prototype-3x5/peg.js

/**
 * A Pattern is a function that matches against a string starting at a given index.
 *
 * If it matches successfully, it returns some captured value, and the index following the match.
 *
 * On success or failure, it returns the furthest point the pattern could make sense of, and a description of what was expected next at that point.
 *
 * For simple patterns, the "furthest point" may just be the following index; however, some more complex patterns might succeed,
 * but consume less input than they would have been able to if some other expected symbol was found. Reporting
 * the furthest a pattern could hypothetically have gotten can help generate better error messages if no valid parse tree is found.
 *
 * @template T
 * @callback Peg.PatternCall
 * @param {string} source - the string being parsed
 * @param {number} index - the index in the string to begin matching from
 * @returns {[[T, number] | null, number, string]} - [successValue, furthest symbol attempted, expected pattern]
 */
/**
 * @template T
 * @typedef {object} Peg.PatternExt
 * @property {<U>(map: (value: T) => U) => Peg.Pattern<U>} map Creates a pattern that wraps another pattern, transforming the returned value on a match
 * @property {string} expectLabel A human-readable annotation describing the pattern for error messages
 * @property {(label: string) => Peg.Pattern<T>} expects Adds a human-readable annotation describing the pattern
 */
/**
 * @template T
 * @typedef {Peg.PatternCall<T> & Peg.PatternExt<T>} Peg.Pattern
 */
var Peg = window.Peg ?? {};

/**
 * Makes a pattern from a function, adding helper methods.
 *
 * @template T
 * @param {(source: string, index: number) => ([[T, number] | null, number, string])} matchFunc
 * @returns {Peg.Pattern<T>}
 */
Peg.WrapPattern = function (matchFunc) {
  const pattern = /** @type {Peg.Pattern<T>} */ (matchFunc);
  pattern.map = function (map) {
    return Peg.WrapPattern(function (source, index) {
      const [value, furthest, expected] = pattern(source, index);
      return [value ? [map(value[0]), value[1]] : null, furthest, expected];
    }).expects(pattern.expectLabel);
  };

  pattern.expectLabel = pattern.name;
  pattern.expects = (label) => {
    pattern.expectLabel = label;
    return pattern;
  };

  return pattern;
};

/**
 * Proxies to a pattern retrieved from an accessor function.
 *
 * Allows using a pattern recursively in its own definition, by returning the value of the const assigned to.
 *
 * @template T
 * @param {() => Peg.Pattern<T>} getPattern
 * @returns {Peg.Pattern<T>}
 */
Peg.Use = function (getPattern) {
  return Peg.WrapPattern(function (source, index) {
    return getPattern()(source, index);
  }).expects(String(getPattern));
};

/**
 * Creates a pattern matching a regex & returning any captures. The regex needs to be sticky (using the //y modifier)
 * @param {RegExp} regex
 * @return {Peg.Pattern<RegExpExecArray>}
 */
Peg.Regex = function (regex) {
  /** @type {Peg.Pattern<RegExpExecArray>} */
  const pattern = Peg.WrapPattern(function (source, index) {
    regex.lastIndex = index;
    const matches = regex.exec(source);
    return matches
      ? [[matches, regex.lastIndex], regex.lastIndex, pattern.expectLabel]
      : [null, index, pattern.expectLabel];
  }).expects(regex.source);
  return pattern;
};

/**
 * Creates a pattern that tries the given patterns, in order, until it finds one that matches at the current index.
 * @template T
 * @param {...Peg.Pattern<T>} patterns
 * @return {Peg.Pattern<T>}
 */
Peg.Choose = function (...patterns) {
  const genericExpected = patterns
    .map((pattern) => pattern.expectLabel)
    .join(" | ");
  return Peg.WrapPattern(function (source, index) {
    let furthestFound = index;
    let furthestExpected = genericExpected;
    for (const pattern of patterns) {
      const [value, furthest, expected] = pattern(source, index);
      if (value) {
        return [value, furthest, expected];
      } else if (furthest > furthestFound) {
        furthestFound = furthest;
        furthestExpected = expected;
      }
    }
    return [null, furthestFound, furthestExpected];
  }).expects(genericExpected);
};

/**
 * Creates a pattern that concatenates the given patterns, returning a tuple of their captured values.
 *
 * For example, if A matches "a" and captures 1, while B matches "b" and captures null,
 * then `Sequence(A,B)` will match "ab" and capture [1, null]
 * @template {unknown[]} T
 * @param {{[K in keyof T]: Peg.Pattern<T[K]>}} patterns
 * @return {Peg.Pattern<T>}
 */
Peg.Sequence = function (...patterns) {
  const genericExpected = patterns[0]?.expectLabel ?? "(nothing)";
  return Peg.WrapPattern(function (source, index) {
    const values = /** @type {T} */ (/** @type {unknown} */ ([]));
    let furthestFound = index;
    let furthestExpected = genericExpected;
    for (const pattern of patterns) {
      const [value, furthest, expected] = pattern(source, index);
      if (furthest >= furthestFound) {
        furthestFound = furthest;
        furthestExpected = expected;
      }
      if (value == null) {
        return [null, furthestFound, furthestExpected];
      }
      values.push(value[0]);
      index = value[1];
    }
    return [[values, index], furthestFound, furthestExpected];
  }).expects(genericExpected);
};

/**
 * Creates a pattern that matches consecutive runs of the given pattern, returning an array of all captures.
 *
 * The match only succeeds if the run is at least {@link min} instances long.
 *
 * If the given pattern does not consume input, the matching will be terminated to prevent an eternal loop.
 *
 * Note that if the minimum run is zero, this pattern will always succeed, but might not consume any input.
 * @template {unknown} T
 * @param {number} min
 * @param {Peg.Pattern<T>} pattern
 * @return {Peg.Pattern<T[]>}
 */
Peg.AtLeast = function (min, pattern) {
  return Peg.WrapPattern(function (source, index) {
    const values = /** @type {T[]} */ ([]);
    let furthestFound = index;
    let furthestExpected = pattern.expectLabel;
    do {
      const [value, furthest, expected] = pattern(source, index);
      if (furthest > furthestFound) {
        furthestFound = furthest;
        furthestExpected = expected;
      }
      if (value == null) {
        break;
      }
      values.push(value[0]);
      if (index == value[1]) {
        break;
      }
      index = value[1];
    } while (true);
    if (values.length >= min) {
      return [[values, index], furthestFound, furthestExpected];
    } else {
      return [null, furthestFound, furthestExpected];
    }
  }).expects(pattern.expectLabel);
};

/**
 * Creates a pattern that matches the end of input
 * @return {Peg.Pattern<true>}
 */
Peg.End = () => {
  /** @type {Peg.Pattern<true>} */
  const end = Peg.WrapPattern(function End(source, index) {
    return [
      source.length == index ? [/** @type {true} */ (true), index] : null,
      index,
      end.expectLabel,
    ];
  }).expects("<eof>");
  return end;
};

/**
 * Creates a pattern that never succeeds, but reports how far its wrapped pattern could match before failing.
 *
 * This is a hack, meant to improve error messages after an AtLeast(). Maybe this can be removed by patterns returning both success and failure
 *
 * Never consumes input, and fails with zero length if the pattern succeeds.
 * @param {Peg.Pattern<unknown>} pattern
 * @return {Peg.Pattern<never>}
 */
Peg.Hint = function (pattern) {
  return /** @type {Peg.Pattern<never>} */ (
    Peg.WrapPattern(function (source, index) {
      const [value, furthest, expected] = pattern(source, index);
      if (value) {
        console.log("oops match", value, furthest, expected);
        return [null, index, pattern.expectLabel];
      } else {
        return [value, furthest, expected];
      }
    })
  ).expects(pattern.expectLabel);
};