#moonrockz/gleam-parser

• Scanner (tokenize): lossless tokens with trivia (whitespace, line breaks, //, /// and //// comments) that rebuild the source byte for byte; spans are UTF-8 byte offsets as in the compiler; a lexical error is a GL-SCAN-001 diagnostic with the compiler's message plus an Error token, and scanning goes on.
• AST and JSON: the compiler's untyped AST (v1.19.2) with a canonical JSON encoder and decoder (docs/ast-json.md); a Rust oracle on gleam-core writes the same JSON, and tests/fixtures/ast prove byte equality.

• Parser (parse_text, parse_module): a recursive-descent port of the compiler's parse.rs; on every fixture the canonical JSON equals the compiler's byte for byte, and the first error has the compiler's kind and span. A definition that does not parse is reported and skipped; the rest of the file is parsed.
• CST (ParseResult.cst): every token grouped per definition; ModuleCst::to_source() is the text byte for byte, also with errors.
• Formatter (format, format_parsed): gleam format's text (80 columns), from a port of the compiler's formatter on moonrockz/pretty; comments and empty lines are kept. Every one of the compiler's 893 formatter test cases and every file of the 233-file corpus formats to the binary's bytes (tests/format).

  inspect(
    @gleam.format("pub fn main(){\n  io.println(  \"hi\" )\n}\n"),
    content="pub fn main() {\n  io.println(\"hi\")\n}\n",
  )

• Dialect (Dialect::gleam_1_19()): what the parser accepts.
• Config (Config::default(), Config::load()): the dialect and the optional nesting limit (Unlimited by default), from a gleam-parser.toml file, GLEAM_PARSER_* variables and the caller, merged with with_fallback.
• Syntax tree (NodeRef, Tree, NodePath): a read-only node model over the AST with parent links, position lookup and seven traversal styles: walk, fold, reduce, Visitor, pull events, push events and TreeCursor; every traversal uses an explicit stack.
• Report (render_plain, render_terminal, encode_diagnostics): the compiler's diagnostic layout, with colours, or as JSON.
• Parity: on a pinned corpus of 25 published packages (233 files), all 233 files give the compiler's AST byte for byte (mise run test:parity); 44 rejection fixtures check the accept and reject verdicts and the first error against the compiler (mise run test:rejection).

test {
  let r = @gleam.parse_text("import gleam/io\n\npub fn main() {\n  io.println(\"hi\")\n}\n")
  inspect(r.module_.definitions.length(), content="2")
  inspect(r.has_errors(), content="false")
}

test {
  let tree = @gleam.Tree(@gleam.parse_text("pub fn add(a, b) {\n  a + b\n}\n"))
  // Byte 25 is the `b` in `a + b`.
  let b = tree.node_at(@gleam.ByteOffset::from_int(25).unwrap()).unwrap()
  inspect("\{b.category()}/\{b.kind()}", content="expression/var")
  inspect(tree.node_path(b).unwrap(), content="definitions[0].body[0].expression[0].right[0]")
}

• Printer and formatter: Gleam source from an AST, gleam format parity.

#Installation

moon add moonrockz/gleam-parser

import {
  "moonrockz/gleam-parser" @gleam,
}

#Development

• mise run check: info, lint, format and all tests.
• mise run test:targets: the library on wasm, wasm-gc, js and native.
• mise run test:parity and mise run test:rejection: parity with the Gleam compiler on the corpus and on the rejection fixtures (see AGENTS.md).
• See AGENTS.md for conventions and docs/cookbook/ for tested examples.

#Optional target profiles

test {
  let orbit = @gleam.Target::extension("orbit")
  let dialect = @gleam.Dialect::gleam_1_19().register_target(orbit)
  let config = { ..@gleam.Config::default(), dialect }
  let result = @gleam.parse_text("@target(orbit) pub const answer = 42", config~)
  assert_eq(result.has_errors(), false)
  let document = @gleam.decode_extended(@gleam.encode_extended(result))
  assert_eq(document.parsed, result.parsed())
}

let dialect = @gleam.Dialect::gleam_1_19().with_profile(@morphir.profile())
let config = { ..@gleam.Config::default(), dialect }
let result = @gleam.parse_text(
  "@external(morphir, \"example/domain\", \"CustomerId\")\npub type CustomerId\n",
  config~,
)