@bcts/dcbor-parse 1.0.0-alpha.23 → 1.0.0-beta.0

package/dist/index.cjs

@@ -224,7 +224,8 @@ function formatMessage(message, source, range) {
 		lineNumber++;
 		lineStart = idx + 1;
 	}
-	const line = source.split("\n")[lineNumber - 1] ?? "";
+	let line = source.split("\n")[lineNumber - 1] ?? "";
+	if (line.endsWith("\r")) line = line.slice(0, -1);
 	const column = Math.max(0, start - lineStart);
 	const underlineLen = Math.max(1, end - start);
 	const caret = " ".repeat(column) + "^".repeat(underlineLen);
@@ -312,8 +313,83 @@ function unwrapErr(result) {
 	if (!result.ok) return result.error;
 	throw new Error("Called unwrapErr on an Ok result");
 }
+/**
+* Renders a {@link Token} the way Rust's
+* `#[derive(Debug)]` on the corresponding enum variant would:
+*
+* - Variant-only tokens (`BraceOpen`, `Comma`, `Null`, `Unit`, `NaN`,
+*   …) print as the bare variant name.
+* - Variant-with-value tokens print as `Variant(value)` where `value`
+*   uses Rust's `Debug` form for the payload type:
+*   `Bool(true)`, `Number(3.14)`, `String("foo")` (with the inner
+*   double quotes preserved — TS keeps them on the slice anyway),
+*   `TagValue(1234)`, `KnownValueNumber(42)`, `TagName("date")`,
+*   `KnownValueName("isA")`, `DateLiteral(2023-02-08T15:30:45.000Z)`,
+*   etc.
+*
+* Mirrors Rust's `Error::UnexpectedToken(Box<Token>, Span)` formatter
+* `#[error("Unexpected token {0:?}")]` so error messages stay
+* byte-identical to Rust.
+*/
 function tokenDebugString(token) {
-	return JSON.stringify(token);
+	switch (token.type) {
+		case "Bool": return `Bool(${token.value ? "true" : "false"})`;
+		case "BraceOpen": return "BraceOpen";
+		case "BraceClose": return "BraceClose";
+		case "BracketOpen": return "BracketOpen";
+		case "BracketClose": return "BracketClose";
+		case "ParenthesisOpen": return "ParenthesisOpen";
+		case "ParenthesisClose": return "ParenthesisClose";
+		case "Colon": return "Colon";
+		case "Comma": return "Comma";
+		case "Null": return "Null";
+		case "NaN": return "NaN";
+		case "Infinity": return "Infinity";
+		case "NegInfinity": return "NegInfinity";
+		case "Unit": return "Unit";
+		case "ByteStringHex": return `ByteStringHex(Ok(${formatBytesDebug(token.value)}))`;
+		case "ByteStringBase64": return `ByteStringBase64(Ok(${formatBytesDebug(token.value)}))`;
+		case "DateLiteral": return `DateLiteral(Ok(${String(token.value)}))`;
+		case "Number": return `Number(${formatNumberDebug(token.value)})`;
+		case "String": return `String(${JSON.stringify(token.value)})`;
+		case "TagValue": return `TagValue(Ok(${tagOrKnownValueDebug(token.value)}))`;
+		case "TagName": return `TagName(${JSON.stringify(token.value)})`;
+		case "KnownValueNumber": return `KnownValueNumber(Ok(${tagOrKnownValueDebug(token.value)}))`;
+		case "KnownValueName": return `KnownValueName(${JSON.stringify(token.value)})`;
+		case "UR": return `UR(Ok(${token.value.string()}))`;
+	}
+}
+/**
+* Renders a `Vec<u8>` the way Rust's `Debug` does:
+* `[0x68, 0x65, 0x6c, 0x6c, 0x6f]`.
+*/
+function formatBytesDebug(bytes) {
+	const parts = [];
+	for (const b of bytes) parts.push(`0x${b.toString(16).padStart(2, "0")}`);
+	return `[${parts.join(", ")}]`;
+}
+/**
+* Renders a JS `number` the way Rust's `f64::Debug` typically prints
+* it — using a decimal point even for integral values (e.g. `42.0`),
+* and `inf` / `-inf` / `NaN` for non-finite numbers. The dCBOR-parse
+* Rust source rarely produces a `Number` token in error messages
+* (numbers normally land in tagged-content contexts), but we still
+* mirror the convention so any error text is consistent with Rust.
+*/
+function formatNumberDebug(n) {
+	if (Number.isNaN(n)) return "NaN";
+	if (!Number.isFinite(n)) return n > 0 ? "inf" : "-inf";
+	if (Number.isInteger(n)) return `${n}.0`;
+	return String(n);
+}
+/**
+* Renders a `u64` payload the way Rust's `Debug` does — a bare digit
+* sequence without trailing `n` for `bigint` values. Mirrors
+* `<u64 as Debug>::fmt` and `<TagValue as Debug>::fmt` (TagValue is a
+* type alias for u64 in `bc-ur` / `dcbor`).
+*/
+function tagOrKnownValueDebug(value) {
+	return typeof value === "bigint" ? value.toString() : String(value);
 }
 
 //#endregion
@@ -487,11 +563,14 @@ var Lexer = class {
 				this._position++;
 				continue;
 			}
-			if (ch === "/" && this._position + 1 < this._source.length && this._source[this._position + 1] !== "/") {
-				this._position++;
-				while (this._position < this._source.length && this._source[this._position] !== "/") this._position++;
-				if (this._position < this._source.length) this._position++;
-				continue;
+			if (ch === "/") {
+				let scan = this._position + 1;
+				while (scan < this._source.length && this._source[scan] !== "/") scan++;
+				if (scan < this._source.length) {
+					this._position = scan + 1;
+					continue;
+				}
+				break;
 			}
 			if (ch === "#") {
 				while (this._position < this._source.length && this._source[this._position] !== "\n") this._position++;
@@ -500,23 +579,33 @@ var Lexer = class {
 			break;
 		}
 	}
+	/**
+	* Matches reserved keywords: `true`, `false`, `null`, `NaN`,
+	* `Infinity`, `-Infinity`, `Unit`.
+	*
+	* Mirrors Rust's `Logos` `#[token(...)]` matcher
+	* (`bc-dcbor-parse-rust/src/token.rs:12-50, 164`), which is greedy
+	* and emits the keyword token *as soon as the literal matches* —
+	* subsequent characters become a separate (likely unrecognized) token
+	* stream. So input like `truex` lexes as `Bool(true)` followed by an
+	* unrecognized run on `x`. Earlier revisions of this port enforced an
+	* identifier boundary check (`!_isIdentifierChar(nextChar)`) and
+	* rejected the whole prefix as a single `UnrecognizedToken`, which
+	* broke span/variant parity with Rust.
+	*/
 	_tryMatchKeyword() {
 		const keywords = [
+			["-Infinity", token.negInfinity()],
 			["true", token.bool(true)],
 			["false", token.bool(false)],
 			["null", token.null()],
 			["NaN", token.nan()],
 			["Infinity", token.infinity()],
-			["-Infinity", token.negInfinity()],
 			["Unit", token.unit()]
 		];
 		for (const [keyword, tok] of keywords) if (this._matchLiteral(keyword)) {
-			const nextChar = this._source[this._position];
-			if (nextChar === void 0 || !this._isIdentifierChar(nextChar)) {
-				this._tokenEnd = this._position;
-				return ok(tok);
-			}
-			this._position = this._tokenStart;
+			this._tokenEnd = this._position;
+			return ok(tok);
 		}
 	}
 	_tryMatchDateLiteral() {
@@ -545,9 +634,9 @@ var Lexer = class {
 			if (this._source[this._position + numStr.length] === "(" && !numStr.includes(".") && !numStr.includes("e") && !numStr.includes("E") && !numStr.startsWith("-")) {
 				this._position += numStr.length + 1;
 				this._tokenEnd = this._position;
-				const tagValue = parseInt(numStr, 10);
-				if (!Number.isSafeInteger(tagValue) || tagValue < 0) return err(parseError.invalidTagValue(numStr, span(this._tokenStart, this._tokenStart + numStr.length)));
-				return ok(token.tagValue(tagValue));
+				const parsed = parseUsize64(numStr);
+				if (parsed === void 0) return err(parseError.invalidTagValue(numStr, span(this._tokenStart, this._tokenStart + numStr.length)));
+				return ok(token.tagValue(parsed));
 			}
 			this._position += numStr.length;
 			this._tokenEnd = this._position;
@@ -579,15 +668,6 @@ var Lexer = class {
 			return ok(token.string(fullMatch));
 		}
 		this._position++;
-		while (this._position < this._source.length) {
-			const ch = this._source[this._position];
-			if (ch === "\"" || ch === "\n") {
-				if (ch === "\"") this._position++;
-				break;
-			}
-			if (ch === "\\") this._position += 2;
-			else this._position++;
-		}
 		this._tokenEnd = this._position;
 		return err(parseError.unrecognizedToken(this.span()));
 	}
@@ -644,8 +724,8 @@ var Lexer = class {
 			const numStr = match[1];
 			this._position += fullMatch.length;
 			this._tokenEnd = this._position;
-			const value = parseInt(numStr, 10);
-			if (!Number.isSafeInteger(value) || value < 0) return err(parseError.invalidKnownValue(numStr, span(this._tokenStart + 1, this._tokenEnd - 1)));
+			const value = parseUsize64(numStr);
+			if (value === void 0) return err(parseError.invalidKnownValue(numStr, span(this._tokenStart + 1, this._tokenEnd - 1)));
 			return ok(token.knownValueNumber(value));
 		}
 		match = /^'([a-zA-Z_][a-zA-Z0-9_-]*)'/.exec(remaining);
@@ -657,8 +737,6 @@ var Lexer = class {
 			return ok(token.knownValueName(name));
 		}
 		this._position++;
-		while (this._position < this._source.length && this._source[this._position] !== "'") this._position++;
-		if (this._position < this._source.length) this._position++;
 		this._tokenEnd = this._position;
 		return err(parseError.unrecognizedToken(this.span()));
 	}
@@ -704,11 +782,39 @@ var Lexer = class {
 		}
 		return false;
 	}
-	_isIdentifierChar(ch) {
-		return /[a-zA-Z0-9_-]/.test(ch);
-	}
 };
 /**
+* Strictly parses a non-negative integer string in the range
+* `[0, 2^64 - 1]`, mirroring Rust `<u64 as FromStr>::from_str`.
+*
+* - Empty input or non-digit characters → `undefined`.
+* - Values that fit in `Number.MAX_SAFE_INTEGER` are returned as plain
+*   `number`s, so callers in the common case (tag values like `40000`,
+*   known values like `1`) never see a `bigint`.
+* - Values in `(2^53-1, 2^64-1]` are returned as `bigint`. dCBOR's
+*   `cbor({ tag, value })` and `KnownValue` constructors both accept
+*   `bigint` natively, so the bigint flows through to wire encoding
+*   without precision loss.
+* - Values strictly greater than `2^64 - 1` (or negative) are rejected
+*   so this parser never produces a tag/known-value outside the
+*   `u64` domain — matches Rust which fails `parse::<u64>()` in that
+*   case.
+*/
+const MAX_U64 = (1n << 64n) - 1n;
+function parseUsize64(s) {
+	if (s.length === 0) return void 0;
+	if (!/^\d+$/.test(s)) return void 0;
+	let value;
+	try {
+		value = BigInt(s);
+	} catch {
+		return;
+	}
+	if (value < 0n || value > MAX_U64) return void 0;
+	if (value <= BigInt(Number.MAX_SAFE_INTEGER)) return Number(value);
+	return value;
+}
+/**
 * Converts a hex string to bytes.
 */
 function hexToBytes(hex) {
@@ -994,12 +1100,12 @@ function parseMap(lexer) {
 		const keySpan = lexer.span();
 		if (map.has(key)) return err(parseError.duplicateMapKey(keySpan));
 		const colonResult = expectToken(lexer);
-		if (!colonResult.ok) return colonResult;
-		if (colonResult.value.type !== "Colon") return err(parseError.expectedColon(lexer.span()));
+		if (!colonResult.ok || colonResult.value.type !== "Colon") return err(parseError.expectedColon(lexer.span()));
 		const valueResult = parseItem(lexer);
 		if (!valueResult.ok) {
 			if (valueResult.error.type === "UnexpectedToken") {
-				if (valueResult.error.token.type === "BraceClose") return err(parseError.expectedMapKey(lexer.span()));
+				const unexpected = valueResult.error;
+				if (unexpected.token.type === "BraceClose") return err(parseError.expectedMapKey(unexpected.span));
 			}
 			return valueResult;
 		}
@@ -1038,12 +1144,20 @@ const composeError = {
 };
 /**
 * Gets the error message for a compose error.
+*
+* Mirrors Rust `Error::Display` (`bc-dcbor-parse-rust/src/compose.rs`):
+* the `ParseError` arm uses `#[error("Invalid CBOR item: {0}")]`, which
+* formats the inner error via its `Display` impl — *not* the variant
+* name. So `Error::ParseError(Error::EmptyInput)` formats as
+* `"Invalid CBOR item: Empty input"`, not
+* `"Invalid CBOR item: EmptyInput"`. We delegate to {@link errorMessage}
+* to get the same `Display`-style text.
 */
 function composeErrorMessage(error) {
 	switch (error.type) {
 		case "OddMapLength": return "Invalid odd map length";
 		case "DuplicateMapKey": return "Duplicate map key";
-		case "ParseError": return `Invalid CBOR item: ${error.error.type}`;
+		case "ParseError": return `Invalid CBOR item: ${errorMessage(error.error)}`;
 	}
 }
 /**