/* * Copyright 2023 dorkbox, llc * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ /******************************************************************************* * Copyright 2011 LibGDX. * Mario Zechner @gmail.com> * Nathan Sweet @gmail.com> * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ package dorkbox.collections import java.util.* import kotlin.math.ceil import kotlin.math.max /** * An unordered set where the keys are objects. Null keys are not allowed. No allocation is done except when growing the table * size. * * * This class performs fast contains and remove (typically O(1), worst case O(n) but that is rare in practice). Add may be * slightly slower, depending on hash collisions. Hashcodes are rehashed to reduce collisions and the need to resize. Load factors * greater than 0.91 greatly increase the chances to resize to the next higher POT size. * * * Unordered sets and maps are not designed to provide especially fast iteration. Iteration is faster with OrderedSet and * OrderedMap. * * * This implementation uses linear probing with the backward shift algorithm for removal. Hashcodes are rehashed using Fibonacci * hashing, instead of the more common power-of-two mask, to better distribute poor hashCodes (see [Malte * Skarupke's blog post](https://probablydance.com/2018/06/16/fibonacci-hashing-the-optimization-that-the-world-forgot-or-a-better-alternative-to-integer-modulo/)). Linear probing continues to work even when all hashCodes collide, just more slowly. * @author Nathan Sweet * @author Tommy Ettinger */ open class ObjectSet : MutableSet { companion object { const val version = Collections.version fun with(vararg array: T): ObjectSet { val set = ObjectSet() set.addAll(*array) return set } fun tableSize(capacity: Int, loadFactor: Float): Int { if (capacity < 0) { throw StateException("capacity must be >= 0: $capacity") } val tableSize: Int = Collections.nextPowerOfTwo( max(2.0, ceil((capacity / loadFactor).toDouble()).toInt().toDouble()).toInt() ) if (tableSize > 1 shl 30) { throw StateException("The required capacity is too large: $capacity") } return tableSize } } override var size = 0 var keyTable: Array var loadFactor: Float var threshold: Int /** * Used by [.place] to bit shift the upper bits of a `long` into a usable range (>= 0 and <= * [.mask]). The shift can be negative, which is convenient to match the number of bits in mask: if mask is a 7-bit * number, a shift of -7 shifts the upper 7 bits into the lowest 7 positions. This class sets the shift > 32 and < 64, * which if used with an int will still move the upper bits of an int to the lower bits due to Java's implicit modulus on * shifts. * * [.mask] can also be used to mask the low bits of a number, which may be faster for some hashcodes, if * [.place] is overridden. */ protected var shift: Int /** * A bitmask used to confine hashcodes to the size of the table. Must be all 1 bits in its low positions, ie a power of two * minus 1. If [.place] is overriden, this can be used instead of [.shift] to isolate usable bits of a * hash. */ protected var mask: Int @Transient private var iterator1: ObjectSetIterator? = null @Transient private var iterator2: ObjectSetIterator? = null /** * Creates a new set with the default capacity of 51 and loadfactor of 0.89 */ constructor(): this(51, 0.8f) /** * Creates a new set with the specified initial capacity and load factor. This set will hold initialCapacity items before * growing the backing table. * * @param initialCapacity The backing array size is initialCapacity / loadFactor, increased to the next power of two. * @param loadFactor The loadfactor used to determine backing array growth */ constructor(initialCapacity: Int = 51, loadFactor: Float = 0.8f) { if ((loadFactor <= 0f || loadFactor >= 1f)) { throw StateException("loadFactor must be > 0 and < 1: $loadFactor") } this.loadFactor = loadFactor val tableSize = tableSize(initialCapacity, loadFactor) threshold = (tableSize * loadFactor).toInt() mask = tableSize - 1 shift = java.lang.Long.numberOfLeadingZeros(mask.toLong()) @Suppress("UNCHECKED_CAST") keyTable = arrayOfNulls(tableSize) as Array } /** * Creates a new set identical to the specified set. */ constructor(set: ObjectSet) : this((set.keyTable.size * set.loadFactor).toInt(), set.loadFactor) { System.arraycopy(set.keyTable, 0, keyTable, 0, set.keyTable.size) @Suppress("LeakingThis") size = set.size } /** * Returns an index >= 0 and <= [.mask] for the specified `item`. * * * The default implementation uses Fibonacci hashing on the item's [Object.hashCode]: the hashcode is multiplied by a * long constant (2 to the 64th, divided by the golden ratio) then the uppermost bits are shifted into the lowest positions to * obtain an index in the desired range. Multiplication by a long may be slower than int (eg on GWT) but greatly improves * rehashing, allowing even very poor hashcodes, such as those that only differ in their upper bits, to be used without high * collision rates. Fibonacci hashing has increased collision rates when all or most hashcodes are multiples of larger * Fibonacci numbers (see [Malte Skarupke's blog post](https://probablydance.com/2018/06/16/fibonacci-hashing-the-optimization-that-the-world-forgot-or-a-better-alternative-to-integer-modulo/)). * * * This method can be overriden to customizing hashing. This may be useful eg in the unlikely event that most hashcodes are * Fibonacci numbers, if keys provide poor or incorrect hashcodes, or to simplify hashing if keys provide high quality * hashcodes and don't need Fibonacci hashing: `return item.hashCode() & mask;` */ protected fun place(item: T): Int { return (item.hashCode() * -0x61c8864680b583ebL ushr shift).toInt() } /** * Returns the index of the key if already present, else -(index + 1) for the next empty index. This can be overridden in this * pacakge to compare for equality differently than [Object.equals]. */ fun locateKey(key: T?): Int { requireNotNull(key) { "key cannot be null." } val keyTable = keyTable var i = place(key) while (true) { val other: T = keyTable[i] ?: return -(i + 1) // Empty space is available. if (other == key) return i // Same key was found. i = i + 1 and mask } } override fun addAll(elements: Collection): Boolean { var added = false elements.forEach { added = add(it) || added } return added } /** * Returns true if the key was added to the set or false if it was already in the set. If this set already contains the key, * the call leaves the set unchanged and returns false. */ override fun add(element: T): Boolean { var i = locateKey(element) if (i >= 0) return false // Existing key was found. i = -(i + 1) // Empty space was found. keyTable[i] = element if (++size >= threshold) resize(keyTable.size shl 1) return true } fun addAll(array: Array) { addAll(array, 0, array.size) } fun addAll(array: Array, offset: Int, length: Int) { if (offset + length > array.size) { throw StateException("offset + length must be <= size: $offset + $length <= ${array.size}") } addAll(array, offset, length) } @Suppress("UNCHECKED_CAST") fun addAll(vararg array: T): Boolean { return addAll(array as Array, 0, array.size) } fun addAll(array: Array, offset: Int, length: Int): Boolean { ensureCapacity(length) val oldSize = size var i = offset val n = i + length while (i < n) { add(array[i]) i++ } return oldSize != size } fun addAll(set: ObjectSet) { ensureCapacity(set.size) val keyTable = set.keyTable var i = 0 val n = keyTable.size while (i < n) { val key: T? = keyTable[i] key?.let { add(it) } i++ } } /** * Skips checks for existing keys, doesn't increment size. */ private fun addResize(key: T) { val keyTable = keyTable var i = place(key) while (true) { if (keyTable[i] == null) { keyTable[i] = key return } i = (i + 1) and mask } } override fun retainAll(elements: Collection): Boolean { var removed = false keyTable.forEach { key -> if (key != null) { if (!elements.contains(key)) { removed = remove(key) || removed } } } return removed } override fun removeAll(elements: Collection): Boolean { var removed = false elements.forEach { if (remove(it)) { removed = true } } return removed } /** * Returns true if the key was removed. */ override fun remove(element: T): Boolean { var i = locateKey(element) if (i < 0) return false val keyTable = keyTable val mask = mask var next = (i + 1) and mask var k: T? while (keyTable[next].also { k = it } != null) { val placement = place(k!!) if ((next - placement and mask) > (i - placement and mask)) { keyTable[i] = k i = next } next = (next + 1) and mask } keyTable[i] = null size-- return true } /** * Returns true if the set has one or more items. */ fun notEmpty(): Boolean { return size > 0 } /** * Returns true if the set is empty. */ override fun isEmpty(): Boolean { return size == 0 } /** * Reduces the size of the backing arrays to be the specified capacity / loadFactor, or less. If the capacity is already less, * nothing is done. If the set contains more items than the specified capacity, the next highest power of two capacity is used * instead. */ fun shrink(maximumCapacity: Int) { if (maximumCapacity < 0) { throw StateException("maximumCapacity must be >= 0: $maximumCapacity") } val tableSize = tableSize(maximumCapacity, loadFactor) if (keyTable.size > tableSize) resize(tableSize) } /** * Clears the set and reduces the size of the backing arrays to be the specified capacity / loadFactor, if they are larger. * The reduction is done by allocating new arrays, though for large arrays this can be faster than clearing the existing * array. */ open fun clear(maximumCapacity: Int) { val tableSize = tableSize(maximumCapacity, loadFactor) if (keyTable.size <= tableSize) { clear() return } size = 0 resize(tableSize) } /** * Clears the set, leaving the backing arrays at the current capacity. When the capacity is high and the population is low, * iteration can be unnecessarily slow. [.clear] can be used to reduce the capacity. */ override fun clear() { if (size == 0) return size = 0 Arrays.fill(keyTable, null) } override fun containsAll(elements: Collection): Boolean { elements.forEach { if (!contains(it)) { return false } } return true } override operator fun contains(element: T): Boolean { return locateKey(element) >= 0 } operator fun get(key: T): T? { val i = locateKey(key) return if (i < 0) null else keyTable[i] } fun first(): T? { val keyTable = keyTable var i = 0 val n = keyTable.size while (i < n) { if (keyTable[i] != null) return keyTable[i] i++ } throw IllegalStateException("ObjectSet is empty.") } /** * Increases the size of the backing array to accommodate the specified number of additional items / loadFactor. Useful before * adding many items to avoid multiple backing array resizes. */ fun ensureCapacity(additionalCapacity: Int) { val tableSize = tableSize(size + additionalCapacity, loadFactor) if (keyTable.size < tableSize) resize(tableSize) } @Suppress("UNCHECKED_CAST") private fun resize(newSize: Int) { val oldCapacity = keyTable.size threshold = (newSize * loadFactor).toInt() mask = newSize - 1 shift = java.lang.Long.numberOfLeadingZeros(mask.toLong()) val oldKeyTable = keyTable keyTable = arrayOfNulls(newSize) as Array if (size > 0) { for (i in 0 until oldCapacity) { val key = oldKeyTable[i] key?.let { addResize(it) } } } } override fun hashCode(): Int { var h = size val keyTable = keyTable var i = 0 val n = keyTable.size while (i < n) { val key: T? = keyTable[i] if (key != null) h += key.hashCode() i++ } return h } @Suppress("UNCHECKED_CAST") override fun equals(other: Any?): Boolean { if (other !is ObjectSet<*>) return false other as ObjectSet if (other.size != size) return false val keyTable = keyTable var i = 0 val n = keyTable.size while (i < n) { if (keyTable[i] != null && !other.contains(keyTable[i])) return false i++ } return true } override fun toString(): String { return '{'.toString() + toString(", ") + '}' } open fun toString(separator: String): String { if (size == 0) return "" val buffer = StringBuilder(32) val keyTable = keyTable var i = keyTable.size while (i-- > 0) { val key: T = keyTable[i] ?: continue buffer.append(if (key === this) "(this)" else key) break } while (i-- > 0) { val key: T = keyTable[i] ?: continue buffer.append(separator) buffer.append(if (key === this) "(this)" else key) } return buffer.toString() } /** * Returns an iterator for the keys in the set. Remove is supported. * * If [Collections.allocateIterators] is false, the same iterator instance is returned each time this method is called. * Use the [ObjectSetIterator] constructor for nested or multithreaded iteration. */ override fun iterator(): ObjectSetIterator { if (Collections.allocateIterators) return ObjectSetIterator(this) if (iterator1 == null) { iterator1 = ObjectSetIterator(this) iterator2 = ObjectSetIterator(this) } if (!iterator1!!.valid) { iterator1!!.reset() iterator1!!.valid = true iterator2!!.valid = false return iterator1 as ObjectSetIterator } iterator2!!.reset() iterator2!!.valid = true iterator1!!.valid = false return iterator2 as ObjectSetIterator } open class ObjectSetIterator(val set: ObjectSet) : Iterable, MutableIterator { var hasNext = false var nextIndex = 0 var currentIndex = 0 var valid = true init { @Suppress("LeakingThis") reset() } open fun reset() { currentIndex = -1 nextIndex = -1 findNextIndex() } private fun findNextIndex() { val keyTable = set.keyTable val n = set.keyTable.size while (++nextIndex < n) { if (keyTable[nextIndex] != null) { hasNext = true return } } hasNext = false } override fun remove() { var i = currentIndex check(i >= 0) { "next must be called before remove." } val keyTable = set.keyTable val mask = set.mask var next = i + 1 and mask var key: K? while (keyTable[next].also { key = it } != null) { val placement = set.place(key!!) if ((next - placement and mask) > (i - placement and mask)) { keyTable[i] = key i = next } next = next + 1 and mask } keyTable[i] = null set.size-- if (i != currentIndex) --nextIndex currentIndex = -1 } override fun hasNext(): Boolean { if (!valid) throw RuntimeException("#iterator() cannot be used nested.") return hasNext } override fun next(): K { if (!hasNext) throw NoSuchElementException() if (!valid) throw RuntimeException("#iterator() cannot be used nested.") val key = set.keyTable[nextIndex] currentIndex = nextIndex findNextIndex() return key!! } override fun iterator(): ObjectSetIterator { return this } /** * Returns a new array containing the remaining values. */ open fun toArray(): Array { @Suppress("UNCHECKED_CAST") return Array(set.size) { next() } as Array } /** * Adds the remaining values to the array. */ open fun toArray(array: Array): Array { var i = 0 while(hasNext) { array[i++] = next() } return array } } }