ConcurrentHashMap核心原理，彻底给整明白了

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"ConcurrentHashMap，它在技术面试中出现的频率相当之高，所以我们必须对它深入理解和掌握。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"谈到 ConcurrentHashMap，就一定会想到 HashMap。HashMap 在我们的代码中使用频率更高，不需要考虑线程安全的地方，我们一般都会使用 HashMap。HashMap 的实现非常经典，如果你读过 HashMap 的源代码，那么对 ConcurrentHashMap 源代码的理解会相对轻松，因为两者采用的数据结构是类似的"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这篇文章主要讲解ConcurrentHashMap的核心原理，并注释详细源码，文章篇幅较长，可收藏再看"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"基本结构"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"ConcurrentHashMap 是一个存储 key/value 对的容器，并且是线程安全的。我们先看 ConcurrentHashMap 的存储结构，如下图："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/ae/aef220bcfa46e904923f844205a74c2b.jpeg","alt":"ConcurrentHashMap核心原理，彻底给整明白了","title":null,"style":null,"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"虽然 ConcurrentHashMap 的底层数据结构，和方法的实现细节和 HashMap 大体一致，但两者在类结构上却没有任何关联，我们看下 ConcurrentHashMap 的类图："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/23/235b9cc134235a4902cf894c4129e113.jpeg","alt":"ConcurrentHashMap核心原理，彻底给整明白了","title":null,"style":null,"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"看 ConcurrentHashMap 源码，我们会发现很多方法和代码和 HashMap 很相似，有的同学可能会问，为什么不继承 HashMap 呢？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"继承的确是个好办法，但ConcurrentHashMap 都是在方法中间进行一些加锁操作，也就是说加锁把方法切割了，继承就很难解决这个问题。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"ConcurrentHashMap和HashMap两者的相同之处："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"数组、链表结构几乎相同，所以底层对数据结构的操作思路是相同的（只是思路相同，底层实现不同）；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"都实现了 Map 接口，继承了 AbstractMap 抽象类，所以大多数的方法也都是相同的，HashMap 有的方法，ConcurrentHashMap 几乎都有，所以当我们需要从 HashMap 切换到 ConcurrentHashMap 时，无需关心两者之间的兼容问题。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"不同之处："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"红黑树结构略有不同，HashMap 的红黑树中的节点叫做 TreeNode，TreeNode 不仅仅有属性，还维护着红黑树的结构，比如说查找，新增等等；ConcurrentHashMap 中红黑树被拆分成两块，TreeNode 仅仅维护的属性和查找功能，新增了 TreeBin，来维护红黑树结构，并负责根节点的加锁和解锁；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"新增 ForwardingNode （转移）节点，扩容的时候会使用到，通过使用该节点，来保证扩容时的线程安全。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这些概念名词文章后面都会依次介绍"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"基本构成"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"重要属性"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们来看看 ConcurrentHashMap 的几个重要属性"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"//这个Node数组就是ConcurrentHashMap用来存储数据的哈希表。\ntransient volatile Node[] table\n//这是默认的初始化哈希表数组大小\nprivate static final int DEFAULT_CAPACITY = 16;\n//转化为红黑树的链表长度阈值\nstatic final int TREEIFY_THRESHOLD = 8\n//这个标识位用于识别扩容时正在转移数据\nstatic final int MOVED = -1\n//计算哈希值时用到的参数，用来去除符号位\nstatic final int HASH_BITS = 0x7fffffff;\n//数据转移时，新的哈希表数组\nprivate transient volatile Node[] nextTable;\n复制代码"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"重要组成元素"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Node"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"链表中的元素为Node对象。他是链表上的一个节点，内部存储了key、value值，以及他的下一个节点的引用。这样一系列的Node就串成一串，组成一个链表。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"”"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"ForwardingNode"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"当进行扩容时，要把链表迁移到新的哈希表，在做这个操作时，会在把数组中的头节点替换为ForwardingNode对象。ForwardingNode中不保存key和value，只保存了扩容后哈希表（nextTable）的引用。此时查找相应node时，需要去nextTable中查找。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"”"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"TreeBin"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"当链表转为红黑树后，数组中保存的引用为 TreeBin，TreeBin 内部不保存 key/value，他保存了 TreeNode的list以及红黑树 root。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"”"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"TreeNode"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"红黑树的节点。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"”"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"下面依次讲解各个核心方法，有详细注释"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"Put方法"}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"public V put(K key, V value) {\n return putVal(key, value, false);\n}\n复制代码"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/7a/7a736d60e8af1e85e444c355193876c6.png","alt":"ConcurrentHashMap核心原理，彻底给整明白了","title":null,"style":null,"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"ConcurrentHashMap 在 put 方法上的整体思路和 HashMap 相同，但在线程安全方面写了很多保障的代码，我们先来看下大体思路："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"1.如果数组为空，初始化，初始化完成之后，走 2；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2.计算当前槽点有没有值，没有值的话，cas 创建，失败继续自旋（for 死循环），直到成功，槽点有值的话，走 3；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"3.如果槽点是转移节点(正在扩容)，就会一直自旋等待扩容完成之后再新增，不是转移节点走 4；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4.槽点有值的，先锁定当前槽点，保证其余线程不能操作，如果是链表，新增值到链表的尾部，如果是红黑树，使用红黑树新增的方法新增；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"5.新增完成之后 check 需不需要扩容，需要的话去扩容。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/4e/4e121cc73f847396bedbb2fcc45ef51f.jpeg","alt":"ConcurrentHashMap核心原理，彻底给整明白了","title":null,"style":null,"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"ConcurrentHashMap在put过程中，采用了哪些手段来保证线程安全呢？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"数组初始化时的线程安全"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"数组初始化时，首先通过自旋来保证一定可以初始化成功，然后通过 CAS 设置 SIZECTL 变量的值，来保证同一时刻只能有一个线程对数组进行初始化，CAS 成功之后，还会再次判断当前数组是否已经初始化完成，如果已经初始化完成，就不会再次初始化，通过自旋 + CAS + 双重 check 等手段保证了数组初始化时的线程安全"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"那么接下来我们就来看看 initTable 方法。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/fa/fa4c5e4f2ba0286bfb281bcd6377e963.jpeg","alt":"ConcurrentHashMap核心原理，彻底给整明白了","title":null,"style":null,"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"注意里面有个关键的值 sizeCtl，这个值有多个含义。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"1、-1 代表有线程正在创建 table；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2、-N 代表有 N-1 个线程正在复制 table；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"3、在 table 被初始化前，代表根据构造函数传入的值计算出的应被初始化的大小；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4、在 table 被初始化后，则被设置为 table 大小 的 75%，代表 table 的容量（数组容量）。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"新增槽点值时的线程安全"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此时为了保证线程安全，做了四处优化："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"1.通过自旋死循环保证一定可以新增成功。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在新增之前，通过 for (Node[] tab = table;;)这样的死循环来保证新增一定可以成功，一旦新增成功，就可以退出当前死循环，新增失败的话，会重复新增的步骤，直到新增成功为止。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2.当前槽点为空时，通过 CAS 新增。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Java 这里的写法非常严谨，没有在判断槽点为空的情况下直接赋值，因为在判断槽点为空和赋值的瞬间，很有可能槽点已经被其他线程复制了，所以我们采用 CAS 算法，能够保证槽点为空的情况下赋值成功，如果恰好槽点已经被其他线程赋值，当前 CAS 操作失败，会再次执行 for 自旋，再走槽点有值的 put 流程，这里就是自选 + CAS 的结合。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"3.当前槽点有值，锁住当前槽点。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"put 时，如果当前槽点有值，就是 key 的 hash 冲突的情况，此时槽点上可能是链表或红黑树，我们通过锁住槽点，来保证同一时刻只会有一个线程能对槽点进行修改"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"V oldVal = null;\n//锁定当前槽点，其余线程不能操作，保证了安全\nsynchronized (f) {\n复制代码"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4.红黑树旋转时，锁住红黑树的根节点，保证同一时刻，当前红黑树只能被一个线程旋转"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"Hash算法"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"spread方法源码分析"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"哈希算法的逻辑，决定 ConcurrentHashMap 保存和读取速度。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"static final int spread(int h) {\n return (h ^ (h >>> 16)) & HASH_BITS;\n}\n复制代码"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"传入的参数h为 key 对象的 hashCode，spreed 方法对 hashCode 进行了加工。重新计算出 hash。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"hash 值是用来映射该 key 值在哈希表中的位置。取出哈希表中该 hash 值对应位置的代码如下。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"tabAt(tab, i = (n - 1) & hash);\n复制代码"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们先看这一行代码的逻辑，第一个参数为哈希表，第二个参数是哈希表中的数组下标。通过 (n - 1) & hash 计算下标。n 为数组长度，我们以默认大小 16 为例，那么 n-1 = 15，我们可以假设 hash 值为 100"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"n的值15转为二进制：\n0000 0000 0000 0000 0000 0000 0000 1111\nhash的值100转为二进制：\n0000 0000 0000 0000 0000 0000 0110 0100。\n计算结果：\n0000 0000 0000 0000 0000 0000 0000 0100\n对应的十进制值为 4\n复制代码"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"15的二进制高位都为0，低位都是1。那么经过&计算后，hash值100的高位全部被清零，低位则保持不变，并且一定是小于（n-1）的。也就是说经过如此计算，通过hash值得到的数组下标绝对不会越界。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"这里提出几个问题："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"1、数组大小可以为 17，或者 18 吗？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2、如果为了保证不越界为什么不直接用 % 计算取余数？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"3、为什么不直接用 key 的 hashCode，而是使用经 spreed 方法加工后的 hash 值？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"数组大小必须为 2 的 n 次方"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第一个问题的答案是数组大小必须为 2 的 n 次方，也就是 16、32、64….不能为其他值。因为如果不是 2 的 n 次方，那么经过计算的数组下标会增大碰撞的机率"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果hash值的二进制是 10000（十进制16）、10010（十进制18）、10001（十进制17），和10100做&计算后，都是10000，也就是都被映射到数组16这个下标上。这三个值会以链表的形式存储在数组16下标的位置。这显然不是我们想要的结果。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"但如果数组长度n为2的n次方，2进制的数值为10，100，1000，10000……n-1后对应二进制为1，11，111，1111……这样和hash值低位&后，会保留原来hash值的低位数值，那么只要hash值的低位不一样，就不会发生碰撞。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"同时(n - 1) & hash等价于 hash%n。那么为什么不直接用hash%n呢？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这是因为按位的操作效率会更高。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"为什么不直接用 key 的 hashCode？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"其实说到底还是为了减少碰撞的概率。我们先看看 spreed 方法中的代码做了什么事情："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"h ^ (h >>> 16)\n复制代码"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这个意思是把 h 的二进制数值向右移动 16 位。我们知道整形为 32 位，那么右移 16 位后，就是把高 16 位移到了低 16 位。而高 16 位清0了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"^为异或操作，二进制按位比较，如果相同则为 0，不同则为 1。这行代码的意思就是把高低16位做异或。如果两个hashCode值的低16位相同，但是高位不同，经过如此计算，低16位会变得不一样了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"为什么要把地位变得不一样呢？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这是由于哈希表数组长度n会是偏小的数值，那么进行(n - 1) & hash运算时，一直使用的是hash较低位的值。那么即使hash值不同，但如果低位相当，也会发生碰撞。而进行h ^ (h >>> 16)加工后的hash值，让hashCode高位的值也参与了哈希运算，因此减少了碰撞的概率。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"(h ^ (h >>> 16)) & HASH_BITS\n复制代码"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"为何高位移到低位和原来低位做异或操作后，还需要和HASH_BITS这个常量做 & 计算呢？HASH_BITS 这个常量的值为 0x7fffffff，转化为二进制为 0111 1111 1111 1111 1111 1111 1111 1111。这个操作后会把最高位转为 0，其实就是消除了符号位，得到的都是正数。这是因为负的 hashCode 在ConcurrentHashMap 中有特殊的含义，因此我们需要得到一个正的 hashCode。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"扩容源码分析"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们大致了解了ConcurrentHashMap 的存储结构，那么我们思考一个问题，当数组中保存的链表越来越多，那么再存储进来的元素大概率会插入到现有的链表中，而不是使用数组中剩下的空位。这样会造成数组中保存的链表越来越长，由此导致哈希表查找速度下降，从 O(1) 慢慢趋近于链表的时间复杂度 O(n/2)，这显然违背了哈希表的初衷。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"所以ConcurrentHashMap 会做一个操作，称为扩容。也就是把数组长度变大，增加更多的空位出来，最终目的就是预防链表过长，这样查找的时间复杂度才会趋向于 O(1)。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"扩容的操作并不会在数组没有空位时才进行，因为在空位快满时，新保存元素更大的概率会命中已经使用的位置，那么可能最后几个桶位很难被使用，而链表却越来越长了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另外 ConcurrentHashMap 还会有链表转红黑树的操作，以提高查找的速度，红黑树时间复杂度为 O(logn)，而链表是 O(n/2)，因此只在 O(logn)= (MAXIMUM_CAPACITY >>> 1)) ?\n MAXIMUM_CAPACITY :\n tableSizeFor(initialCapacity + (initialCapacity >>> 1) + 1));\n this.sizeCtl = cap;\n}\n复制代码"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这是一个有参数的构造方法。如果你对未来存储的数据量有预估，我们可以指定哈希表的大小，避免频繁的扩容操作。tableSizeFor 这个方法确保了哈希表的大小永远都是 2 的 n 次方。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"注意这里传入的参数不是 initialCapacity，而是 initialCapacity 的 1.5 倍 + 1。这样做是为了保证在默认 75% 的负载因子下，能够足够容纳 initialCapacity 数量的元素。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"ConcurrentHashMap (int initialCapacity) 构造函数总结下："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"1、构造函数中并不会初始化哈希表；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2、构造函数中仅设置哈希表大小的变量 sizeCtl；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"3、initialCapacity 并不是哈希表大小；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4、哈希表大小为 initialCapacity*1.5+1 后，向上取最小的 2 的 n 次方。如果超过最大容量一半，那么就是最大容量。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"tableSizeFor 是如何实现向上取得最接近入参 2 的 n 次方的。下面我们来看 tableSizeFor 源代码："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"private static final int tableSizeFor(int c) {\n int n = c - 1;\n n |= n >>> 1;\n n |= n >>> 2;\n n |= n >>> 4;\n n |= n >>> 8;\n n |= n >>> 16;\n return (n < 0) ? 1 : (n >= MAXIMUM_CAPACITY) ? MAXIMUM_CAPACITY : n + 1;\n}\n复制代码"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"依旧是二进制按位操作，这样一顿操作后，得到的数值就是大于 c 的最小 2 的 n 次。我们推演下过程，假设 c 是 9："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"1、int n = 9 - 1\nn=8\n2、n |= n >>> 1\nn=1000\nn >>> 1=0100\n两个值按位或后\nn=1100\n3、n |= n >>> 2\nn=1100\nn >>> 2=0011\nn=1111\n复制代码"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"到这里可以看出规律来了。如果 c 足够大，使得 n 很大，那么运算到 n |= n >>> 16 时，n 的 32 位都为 1。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"总结一下这一段逻辑，其实就是把 n 有数值的 bit 位全部置为 1。这样就得到了一个肯定大于等于 n 的值。我们再看最后一行代码，最终返回的是 n+1，那么一个所有位都是 1 的二进制数字，+1 后得到的就是一个 2 的 n 次方数值。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"看完三件事❤️"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你觉得这篇内容对你还蛮有帮助，我想邀请你帮我三个小忙："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"numberedlist","attrs":{"start":null,"normalizeStart":1},"content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"点赞，转发，有你们的 『点赞和评论』，才是我创造的动力。"}]}]},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"关注公众号 『 "},{"type":"text","marks":[{"type":"strong"}],"text":"java烂猪皮"},{"type":"text","text":" 』，不定期分享原创知识。"}]}]},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"同时可以期待后续文章ing🚀"}]}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/85/8518f1f13ab07e46122c84420bbf39a8.png","alt":null,"title":"","style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":"","fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}}]}