Added LockFreeLongMap
Robinson
parent
a8db6b42e4
commit
50a9bf4310
|
|
@ -0,0 +1,224 @@
|
|||
/*
|
||||
* 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.
|
||||
*/
|
||||
package dorkbox.collections
|
||||
|
||||
import java.io.Serializable
|
||||
import java.util.concurrent.atomic.*
|
||||
|
||||
/**
|
||||
* This class uses the "single-writer-principle" for lock-free publication.
|
||||
*
|
||||
*
|
||||
* Since there are only 2 methods to guarantee that modifications can only be called one-at-a-time (either it is only called by
|
||||
* one thread, or only one thread can access it at a time) -- we chose the 2nd option -- and use 'synchronized' to make sure that only
|
||||
* one thread can access this modification methods at a time. Getting or checking the presence of values can then happen in a lock-free
|
||||
* manner.
|
||||
*
|
||||
*
|
||||
* According to my benchmarks, this is approximately 25% faster than ConcurrentHashMap for (all types of) reads, and a lot slower for
|
||||
* contended writes.
|
||||
*
|
||||
*
|
||||
* This data structure is for many-read/few-write scenarios
|
||||
*
|
||||
*
|
||||
* An unordered map. This implementation is a cuckoo hash map using 3 hashes, random walking, and a small stash for problematic
|
||||
* keys. Null keys are not allowed. Null values are allowed. No allocation is done except when growing the table size. <br></br>
|
||||
*
|
||||
*
|
||||
* This map performs very fast get, containsKey, and remove (typically O(1), worst case O(log(n))). Put may be a bit slower,
|
||||
* depending on hash collisions. Load factors greater than 0.91 greatly increase the chances the map will have to rehash to the
|
||||
* next higher POT size.
|
||||
*
|
||||
*
|
||||
* Iteration can be very slow for a map with a large capacity. [.clear] and [.shrink] can be used to reduce
|
||||
* the capacity. [OrderedMap] provides much faster iteration.
|
||||
*/
|
||||
class LockFreeLongMap<V> : MutableMap<Long, V>, Cloneable, Serializable {
|
||||
|
||||
@Volatile
|
||||
private var hashMap: LongMap<V>
|
||||
|
||||
// synchronized is used here to ensure the "single writer principle", and make sure that ONLY one thread at a time can enter this
|
||||
// section. Because of this, we can have unlimited reader threads all going at the same time, without contention (which is our
|
||||
// use-case 99% of the time)
|
||||
/**
|
||||
* Constructs an empty <tt>HashMap</tt> with the default initial capacity
|
||||
* (16) and the default load factor (0.75).
|
||||
*/
|
||||
constructor() {
|
||||
hashMap = LongMap()
|
||||
}
|
||||
|
||||
/**
|
||||
* Constructs an empty <tt>HashMap</tt> with the specified initial
|
||||
* capacity and the default load factor (0.75).
|
||||
*
|
||||
* @param initialCapacity the initial capacity.
|
||||
*
|
||||
* @throws StateException if the initial capacity is negative.
|
||||
*/
|
||||
constructor(initialCapacity: Int) {
|
||||
hashMap = LongMap(initialCapacity)
|
||||
}
|
||||
|
||||
/**
|
||||
* Constructs an empty <tt>HashMap</tt> with the specified initial
|
||||
* capacity and load factor.
|
||||
*
|
||||
* @param initialCapacity the initial capacity
|
||||
* @param loadFactor the load factor
|
||||
*
|
||||
* @throws StateException if the initial capacity is negative
|
||||
* or the load factor is nonpositive
|
||||
*/
|
||||
constructor(initialCapacity: Int, loadFactor: Float) {
|
||||
hashMap = LongMap(initialCapacity, loadFactor)
|
||||
}
|
||||
|
||||
override val size: Int
|
||||
get() {
|
||||
// use the SWP to get a lock-free get of the value
|
||||
return mapREF[this].size
|
||||
}
|
||||
|
||||
override fun isEmpty(): Boolean {
|
||||
// use the SWP to get a lock-free get of the value
|
||||
return mapREF[this].size == 0
|
||||
}
|
||||
|
||||
@Suppress("UNCHECKED_CAST")
|
||||
override fun containsKey(key: Long): Boolean {
|
||||
// use the SWP to get a lock-free get of the value
|
||||
val value = mapREF[this] as LongMap<V>
|
||||
return value.containsKey(key)
|
||||
}
|
||||
|
||||
override fun containsValue(value: V): Boolean {
|
||||
return containsValue(value, false)
|
||||
}
|
||||
|
||||
fun containsValue(value: Any?, identity: Boolean): Boolean {
|
||||
// use the SWP to get a lock-free get of the value
|
||||
return mapREF[this].containsValue(value, identity)
|
||||
}
|
||||
|
||||
@Suppress("UNCHECKED_CAST")
|
||||
override operator fun get(key: Long): V? {
|
||||
// use the SWP to get a lock-free get of the value
|
||||
val value = mapREF[this] as LongMap<V>
|
||||
return value.get(key)
|
||||
}
|
||||
|
||||
@Synchronized
|
||||
override fun put(key: Long, value: V): V? {
|
||||
return hashMap.put(key, value)
|
||||
}
|
||||
|
||||
@Synchronized
|
||||
override fun remove(key: Long): V? {
|
||||
return hashMap.remove(key)
|
||||
}
|
||||
|
||||
@Synchronized
|
||||
override fun putAll(from: Map<out Long, V>) {
|
||||
hashMap.putAll(from)
|
||||
}
|
||||
|
||||
@Synchronized
|
||||
override fun clear() {
|
||||
hashMap.clear()
|
||||
}
|
||||
|
||||
/**
|
||||
* DO NOT MODIFY THE MAP VIA THIS (unless you synchronize around it!) It will result in unknown object visibility!
|
||||
*
|
||||
* Returns an iterator for the keys in the map. Remove is supported. Note that the same iterator instance is returned each
|
||||
* time this method is called. Use the [ObjectMap.Entries] constructor for nested or multithreaded iteration.
|
||||
*/
|
||||
override val keys: LongMap.Keys
|
||||
get() {
|
||||
return mapREF[this].keys()
|
||||
}
|
||||
|
||||
/**
|
||||
* DO NOT MODIFY THE MAP VIA THIS (unless you synchronize around it!) It will result in unknown object visibility!
|
||||
*
|
||||
* Returns an iterator for the values in the map. Remove is supported. Note that the same iterator instance is returned each
|
||||
* time this method is called. Use the [ObjectMap.Entries] constructor for nested or multithreaded iteration.
|
||||
*/
|
||||
@Suppress("UNCHECKED_CAST")
|
||||
override val values: LongMap.Values<V>
|
||||
get() {
|
||||
return mapREF[this].values() as LongMap.Values<V>
|
||||
}
|
||||
|
||||
/**
|
||||
* DO NOT MODIFY THE MAP VIA THIS (unless you synchronize around it!) It will result in unknown object visibility!
|
||||
*
|
||||
* Returns an iterator for the entries in the map. Remove is supported. Note that the same iterator instance is returned each
|
||||
* time this method is called. Use the [ObjectMap.Entries] constructor for nested or multithreaded iteration.
|
||||
*/
|
||||
@Suppress("UNCHECKED_CAST")
|
||||
override val entries: MutableSet<MutableMap.MutableEntry<Long, V>>
|
||||
get() {
|
||||
return mapREF[this].entries() as MutableSet<MutableMap.MutableEntry<Long, V>>
|
||||
}
|
||||
|
||||
override fun equals(other: Any?): Boolean {
|
||||
return mapREF[this] == other
|
||||
}
|
||||
|
||||
fun equalsIdentity(other: Any?): Boolean {
|
||||
return mapREF[this].equalsIdentity(other)
|
||||
}
|
||||
|
||||
override fun hashCode(): Int {
|
||||
return mapREF[this].hashCode()
|
||||
}
|
||||
|
||||
override fun toString(): String {
|
||||
return mapREF[this].toString()
|
||||
}
|
||||
|
||||
/**
|
||||
* Clears the map and reduces the size of the backing arrays to be the specified capacity, 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.
|
||||
*/
|
||||
@Synchronized
|
||||
fun clear(maximumCapacity: Int) {
|
||||
mapREF[this].clear(maximumCapacity)
|
||||
}
|
||||
|
||||
/**
|
||||
* Reduces the size of the backing arrays to be the specified capacity or less. If the capacity is already less, nothing is
|
||||
* done.
|
||||
* If the map contains more items than the specified capacity, the next highest power of two capacity is used instead.
|
||||
*/
|
||||
@Synchronized
|
||||
fun shrink(maximumCapacity: Int) {
|
||||
mapREF[this].shrink(maximumCapacity)
|
||||
}
|
||||
|
||||
companion object {
|
||||
const val version = Collections.version
|
||||
|
||||
// Recommended for best performance while adhering to the "single writer principle". Must be static-final
|
||||
private val mapREF = AtomicReferenceFieldUpdater.newUpdater(
|
||||
LockFreeLongMap::class.java, LongMap::class.java, "hashMap"
|
||||
)
|
||||
}
|
||||
}
|
||||
Loading…
Reference in New Issue