Integer range specification

[Kristóf Poduszló]

seedrandom('seed').int32(0, 5) should return a value between 0 (inclusive) and 5 (exclusive): 0, 1, 2, 3, 4. I have already implemented the functionality myself, but feel that this should be a built-in method.

/**
 * A randomization engine which can be made deterministic by using a
 * custom-seeded engine.
 */
export default class Randomizer {
  /**
   * Engine to be used for randomization.
   * @type {Object}
   */
  engine;

  constructor(engine) {
    this.engine = engine;
  }

  /**
   * Returns a random 32-bit integer (signed or unsigned) in the given range.
   * @param {number} min Minimum value (included).
   * @param {number} max Maximum value (excluded).
   * @returns {number}
   */
  getInt(min, max) {
    const range = max - min;

    // Get the lowest integer which is not part of the equal distribution range
    const firstTooHighValue =
      Config.UINT32_RANGE - Config.UINT32_RANGE % range;

    let result;
    do {
      // Generate a random 32-bit unsigned integer
      result = this.engine.int32() - Config.INT32_MIN_VALUE;
    } while (
      // Ensure equal distribution
      result >= firstTooHighValue
    );

    // Place the result in the range and offset it by the minimum value
    return result % range + min;
  }

  /**
   * Shuffles the elements of an array.
   * @param {Object[]} array Array to be shuffled.
   * @returns {Object[]}
   */
  shuffleArray(array) {
    const result = [...array];

    // Perform Durstenfeld shuffle
    for (let i = array.length - 1; i > 0; --i) {
      // Generate a random integer in [0, i] deterministically
      const j = this.getInt(0, i);

      // Swap result[i] and result[j]
      [result[i], result[j]] = [result[j], result[i]];
    }

    return result;
  }
}

[David Bau]

In the spirit of keeping this library as small as possible, I think we should consider convenience functions "out of scope" - there are a large number of convenience functions that should be supplied by a full random number library: uniform, normal, exponential, integer ranges, shuffles, etc. We just provide raw access to a PRNG source and leave it to other libraries to supply more specific distributions.