?, alpha?, config?). Update any direct ColorSymbol constructor calls to use new parameter order.

What's Changed

• Add alpha property by @floscr in #88

Full Changelog: v0.16.1...v0.17.0

v0.16.1

Added

• Script mode for REPL: Two execution modes for interactive development

  • Inline mode (default): Executes each line immediately, like traditional REPLs
  • Script mode (--mode script): Accumulates lines into a multi-statement script for complex workflows
  • Smart output handling: Only prints assigned/reassigned values and expression results
  • New commands: exit() to quit, clear() to reset script buffer

• Interpreter

  • Public getSymbol() method on Interpreter for accessing symbol table values

What's Changed

• feat: adds script mode to repl by @floscr in #89

Full Changelog: v0.16.0...v0.16.1

v0.16.0

Changed

• BREAKING: Cleaned up processor output interface: Removed internal properties from public API

  • Removed unresolved property from ProcessorResult - internal tracking no longer exposed
  • Removed subFieldPaths property from ProcessorResult - internal tracking no longer exposed
  • Added getSubFieldPaths() method on PrefixResolver for internal use
  • Simplified ProcessorOutput by removing redundant filtering logic in builder
  • Users should rely on tokens Map for resolved values and issues Map for problems

• BREAKING: Unified error tracking in issues Map: Removed separate errors Map from processor output

  • ProcessorOutput no longer includes errors property - all errors now tracked in issues Map
  • Added getTokenError() helper to extract Error objects from issues
  • Added tokenHasError() helper to check if token has error vs lint issue
  • Errors in issues can be identified by presence of code property (vs lint issues)
  • Simplifies error handling with single source of truth for all token problems
  • Migration: Replace result.errors.get(token) with getTokenError(result.issues, token)
  • Migration: Replace result.errors.has(token) with tokenHasError(result.issues?.get(token))

What's Changed

• Remove errors from output by @floscr in #87

Full Changelog: v0.15.1...v0.16.0

v0.15.1

Fixed

• Phantom Token Resolution: Resolved type-safety issue in missing dependency tracking
  • Removed unsafe type casts (error as unknown as InterpreterResult) in referenceCache
  • Added dedicated missingDependencies set to track references to non-existent tokens
  • Ensured referenceCache only contains valid InterpreterResult values (ISymbolType | string | null)
  • Fixed circular dependency detection to correctly report CIRCULAR_DEPENDENCY instead of TOKEN_NOT_FOUND
  • Prevents phantom tokens (missing dependencies) from appearing in resolved output

What's Changed

• Fix phantom tokens in token output by @floscr in #86

Full Changelog: v0.15.0...v0.15.1

v0.15.0

Added

• CSS and Penpot Preset Validators: Pre-built validation rulesets for token types

  • css.createLintRunner(): Standards-compliant CSS validators (opacity, fontWeight, borderRadius, boxShadow, lineHeight, letterSpacing, textTransform, textDecoration)
  • penpot.createLintRunner(): Penpot-specific validators extending CSS rules (typography, shadow, strokeWidth)
  • Composable primitives: number(), string(), boolean(), color(), numberWithUnit()
  • Combinators: or(), all(), oneOrList(), list(), struct(), arrayOf()

• LintRunner Extension API: Composable rule system with .extend() method

  • Override specific validators while preserving others
  • Combine validators using all() combinator
  • Immutable API returns new instances

• Structured Token Field-Level Linting: Validation with field-level granularity

  • path property on LintIssue identifies specific fields (e.g., ["fontSize"] or [0, "blur"])
  • Supports Map-based (typography) and array-based (box-shadow) tokens
  • Cross-field validation (e.g., lineHeight requires fontSize)

• TokenSymbol Validator Support: Validators can use .get(), .keys(), .values(), .length() methods

• Helper Functions: getAffectedTokens(), getBrokenReferences(), getModifiedDependants(), getRenamedReferences()

Changed

• BREAKING: LintResult Type: Changed from object to Map<RefPath, LintIssue[]>

  • Before: result.lint.errors, result.lint.warnings, result.lint.hasErrors
  • After: result.issues (Map), filter by issue.severity

• BREAKING: LintIssue Structure:

  • Removed ruleId property (use code instead)
  • Changed tokenName type from string to RefPath
  • Added optional path property: (string | number)[]

• BREAKING: CreateIssueFn Signature: Now accepts full issue object instead of individual parameters

• BREAKING: Removed aggregateResults(): Work directly with Map<RefPath, LintIssue[]>

• ProcessorOutput: Renamed lintIssues to issues (now includes lint issues and language errors)

• Error Handling: DependencyError class replaced with createDependencyError() helper returning ProcessorError

• CRUD Result Types: Unified to TokenOperationResult with tokens, resolved, issues, dependants properties

• Circular dependency handling: Circular dependencies no longer throw errors but are handled gracefully

What's Changed

• Linter setup by @floscr in #85

Full Changelog: v0.14.0...v0.15.0

v0.14.0

Changed

• BREAKING: Builder output refactoring: Simplified builder architecture and separated string vs symbol outputs

  • Any output is now defined by the builder property
  • Migration: Replace { output: "string" } with { builder: new StringMapBuilder() } for string Map output
  • Migration: Replace { output: "symbols" } by removing the option (symbols are now the default)

• Enhanced CRUD operations with linter support: All TokenResolver CRUD methods now return lint results

  • createToken(), updateToken(), and deleteToken() now return lintIssues array in their results
  • Linter is preserved across CRUD operations when passed to buildTokens() or processTokens()
  • Enables real-time validation feedback during token editing workflows

Added

• buildTokens() export: Core token processing function now available in public API

  • Skips normalization overhead when you already have Map<string, TokenData>
  • Lower-level API for performance-critical applications

• Error class exports: All error classes now exported from main library entry point

  • New errorClasses export includes: LanguageError, LexerError, ParserError, InterpreterError, ProcessorError, DependencyError
  • Enables consumers to handle and throw interpreter errors properly
  • Useful for custom validators and error handling in applications

• isList, isDictionary type guard helpers: Added type checking utilities for List and Dictionary symbols

What's Changed

• Use library errors by @floscr in #82
• Crud refinement by @floscr in #84

Full Changelog: v0.13.3...v0.14.0

v0.13.3

• Exported linter

Full Changelog: v0.13.2...v0.13.3

v0.13.2

Added

• Exported linter

What's Changed

• Export linter in processor by @floscr in #83

Full Changelog: v0.13.1...v0.13.2

v0.13.1

Added

• Exported renameReferences utility function from library

What's Changed

• Use crud operations in runtime UI by @floscr in #81

Full Changelog: v0.13.0...v0.13.1

v0.13.0

Added

• Symbol toJs() method: All symbol types now have a toJs() method for converting to plain JavaScript values
  • Supports ToJsOptions with recursive (default: true) and stringify (default: false) options
  • NumberWithUnitSymbol: Returns { value: number | null, unit: string } by default, or string when stringify: true
  • ColorSymbol: Returns structured object with type field and component values by default
    • Hex colors: { type: "hex", value: "#ff0000" }
    • Dynamic colors: { type: "rgb", r: 255, g: 0, b: 0, ... }
    • String representation when stringify: true
  • ListSymbol, DictionarySymbol, TokenSymbol: Recursively convert nested values with recursive option
  • Value types extracted and reused: NumberValue, StringValue, BooleanValue, NumberWithUnitValue, ColorValue
  • Comprehensive test coverage across all symbol types

Changed

• Removed symbolTypeToJsValue() function: Developers should call symbol.toJs() directly instead
• Updated serializeInterpreterResult(): Now accepts ToJsOptions parameter for customization
• Builder serialization: Uses stringify: true option to maintain backward-compatible string output for NumberWithUnitSymbol and ColorSymbol

Added

• TokenResolver CRUD operations: Complete token lifecycle management with dependency tracking

  • createToken(): Add new tokens with automatic dependency graph updates
  • updateToken(): Modify existing tokens with options for renaming and reference updates
    • updateReferences option: Automatically updates all dependent tokens when renaming
    • Returns renamedReferences set tracking which tokens had references updated
    • Returns brokenReferences set when renaming without updateReferences
  • deleteToken(): Remove tokens with broken reference detection
    • Returns brokenReferences set identifying tokens with missing dependencies
  • All operations return affected tokens and dependency subgraphs for UI updates
  • Full integration test coverage for complex CRUD workflows

• AST utilities for reference management

  • walkAST(): Traverse AST nodes with visitor pattern
  • filterAST(): Collect matching nodes into flat array
  • collectReferenceNodes(): Extract all reference nodes, optionally filtered by name
  • renameReferences(): Safely rename token references in expressions using AST
  • Token positions (pos, endPos) now tracked on all tokens for precise text manipulation

• Enhanced token position tracking

  • Lexer now records start and end positions for all tokens
  • Enables precise source code manipulation for renames and updates
  • Reference tokens include exact character ranges in original string

Changed

• BREAKING: Removed @tokenscript/stencil-components package

  • Stencil web components moved out of monorepo to reduce dependencies
  • Removed from CI/CD pipeline and build scripts
  • Examples no longer depend on components package
  • Runtime UI example now uses native form implementation

• Enhanced error codes

  • Added ProcessorErrorCode.TOKEN_ALREADY_EXISTS for duplicate token detection
  • Improved error messages with structured data for better debugging

Fixed

• Build configuration preserves class names in production bundles

  • Added keepNames: true to tsup build configs to prevent Symbol class name mangling
  • Ensures runtime type checking and debugging remain accurate in minified builds

• Test suite cleanup

  • Removed debug console.log statements from reference method tests

Removed

• Stencil Components package (packages/stencil-components/)
  • <token-form> web component and related infrastructure
  • Stencil build configuration and dependencies
  • CI/CD integration for component builds and tests
  • Example integrations using Stencil components

What's Changed

• Add token crud operations by @floscr in #75
• Fix vercel build by @floscr in #76
• DevExp Fixes by @floscr in #77
• Add .toJs interface method by @floscr in #78
• Add release script by @floscr in #79
• Prevent library rebuild on CI by @floscr in #80

Full Changelog: v0.12.1...v0.13.0