Brief introduction
ConcurrentHashMap in Java is a thread-safe and efficient HashMap achieved . Usually involving high concurrency If we want to map the structure, and that the first thought is it.
Then I would understand several aspects about ConcurrentHashMap:
1) ConcurrentHashMap JDK8 in the structure
2) ConcurrentHashMap put the method, szie method
3) ConcurrentHashMap expansion
4) the difference between HashMap, Hashtable, ConccurentHashMap three
5) ConcurrentHashMap JDK8 and the difference JDK7
Source code analysis
ConcurrentHashMap in JDK8 in structure
First, look at the underlying structure composed of:
In fact, and 1.8 HashMap structurally similar, when the list of nodes exceeds a specified threshold value , then, is converted into a red-black tree , the general structure is the same.
So in the end it is how to achieve thread-safe?
The answer: which abandoned the original Segment segment lock, while the use of CAS + synchronized to ensure the safety of concurrency . As for how, I continue to look at the logic put method
Logic put method
ConcurrentHashMap most commonly used method is put and get methods, here mainly to see the code comments, ease of understanding.
/** Implementation for put and putIfAbsent */
final V putVal(K key, V value, boolean onlyIfAbsent) {
if (key == null || value == null) throw new NullPointerException();
//1. 计算key的hash值
int hash = spread(key.hashCode());
int binCount = 0;
for (Node<K,V>[] tab = table;;) {
Node<K,V> f; int n, i, fh;
//2. 如果当前table还没有初始化先调用initTable方法将tab进行初始化
if (tab == null || (n = tab.length) == 0)
tab = initTable();
//3. tab中索引为i的位置的元素为null，则直接使用CAS将值插入即可
else if ((f = tabAt(tab, i = (n - 1) & hash)) == null) {
if (casTabAt(tab, i, null,
new Node<K,V>(hash, key, value, null)))
break; // no lock when adding to empty bin
}
//4. 当前正在扩容
else if ((fh = f.hash) == MOVED)
tab = helpTransfer(tab, f);
else {
V oldVal = null;
synchronized (f) {
if (tabAt(tab, i) == f) {
//5. 当前为链表，在链表中插入新的键值对
if (fh >= 0) {
binCount = 1;
for (Node<K,V> e = f;; ++binCount) {
K ek;
if (e.hash == hash &&
((ek = e.key) == key ||
(ek != null && key.equals(ek)))) {
oldVal = e.val;
if (!onlyIfAbsent)
e.val = value;
break;
}
Node<K,V> pred = e;
if ((e = e.next) == null) {
pred.next = new Node<K,V>(hash, key,
value, null);
break;
}
}
}
// 6.当前为红黑树，将新的键值对插入到红黑树中
else if (f instanceof TreeBin) {
Node<K,V> p;
binCount = 2;
if ((p = ((TreeBin<K,V>)f).putTreeVal(hash, key,
value)) != null) {
oldVal = p.val;
if (!onlyIfAbsent)
p.val = value;
}
}
}
}
// 7.插入完键值对后再根据实际大小看是否需要转换成红黑树
if (binCount != 0) {
if (binCount >= TREEIFY_THRESHOLD)
treeifyBin(tab, i);
if (oldVal != null)
return oldVal;
break;
}
}
}
//8.对当前容量大小进行检查，如果超过了临界值（实际大小*加载因子）就需要扩容
addCount(1L, binCount);
return null;
}
Please read the code comments, there is a general understanding, then we learn about in more detail:
This process clearly put on the table for the current unconditional put from the circulation until successful , can be divided into the following six-step process outlined by:
1, it is determined Node [] array is initialized, is not initialize operation
2, the index coordinate positioning hash array , whether a node Node, if not using add CAS (head node list), the process proceeds to the next addition failed cycle.
3, check the inside is expansion, helped it an expansion.
4, if f! = null, then the use of synchronized locking f element (list / black tree head element). If the Node (chain structure) to add the list operation is executed; if the TreeNode (tree structure) to add tree operation is performed.
5, it is judged chain length 8 has reached a critical value (default value), when the node value exceeds the need to convert the list to a tree structure .
6, if added successfully invoked addCount () method of statistical size, and check whether you need expansion
1.spread(key,hashCode())
This method of action heavy Hash , Hash to reduce conflict
static final int spread(int h) {
return (h ^ (h >>> 16)) & HASH_BITS;
}
The method is mainly a low of 16 hashCode key is in the high 16-bit XOR operation , so that the hash value can not only be dispersed uniformly reduce the probability of hash collision, only used the additional XOR operation, and the performance overhead able to combine.
2.initTable method
The main role will be initialized tab
private final Node<K,V>[] initTable() {
Node<K,V>[] tab; int sc;
while ((tab = table) == null || tab.length == 0) {
if ((sc = sizeCtl) < 0)
// 1. 保证只有一个线程正在进行初始化操作
Thread.yield(); // lost initialization race; just spin
else if (U.compareAndSwapInt(this, SIZECTL, sc, -1)) {
try {
if ((tab = table) == null || tab.length == 0) {
// 2. 得出数组的大小
int n = (sc > 0) ? sc : DEFAULT_CAPACITY;
@SuppressWarnings("unchecked")
// 3. 这里才真正的初始化数组
Node<K,V>[] nt = (Node<K,V>[])new Node<?,?>[n];
table = tab = nt;
// 4. 计算数组中可用的大小：实际大小n*0.75（加载因子）
sc = n - (n >>> 2);
}
} finally {
sizeCtl = sc;
}
break;
}
}
return tab;
}
To ensure that properly initialized, in step 1 will first determine if carried through, if the current thread is initializing have a value that is sizeCtl becomes -1 , this time in the other thread If determined to be true to invoke Thread.yield () let the CPU time slice. Being initialized thread calls U.compareAndSwapInt method sizeCtl change that is being initialized state 1. Further things should also be noted that, in the fourth step further calculates the size of the array is the array available in the actual size of the load factor is 0.75 multiplied by n. Here you can see how the count multiplied by 0.75, and 0.75 for the quarter Third, where n - (n >>> 2) is not exactly the N-(1/4) the n-= (3/4) the n- , very interesting it :). If no parameters are selected constructor, here will be used in new Node is the default size when the array DEFAULT_CAPACITY (16), and then multiplied by the load factor is 0.75 12, that is to say the available size of the array 12.
3.CAS critical operations
tabAt () This method is used to obtain the array index table Node element i .
casTabAt () using the CAS operation setting table index array element i
setTabAt () method is used to set the table index in the array element i
4.ConcurrentHashMap expansion
By determining a hash value of the node is not equal to -1 (MOVED), code (fh = f.hash) == MOVED, Map is described expansion . Then helped Map for expansion. To speed.
How to help the expansion of it? To achieve that look helpTransfer methods.
/**
* Helps transfer if a resize is in progress.
*/
final Node<K,V>[] helpTransfer(Node<K,V>[] tab, Node<K,V> f) {
Node<K,V>[] nextTab; int sc;
// 如果 table 不是空 且 node 节点是转移类型，数据检验
// 且 node 节点的 nextTable（新 table） 不是空，同样也是数据校验
// 尝试帮助扩容
if (tab != null && (f instanceof ForwardingNode) &&
(nextTab = ((ForwardingNode<K,V>)f).nextTable) != null) {
// 根据 length 得到一个标识符号
int rs = resizeStamp(tab.length);
// 如果 nextTab 没有被并发修改 且 tab 也没有被并发修改
// 且 sizeCtl < 0 （说明还在扩容）
while (nextTab == nextTable && table == tab &&
(sc = sizeCtl) < 0) {
// 如果 sizeCtl 无符号右移 16 不等于 rs （ sc前 16 位如果不等于标识符，则标识符变化了）
// 或者 sizeCtl == rs + 1 （扩容结束了，不再有线程进行扩容）（默认第一个线程设置 sc ==rs 左移 16 位 + 2，当第一个线程结束扩容了，就会将 sc 减一。这个时候，sc 就等于 rs + 1）
// 或者 sizeCtl == rs + 65535 （如果达到最大帮助线程的数量，即 65535）
// 或者转移下标正在调整 （扩容结束）
// 结束循环，返回 table
if ((sc >>> RESIZE_STAMP_SHIFT) != rs || sc == rs + 1 ||
sc == rs + MAX_RESIZERS || transferIndex <= 0)
break;
// 如果以上都不是, 将 sizeCtl + 1, （表示增加了一个线程帮助其扩容）
if (U.compareAndSwapInt(this, SIZECTL, sc, sc + 1)) {
// 进行转移
transfer(tab, nextTab);
// 结束循环
break;
}
}
return nextTab;
}
return table;
}
The basic logic is in code comments, where the key transfer (), then we continue in-depth look
private final void transfer(Node<K,V>[] tab, Node<K,V>[] nextTab) {
int n = tab.length, stride;
// 每核处理的量小于16，则强制赋值16
if ((stride = (NCPU > 1) ? (n >>> 3) / NCPU : n) < MIN_TRANSFER_STRIDE)
stride = MIN_TRANSFER_STRIDE; // subdivide range
if (nextTab == null) { // initiating
try {
@SuppressWarnings("unchecked")
Node<K,V>[] nt = (Node<K,V>[])new Node<?,?>[n << 1]; //构建一个nextTable对象，其容量为原来容量的两倍
nextTab = nt;
} catch (Throwable ex) { // try to cope with OOME
sizeCtl = Integer.MAX_VALUE;
return;
}
nextTable = nextTab;
transferIndex = n;
}
int nextn = nextTab.length;
// 连接点指针，用于标志位（fwd的hash值为-1，fwd.nextTable=nextTab）
ForwardingNode<K,V> fwd = new ForwardingNode<K,V>(nextTab);
// 当advance == true时，表明该节点已经处理过了
boolean advance = true;
boolean finishing = false; // to ensure sweep before committing nextTab
for (int i = 0, bound = 0;;) {
Node<K,V> f; int fh;
// 控制 --i ,遍历原hash表中的节点
while (advance) {
int nextIndex, nextBound;
if (--i >= bound || finishing)
advance = false;
else if ((nextIndex = transferIndex) <= 0) {
i = -1;
advance = false;
}
// 用CAS计算得到的transferIndex
else if (U.compareAndSwapInt
(this, TRANSFERINDEX, nextIndex,
nextBound = (nextIndex > stride ?
nextIndex - stride : 0))) {
bound = nextBound;
i = nextIndex - 1;
advance = false;
}
}
if (i < 0 || i >= n || i + n >= nextn) {
int sc;
// 已经完成所有节点复制了
if (finishing) {
nextTable = null;
table = nextTab; // table 指向nextTable
sizeCtl = (n << 1) - (n >>> 1); // sizeCtl阈值为原来的1.5倍
return; // 跳出死循环，
}
// CAS 更扩容阈值，在这里面sizectl值减一，说明新加入一个线程参与到扩容操作
if (U.compareAndSwapInt(this, SIZECTL, sc = sizeCtl, sc - 1)) {
if ((sc - 2) != resizeStamp(n) << RESIZE_STAMP_SHIFT)
return;
finishing = advance = true;
i = n; // recheck before commit
}
}
// 遍历的节点为null，则放入到ForwardingNode 指针节点
else if ((f = tabAt(tab, i)) == null)
advance = casTabAt(tab, i, null, fwd);
// f.hash == -1 表示遍历到了ForwardingNode节点，意味着该节点已经处理过了
// 这里是控制并发扩容的核心
else if ((fh = f.hash) == MOVED)
advance = true; // already processed
else {
// 节点加锁
synchronized (f) {
// 节点复制工作
if (tabAt(tab, i) == f) {
Node<K,V> ln, hn;
// fh >= 0 ,表示为链表节点
if (fh >= 0) {
// 构造两个链表 一个是原链表 另一个是原链表的反序排列
int runBit = fh & n;
Node<K,V> lastRun = f;
for (Node<K,V> p = f.next; p != null; p = p.next) {
int b = p.hash & n;
if (b != runBit) {
runBit = b;
lastRun = p;
}
}
if (runBit == 0) {
ln = lastRun;
hn = null;
}
else {
hn = lastRun;
ln = null;
}
for (Node<K,V> p = f; p != lastRun; p = p.next) {
int ph = p.hash; K pk = p.key; V pv = p.val;
if ((ph & n) == 0)
ln = new Node<K,V>(ph, pk, pv, ln);
else
hn = new Node<K,V>(ph, pk, pv, hn);
}
// 在nextTable i 位置处插上链表
setTabAt(nextTab, i, ln);
// 在nextTable i + n 位置处插上链表
setTabAt(nextTab, i + n, hn);
// 在table i 位置处插上ForwardingNode 表示该节点已经处理过了
setTabAt(tab, i, fwd);
// advance = true 可以执行--i动作，遍历节点
advance = true;
}
// 如果是TreeBin，则按照红黑树进行处理，处理逻辑与上面一致
else if (f instanceof TreeBin) {
TreeBin<K,V> t = (TreeBin<K,V>)f;
TreeNode<K,V> lo = null, loTail = null;
TreeNode<K,V> hi = null, hiTail = null;
int lc = 0, hc = 0;
for (Node<K,V> e = t.first; e != null; e = e.next) {
int h = e.hash;
TreeNode<K,V> p = new TreeNode<K,V>
(h, e.key, e.val, null, null);
if ((h & n) == 0) {
if ((p.prev = loTail) == null)
lo = p;
else
loTail.next = p;
loTail = p;
++lc;
}
else {
if ((p.prev = hiTail) == null)
hi = p;
else
hiTail.next = p;
hiTail = p;
++hc;
}
}
// 扩容后树节点个数若<=6，将树转链表
ln = (lc <= UNTREEIFY_THRESHOLD) ? untreeify(lo) :
(hc != 0) ? new TreeBin<K,V>(lo) : t;
hn = (hc <= UNTREEIFY_THRESHOLD) ? untreeify(hi) :
(lc != 0) ? new TreeBin<K,V>(hi) : t;
setTabAt(nextTab, i, ln);
setTabAt(nextTab, i + n, hn);
setTabAt(tab, i, fwd);
advance = true;
}
}
}
}
}
}
Expansion process is a bit complicated, you can view the top notes. Here mainly related to the expansion of multi-threaded, ForwardingNode role is to support the expansion of operations, the processed node and an empty node set ForwardingNode, when processing multiple concurrent threads through ForwardingNode it said it had traversed, we traverse the back, bottom multithreading is a process of expansion of cooperation:
5.treeifyBin () processes the list of red-black tree revolutions
The basic logic in the code comments.
private final void treeifyBin(Node<K,V>[] tab, int index) {
Node<K,V> b; int n, sc;
if (tab != null) {
//如果整个table的数量小于64，就扩容至原来的一倍，不转红黑树了
//因为这个阈值扩容可以减少hash冲突，不必要去转红黑树
if ((n = tab.length) < MIN_TREEIFY_CAPACITY)
tryPresize(n << 1);
else if ((b = tabAt(tab, index)) != null && b.hash >= 0) {
synchronized (b) {
if (tabAt(tab, index) == b) {
TreeNode<K,V> hd = null, tl = null;
for (Node<K,V> e = b; e != null; e = e.next) {
//封装成TreeNode
TreeNode<K,V> p =
new TreeNode<K,V>(e.hash, e.key, e.val,
null, null);
if ((p.prev = tl) == null)
hd = p;
else
tl.next = p;
tl = p;
}
//通过TreeBin对象对TreeNode转换成红黑树
setTabAt(tab, index, new TreeBin<K,V>(hd));
}
}
}
}
}
addCount () method of calculating the size of ConcurrentHashMap
private final void addCount(long x, int check) {
CounterCell[] as; long b, s;
//更新baseCount，table的数量，counterCells表示元素个数的变化
if ((as = counterCells) != null ||
!U.compareAndSwapLong(this, BASECOUNT, b = baseCount, s = b + x)) {
CounterCell a; long v; int m;
boolean uncontended = true;
//如果多个线程都在执行，则CAS失败，执行fullAddCount，全部加入count
if (as == null || (m = as.length - 1) < 0 ||
(a = as[ThreadLocalRandom.getProbe() & m]) == null ||
!(uncontended =
U.compareAndSwapLong(a, CELLVALUE, v = a.value, v + x))) {
fullAddCount(x, uncontended);
return;
}
if (check <= 1)
return;
s = sumCount();
}
//check>=0表示需要进行扩容操作
if (check >= 0) {
Node<K,V>[] tab, nt; int n, sc;
while (s >= (long)(sc = sizeCtl) && (tab = table) != null &&
(n = tab.length) < MAXIMUM_CAPACITY) {
int rs = resizeStamp(n);
if (sc < 0) {
if ((sc >>> RESIZE_STAMP_SHIFT) != rs || sc == rs + 1 ||
sc == rs + MAX_RESIZERS || (nt = nextTable) == null ||
transferIndex <= 0)
break;
if (U.compareAndSwapInt(this, SIZECTL, sc, sc + 1))
transfer(tab, nt);
}
//当前线程发起库哦哦让操作，nextTable=null
else if (U.compareAndSwapInt(this, SIZECTL, sc,
(rs << RESIZE_STAMP_SHIFT) + 2))
transfer(tab, null);
s = sumCount();
}
}
}
put process has now been analyzed over, you may find that he used in the concurrent processing is optimistic locking, when there are conflicts have come to concurrent processing, and process steps are clear, but the details of the design is very complicated, after all multi-threaded the scene also complicated.
get method
ConcurrentHashMap the get operation process is very simple, can be described in three steps:
1. Calculate hash value, to locate the position of the index table, if the first node is in line with returns.
2. If you have the time expansion, will call signs node ForwardingNode the find method is expansion and look for the node, match returns.
3. The above do not meet, then it is down to traverse the node, match returns, otherwise it returns null final
public V get(Object key) {
Node<K,V>[] tab; Node<K,V> e, p; int n, eh; K ek;
int h = spread(key.hashCode()); //计算两次hash
if ((tab = table) != null && (n = tab.length) > 0 &&
(e = tabAt(tab, (n - 1) & h)) != null) {//读取首节点的Node元素
if ((eh = e.hash) == h) { //如果该节点就是首节点就返回
if ((ek = e.key) == key || (ek != null && key.equals(ek)))
return e.val;
}
//hash值为负值表示正在扩容，这个时候查的是ForwardingNode的find方法来定位到nextTable来
//查找，查找到就返回
else if (eh < 0)
return (p = e.find(h, key)) != null ? p.val : null;
while ((e = e.next) != null) {//既不是首节点也不是ForwardingNode，那就往下遍历
if (e.hash == h &&
((ek = e.key) == key || (ek != null && key.equals(ek))))
return e.val;
}
}
return null;
}
Method logical szie
For the calculation of the size, the expansion and addCount () method would have been dealt with, can take note Put function, which will have addCount () function , already computed, and then when you size directly to you.
public int size() {
long n = sumCount();
return ((n < 0L) ? 0 :
(n > (long)Integer.MAX_VALUE) ? Integer.MAX_VALUE :
(int)n);
}
final long sumCount() {
CounterCell[] as = counterCells; CounterCell a; //变化的数量
long sum = baseCount;
if (as != null) {
for (int i = 0; i < as.length; ++i) {
if ((a = as[i]) != null)
sum += a.value;
}
}
return sum;
}
The difference between HashMap, Hashtable, ConccurentHashMap three
HashMap thread-safe, an array + + list red-black tree
Hashtable thread-safe, lock the entire object, an array + list
ConccurentHashMap thread-safe, CAS + synchronization lock, an array + + list red-black tree
HashMap of key, value can be null, the other two not.
Difference in the JDK1.7 and JDK1.8
JDK1.8 major improvements in design are the following:
1, without using employed node segment, to achieve a reduction locked lock granularity node .
2, in the process of designing a MOVED state when resize the thread 2 also put data, thread 2 will help resize.
3, using 3 CAS operation node to ensure some of the atoms of the operation , instead of the lock in this way.
4, different values sizeCtl to represent different meanings, acts to control.
Rather than using synchronized ReentrantLock
to sum up
ConcurrentHashMap basic principle is summarized here, where there are missing can leave a message to tell me ha.
