Advent of Code 2021 in Kotlin - Day 11


The Day 11 problem requires form us writing some simulation of flashing in described manner. We can implement this in some readable way by using custom data structures and redefining some operators in Kotlin - let’s begin and see how to express some complicated code in a few lines of text.


We defined the operators for modifying internal values of the map to get extra functionality by calculating posOf i.e. the map from values to the positions where we can find every of them on the map. Thanks to this function we can always ask about the places with energy equal to 10 (in constant time) and propagate the energy from them in single flash execution.

Let’s notice that we defined the flash function as tailrec fun so it is optimized by the compiler to standard loop. We did that because it’s usually more readable to see that was our intention when writing this code - we want to mark current positions as flashed, then for every of them propagate the energy to its neighbours and then try to flash again from the positions on which the maxVal appeared in current step of flashing.

It’s worth mentioning also how we defined the LazyDefaultMap<K, V> (and modified DefaultMap<K, V>) when solving this task. Now we ended up with the definition, that allows to compute the default value lazily, but also it sets the computed value to the backing map. It seemed to be more reasonable approach as we think about the value extracted from map as it would be in it - so when it’s a mutable value, then after its modifications, we should observe this modification in map.

class LazyDefaultMap<K, V>(
  private val default: () -> V,
  private val map: MutableMap<K, V> = HashMap()
) : MutableMap<K, V> by map {
  override fun get(key: K): V = map.getOrDefault(key, default()).also { map[key] = it }

Don’t miss the definitions that uses Sequence<T> in today code - they both were defined to improve code readability and keep the performance in my opinion - let me know what you think about such “functional” approaching to the problems 😉.


object Day11 : AdventDay() {
  override fun solve() {
    val map = reads<String>()?.toEnergyMap() ?: return

    with(map.copy()) {
      (1..100).fold(0) { sum, _ -> sum + simulateStep() }.printIt()
    with(map.copy()) {
      generateSequence(1, Int::inc).first { simulateStep() == 100 }.printIt()

private fun List<String>.toEnergyMap() =
  EnergyMap(maxVal = 10, map { line -> { it.digitToInt() }.toMutableList() })

private data class Pos(val x: Int, val y: Int)

private data class EnergyMap(val maxVal: Int, private val values: List<MutableList<Int>>) {

  val indices = values.flatMapIndexed { y, row -> { Pos(it, y) } }
  private val posOf = LazyDefaultMap(::mutableSetOf,
    indices.groupBy { this[it] }.mapValues { it.value.toMutableSet() }.toMutableMap()

  fun copy() = EnergyMap(maxVal, { it.toMutableList() })
  operator fun get(p: Pos): Int = with(p) { values[y][x] }
  operator fun set(p: Pos, v: Int) {
    val newVal = v.coerceAtMost(maxVal)
    values[p.y][p.x] = newVal

  fun neighbours(of: Pos) = sequence {
    for (x in -1..1) for (y in -1..1) yield(Pair(x, y))
    .filterNot { (x, y) -> x == 0 && y == 0 }
    .map { (x, y) -> Pos(of.x + x, of.y + y) }
    .filter { it.isValid() }

  fun simulateStep(): Int {
    val flashed = mutableSetOf<Pos>()
    tailrec fun flash(flash: Set<Pos>) {
      flashed += flash
      flash.toList().forEach { pos -> neighbours(pos).forEach { this[it] = this[it] + 1 } }
      if (flash.isNotEmpty()) flash(posOf[maxVal] - flashed)
    indices.forEach { this[it] = this[it] + 1 }
    indices.forEach { this[it] = this[it] % maxVal }
    return flashed.size

  private fun Pos.isValid() = y in values.indices && x in values[y].indices

Extra notes

Let’s see that we solved each of the parts of the task on the copy of the input map. That’s because the map internal state is mutable (it changes during the simulation) but we need to have a fresh map when trying to solve the second part of the problem.

We used with Kotlin construct here, that simplified our code and introduced the context of current copy of EnergyMap. That’s one of the really cool Kotlin constructs that is in fact an inline function, so it brings no extra overhead on runtime while really simplifying our code.

Also in this day we introduced tests running with expected outputs of days. We used for it some tricky and simple function

fun catchSystemOut(action: () -> Unit) = ByteArrayOutputStream().also {
  val originalOut = System.out

which is capable of catching the System.out value and returning it as simple String (that is later cmopared with the expected output in tests and Advent days are defined to print their solutions).

Student of Computer Science

My interests include robotics (mainly with Arduino), mobile development for Android (love Kotlin) and Java SE/EE applications development.