org.apache.myfaces.shared_impl.util.BiLevelCacheMap Maven / Gradle / Ivy
/*
* 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.myfaces.shared_impl.util;
import java.util.Collection;
import java.util.Collections;
import java.util.HashMap;
import java.util.Map;
import java.util.Set;
/**
* A bi-level cache based on HashMap for caching objects with minimal sychronization
* overhead. The limitation is that remove() is very expensive.
*
* Access to L1 map is not sychronized, to L2 map is synchronized. New values
* are first stored in L2. Once there have been more that a specified mumber of
* misses on L1, L1 and L2 maps are merged and the new map assigned to L1
* and L2 cleared.
*
*
* IMPORTANT:entrySet(), keySet(), and values() return unmodifiable snapshot collections.
*
*
* @author Anton Koinov (latest modification by $Author: matzew $)
* @version $Revision: 557350 $ $Date: 2007-07-18 13:19:50 -0500 (Wed, 18 Jul 2007) $
*/
public abstract class BiLevelCacheMap implements Map
{
//~ Instance fields ----------------------------------------------------------------------------
private static final int INITIAL_SIZE_L1 = 32;
/** To preinitialize _cacheL1 with default values use an initialization block */
protected Map _cacheL1;
/** Must be final because it is used for synchronization */
private final Map _cacheL2;
private final int _mergeThreshold;
private int _missCount;
//~ Constructors -------------------------------------------------------------------------------
public BiLevelCacheMap(int mergeThreshold)
{
_cacheL1 = new HashMap(INITIAL_SIZE_L1);
_cacheL2 = new HashMap(HashMapUtils.calcCapacity(mergeThreshold));
_mergeThreshold = mergeThreshold;
}
//~ Methods ------------------------------------------------------------------------------------
public boolean isEmpty()
{
synchronized (_cacheL2) {
return _cacheL1.isEmpty() && _cacheL2.isEmpty();
}
}
public void clear()
{
synchronized (_cacheL2) {
_cacheL1 = new HashMap(); // dafault size
_cacheL2.clear();
}
}
public boolean containsKey(Object key)
{
synchronized (_cacheL2) {
return _cacheL1.containsKey(key) || _cacheL2.containsKey(key);
}
}
public boolean containsValue(Object value)
{
synchronized (_cacheL2) {
return _cacheL1.containsValue(value) || _cacheL2.containsValue(value);
}
}
public Set entrySet()
{
synchronized (_cacheL2)
{
mergeIfL2NotEmpty();
return Collections.unmodifiableSet(_cacheL1.entrySet());
}
}
public Object get(Object key)
{
Map cacheL1 = _cacheL1;
Object retval = cacheL1.get(key);
if (retval != null)
{
return retval;
}
synchronized (_cacheL2)
{
// Has another thread merged caches while we were waiting on the mutex? Then check L1 again
if (cacheL1 != _cacheL1)
{
if ((retval = _cacheL1.get(key)) != null)
{
// do not update miss count (it is not a miss anymore)
return retval;
}
}
if ((retval = _cacheL2.get(key)) == null)
{
retval = newInstance(key);
if (retval != null)
{
put(key, retval);
mergeIfNeeded();
}
}
else
{
mergeIfNeeded();
}
}
return retval;
}
public Set keySet()
{
synchronized (_cacheL2)
{
mergeIfL2NotEmpty();
return Collections.unmodifiableSet(_cacheL1.keySet());
}
}
/**
* If key is already in cacheL1, the new value will show with a delay,
* since merge L2->L1 may not happen immediately. To force the merge sooner,
* call size().
*/
public Object put(Object key, Object value)
{
synchronized (_cacheL2)
{
_cacheL2.put(key, value);
// not really a miss, but merge to avoid big increase in L2 size
// (it cannot be reallocated, it is final)
mergeIfNeeded();
}
return value;
}
public void putAll(Map map)
{
synchronized (_cacheL2)
{
mergeIfL2NotEmpty();
// sepatare merge to avoid increasing L2 size too much
// (it cannot be reallocated, it is final)
merge(map);
}
}
/** This operation is very expensive. A full copy of the Map is created */
public Object remove(Object key)
{
synchronized (_cacheL2)
{
if (!_cacheL1.containsKey(key) && !_cacheL2.containsKey(key))
{
// nothing to remove
return null;
}
Object retval;
Map newMap;
synchronized (_cacheL1)
{
// "dummy" synchronization to guarantee _cacheL1 will be assigned after fully initialized
// at least until JVM 1.5 where this should be guaranteed by the volatile keyword
newMap = HashMapUtils.merge(_cacheL1, _cacheL2);
retval = newMap.remove(key);
}
_cacheL1 = newMap;
_cacheL2.clear();
_missCount = 0;
return retval;
}
}
public int size()
{
// Note: cannot simply return L1.size + L2.size
// because there might be overlaping of keys
synchronized (_cacheL2)
{
mergeIfL2NotEmpty();
return _cacheL1.size();
}
}
public Collection values()
{
synchronized (_cacheL2)
{
mergeIfL2NotEmpty();
return Collections.unmodifiableCollection(_cacheL1.values());
}
}
private void mergeIfL2NotEmpty()
{
if (!_cacheL2.isEmpty())
{
merge(_cacheL2);
}
}
private void mergeIfNeeded()
{
if (++_missCount >= _mergeThreshold)
{
merge(_cacheL2);
}
}
private void merge(Map map)
{
Map newMap;
synchronized (_cacheL1)
{
// "dummy" synchronization to guarantee _cacheL1 will be assigned after fully initialized
// at least until JVM 1.5 where this should be guaranteed by the volatile keyword
// But is this enough (in our particular case) to resolve the issues with DCL?
newMap = HashMapUtils.merge(_cacheL1, map);
}
_cacheL1 = newMap;
_cacheL2.clear();
_missCount = 0;
}
/**
* Subclasses must implement to have automatic creation of new instances
* or alternatively can use put to add new items to the cache.
*
* Implementing this method is prefered to guarantee that there will be only
* one instance per key ever created. Calling put() to add items in a multi-
* threaded situation will require external synchronization to prevent two
* instances for the same key, which defeats the purpose of this cache
* (put() is useful when initialization is done during startup and items
* are not added during execution or when (temporarily) having possibly two
* or more instances of the same key is not of concern).
*
* @param key lookup key
* @return new instace for the requested key
*/
protected abstract Object newInstance(Object key);
}