ConcurrentHashMap核心原理，徹底給整明白了

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"ConcurrentHashMap，它在技術面試中出現的頻率相當之高，所以我們必須對它深入理解和掌握。"}]},{"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":"談到 ConcurrentHashMap，就一定會想到 HashMap。HashMap 在我們的代碼中使用頻率更高，不需要考慮線程安全的地方，我們一般都會使用 HashMap。HashMap 的實現非常經典，如果你讀過 HashMap 的源代碼，那麼對 ConcurrentHashMap 源代碼的理解會相對輕鬆，因爲兩者採用的數據結構是類似的"}]},{"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":"這篇文章主要講解ConcurrentHashMap的核心原理，並註釋詳細源碼，文章篇幅較長，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":"基本結構"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"ConcurrentHashMap 是一個存儲 key/value 對的容器，並且是線程安全的。我們先看 ConcurrentHashMap 的存儲結構，如下圖："}]},{"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，徹底給整明白了","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":"雖然 ConcurrentHashMap 的底層數據結構，和方法的實現細節和 HashMap 大體一致，但兩者在類結構上卻沒有任何關聯，我們看下 ConcurrentHashMap 的類圖："}]},{"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，徹底給整明白了","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 源碼，我們會發現很多方法和代碼和 HashMap 很相似，有的同學可能會問，爲什麼不繼承 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":"繼承的確是個好辦法，但ConcurrentHashMap 都是在方法中間進行一些加鎖操作，也就是說加鎖把方法切割了，繼承就很難解決這個問題。"}]},{"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兩者的相同之處："}]},{"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":"數組、鏈表結構幾乎相同，所以底層對數據結構的操作思路是相同的（只是思路相同，底層實現不同）；"}]},{"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":"都實現了 Map 接口，繼承了 AbstractMap 抽象類，所以大多數的方法也都是相同的，HashMap 有的方法，ConcurrentHashMap 幾乎都有，所以當我們需要從 HashMap 切換到 ConcurrentHashMap 時，無需關心兩者之間的兼容問題。"}]},{"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":"不同之處："}]},{"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":"紅黑樹結構略有不同，HashMap 的紅黑樹中的節點叫做 TreeNode，TreeNode 不僅僅有屬性，還維護着紅黑樹的結構，比如說查找，新增等等；ConcurrentHashMap 中紅黑樹被拆分成兩塊，TreeNode 僅僅維護的屬性和查找功能，新增了 TreeBin，來維護紅黑樹結構，並負責根節點的加鎖和解鎖；"}]},{"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 （轉移）節點，擴容的時候會使用到，通過使用該節點，來保證擴容時的線程安全。"}]},{"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":"這些概念名詞文章後面都會依次介紹"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"基本構成"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"重要屬性"}]},{"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":"我們來看看 ConcurrentHashMap 的幾個重要屬性"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"//這個Node數組就是ConcurrentHashMap用來存儲數據的哈希表。\ntransient volatile Node[] table\n//這是默認的初始化哈希表數組大小\nprivate static final int DEFAULT_CAPACITY = 16;\n//轉化爲紅黑樹的鏈表長度閾值\nstatic final int TREEIFY_THRESHOLD = 8\n//這個標識位用於識別擴容時正在轉移數據\nstatic final int MOVED = -1\n//計算哈希值時用到的參數，用來去除符號位\nstatic final int HASH_BITS = 0x7fffffff;\n//數據轉移時，新的哈希表數組\nprivate transient volatile Node[] nextTable;\n複製代碼"}]},{"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":"重要組成元素"}]},{"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":"鏈表中的元素爲Node對象。他是鏈表上的一個節點，內部存儲了key、value值，以及他的下一個節點的引用。這樣一系列的Node就串成一串，組成一個鏈表。"}]},{"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":"當進行擴容時，要把鏈表遷移到新的哈希表，在做這個操作時，會在把數組中的頭節點替換爲ForwardingNode對象。ForwardingNode中不保存key和value，只保存了擴容後哈希表（nextTable）的引用。此時查找相應node時，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":"當鏈表轉爲紅黑樹後，數組中保存的引用爲 TreeBin，TreeBin 內部不保存 key/value，他保存了 TreeNode的list以及紅黑樹 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":"紅黑樹的節點。"}]},{"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":"下面依次講解各個核心方法，有詳細註釋"}]},{"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複製代碼"}]},{"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，徹底給整明白了","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 方法上的整體思路和 HashMap 相同，但在線程安全方面寫了很多保障的代碼，我們先來看下大體思路："}]},{"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.如果數組爲空，初始化，初始化完成之後，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.計算當前槽點有沒有值，沒有值的話，cas 創建，失敗繼續自旋（for 死循環），直到成功，槽點有值的話，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.如果槽點是轉移節點(正在擴容)，就會一直自旋等待擴容完成之後再新增，不是轉移節點走 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.槽點有值的，先鎖定當前槽點，保證其餘線程不能操作，如果是鏈表，新增值到鏈表的尾部，如果是紅黑樹，使用紅黑樹新增的方法新增；"}]},{"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.新增完成之後 check 需不需要擴容，需要的話去擴容。"}]},{"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，徹底給整明白了","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過程中，採用了哪些手段來保證線程安全呢？"}]},{"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":"數組初始化時的線程安全"}]},{"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":"數組初始化時，首先通過自旋來保證一定可以初始化成功，然後通過 CAS 設置 SIZECTL 變量的值，來保證同一時刻只能有一個線程對數組進行初始化，CAS 成功之後，還會再次判斷當前數組是否已經初始化完成，如果已經初始化完成，就不會再次初始化，通過自旋 + CAS + 雙重 check 等手段保證了數組初始化時的線程安全"}]},{"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":"那麼接下來我們就來看看 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，徹底給整明白了","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":"注意裏面有個關鍵的值 sizeCtl，這個值有多個含義。"}]},{"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 代表有線程正在創建 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 個線程正在複製 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，代表根據構造函數傳入的值計算出的應被初始化的大小；"}]},{"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 被初始化後，則被設置爲 table 大小 的 75%，代表 table 的容量（數組容量）。"}]},{"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":"新增槽點值時的線程安全"}]},{"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":"此時爲了保證線程安全，做了四處優化："}]},{"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.通過自旋死循環保證一定可以新增成功。"}]},{"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，通過 for (Node[] tab = table;;)這樣的死循環來保證新增一定可以成功，一旦新增成功，就可以退出當前死循環，新增失敗的話，會重複新增的步驟，直到新增成功爲止。"}]},{"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.當前槽點爲空時，通過 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 這裏的寫法非常嚴謹，沒有在判斷槽點爲空的情況下直接賦值，因爲在判斷槽點爲空和賦值的瞬間，很有可能槽點已經被其他線程複製了，所以我們採用 CAS 算法，能夠保證槽點爲空的情況下賦值成功，如果恰好槽點已經被其他線程賦值，當前 CAS 操作失敗，會再次執行 for 自旋，再走槽點有值的 put 流程，這裏就是自選 + CAS 的結合。"}]},{"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.當前槽點有值，鎖住當前槽點。"}]},{"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 時，如果當前槽點有值，就是 key 的 hash 衝突的情況，此時槽點上可能是鏈表或紅黑樹，我們通過鎖住槽點，來保證同一時刻只會有一個線程能對槽點進行修改"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"V oldVal = null;\n//鎖定當前槽點，其餘線程不能操作，保證了安全\nsynchronized (f) {\n複製代碼"}]},{"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.紅黑樹旋轉時，鎖住紅黑樹的根節點，保證同一時刻，當前紅黑樹只能被一個線程旋轉"}]},{"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方法源碼分析"}]},{"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":"哈希算法的邏輯，決定 ConcurrentHashMap 保存和讀取速度。"}]},{"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複製代碼"}]},{"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":"傳入的參數h爲 key 對象的 hashCode，spreed 方法對 hashCode 進行了加工。重新計算出 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 值是用來映射該 key 值在哈希表中的位置。取出哈希表中該 hash 值對應位置的代碼如下。"}]},{"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複製代碼"}]},{"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":"我們先看這一行代碼的邏輯，第一個參數爲哈希表，第二個參數是哈希表中的數組下標。通過 (n - 1) & hash 計算下標。n 爲數組長度，我們以默認大小 16 爲例，那麼 n-1 = 15，我們可以假設 hash 值爲 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轉爲二進制：\n0000 0000 0000 0000 0000 0000 0000 1111\nhash的值100轉爲二進制：\n0000 0000 0000 0000 0000 0000 0110 0100。\n計算結果：\n0000 0000 0000 0000 0000 0000 0000 0100\n對應的十進制值爲 4\n複製代碼"}]},{"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的二進制高位都爲0，低位都是1。那麼經過&計算後，hash值100的高位全部被清零，低位則保持不變，並且一定是小於（n-1）的。也就是說經過如此計算，通過hash值得到的數組下標絕對不會越界。"}]},{"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":"這裏提出幾個問題："}]},{"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、數組大小可以爲 17，或者 18 嗎？"}]},{"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、如果爲了保證不越界爲什麼不直接用 % 計算取餘數？"}]},{"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、爲什麼不直接用 key 的 hashCode，而是使用經 spreed 方法加工後的 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":"數組大小必須爲 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":"第一個問題的答案是數組大小必須爲 2 的 n 次方，也就是 16、32、64….不能爲其他值。因爲如果不是 2 的 n 次方，那麼經過計算的數組下標會增大碰撞的機率"}]},{"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值的二進制是 10000（十進制16）、10010（十進制18）、10001（十進制17），和10100做&計算後，都是10000，也就是都被映射到數組16這個下標上。這三個值會以鏈表的形式存儲在數組16下標的位置。這顯然不是我們想要的結果。"}]},{"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":"但如果數組長度n爲2的n次方，2進制的數值爲10，100，1000，10000……n-1後對應二進制爲1，11，111，1111……這樣和hash值低位&後，會保留原來hash值的低位數值，那麼只要hash值的低位不一樣，就不會發生碰撞。"}]},{"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":"同時(n - 1) & hash等價於 hash%n。那麼爲什麼不直接用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":"這是因爲按位的操作效率會更高。"}]},{"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":"爲什麼不直接用 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":"其實說到底還是爲了減少碰撞的概率。我們先看看 spreed 方法中的代碼做了什麼事情："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"h ^ (h >>> 16)\n複製代碼"}]},{"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":"這個意思是把 h 的二進制數值向右移動 16 位。我們知道整形爲 32 位，那麼右移 16 位後，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":"^爲異或操作，二進制按位比較，如果相同則爲 0，不同則爲 1。這行代碼的意思就是把高低16位做異或。如果兩個hashCode值的低16位相同，但是高位不同，經過如此計算，低16位會變得不一樣了。"}]},{"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":"爲什麼要把地位變得不一樣呢？"}]},{"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":"這是由於哈希表數組長度n會是偏小的數值，那麼進行(n - 1) & hash運算時，一直使用的是hash較低位的值。那麼即使hash值不同，但如果低位相當，也會發生碰撞。而進行h ^ (h >>> 16)加工後的hash值，讓hashCode高位的值也參與了哈希運算，因此減少了碰撞的概率。"}]},{"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複製代碼"}]},{"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":"爲何高位移到低位和原來低位做異或操作後，還需要和HASH_BITS這個常量做 & 計算呢？HASH_BITS 這個常量的值爲 0x7fffffff，轉化爲二進制爲 0111 1111 1111 1111 1111 1111 1111 1111。這個操作後會把最高位轉爲 0，其實就是消除了符號位，得到的都是正數。這是因爲負的 hashCode 在ConcurrentHashMap 中有特殊的含義，因此我們需要得到一個正的 hashCode。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"擴容源碼分析"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我們大致瞭解了ConcurrentHashMap 的存儲結構，那麼我們思考一個問題，當數組中保存的鏈表越來越多，那麼再存儲進來的元素大概率會插入到現有的鏈表中，而不是使用數組中剩下的空位。這樣會造成數組中保存的鏈表越來越長，由此導致哈希表查找速度下降，從 O(1) 慢慢趨近於鏈表的時間複雜度 O(n/2)，這顯然違背了哈希表的初衷。"}]},{"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 會做一個操作，稱爲擴容。也就是把數組長度變大，增加更多的空位出來，最終目的就是預防鏈表過長，這樣查找的時間複雜度纔會趨向於 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":"擴容的操作並不會在數組沒有空位時才進行，因爲在空位快滿時，新保存元素更大的概率會命中已經使用的位置，那麼可能最後幾個桶位很難被使用，而鏈表卻越來越長了。"}]},{"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 還會有鏈表轉紅黑樹的操作，以提高查找的速度，紅黑樹時間複雜度爲 O(logn)，而鏈表是 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複製代碼"}]},{"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":"這是一個有參數的構造方法。如果你對未來存儲的數據量有預估，我們可以指定哈希表的大小，避免頻繁的擴容操作。tableSizeFor 這個方法確保了哈希表的大小永遠都是 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":"注意這裏傳入的參數不是 initialCapacity，而是 initialCapacity 的 1.5 倍 + 1。這樣做是爲了保證在默認 75% 的負載因子下，能夠足夠容納 initialCapacity 數量的元素。"}]},{"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) 構造函數總結下："}]},{"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、構造函數中並不會初始化哈希表；"}]},{"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、構造函數中僅設置哈希表大小的變量 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 並不是哈希表大小；"}]},{"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、哈希表大小爲 initialCapacity*1.5+1 後，向上取最小的 2 的 n 次方。如果超過最大容量一半，那麼就是最大容量。"}]},{"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 是如何實現向上取得最接近入參 2 的 n 次方的。下面我們來看 tableSizeFor 源代碼："}]},{"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複製代碼"}]},{"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":"依舊是二進制按位操作，這樣一頓操作後，得到的數值就是大於 c 的最小 2 的 n 次。我們推演下過程，假設 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兩個值按位或後\nn=1100\n3、n |= n >>> 2\nn=1100\nn >>> 2=0011\nn=1111\n複製代碼"}]},{"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":"到這裏可以看出規律來了。如果 c 足夠大，使得 n 很大，那麼運算到 n |= n >>> 16 時，n 的 32 位都爲 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":"總結一下這一段邏輯，其實就是把 n 有數值的 bit 位全部置爲 1。這樣就得到了一個肯定大於等於 n 的值。我們再看最後一行代碼，最終返回的是 n+1，那麼一個所有位都是 1 的二進制數字，+1 後得到的就是一個 2 的 n 次方數值。"}]},{"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":"如果你覺得這篇內容對你還蠻有幫助，我想邀請你幫我三個小忙："}]},{"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":"點贊，轉發，有你們的 『點贊和評論』，纔是我創造的動力。"}]}]},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"關注公衆號 『 "},{"type":"text","marks":[{"type":"strong"}],"text":"java爛豬皮"},{"type":"text","text":" 』，不定期分享原創知識。"}]}]},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"同時可以期待後續文章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}}]}