com.tangosol.util.SegmentedHashMap Maven / Gradle / Ivy
Show all versions of coherence Show documentation
/*
* Copyright (c) 2000, 2020, Oracle and/or its affiliates.
*
* Licensed under the Universal Permissive License v 1.0 as shown at
* http://oss.oracle.com/licenses/upl.
*/
package com.tangosol.util;
import com.oracle.coherence.common.base.Blocking;
import com.oracle.coherence.common.collections.AbstractStableIterator;
import java.lang.ref.WeakReference;
import java.util.AbstractCollection;
import java.util.AbstractSet;
import java.util.ArrayList;
import java.util.Collection;
import java.util.Iterator;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.WeakHashMap;
import java.util.concurrent.atomic.AtomicLong;
/**
* An implementation of java.util.Map that is optimized for heavy concurrent use.
*
* Retrieval and update operations to the map (e.g. get, put)
* are non-blocking and uncontended and will reflect some state of the map.
* Insert and remove operations to the map (e.g. put, remove)
* do require internal locking.
*
* The entries in the map are internally segmented so as to permit a high level
* of concurrent "locked" operations without contention.
*
* Retrievals and updates that run concurrently with bulk operations
* (e.g. putAll, clear may reflect insertion or removal of
* only some entries. Iterators on the Map may also reflect concurrent updates
* made since the Iterator was created. However, Iterators will not throw
* ConcurrentModificationException.
*
* @since Coherence 3.5
* @author rhl 2008.12.01
*/
public class SegmentedHashMap
extends Base
implements Map
{
// ----- constructors ---------------------------------------------------
/**
* Default constructor.
*/
public SegmentedHashMap()
{
this(DEFAULT_INITIALSIZE, DEFAULT_LOADFACTOR, DEFAULT_GROWTHRATE);
}
/**
* Construct a thread-safe hash map using the specified settings.
*
* @param cInitialBuckets the initial number of hash buckets, 0 < n
* @param flLoadFactor the acceptable load factor before resizing
* occurs, 0 < n, such that a load factor of
* 1.0 causes resizing when the number of entries
* exceeds the number of buckets
* @param flGrowthRate the rate of bucket growth when a resize occurs,
* 0 < n, such that a growth rate of 1.0 will
* double the number of buckets:
* bucketcount = bucketcount * (1 + growthrate)
*/
public SegmentedHashMap(int cInitialBuckets, float flLoadFactor,
float flGrowthRate)
{
if (cInitialBuckets <= 0)
{
throw new IllegalArgumentException("SegmentedHashMap: "
+ "Initial number of buckets must be greater than zero.");
}
if (flLoadFactor <= 0)
{
throw new IllegalArgumentException("SegmentedHashMap: "
+ "Load factor must be greater than zero.");
}
if (flGrowthRate <= 0)
{
throw new IllegalArgumentException("SegmentedHashMap: "
+ "Growth rate must be greater than zero.");
}
// initialize the hash map data structure
m_aeBucket = new Entry[cInitialBuckets];
m_cSegmentCapacity =
Math.max((int) (cInitialBuckets * flLoadFactor) / SEGMENT_COUNT,
MIN_SEGMENT_CAPACITY);
m_flLoadFactor = flLoadFactor;
m_flGrowthRate = flGrowthRate;
// Initialize the segment control structures
m_aSegment = new Segment[LOCK_COUNT];
for (int i = 0; i < LOCK_COUNT; i++)
{
m_aSegment[i] = new Segment();
}
initializeActions();
}
// ----- accessors ------------------------------------------------------
/**
* Return the registered action for insert.
*
* @return the registered action for insert
*/
protected InsertAction getInsertAction()
{
return m_actionInsert;
}
/**
* Specify the action for insert.
*
* @param action the action for insert
*/
protected void setInsertAction(InsertAction action)
{
m_actionInsert = action;
}
/**
* Return the registered action for getEntryInternal.
*
* @return the registered action for getEntryInternal
*/
protected GetEntryAction getGetEntryAction()
{
return m_actionGetEntry;
}
/**
* Specify the action for getEntryInternal.
*
* @param action the action for getEntryInternal
*/
protected void setGetEntryAction(GetEntryAction action)
{
m_actionGetEntry = action;
}
/**
* Return the registered action for remove().
*
* @return the registered action for remove()
*/
protected RemoveAction getRemoveAction()
{
return m_actionRemove;
}
/**
* Specify the action for remove().
*
* @param action the action for remove()
*/
protected void setRemoveAction(RemoveAction action)
{
m_actionRemove = action;
}
/**
* Return the registered action for containsValue().
*
* @return the registered action for containsValue()
*/
protected ContainsValueAction getContainsValueAction()
{
return m_actionContainsValue;
}
/**
* Specify the action for containsValue().
*
* @param action the action for containsValue()
*/
protected void setContainsValueAction(ContainsValueAction action)
{
m_actionContainsValue = action;
}
// ----- Object methods -------------------------------------------------
/**
* Compares the specified object with this map for equality. Returns
* true if the given object is also a map and the two maps
* represent the same mappings. More formally, two maps t1 and
* t2 represent the same mappings if
* t1.keySet().equals(t2.keySet()) and for every key k
* in t1.keySet(), (t1.get(k)==null ? t2.get(k)==null :
* t1.get(k).equals(t2.get(k))) . This ensures that the
* equals method works properly across different implementations
* of the map interface.
*
* @param oThat object to be compared for equality with this Map
*
* @return true if the specified object is equal to this Map
*/
public boolean equals(Object oThat)
{
if (oThat == this)
{
return true;
}
if (!(oThat instanceof Map))
{
return false;
}
Map mapThat = (Map) oThat;
if (mapThat.size() != this.size())
{
return false;
}
for (Iterator iter = mapThat.entrySet().iterator(); iter.hasNext(); )
{
Map.Entry entryThat = (Map.Entry) iter.next();
Entry entryThis = getEntryInternal(entryThat.getKey());
if (!Base.equals(entryThis, entryThat))
{
return false;
}
}
return true;
}
/**
* Returns the hash code value for this Map. The hash code of a Map is
* defined to be the sum of the hash codes of each entry in the Map's
* entrySet() view. This ensures that t1.equals(t2)
* implies that t1.hashCode()==t2.hashCode() for any two maps
* t1 and t2, as required by the general contract of
* Object.hashCode.
*
* @return the hash code value for this Map
*/
public int hashCode()
{
int nHash = 0;
for (Iterator iter = entrySet().iterator(); iter.hasNext(); )
{
nHash += iter.next().hashCode();
}
return nHash;
}
/**
* Returns a String representation of this map.
*
* @return a String representation of this map
*/
public String toString()
{
StringBuffer buf = new StringBuffer();
buf.append("{");
Iterator iter = entrySet().iterator();
boolean fHasNext = iter.hasNext();
while (fHasNext)
{
Entry entry = (Entry) iter.next();
Object oKey = entry.getKey();
Object oValue = entry.getValue();
buf.append(oKey == this ? "(this Map)" : oKey);
buf.append("=");
buf.append(oValue == this ? "(this Map)" : oValue);
if (fHasNext = iter.hasNext())
{
buf.append(", ");
}
}
buf.append("}");
return buf.toString();
}
// ----- Map interface --------------------------------------------------
/**
* Returns the number of key-value mappings in this map.
*
* Note: Unlike some Map implementations, the size() operation on
* this map may be relatively expensive.
*
* @return the number of key-value mappings in this map
*/
public int size()
{
// On a JSR-133 compliant JVM, doing a read from the atomic counter
// will guarantee up-to-date values for the cEntries field on all
// segments. On non-compliant JVMs, we accept that the value returned
// may be stale.
m_atomicLocks.get();
// If there are no synthetic entries, just sum the Entry counters.
// Subclasses that may introduce synthetic entries must override
// size() to provide a suitable implementation to account for
// synthetics.
int cEntries = 0;
Segment[] aSegment = m_aSegment;
for (int i = 0; i < SEGMENT_COUNT; i++)
{
cEntries += aSegment[i].cEntries;
}
return cEntries;
}
/**
* Returns true if this map contains no key-value mappings.
*
* @return true if this map contains no key-value mappings
*/
public boolean isEmpty()
{
return size() == 0;
}
/**
* Returns true iff this map contains a mapping for the specified
* key.
*
* @param oKey key whose presence in this map is to be tested
*
* @return true iff this map contains a mapping for the specified
* key
*/
public boolean containsKey(Object oKey)
{
return getEntryInternal(oKey) != null;
}
/**
* Returns true if this map maps one or more keys to the
* specified value.
*
* @param oValue value whose presence in this map is to be tested
*
* @return true if this map maps one or more keys to the
* specified value
*/
public boolean containsValue(Object oValue)
{
ContainsValueAction actionEntry = getContainsValueAction();
// if there was an intervening resize and the value was not found,
// apply the action again as the Entry may have been missed.
Entry[] aeBucket;
do
{
Object oContext = actionEntry.instantiateContext(oValue);
aeBucket = getStableBucketArray();
invokeOnAllKeys(oContext, /*fLock*/ false, actionEntry);
if (actionEntry.isFound(oContext))
{
return true;
}
}
while (aeBucket != getStableBucketArray());
return false;
}
/**
* Returns the value to which this map maps the specified key. Returns
* null if the map contains no mapping for this key. A return
* value of null does not necessarily indicate that the map
* contains no mapping for the key; it's also possible that the map
* explicitly maps the key to null. The containsKey
* operation may be used to distinguish these two cases.
*
* @param oKey key whose associated value is to be returned
*
* @return the value to which this map maps the specified key, or
* null if the map contains no mapping for this key
*/
public Object get(Object oKey)
{
Entry entry = getEntryInternal(oKey);
return (entry == null ? null : entry.getValue());
}
/**
* Locate an Entry in the this map based on its key.
*
* @param key the key object to search for
*
* @return the Entry or null if the entry does not exist
*/
public Map.Entry getEntry(Object key)
{
return getEntryInternal(key);
}
/**
* Associates the specified value with the specified key in this map. If
* the map previously contained a mapping for this key, the old value is
* replaced.
*
* @param oKey key with which the specified value is to be associated
* @param oValue value to be associated with the specified key
*
* @return previous value associated with specified key, or null
* if there was no mapping for key. A null return can
* also indicate that the map previously associated null
* with the specified key
*/
public Object put(Object oKey, Object oValue)
{
Object oOrig = putInternal(oKey, oValue);
return oOrig == NO_VALUE ? null : oOrig;
}
/**
* Copies all of the mappings from the specified map to this map.
* putAll is semantically equivalent to:
*
* for (Iterator iter = mapOther.entrySet().iterator(); iter.hasNext(); )
* {
* Map.Entry entry = (Map.Entry) iter.next();
* put(entry.getKey(), entry.getValue());
* }
*
*
* @param mapOther mappings to be stored in this map
*/
public void putAll(Map mapOther)
{
// If the insert is fairly small, take the naive approach of iterating
// over the supplied Map's entrySet and inserting each entry.
//
// For a large insert, we would expect to have to lock each segment
// one or more times with the naive approach. Instead lock all of the
// segments at once and do the updates/inserts while holding all of
// the segment-locks.
int nSizeOther = mapOther.size();
if (nSizeOther < PUTALL_THRESHOLD)
{
// take the naive approach
for (Iterator iter = mapOther.entrySet().iterator();
iter.hasNext();)
{
Map.Entry entryOther = (Map.Entry) iter.next();
putInternal(entryOther.getKey(), entryOther.getValue());
}
}
else
{
// Define an action to update or insert.
// The action will run while all segments are locked.
EntryAction actionPut = new InsertAction()
{
public Object invokeFound(Object oKey,
Object oContext,
Entry[] aeBucket,
int nBucket,
Entry entryPrev,
Entry entryCur)
{
Object oResult = super.invokeFound(oKey,
oContext,
aeBucket,
nBucket,
entryPrev,
entryCur);
if (oResult == NO_VALUE)
{
// update
entryCur.setValueInternal(oContext);
}
return oResult;
}
};
// lock all of the buckets, and perform the updates/inserts while
// holding all of the segment locks. For a big putAll(), this is
// faster than locking each key individually.
lockAllBuckets();
try
{
if (nSizeOther > m_cSegmentCapacity * SEGMENT_COUNT)
{
// grow to at least the size of the user-supplied map
grow(nSizeOther);
}
for (Iterator iter = mapOther.entrySet().iterator();
iter.hasNext();)
{
Map.Entry entryOther = (Map.Entry) iter.next();
// invoke the put action without additional locking
// because we have already locked all buckets
invokeOnKey(entryOther.getKey(), entryOther.getValue(),
/*fLock*/ false, actionPut);
}
}
finally
{
unlockAllBuckets();
}
}
}
/**
* Removes the mapping for this key from this map if present.
*
* @param oKey key whose mapping is to be removed from the map
*
* @return previous value associated with specified key, or null
* if there was no mapping for key. A null return can
* also indicate that the map previously associated null
* with the specified key
*/
public Object remove(Object oKey)
{
Object oOrig = removeInternal(oKey, getRemoveAction(), /*oContext*/null);
return oOrig == NO_VALUE ? null : oOrig;
}
/**
* Removes all mappings from this map.
*/
public void clear()
{
// If there are no synthetic entries, just clear and reset the bucket
// array. Subclasses that may introduce synthetic entries must
// override clear() to provide a suitable implementation to account
// for synthetics.
lockAllBuckets();
try
{
m_aeBucket = new Entry[DEFAULT_INITIALSIZE];
// clear the per-segment entry counters
Segment[] aSegment = m_aSegment;
for (int i = 0; i < SEGMENT_COUNT; i++)
{
aSegment[i].cEntries = 0;
}
m_cSegmentCapacity =
Math.max((int) (DEFAULT_INITIALSIZE * m_flLoadFactor) / SEGMENT_COUNT,
MIN_SEGMENT_CAPACITY);
}
finally
{
unlockAllBuckets();
}
}
/**
* Returns a set view of the mappings contained in this map. Each element
* in the returned set is a Map.Entry. The set is backed by the
* map, so changes to the map are reflected in the set, and vice-versa.
* If the map is modified while an iteration over the set is in progress,
* the results of the iteration are undefined. The set supports element
* removal, which removes the corresponding mapping from the map, via the
* Iterator.remove, Set.remove, removeAll,
* retainAll and clear operations. It does not support
* the add or addAll operations.
*
* @return a set view of the mappings contained in this map.
*/
public Set entrySet()
{
// no need to synchronize; it is acceptable that two threads would
// instantiate an entry set
EntrySet set = m_setEntries;
if (set == null)
{
m_setEntries = set = instantiateEntrySet();
}
return set;
}
/**
* Returns a Set view of the keys contained in this map. The Set is
* backed by the map, so changes to the map are reflected in the Set,
* and vice-versa. (If the map is modified while an iteration over
* the Set is in progress, the results of the iteration are undefined.)
* The Set supports element removal, which removes the corresponding entry
* from the map, via the Iterator.remove, Set.remove, removeAll
* retainAll, and clear operations. It does not support the add or
* addAll operations.
*
* @return a Set view of the keys contained in this map
*/
public Set keySet()
{
// no need to synchronize; it is acceptable that two threads would
// instantiate a key set
KeySet set = m_setKeys;
if (set == null)
{
m_setKeys = set = instantiateKeySet();
}
return set;
}
/**
* Returns a collection view of the values contained in this map. The
* collection is backed by the map, so changes to the map are reflected in
* the collection, and vice-versa. If the map is modified while an
* iteration over the collection is in progress, the results of the
* iteration are undefined. The collection supports element removal,
* which removes the corresponding mapping from the map, via the
* Iterator.remove, Collection.remove,
* removeAll, retainAll and clear operations.
* It does not support the add or addAll operations.
*
* @return a collection view of the values contained in this map
*/
public Collection values()
{
// no need to synchronize; it is acceptable that two threads would
// instantiate a values collection
ValuesCollection col = m_colValues;
if (col == null)
{
m_colValues = col = instantiateValuesCollection();
}
return col;
}
// ----- helpers --------------------------------------------------------
/**
* Initialize the EntryAction's for this map.
*/
protected void initializeActions()
{
/* getEntryInternal action */
setGetEntryAction(instantiateGetEntryAction());
/* insert action */
setInsertAction(instantiateInsertAction());
/* remove action */
setRemoveAction(instantiateRemoveAction());
/* containsValue action */
setContainsValueAction(instantiateContainsValueAction());
}
/**
* Locate an Entry in the hash map based on its key.
*
* @param oKey the key object to search for
*
* @return the Entry or null
*/
protected Entry getEntryInternal(Object oKey)
{
return getEntryInternal(oKey, /*fSynthetic*/ false);
}
/**
* Locate an Entry in the hash map based on its key.
*
* @param oKey the key object to search for
* @param fSynthetic include synthetic Entry objects representing keys
* that are not contained in the map
*
* @return the Entry or null
*/
protected Entry getEntryInternal(Object oKey, boolean fSynthetic)
{
Object oResult;
Boolean FSynthetic = fSynthetic ? Boolean.TRUE : Boolean.FALSE;
do
{
oResult = invokeOnKey(oKey, FSynthetic,
/*fLock*/ false, getGetEntryAction());
}
while (oResult == NO_VALUE);
return (Entry) oResult;
}
/**
* Associates the specified value with the specified key in this map. If
* the map previously contained a mapping for this key, the old value is
* replaced.
*
* @param oKey key with which the specified value is to be associated
* @param oValue value to be associated with the specified key
*
* @return previous value associated with specified key, or
* NO_VALUE if there was no mapping for key. A
* null return indicates that the map previously
* associated null with the specified key
*/
protected Object putInternal(Object oKey, Object oValue)
{
return putInternal(oKey, oValue, /*fOnlyIfAbsent*/false);
}
/**
* Associates the specified value with the specified key in this map. If
* the map previously contained a mapping for this key, the old value is
* replaced.
*
* @param oKey key with which the specified value is to be associated
* @param oValue value to be associated with the specified key
* @param fOnlyIfAbsent if true, perform a logical insert only; no action
* if the key already exists
*
* @return previous value associated with specified key, or
* NO_VALUE if there was no mapping for key. A
* null return indicates that the map previously
* associated null with the specified key
*/
protected Object putInternal(Object oKey, Object oValue, boolean fOnlyIfAbsent)
{
while (true)
{
Entry entry = getEntryInternal(oKey, /*fSynthetic*/ true);
if (entry == null || entry.isSynthetic())
{
// If an entry is not found, or the entry is synthetic, it is a
// logical insert (and must be done under a segment lock).
Object oResult = invokeOnKey(oKey, oValue, /*fLock*/ true,
getInsertAction());
if (oResult == NO_VALUE)
{
// another thread inserted; try again
continue;
}
if (entry == null)
{
// An Entry was added. Check to see if the map should be
// grown in order to encourage balanced buckets.
ensureLoadFactor(getSegmentForKey(oKey));
}
// successfully inserted
return NO_VALUE;
}
// In the case of an update, it is possible that the Entry is
// being removed or the value is being updated by another thread
return !fOnlyIfAbsent ? entry.setValueInternal(oValue) : NO_VALUE;
}
}
/**
* Removes the mapping for this key from this map if present.
*
* @param oKey key whose mapping is to be removed from the map
* @param actionRemove the EntryAction to apply
* @param oContext the context for the remove action
*
* @return previous value associated with specified key, or
* NO_VALUE if there was no mapping for key. A
* null return indicates that the map previously
* associated null with the specified key.
*/
protected Object removeInternal(Object oKey, EntryAction actionRemove, Object oContext)
{
return invokeOnKey(oKey, oContext, /*fLock*/ true, actionRemove);
}
/**
* Apply the specified toArray() action to the entries in the map. The
* toArray() action is not applied under any segment lock and is expected
* to accept a List instance as a context.
*
* @param action the toArray() action
* @param a the array into which the elements of the Collection
* are to be stored, if it is big enough; otherwise, a
* new array of the same runtime type is allocated for
* this purpose
*
* @return an array containing the elements returned by the specified
* action
*
* @throws ArrayStoreException if the runtime type of the specified array
* is not a supertype of the runtime type of the elements returned
* by the specified action
*/
protected Object[] toArrayInternal(IterableEntryAction action, Object[] a)
{
List list = new ArrayList();
Entry[] aeBucket;
// if there was an intervening resize, apply the action again as
// Entry's may have been missed or repeated during the resize.
do
{
list.clear();
aeBucket = getStableBucketArray();
invokeOnAllKeys(list, /*fLock*/ false, action);
}
while (aeBucket != getStableBucketArray());
return list.toArray(a == null ? new Object[list.size()] : a);
}
/**
* Check whether or not the specified segment is overloaded and if so,
* grow the bucket array (which suggests with high probability that the
* per-segment load will decrease).
*
* @param segment the segment to ensure the load-factor for
*/
protected void ensureLoadFactor(Segment segment)
{
// Note: reading the segment size is a dirty read. The caller is
// expected to have recently held the segment lock.
if (segment.cEntries > m_cSegmentCapacity)
{
Entry[] aeBucket = m_aeBucket;
lockAllBuckets();
try
{
// check that some other thread hasn't already grown
if (m_aeBucket == aeBucket)
{
grow();
}
}
finally
{
unlockAllBuckets();
}
}
}
/**
* Resize the bucket array, rehashing all Entries.
*
* Note: caller of this method is expected to hold locks on all segments of
* the map while making this call.
*/
protected void grow()
{
Entry[] aeOld = m_aeBucket;
int cOld = aeOld.length;
// check if there is no more room to grow
if (cOld >= BIGGEST_MODULO)
{
return;
}
// calculate growth
int cNew = (int) Math.min((long) (cOld * (1F + m_flGrowthRate)),
BIGGEST_MODULO);
grow(cNew);
}
/**
* Resize the bucket array to the specified size, rehashing all Entries.
*
* Note: caller of this method is expected to hold locks on all segments of
* the map while making this call.
*
* @param cNew the minimum size to attempt to grow to
*/
protected synchronized void grow(int cNew)
{
if (isActiveIterator())
{
// don't grow if there are active iterators
return;
}
synchronized (RESIZING)
{
// store off the old bucket array
Entry[] aeOld = m_aeBucket;
int cOld = aeOld.length;
// use NO_ENTRIES to signify that a resize is taking place
m_aeBucket = NO_ENTRIES;
if (cNew <= cOld)
{
// very low growth rate or very low initial size. In either
// case, ensure that at least some growth happens.
cNew = cOld + 1;
}
// use a prime number at least as big than the new size
int[] aiModulo = PRIME_MODULO;
int cModulos = aiModulo.length;
for (int i = 0; i < cModulos; ++i)
{
int iModulo = aiModulo[i];
if (iModulo >= cNew)
{
cNew = iModulo;
break;
}
}
// create a new bucket array; in the case of an OutOfMemoryError
// be sure to restore the old bucket array
Entry[] aeNew;
try
{
aeNew = new Entry[cNew];
}
catch (OutOfMemoryError e)
{
m_aeBucket = aeOld;
throw e;
}
// rehash
Segment[] aSegment = m_aSegment;
for (int i = 0; i < cOld; ++i)
{
Entry entry = aeOld[i];
int nSegmentOld = getSegmentIndex(i);
while (entry != null)
{
// store off the next Entry
// (it is going to get hammered otherwise)
Entry entryNext = entry.nextEntry(true);
// rehash the Entry into the new bucket array
Object oKey = entry.getKey();
int nHash = oKey == null ? 0 : oKey.hashCode();
int nBucket = getBucketIndex(nHash, cNew);
int nSegmentNew = getSegmentIndex(nBucket);
entry.setNext(aeNew[nBucket]);
aeNew[nBucket] = entry;
// update the counters
--aSegment[nSegmentOld].cEntries;
++aSegment[nSegmentNew].cEntries;
// process next Entry in the old list
entry = entryNext;
}
}
// store updated bucket array
m_cSegmentCapacity =
Math.max((int) (cNew * m_flLoadFactor) / SEGMENT_COUNT,
MIN_SEGMENT_CAPACITY);
m_aeBucket = aeNew;
}
}
/**
* Perform an action on all Entries in the map. The action is provided as
* an EntryAction and (if the fLock is specified) it are
* invoked while holding the locks for all segments.
*
* The semantics of invokeOnAllKeys are equivalent to:
*
* for (Iterator iter = entrySet().iterator(); iter.hasNext(); )
* {
* Entry entry = (Entry) iter.next();
* actionEntry.invokeFound(...);
* }
* return oContext;
*
* Except that if fLock is specified, it is performed atomically
* while holding all segment-locks.
*
* @param oContext opaque context for the specified action
* @param fLock true if all segment-locks should be
* acquired before invoking the specified action
* @param actionEntry the action to perform for each entry
*
* @return the specified opaque context
*/
protected Object invokeOnAllKeys(Object oContext, boolean fLock,
IterableEntryAction actionEntry)
{
if (fLock)
{
lockAllBuckets();
}
try
{
Entry[] aeBucket = getStableBucketArray();
int cBuckets = aeBucket.length;
for (int iBucket = 0; iBucket < cBuckets; ++iBucket)
{
// walk all entries in the bucket
Entry entry = aeBucket[iBucket];
Entry entryPrev = null;
while (entry != null)
{
// invoke the action on each Entry
Entry entryNext = entry.nextEntry(/*fSynthetic*/ true);
actionEntry.invokeFound(entry.getKey(), oContext, aeBucket,
iBucket, entryPrev, entry);
if (actionEntry.isComplete(oContext))
{
return oContext;
}
// recalculate entryPrev for the next Entry. Have to
// consider that the action may have removed this Entry
// from the bucket-list.
entryPrev = ((entryPrev == null ?
aeBucket[iBucket] != entry :
entryPrev.nextEntry(true) != entry) ?
entryPrev : entry);
entry = entryNext;
}
}
}
finally
{
if (fLock)
{
unlockAllBuckets();
}
}
return oContext;
}
/**
* Perform an action on the specified key. The action operation is
* provided as an EntryAction and (if fLock is specified), is
* invoked while holding the appropriate segment lock for the key.
*
* The semantics of invokeOnKey are equivalent to:
*
* Object oResult;
* if (containsKey(oKey))
* {
* oResult = action.invokeFound(...);
* }
* else
* {
* oResult = action.invokeNotFound(...);
* }
* return oResult;
*
* Except that if fLock is specified, it is performed atomically
* while holding the segment-lock.
*
* @param oKey the key to act on
* @param oContext opaque context for the specified action
* @param fLock true iff the segment should be locked before invoking
* the specified action
* @param action the action to invoke
*
* @return the result of performing the action
*/
protected Object invokeOnKey(Object oKey, Object oContext,
boolean fLock, EntryAction action)
{
int nHash = (oKey == null ? 0 : oKey.hashCode());
while (true)
{
Entry[] aeBucket = getStableBucketArray();
int cBuckets = aeBucket.length;
int nBucket = getBucketIndex(nHash, cBuckets);
if (fLock)
{
lockBucket(nBucket);
}
try
{
if (aeBucket != m_aeBucket)
{
continue;
}
// walk the linked list of entries (open hash) in the bucket
Entry entryCur = aeBucket[nBucket];
Entry entryPrev = null;
while (entryCur != null)
{
Entry entryNext = entryCur.nextEntry(true);
// optimization: check hash first
if (nHash == entryCur.m_nHash &&
Base.equals(oKey, entryCur.getKey()))
{
return action.invokeFound(oKey, oContext, aeBucket,
nBucket, entryPrev, entryCur);
}
entryPrev = entryCur;
entryCur = entryNext;
}
return action.invokeNotFound(oKey, oContext, aeBucket, nBucket);
}
finally
{
if (fLock)
{
unlockBucket(nBucket);
}
}
}
}
/**
* Calculate the bucket number for a particular hash code.
*
* @param nHash the hash code
* @param cBuckets the number of buckets
*
* @return the bucket index for the specified hash code
*/
protected int getBucketIndex(int nHash, int cBuckets)
{
return (int) ((((long) nHash) & 0xFFFFFFFFL) % ((long) cBuckets));
}
/**
* Calculate the segment index for the the specified bucket.
*
* @param nBucket the bucket number
*
* @return the segment index
*/
protected int getSegmentIndex(int nBucket)
{
return nBucket % SEGMENT_COUNT;
}
/**
* Return the Segment object for the specified key.
*
* @param oKey the key
*
* @return the Segment for the specified key
*/
protected Segment getSegmentForKey(Object oKey)
{
Entry[] aeBucket = getStableBucketArray();
int cBuckets = aeBucket.length;
int nHash = (oKey == null ? 0 : oKey.hashCode());
int nBucket = getBucketIndex(nHash, cBuckets);
return m_aSegment[getSegmentIndex(nBucket)];
}
/**
* Get the bucket array, or if a resize is occurring, wait for the resize
* to complete and return the new bucket array.
*
* @return the latest bucket array
*/
protected Entry[] getStableBucketArray()
{
// get the bucket array
Entry[] aeBucket = m_aeBucket;
if (aeBucket == NO_ENTRIES)
{
// wait for the ongoing resize to complete
synchronized (RESIZING)
{
return m_aeBucket;
}
}
return aeBucket;
}
/**
* Register the activation of an Iterator.
*
* Note: The map will not grow while there are any active iterators.
*
* @param iter the activated iterator
*/
protected synchronized void iteratorActivated(Iterator iter)
{
Object oIterActive = m_oIterActive;
if (oIterActive == null)
{
// optimize for single active iterator
m_oIterActive = new WeakReference(iter);
}
else if (oIterActive instanceof WeakReference)
{
Object oPrev = ((WeakReference) oIterActive).get();
if (oPrev == null)
{
// previous Iterator was GC'd, replace it
m_oIterActive = new WeakReference(iter);
}
else
{
// switch from single to multiple active iterators;
// WeakHashMap will automatically clean up any GC'd iterators
Map map = new WeakHashMap();
m_oIterActive = map;
map.put(oPrev, null);
map.put(iter, null);
}
}
else
{
// it's a map, add additional active iterator
((Map) oIterActive).put(iter, null);
}
}
/**
* Release the (formerly-active) Iterator.
*
* Note: This method could be used to accelerate the destruction of
* unexhausted iterators.
*
* @param iter the iterator to be released
*/
public synchronized void releaseIterator(Iterator iter)
{
Object oIterActive = m_oIterActive;
if (oIterActive instanceof WeakReference)
{
// there is only one active iterator, it must be this one
if (((WeakReference) oIterActive).get() == iter)
{
m_oIterActive = null;
}
}
else
{
// remove one of many active iterators
((Map) oIterActive).remove(iter);
// once we go to an Iterator Map we don't switch back; the
// assumption is that if multiple concurrent iterators are
// allocated once, they will be allocated again, so avoid
// thrashing on frequent WeakHashMap creation
}
}
/**
* Determine if there are any active Iterators, which may mean that they
* are in the middle of iterating over the Map.
*
* @return true iff there is at least one active Iterator
*/
protected synchronized boolean isActiveIterator()
{
Object oIterActive = m_oIterActive;
if (oIterActive == null)
{
return false;
}
else if (oIterActive instanceof WeakReference)
{
return ((WeakReference) oIterActive).get() != null;
}
else
{
return !((Map) oIterActive).isEmpty();
}
}
/**
* Attempt to lock the segment corresponding to the specified bucket.
*
* @param nBucket the bucket index
*
* @return true iff the segment was successfully locked
*/
protected boolean lockBucket(int nBucket)
{
return lockSegment(getSegmentIndex(nBucket), true);
}
/**
* Attempt to lock the specified segment.
*
* @param nSegment the segment to lock
* @param fBlock should we block on trying to lock the segment
*
* @return true iff the segment was successfully locked
*/
protected boolean lockSegment(int nSegment, boolean fBlock)
{
long maskLock = 1L << nSegment;
AtomicLong atomicLocks = m_atomicLocks;
// optimization: assume nothing is already locked
if (atomicLocks.compareAndSet(LOCKS_NONE, maskLock))
{
// optimistic case worked; we locked the segment
return true;
}
// check to see if everything is locked or a lock-all is pending, or
// if the desired segment is locked
for (int cAttempts = 0; true;)
{
long lLocks = atomicLocks.get();
if ((lLocks & (LOCK_ALL_PENDING | maskLock)) != 0L)
{
if (fBlock)
{
// spin a few times before waiting
if ((++cAttempts % SEGMENT_LOCK_MAX_SPIN) == 0)
{
contendForSegment(nSegment);
}
continue;
}
else
{
return false;
}
}
if (atomicLocks.compareAndSet(lLocks, lLocks | maskLock))
{
return true;
}
}
}
/**
* Unlock the segment corresponding to the specified bucket that was
* previously locked using the {@link #lockBucket} method.
*
* @param nBucket the bucket to unlock
*/
protected void unlockBucket(int nBucket)
{
unlockSegment(getSegmentIndex(nBucket));
}
/**
* Unlock the specified segment previously locked using the {@link
* #lockSegment} method.
*
* @param nSegment the segment to unlock
*/
protected void unlockSegment(int nSegment)
{
long maskLock = 1L << nSegment;
AtomicLong atomicLocks = m_atomicLocks;
long lLocks = maskLock;
// make an optimistic attempt (first loop-iteration) in assuming that
// nothing other than nSegment is locked
while (!atomicLocks.compareAndSet(lLocks, lLocks & ~maskLock))
{
lLocks = atomicLocks.get();
}
// Check to see if there are other threads contending for the segment
// that we just unlocked that need to be notified
Segment segment = m_aSegment[nSegment];
if (segment.fContend)
{
synchronized (segment)
{
if (segment.fContend)
{
segment.fContend = false;
segment.notifyAll();
}
}
}
}
/**
* Wait for a segment to be unlocked.
*
* @param nSegment the segment-lock to be waited for
*/
protected void contendForSegment(int nSegment)
{
long maskLock = 1L << nSegment;
Segment segment = m_aSegment[nSegment];
synchronized (segment)
{
// set the contended bit for the segment
segment.fContend = true;
// check that the segment is still locked
if ((m_atomicLocks.get() & maskLock) == 0)
{
// The segment was unlocked; try to lock again.
return;
}
// At this point, we know the lock and contend bits are set, so we
// know that there is a segment-lock holder that should notify us
// when they unlock it.
try
{
Blocking.wait(segment, 1000);
}
catch (InterruptedException e)
{
Thread.currentThread().interrupt();
throw Base.ensureRuntimeException(e, "Segment lock interrupted");
}
}
}
/**
* Lock everything. This method will not return until the lock is placed.
* It must not be called on a thread that could already hold a lock.
*/
protected void lockAllBuckets()
{
int cAttempts = 0;
while (!lockAllSegments(LOCKS_NONE))
{
// wait for (possible) resize to finish
getStableBucketArray();
// spin a few times before waiting
if ((++cAttempts % SEGMENT_LOCK_MAX_SPIN) == 0)
{
contendForSegment(LOCK_ALL_PENDING_IDX);
}
}
}
/**
* Lock everything, assuming that the segment for the specified bucket has
* already been locked.
*
* @param nBucketAlreadyLocked the bucket that was already locked.
*
* @return false if the operation failed because another thread
* was also trying to lock everything (indicating potential
* deadlock)
*/
protected boolean lockAllBuckets(int nBucketAlreadyLocked)
{
return lockAllSegments(1L << (getSegmentIndex(nBucketAlreadyLocked)));
}
/**
* Unlock everything.
*/
protected void unlockAllBuckets()
{
unlockAllSegments(LOCKS_NONE);
}
/**
* Unlock everything, leaving only the segment for the specified bucket
* locked.
*
* @param nBucketLeaveLocked the bucket that was already locked
*/
protected void unlockAllBuckets(int nBucketLeaveLocked)
{
unlockAllSegments(1L << (getSegmentIndex(nBucketLeaveLocked)));
}
/**
* Lock all segments except for the specified segments that have already
* been locked by the calling thread.
*
* @param lLocksHeld the bit-mask representing all segment-locks that the
* calling thread already holds
*
* @return false if the operation failed because another thread was also
* trying to lock everything (indicating potential deadlock)
*/
protected boolean lockAllSegments(long lLocksHeld)
{
AtomicLong atomicLocks = m_atomicLocks;
// the passed-in mask must already be locked
/*
Base.azzert(lLocksHeld == LOCKS_NONE ||
(atomicLocks.getCount() & lLocksHeld) == lLocksHeld);
*/
// optimistic attempt: lock everything
if (atomicLocks.compareAndSet(lLocksHeld, LOCKS_ALL))
{
// everything else was available and is now locked
return true;
}
// pessimistic attempt: attempt to lock the "intent" lock
while (true)
{
long lLocks = atomicLocks.get();
if ((lLocks & LOCK_ALL_PENDING) != 0L)
{
return false;
}
if (atomicLocks.compareAndSet(lLocks, lLocks | LOCK_ALL_PENDING))
{
lLocksHeld |= LOCK_ALL_PENDING;
break;
}
}
// lock all remaining unlocked segments
int cAttempts = 0;
while (lLocksHeld != LOCKS_ALL)
{
// find what locks are held in total by this and all other threads
long lLocks = atomicLocks.get();
// request all the locks that are not currently held by any thread
if (atomicLocks.compareAndSet(lLocks, LOCKS_ALL))
{
// we just locked all of the available segment-locks; continue
// until we have locked all of the segment-locks.
lLocksHeld |= ~lLocks;
}
else
{
// occasionally stop spinning and yield
if ((++cAttempts % SEGMENT_LOCK_MAX_SPIN) == 0)
{
Thread.yield();
}
}
}
return true;
}
/**
* Unlock all segments, except the segment-locks indicated by the specified
* bit-vector. This method must only be called by a thread if that thread
* has successfully called {@link #lockAllSegments}.
*
* @param lLocksKeep the segment-locks to keep locked
*/
protected void unlockAllSegments(long lLocksKeep)
{
if (!m_atomicLocks.compareAndSet(LOCKS_ALL, lLocksKeep))
{
throw new IllegalStateException();
}
// check to see if any threads are contending for segment-locks that
// we just unlocked (or the lock-all intention bit).
Segment[] aSegment = m_aSegment;
for (int i = 0; i < LOCK_COUNT; i++)
{
long maskLock = 1L << i;
Segment segment = aSegment[i];
// if we are not keeping this segment-lock and it is being waited
// on, we need to notify the waiters
if ((lLocksKeep & maskLock) == 0 && segment.fContend)
{
synchronized (segment)
{
if (segment.fContend)
{
segment.fContend = false;
segment.notifyAll();
}
}
}
}
}
// ----- inner interface: EntryAction -----------------------------------
/**
* An EntryAction encapsulates a logical action to be executed in the
* context of a key (that may or may not exist in the map). If an Entry
* exists in the map for the key, invokeFound is called; otherwise
* invokeNotFound is called.
*
* EntryAction instances are state-aware and are invoked with an opaque
* context object supplied by the action client. The meanings of the
* supplied context and return-values are polymorphic in EntryAction type.
* Depending on the EntryAction type, it may or may not be invoked while
* holding the segment lock for the key.
*
* An EntryAction should not acquire additional segment-locks.
*/
protected static interface EntryAction
{
/**
* Invoke some action, holding the segment lock, when a matching Entry
* is found.
*
* @param oKey the key to which the action is applied
* @param oContext opaque context specific to the action
* @param aeBucket the bucket array
* @param nBucket the index into the bucket array
* @param entryPrev the Entry object immediately preceding the
* Entry that was found, or null
* @param entryCur the Entry object that was found
*
* @return an opaque result value
*/
public Object invokeFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket,
Entry entryPrev, Entry entryCur);
/**
* Invoke some action, holding the segment lock, when no matching Entry
* is found.
*
* @param oKey the key to which the action is applied
* @param oContext opaque context specific to the action
* @param aeBucket the bucket array
* @param nBucket the index into the bucket array
*
* @return an opaque result value
*/
public Object invokeNotFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket);
}
// ----- inner interface: IterableEntryAction ---------------------------
/**
* IterableEntryAction is an EntryAction that is suitable for applying to
* all keys in a map.
*/
protected static interface IterableEntryAction
extends EntryAction
{
/**
* Return true iff further key evaluations for the given context are
* known to be unnecessary.
*
* @param oContext the action context
*
* @return true iff further evaluations are unnecessary
*/
public boolean isComplete(Object oContext);
}
// ----- inner class: EntryActionAdapter --------------------------------
/**
* EntryActionAdapter is a convenience class that provides default
* implementations for the EntryAction and IterableEntryAction interface
* methods.
*/
protected static abstract class EntryActionAdapter
implements EntryAction, IterableEntryAction
{
/**
* {@inheritDoc}
*/
public Object invokeFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket,
Entry entryPrev, Entry entryCur)
{
return NO_VALUE;
}
/**
* {@inheritDoc}
*/
public Object invokeNotFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket)
{
return NO_VALUE;
}
/**
* {@inheritDoc}
*/
public boolean isComplete(Object oContext)
{
return false;
}
}
// ----- inner class: GetEntryAction ------------------------------------
/**
* Factory for GetEntryAction
*
* @return a GetEntryAction
*/
protected GetEntryAction instantiateGetEntryAction()
{
return new GetEntryAction();
}
/**
* Action support for getEntryInternal. The action performs an Entry
* lookup by key and is not required to run while holding segment locks.
*
* The context object for a GetEntryAction is either Boolean.TRUE
* or Boolean.FALSE indicating whether or not to return synthetic
* entries.
*
* The result of invoking a GetEntryAction is the (possibly synthetic)
* Entry corresponding to a given key, null if no matching Entry
* is found, or NO_VALUE indicating that a concurrent resize
* occurred, and the operation must be repeated.
*/
protected class GetEntryAction
implements EntryAction
{
/**
* {@inheritDoc}
*/
public Object invokeFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket,
Entry entryPrev, Entry entryCur)
{
// An Entry object was found. If synthetic entries are to be
// returned, or the found entry is not synthetic, then return it.
boolean fSynthetic = oContext == Boolean.TRUE;
return (fSynthetic || !entryCur.isSynthetic()) ? entryCur : null;
}
/**
* {@inheritDoc}
*/
public Object invokeNotFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket)
{
// If a resize occurred, the oKey may be in the map and was
// reshuffled while we were looking for it; we know a resize
// occurred iff the hash bucket array changed, in which case we
// must return NO_VALUE so that the lookup is retried.
return aeBucket == m_aeBucket ? null : NO_VALUE;
}
}
// ----- inner class: InsertAction --------------------------------------
/**
* Factory for InsertAction
*
* @return an InsertAction
*/
protected InsertAction instantiateInsertAction()
{
return new InsertAction();
}
/**
* Action support for insert. The action performs locked insert, and is
* expected to run while holding the segment-lock for the specified key.
*
* The context object for an InsertAction is the value object to insert in
* the map, or NO_VALUE to insert a synthetic Entry.
*
* The result of invoking an InsertAction is the (possibly synthetic) Entry
* object that was inserted for the specified key, or NO_VALUE
* indicating that a mapping for the key already exists in the map.
*/
protected class InsertAction
implements EntryAction
{
/**
* {@inheritDoc}
*/
public Object invokeFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket,
Entry entryPrev, Entry entryCur)
{
if (entryCur.isSynthetic())
{
// logical insert
entryCur.setValueInternal(oContext);
// return the (logically) newly inserted entry
return entryCur;
}
// An entry was found under put(); return NO_VALUE
return NO_VALUE;
}
/**
* {@inheritDoc}
*/
public Object invokeNotFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket)
{
// No entry object was found under put(); having locked the
// segment, we are now free to insert a new Entry.
Object oValue = oContext;
int nHash = (oKey == null ? 0 : oKey.hashCode());
// instantiate and configure a new Entry
Entry entry = instantiateEntry(oKey, oValue, nHash);
// put the Entry in at the front of the list of entries for that
// bucket
entry.setNext(aeBucket[nBucket]);
aeBucket[nBucket] = entry;
// increment the entry counter for the segment
++m_aSegment[getSegmentIndex(nBucket)].cEntries;
// return the newly created entry
return entry;
}
}
// ----- inner class: RemoveAction --------------------------------------
/**
* Factory for RemoveAction
*
* @return a RemoveAction
*/
protected RemoveAction instantiateRemoveAction()
{
return new RemoveAction();
}
/**
* Action support for remove(). The action performs a locked remove, and is
* expected to run while holding the segment-lock for the specified key.
*
* The context object for a RemoveAction is unused.
*
* The result of invoking a RemoveAction is the previous value associated
* with the specified key, or NO_VALUE if no mapping for the key
* exists in the map. Note that a synthetic Entry does not represent a
* key-value mapping, so NO_VALUE is returned if a matching
* synthetic Entry is found.
*/
protected class RemoveAction
extends EntryActionAdapter
{
/**
* {@inheritDoc}
*/
public Object invokeFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket,
Entry entryPrev, Entry entryCur)
{
// Remove the entry from the bucket-list
Entry entryNext = entryCur.nextEntry(true);
if (entryPrev == null)
{
aeBucket[nBucket] = entryNext;
}
else
{
entryPrev.setNext(entryNext);
}
--m_aSegment[getSegmentIndex(nBucket)].cEntries;
return entryCur.getValueInternal();
}
}
// ----- inner class: ContainsValueAction -------------------------------
/**
* Factory for ContainsValueAction
*
* @return a ContainsValueAction
*/
protected ContainsValueAction instantiateContainsValueAction()
{
return new ContainsValueAction();
}
/**
* Action support for containsValue(). The action performs a lookup by
* value and is not required to run while holding any segment-locks.
*
* The context object for a ContainsValueAction is an opaque context
* created by instantiateContext.
*
* The result of invoking a ContainsValueAction is Boolean.TRUE if
* the value is found in the map or Boolean.FALSE if the value is
* not found in the map.
*/
protected static class ContainsValueAction
extends EntryActionAdapter
{
/**
* {@inheritDoc}
*/
public Object invokeFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket,
Entry entryPrev, Entry entryCur)
{
ContainsValueContext context = (ContainsValueContext) oContext;
if (Base.equals(entryCur.getValue(), context.m_oValue))
{
context.m_fFound = true;
}
return NO_VALUE;
}
/**
* {@inheritDoc}
*/
public boolean isComplete(Object oContext)
{
return isFound(oContext);
}
/**
* Return true iff the value was found
*
* @param oContext the ContainsValueContext object
*
* @return true iff the value was found
*/
public boolean isFound(Object oContext)
{
return ((ContainsValueContext) oContext).m_fFound;
}
/**
* Instantiate a context appropriate for applying ContainsValueAction
* to lookup oValue in the map.
*
* @param oValue the value to test the existence of
*
* @return a context to use with a ContainsValueAction
*/
public Object instantiateContext(Object oValue)
{
ContainsValueContext oContext = new ContainsValueContext();
oContext.m_oValue = oValue;
return oContext;
}
/**
* Context for ContainsValueAction.
*/
private static class ContainsValueContext
{
/**
* Has the value been found?
*/
private boolean m_fFound;
/**
* The value being searched for.
*/
private Object m_oValue;
}
}
// ----- inner class: Segment -------------------------------------------
/**
* Segment metadata.
*/
protected static class Segment
{
/**
* The number of Entry objects (including synthetics) in this segment.
*
* Note: On a JSR-133 non-compliant JVM, this value may be stale to
* some extent. It is only used as an internal growth statistic, so
* this is acceptable. On JSR-133 compliant JVMs, the segment lock
* under which this value is modified performs an atomic CAS which
* would guarantee that writes get flushed.
*/
protected int cEntries;
/**
* Are any threads contending to lock this segment?
*/
protected volatile boolean fContend;
}
// ----- inner class: Entry ---------------------------------------------
/**
* Factory for Entry.
*
* @param oKey the key
* @param oValue the value
* @param nHash the hashCode value of the key
*
* @return a new instance of the Entry class (or a subclass thereof)
*/
protected Entry instantiateEntry(Object oKey, Object oValue, int nHash)
{
return new Entry(oKey, oValue, nHash);
}
/**
* Return the first non-synthetic Entry object contained by in the
* specified bucket.
*
* @param aeBucket the array of hash buckets
* @param nBucket the bucket index
*
* @return the first non-synthetic Entry in the specified bucket or null
*/
protected static Entry entryFromBucket(Entry[] aeBucket, int nBucket)
{
Entry entry = aeBucket[nBucket];
return (entry != null && entry.isSynthetic()) ?
entry.nextEntry() : entry;
}
/**
* A map entry (key-value pair). The Map.entrySet method returns
* a collection-view of the map, whose elements are of this class.
*/
protected static class Entry
extends Base
implements Map.Entry
{
/**
* Construct an Entry object with the specified key, value and hash.
*
* @param oKey key with which the specified value is to be associated
* @param oValue value to be associated with the specified key
* @param nHash the hashCode for the specified key
*/
protected Entry(Object oKey, Object oValue, int nHash)
{
m_oKey = oKey;
m_oValue = oValue;
m_nHash = nHash;
}
/**
* Returns the key corresponding to this entry.
*
* @return the key corresponding to this entry
*/
public Object getKey()
{
return m_oKey;
}
/**
* Returns the value corresponding to this entry. If the mapping
* has been removed from the backing map (by the iterator's
* remove operation), the results of this call are undefined.
*
* @return the value corresponding to this entry
*/
public Object getValue()
{
Object oValue = m_oValue;
return oValue == SegmentedHashMap.NO_VALUE ? null : oValue;
}
/**
* Returns the value corresponding to this entry, or NO_VALUE
* if this Entry is synthetid. If the mapping has been removed from
* the backing map (by the iterator's remove operation), the
* results of this call are undefined.
*
* @return the value corresponding to this entry, or NO_VALUE
* if the Entry is synthetic
*/
protected Object getValueInternal()
{
return m_oValue;
}
/**
* Replaces the value corresponding to this entry with the specified
* value (writes through to the map). The behavior of this call is
* undefined if the mapping has already been removed from the map (by
* the iterator's remove operation).
*
* @param oValue new value to be stored in this entry
*
* @return old value corresponding to the entry
*/
public Object setValue(Object oValue)
{
Object oPrev = setValueInternal(oValue);
return oPrev == SegmentedHashMap.NO_VALUE ? null : oPrev;
}
/**
* Replaces the value corresponding to this entry with the specified
* value (writes through to the map). The behavior of this call is
* undefined if the mapping has already been removed from the map.
*
* @param oValue new value to be stored in this entry
*
* @return old value corresponding to the entry, or NO_VALUE
* if the Entry was synthetic
*/
protected Object setValueInternal(Object oValue)
{
Object oPrev = m_oValue;
m_oValue = oValue;
return oPrev;
}
/**
* Compares the specified object with this entry for equality.
* Returns true if the given object is also a map entry and
* the two entries represent the same mapping. More formally, two
* entries e1 and e2 represent the same mapping
* if:
*
* (e1.getKey()==null ?
* e2.getKey()==null : e1.getKey().equals(e2.getKey())) &&
* (e1.getValue()==null ?
* e2.getValue()==null : e1.getValue().equals(e2.getValue()))
*
* This ensures that the equals method works properly across
* different implementations of the Map.Entry interface.
*
* @param o object to be compared for equality with this map entry
*
* @return true if the specified object is equal to this map
* entry
*/
public boolean equals(Object o)
{
if (o instanceof Map.Entry)
{
Map.Entry that = (Map.Entry) o;
return this == that ||
Base.equals(this.getKey(), that.getKey()) &&
Base.equals(this.getValue(), that.getValue());
}
return false;
}
/**
* Returns the hash code value for this map entry. The hash code
* of a map entry e is defined to be:
*
* (e.getKey()==null ? 0 : e.getKey().hashCode()) ^
* (e.getValue()==null ? 0 : e.getValue().hashCode())
*
* This ensures that e1.equals(e2) implies that
* e1.hashCode()==e2.hashCode() for any two Entries
* e1 and e2, as required by the general contract of
* Object.hashCode.
*
* @return the hash code value for this map entry
*/
public int hashCode()
{
Object oKey = m_oKey;
Object oValue = getValue();
return (oKey == null ? 0 : m_nHash) ^
(oValue == null ? 0 : oValue.hashCode());
}
/**
* Render the map entry as a String.
*
* @return the details about this entry
*/
public String toString()
{
return "key=\"" + getKey() + '\"' + ", value=\"" + getValue() + '\"';
}
/**
* Is this Entry synthetic?
*
* @return true iff this Entry is synthetic
*/
protected boolean isSynthetic()
{
return false;
}
/**
* Set the next entry in the linked list (open hash)
*
* @param eNext the next Entry
*/
protected void setNext(Entry eNext)
{
m_eNext = eNext;
}
/**
* Get the next non-synthetic entry in the linked list (open hash)
*
* @return the next non-synthetic entry in the linked list
*/
protected Entry nextEntry()
{
return nextEntry(false);
}
/**
* Get the next entry in the linked list (open hash). If
* fSynthetic is specified, also return synthetic Entry
* objects. Synethetic entries are Entry objects logically associated
* with a given key, but do not represent a key-value mapping in this
* map.
*
* @param fSynthetic include synthetic Entry objects?
*
* @return the next Entry in the linked list
*/
protected Entry nextEntry(boolean fSynthetic)
{
Entry eNext = m_eNext;
while (eNext != null && !fSynthetic && eNext.isSynthetic())
{
eNext = eNext.m_eNext;
}
return eNext;
}
// ----- data members -----------------------------------------------
/**
* The key. This object reference will not change for the life of
* the Entry.
*/
protected final Object m_oKey;
/**
* The value. This object reference can change within the life of the
* Entry. This field is declared volatile so that get() and put() can
* proceed without performing any synchronization.
*/
protected volatile Object m_oValue;
/**
* The key's hash code. This value will not change for the life of the
* Entry.
*/
protected final int m_nHash;
/**
* The next entry in the linked list (an open hashing implementation).
* This field is declared volatile so the entry iterator can safely
* operate without performing any synchronization.
*/
protected volatile Entry m_eNext;
}
// ----- inner class: EntrySet ------------------------------------------
/**
* Factory for EntrySet
*
* @return a new instance of the EntrySet class (or a subclass thereof)
*/
protected EntrySet instantiateEntrySet()
{
return new EntrySet();
}
/**
* A set of entries backed by this map.
*/
protected class EntrySet
extends AbstractSet
{
// ----- Set interface ------------------------------------------
/**
* Returns an iterator over the elements contained in this collection.
*
* @return an iterator over the elements contained in this collection
*/
public Iterator iterator()
{
return SegmentedHashMap.this.isEmpty()
? NullImplementation.getIterator()
: instantiateIterator();
}
/**
* Returns the number of elements in this collection. If the collection
* contains more than Integer.MAX_VALUE elements, returns
* Integer.MAX_VALUE.
*
* @return the number of elements in this collection
*/
public int size()
{
return SegmentedHashMap.this.size();
}
/**
* Returns true if this collection contains the specified
* element. More formally, returns true if and only if this
* collection contains at least one element e such that
* (o==null ? e==null : o.equals(e)).
*
* @param o object to be checked for containment in this collection
*
* @return true if this collection contains the specified
* element
*/
public boolean contains(Object o)
{
if (o instanceof Map.Entry)
{
Map.Entry thatEntry = (Map.Entry) o;
Map.Entry thisEntry = SegmentedHashMap.this.getEntryInternal(thatEntry.getKey());
return thisEntry != null && thisEntry.equals(thatEntry);
}
return false;
}
/**
* Removes the specified element from this Set of entries if it is
* present by removing the associated entry from the underlying
* Map.
*
* @param o object to be removed from this set, if present
*
* @return true if the set contained the specified element
*/
public boolean remove(Object o)
{
if (contains(o))
{
SegmentedHashMap.this.remove(((Map.Entry) o).getKey());
return true;
}
else
{
return false;
}
}
/**
* Removes all of the elements from this set of Keys by clearing the
* underlying Map.
*/
public void clear()
{
SegmentedHashMap.this.clear();
}
/**
* Returns an array containing all of the elements in this collection.
* If the collection makes any guarantees as to what order its elements
* are returned by its iterator, this method must return the elements
* in the same order. The returned array will be "safe" in that no
* references to it are maintained by the collection. (In other words,
* this method must allocate a new array even if the collection is
* backed by an Array). The caller is thus free to modify the returned
* array.
*
* @return an array containing all of the elements in this collection
*/
public Object[] toArray()
{
return toArray((Object[]) null);
}
/**
* Returns an array with a runtime type is that of the specified array
* and that contains all of the elements in this collection. If the
* collection fits in the specified array, it is returned therein.
* Otherwise, a new array is allocated with the runtime type of the
* specified array and the size of this collection.
*
* If the collection fits in the specified array with room to spare
* (i.e., the array has more elements than the collection), the element
* in the array immediately following the end of the collection is set
* to null. This is useful in determining the length of the
* collection only if the caller knows that the collection does
* not contain any null elements.)
*
* @param ao the array into which the elements of the collection are to
* be stored, if it is big enough; otherwise, a new array of
* the same runtime type is allocated for this purpose
*
* @return an array containing the elements of the collection
*
* @throws ArrayStoreException if the runtime type of the specified
* array is not a supertype of the runtime type of every
* element in this collection
*/
public Object[] toArray(Object[] ao)
{
SegmentedHashMap map = SegmentedHashMap.this;
IterableEntryAction actionEntry = new EntryActionAdapter()
{
public Object invokeFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket,
Entry entryPrev, Entry entryCur)
{
if (!entryCur.isSynthetic())
{
((List) oContext).add(entryCur);
}
return NO_VALUE;
}
};
return map.toArrayInternal(actionEntry, ao);
}
// ----- inner class: Entry Set Iterator ------------------------
/**
* Factory for EntrySetIterator.
*
* @return a new instance of an Iterator over the EntrySet
*/
protected Iterator instantiateIterator()
{
return new EntrySetIterator();
}
/**
* An Iterator over the EntrySet that is backed by the SegmentedHashMap.
*/
protected class EntrySetIterator
extends AbstractStableIterator
{
// ----- constructors -----------------------------------
/**
* Construct an Iterator over the Entries in the SegmentedHashMap.
*/
protected EntrySetIterator()
{
}
// ----- internal -------------------------------------------
/**
* Advance to the next object in the iteration.
*/
protected void advance()
{
if (m_fDeactivated)
{
// the Iterator has already reached the end on a previous
// call to advance()
return;
}
Entry[] aeBucket = this.m_aeBucket;
if (aeBucket == null)
{
// Activate the iterator upon its first use. The map will
// not grow while there are any active iterators.
SegmentedHashMap map = SegmentedHashMap.this;
map.iteratorActivated(this);
aeBucket = this.m_aeBucket = map.getStableBucketArray();
}
Entry entry = m_entryPrev;
int iBucket = -1; // -1 indicates no change
int cBuckets = aeBucket.length;
while (true)
{
if (entry != null)
{
// advance within the currrent bucket
entry = entry.nextEntry();
}
// check if the current bucket has been exhausted, and if
// it has, then advance to the first non-empty bucket
if (entry == null)
{
iBucket = m_iBucket;
do
{
if (++iBucket >= cBuckets)
{
// the iterator has been exhausted
deactivate();
return;
}
entry = entryFromBucket(aeBucket, iBucket);
}
while (entry == null);
}
// update the current bucket index if the iterator
// advanced to a new bucket
if (iBucket >= 0)
{
m_iBucket = iBucket;
}
// remember the entry being iterated next
m_entryPrev = entry;
// report back the actual entry that exists in the Map
// that is being iterated next
setNext(entry);
return;
}
}
/**
* Remove the specified item from the underlying Map.
*
* @param oPrev the previously iterated object to remove
*/
protected void remove(Object oPrev)
{
SegmentedHashMap.this.remove(((Map.Entry) oPrev).getKey());
}
// ----- internal ---------------------------------------
/**
* Shut down the Iterator. This is done on exhaustion of the
* contents of the Iterator, or on finalization of the Iterator.
*/
protected void deactivate()
{
if (!m_fDeactivated)
{
// No more entries to iterate; notify the containing Map
// that this Iterator is no longer active.
SegmentedHashMap.this.releaseIterator(this);
m_fDeactivated = true;
// clean up references (no longer needed)
m_aeBucket = null;
m_entryPrev = null;
}
}
// ----- data members -----------------------------------
/**
* Array of buckets in the underlying map.
*/
private Entry[] m_aeBucket;
/**
* Current bucket being iterated.
*/
private int m_iBucket = -1;
/**
* The most recent Entry object internally iterated.
*/
private Entry m_entryPrev;
/**
* Set to true when the Iterator is complete.
*/
private boolean m_fDeactivated;
}
}
// ----- inner class: KeySet --------------------------------------------
/**
* Factory for KeySet.
*
* @return a new instance of the KeySet class (or subclass thereof)
*/
protected KeySet instantiateKeySet()
{
return new KeySet();
}
/**
* A set of entries backed by this map.
*/
protected class KeySet
extends AbstractSet
{
// ----- Set interface ------------------------------------------
/**
* Obtain an iterator over the keys in the Map.
*
* @return an Iterator that provides a live view of the keys in the
* underlying Map object
*/
public Iterator iterator()
{
return new Iterator()
{
private Iterator m_iter = SegmentedHashMap.this.entrySet().iterator();
public boolean hasNext()
{
return m_iter.hasNext();
}
public Object next()
{
return ((Map.Entry) m_iter.next()).getKey();
}
public void remove()
{
m_iter.remove();
}
};
}
/**
* Determine the number of keys in the Set.
*
* @return the number of keys in the Set, which is the same as the
* number of entries in the underlying Map
*/
public int size()
{
return SegmentedHashMap.this.size();
}
/**
* Determine if a particular key is present in the Set.
*
* @return true iff the passed key object is in the key Set
*/
public boolean contains(Object oKey)
{
return SegmentedHashMap.this.containsKey(oKey);
}
/**
* Removes the specified element from this Set of keys if it is present
* by removing the associated entry from the underlying Map.
*
* @param o object to be removed from this set, if present
*
* @return true if the set contained the specified element
*/
public boolean remove(Object o)
{
SegmentedHashMap map = SegmentedHashMap.this;
if (map.containsKey(o))
{
map.remove(o);
return true;
}
else
{
return false;
}
}
/**
* Removes all of the elements from this set of Keys by clearing the
* underlying Map.
*/
public void clear()
{
SegmentedHashMap.this.clear();
}
/**
* Returns an array containing all of the keys in this set.
*
* @return an array containing all of the keys in this set
*/
public Object[] toArray()
{
return toArray((Object[]) null);
}
/**
* Returns an array with a runtime type is that of the specified array
* and that contains all of the keys in this Set. If the Set fits in
* the specified array, it is returned therein. Otherwise, a new array
* is allocated with the runtime type of the specified array and the
* size of this collection.
*
* If the Set fits in the specified array with room to spare (i.e., the
* array has more elements than the Set), the element in the array
* immediately following the end of the Set is set to null.
* This is useful in determining the length of the Set only if
* the caller knows that the Set does not contain any null
* keys.)
*
* @param a the array into which the elements of the Set are to
* be stored, if it is big enough; otherwise, a new array
* of the same runtime type is allocated for this purpose
*
* @return an array containing the elements of the Set
*
* @throws ArrayStoreException if the runtime type of the specified
* array is not a supertype of the runtime type of every
* element in this Set of keys
*/
public Object[] toArray(Object a[])
{
SegmentedHashMap map = SegmentedHashMap.this;
IterableEntryAction actionEntry = new EntryActionAdapter()
{
public Object invokeFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket,
Entry entryPrev, Entry entryCur)
{
if (!entryCur.isSynthetic())
{
((List) oContext).add(entryCur.getKey());
}
return NO_VALUE;
}
};
return map.toArrayInternal(actionEntry, a);
}
}
// ----- inner class: ValuesCollection ----------------------------------
/**
* Factory for ValuesCollection.
*
* @return a new instance of the ValuesCollection class (or subclass
* thereof)
*/
protected ValuesCollection instantiateValuesCollection()
{
return new ValuesCollection();
}
/**
* A collection of values backed by this map.
*/
protected class ValuesCollection
extends AbstractCollection
{
// ----- Collection interface -----------------------------------
/**
* Obtain an iterator over the values in the Map.
*
* @return an Iterator that provides a live view of the values in the
* underlying Map object
*/
public Iterator iterator()
{
return new Iterator()
{
private Iterator m_iter =
SegmentedHashMap.this.entrySet().iterator();
public boolean hasNext()
{
return m_iter.hasNext();
}
public Object next()
{
return ((Entry) m_iter.next()).getValue();
}
public void remove()
{
m_iter.remove();
}
};
}
/**
* Determine the number of values in the Collection.
*
* @return the number of values in the Collection, which is the same
* as the number of entries in the underlying Map
*/
public int size()
{
return SegmentedHashMap.this.size();
}
/**
* Removes all of the elements from this Collection of values by
* clearing the underlying Map.
*/
public void clear()
{
SegmentedHashMap.this.clear();
}
/**
* Returns an array containing all of the values in the Collection.
*
* @return an array containing all of the values in the Collection
*/
public Object[] toArray()
{
return toArray((Object[]) null);
}
/**
* Returns an array with a runtime type is that of the specified array
* and that contains all of the values in the Collection. If the
* Collection fits in the specified array, it is returned therein.
* Otherwise, a new array is allocated with the runtime type of the
* specified array and the size of this collection.
*
* If the Collection fits in the specified array with room to spare
* (i.e., the array has more elements than the Collection), the element
* in the array immediately following the end of the Collection is set
* to null. This is useful in determining the length of the
* Collection only if the caller knows that the Collection does
* not contain any null values.)
*
* @param ao the array into which the elements of the Collection are
* to be stored, if it is big enough; otherwise, a new
* array of the same runtime type is allocated for this
* purpose
*
* @return an array containing the elements of the Collection
*
* @throws ArrayStoreException if the runtime type of the specified
* array is not a supertype of the runtime type of every
* element in this Collection of values
*/
public Object[] toArray(Object ao[])
{
SegmentedHashMap map = SegmentedHashMap.this;
IterableEntryAction actionEntry = new EntryActionAdapter()
{
public Object invokeFound(Object oKey, Object oContext,
Entry[] aeBucket, int nBucket,
Entry entryPrev, Entry entryCur)
{
if (!entryCur.isSynthetic())
{
((List) oContext).add(entryCur.getValue());
}
return NO_VALUE;
}
};
return map.toArrayInternal(actionEntry, ao);
}
}
// ----- constants ------------------------------------------------------
/**
* When resizing, the entries array is replaced with this special empty
* array to signify a resize.
*/
private static final Entry[] NO_ENTRIES = new Entry[0];
/**
* A list of possible modulos to use.
*/
protected static final int[] PRIME_MODULO =
{
61,127,197,277,397,457,509,587,641,701,761,827,883,953,1019,1129,
1279,1427,1543,1733,1951,2143,2371,2671,2927,3253,3539,3907,4211,
4591,4973,5393,5743,6143,6619,6997,7529,8009,8423,8819,9311,9929,
10069,11087,12203,13003,14051,15017,16007,17027,18061,19013,20063,
23011,27011,30011,35023,40009,45007,50021,60013,70001,80021,90001,
100003,120011,140009,160001,180001,200003,233021,266003,300007,
350003,400009,450001,500009,550007,600011,650011,700001,800011,
850009,900001,950009,1000003,1100009,1200007,1300021,1400017,
1500007,1600033,1700021,1800017,1900009,2000003,2500009,3000017,
3500017,4000037,4500007,5000011,6000011,7000003,8000009,9000011,
10000019,12000017,14000029,16000057,18000041,20000003,25000009,
30000001,35000011,40000003,45000017,50000017,60000011,70000027,
80000023,90000049,100000007,150000001,200000033,300000007,400000009,
500000003,600000001,700000001,800000011,900000011,1000000007,
1100000009,1200000041,1300000003,1400000023,1500000001,1600000009,
1700000009,1800000011,1900000043,
2147483647 // Integer.MAX_VALUE is a prime!
};
/**
* Default initial size provides a prime modulo and is large enough that
* resize is not immediate. (A hash map probably uses less than 128 bytes
* initially.)
*/
public static final int DEFAULT_INITIALSIZE = PRIME_MODULO[0];
/**
* Biggest possible modulo.
*/
protected static final int BIGGEST_MODULO = PRIME_MODULO[PRIME_MODULO.length-1];
/**
* The default load factor is 100%, which means that the hash map will not
* resize until there is (on average) one entry in every bucket. The cost
* of scanning a linked list in a particular bucket is very low, so there
* is little reason for having this value below 1.0, and the goal is
* constant order access, so assuming a perfect hash this will provide the
* optimal access speed relative to size.
*/
public static final float DEFAULT_LOADFACTOR = 1.0F;
/**
* Using the default growth rate, the bucket array will grow by a factor
* of four. The relatively high growth rate helps to ensure less resize
* operations, an important consideration in a high-concurrency map.
*/
public static final float DEFAULT_GROWTHRATE = 3.0F;
/**
* The minimum segment capacity.
*/
protected static final int MIN_SEGMENT_CAPACITY = 2;
/**
* The number of segments to partition the hash buckets into. There is a
* single lock for each segment. This number is specially chosen as 61 is
* the largest prime number smaller than 64 (the size of the datatype
* used to represent the lock).
*/
protected static final int SEGMENT_COUNT = 61;
/**
* The number of segment-locks. Each segment has its own lock and
* there is a global "intention" lock.
*/
protected static final int LOCK_COUNT = SEGMENT_COUNT + 1;
/**
* The lock representation used to indicate that no locks are set.
*/
protected static final long LOCKS_NONE = 0L;
/**
* The lock representation used to indicate that all mutexes are locked.
*/
protected static final long LOCKS_ALL = 0xFFFFFFFFFFFFFFFFL;
/**
* The mutex number used to indicate that a lock-all is pending.
*/
protected static final int LOCK_ALL_PENDING_IDX = 61;
/**
* The bit-mask used to indicate that a lock-all is pending.
*/
protected static final long LOCK_ALL_PENDING = 1L << LOCK_ALL_PENDING_IDX;
/**
* Maximum number of times to spin while trying to acquire a segment lock
* before waiting.
*/
protected static final int SEGMENT_LOCK_MAX_SPIN = 0xF;
/**
* Size threshold used by the putAll operation.
*/
protected static final int PUTALL_THRESHOLD = (int) (1.5 * SEGMENT_COUNT);
/**
* Object to be used as a value representing that the Entry object is
* "synthetic" and while logically associated with a key, does not
* represent a key-value mapping in the Map.
*/
protected static final Object NO_VALUE = new Object();
// ----- data members ---------------------------------------------------
/**
* When resizing completes, a notification is issued against this object.
*/
protected final Object RESIZING = new Object();
/**
* The "segment-locks". This AtomicCounter is actually used just as an
* "Atomic Long" value, with each of the first 61 bits being used to
* represent a segment-lock. In this case, the number of "buckets" is
* fixed to 61 (the largest prime smaller than 64).
*/
protected final AtomicLong m_atomicLocks = new AtomicLong();
/**
* An array of the control-structures for the Map's segments.
*/
protected final Segment[] m_aSegment;
/**
* The array of hash buckets. This field is declared volatile in order to
* reduce synchronization.
*/
protected volatile Entry[] m_aeBucket;
/**
* The capacity of each segment (the point at which we should resize).
*/
protected int m_cSegmentCapacity;
/**
* The determining factor for the hash map capacity given a certain number
* of buckets, such that capacity = bucketcount * loadfactor.
*/
protected final float m_flLoadFactor;
/**
* The rate of growth as a fraction of the current number of buckets,
* 0 < n, such that the hash map grows to bucketcount * (1 + growth-rate).
*/
protected final float m_flGrowthRate;
/**
* The set of entries backed by this map.
*/
protected EntrySet m_setEntries;
/**
* The set of keys backed by this map.
*/
protected KeySet m_setKeys;
/**
* The collection of values backed by this map.
*/
protected ValuesCollection m_colValues;
/**
* A holder for active Iterator(s): either WeakReference(<Iterator>) or
* WeakHashMap(<Iterator> null)
*/
protected Object m_oIterActive;
/**
* The singleton action for getEntryInternal support.
*/
protected GetEntryAction m_actionGetEntry;
/**
* The singleton action for insert support.
*/
protected InsertAction m_actionInsert;
/**
* The singleton action for remove support.
*/
protected RemoveAction m_actionRemove;
/**
* The singleton action for containsValue support.
*/
protected ContainsValueAction m_actionContainsValue;
}