#moonrockz/wit

WIT language tooling for MoonBit.

Source positions, structured diagnostics, lossless lexical analysis, typed immutable ASTs/CSTs, recovering syntax parsing, immutable traversal, rewrites, canonical printing and comment-preserving formatting are implemented for the documented grammar profile. Install with moon add moonrockz/wit@0.1.0.

@scanner.scan(source) returns immutable tokens and diagnostics. Scan::to_source() reconstructs the exact input. Lexical success does not imply syntactic validity. See the scanner package for the lexical profile and diagnostic codes.

@parser.parse(source) returns a typed AST arena, a lossless CST arena and diagnostics. Parse::is_valid() checks for errors. Parsing performs no I/O, never raises, and uses explicit grammar continuations rather than recursive descent. See the parser package for supported syntax, proposal switches and remaining compatibility validation.

#Scope

Package declarations, interfaces, worlds, types, functions, resources, use/include declarations and explicitly selected proposal-gated constructs.

Syntax parsing is separate from semantic resolution. Cross-file/package loading, name resolution, registry access and binding generation belong in later semantic tooling. The syntax parser must not perform I/O.

#Packages

Package	Intended responsibility
span	Implemented: source text, UTF-16 positions, checked slices and span containment
diagnostic	Implemented: codes, severity, immutable source labels and recovery details
scanner	Implemented: lossless tokens, comments, whitespace and lexical diagnostics
ast	Implemented: typed immutable nodes, named child fields and structural validation
cst	Implemented: lossless grammar hierarchy, token ranges and subtree reconstruction
parser	Implemented: recovering grammar parser and explicit proposal switches
traversal	Implemented: AST/CST walks, folds, reductions, events, immutable cursors and rewrites
printer	Implemented: stack-safe canonical AST emission with structural round-trip checks
formatter	Implemented: CST-aware layout preserving comment text, token order and meaning

#Validation

The lexical layer has acceptance examples, 2,400 seeded property cases per target, exact token/span reconstruction laws and a 30,000-level nested comment stress case. Tests also exercise malformed UTF-16 and string escapes. All run on wasm, wasm-gc, js and native.

The parser adds 68 pinned specification examples and 321 upstream WIT files with a licensed fixture manifest, 1,900 seeded property cases, 20,000 nested types, 10,000 missing type closers and 2,000 nested packages per target. AST/CST validation checks postorder references, unique parents, child order and span containment.

Traversal adds 800 seeded equivalence/rewrite cases per target and a 16,000-level stress case for readers, cursors, reductions, AST rewrites and CST source edits.

Printing and formatting run all 389 pinned fixture sources, 1,000 seeded type/comment properties, 18,000 nested type constructors and 1,200 nested package blocks per target. Printer output preserves AST structure apart from locations and arena numbering. Formatter output retains non-whitespace tokens and raw comments and is idempotent. Malformed source returns diagnostics without partial formatted output.

@printer.print(tree) emits canonical source. @formatter.format(source) and format_cst(tree) normalize layout using the selected parser feature profile. Both accept indent_width, defaulting to two spaces, and return explicit errors. The printer supports all implemented proposal forms; the formatter accepts the same feature switches as parsing. Read each package for recovery and invalid-input behavior.

See root Beads epic wasm-devkit-p5m for implementation work.