Skip to content

v0.2.4 — trivia foundation + error-message DSL + docs

Choose a tag to compare

View all tags
@siy siy released this 22 Apr 01:36
· 455 commits to main since this release
v0.2.4
f32023f
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Three-part release.

Trivia attribution foundation

Intra-sequence trivia now attaches to the following sibling's leadingTrivia rather than being dropped. Threaded through the engine and generator via a pendingLeadingTrivia field with save/restore across backtracking combinators (Choice / Optional / And / Not / ZeroOrMore / OneOrMore / Repetition).

Deferred: full source round-trip (RoundTripTest 22/22) still requires rule-exit pos rewind, which would shift NonTerminal span ends and regenerate CstHash baselines. Intentionally out of scope for this release; scheduled for a subsequent patch. RoundTripTest remains @Disabled. See docs/TRIVIA-ATTRIBUTION.md.

Error-message DSL

Four new grammar directives, all optional:

  • %expected "label" (rule-level) — semantic expected-label for failure diagnostics.
  • %recover <terminator> (rule-level) — ADVANCED-recovery point override.
  • %suggest <RuleName> (grammar-level) — Levenshtein-distance hints from a designated vocabulary.
  • %tag "tag" (rule-level) — machine-readable tag on emitted diagnostics.

Diagnostic#tag() field added. Built-in tags: error.unexpected-input, error.expected, error.unclosed.

Suggestion vocabulary is cached at ParsingContext construction for forward compatibility with 0.3.1 incremental parsing.

Documentation pass

  • docs/TRIVIA-ATTRIBUTION.md — attribution rule + deferred round-trip.
  • docs/GRAMMAR-DSL.md — cut-operator edge cases + %expected / %recover / %suggest / %tag.
  • docs/PERF-FLAGS.md — tabular reference for all 10 ParserConfig perf flags from 0.2.2.
  • docs/BENCHMARKING.md — JMH harness usage and extension guide.
  • docs/ERROR_RECOVERY.md expanded with diagnostic tags and %recover override.
  • README cross-linked to all new reference docs.

Breaking changes

None. Grammars without the new directives produce byte-identical generator output (GeneratorFlagInertnessTest 3/3 green). ParserConfig API unchanged. All ten corpus-parity suites stay 22/22 against unchanged baselines.

Tests

587 → 601 passing, 1 skipped (RoundTripTest), 0 failures, 0 errors. New suites: SemanticExpectedTest (3), RuleRecoveryTest (3), SuggestionTest (3), DiagnosticTagTest (5).

Assets 2
Loading