org.apache.commons.lang3.concurrent.AtomicSafeInitializer Maven / Gradle / Ivy
Show all versions of virtdata-lib-curves4 Show documentation
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You 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 org.apache.commons.lang3.concurrent;
import java.util.concurrent.atomic.AtomicReference;
/**
*
* A specialized {@code ConcurrentInitializer} implementation which is similar
* to {@link AtomicInitializer}, but ensures that the {@link #initialize()}
* method is called only once.
*
*
* As {@link AtomicInitializer} this class is based on atomic variables, so it
* can create an object under concurrent access without synchronization.
* However, it implements an additional check to guarantee that the
* {@link #initialize()} method which actually creates the object cannot be
* called multiple times.
*
*
* Because of this additional check this implementation is slightly less
* efficient than {@link AtomicInitializer}, but if the object creation in the
* {@code initialize()} method is expensive or if multiple invocations of
* {@code initialize()} are problematic, it is the better alternative.
*
*
* From its semantics this class has the same properties as
* {@link LazyInitializer}. It is a "save" implementation of the lazy
* initializer pattern. Comparing both classes in terms of efficiency is
* difficult because which one is faster depends on multiple factors. Because
* {@code AtomicSafeInitializer} does not use synchronization at all it probably
* outruns {@link LazyInitializer}, at least under low or moderate concurrent
* access. Developers should run their own benchmarks on the expected target
* platform to decide which implementation is suitable for their specific use
* case.
*
*
* @since 3.0
* @param the type of the object managed by this initializer class
*/
public abstract class AtomicSafeInitializer implements
ConcurrentInitializer {
/** A guard which ensures that initialize() is called only once. */
private final AtomicReference> factory =
new AtomicReference<>();
/** Holds the reference to the managed object. */
private final AtomicReference reference = new AtomicReference<>();
/**
* Get (and initialize, if not initialized yet) the required object
*
* @return lazily initialized object
* @throws ConcurrentException if the initialization of the object causes an
* exception
*/
@Override
public final T get() throws ConcurrentException {
T result;
while ((result = reference.get()) == null) {
if (factory.compareAndSet(null, this)) {
reference.set(initialize());
}
}
return result;
}
/**
* Creates and initializes the object managed by this
* {@code AtomicInitializer}. This method is called by {@link #get()} when
* the managed object is not available yet. An implementation can focus on
* the creation of the object. No synchronization is needed, as this is
* already handled by {@code get()}. This method is guaranteed to be called
* only once.
*
* @return the managed data object
* @throws ConcurrentException if an error occurs during object creation
*/
protected abstract T initialize() throws ConcurrentException;
}