1 package java.util; 2 3 import sun.misc.SharedSecrets; 4 5 import java.io.IOException; 6 import java.io.InvalidObjectException; 7 import java.io.Serializable; 8 import java.lang.reflect.ParameterizedType; 9 import java.lang.reflect.Type; 10 import java.util.function.BiConsumer; 11 import java.util.function.BiFunction; 12 import java.util.function.Consumer; 13 import java.util.function.Function; 14 15 /** 16 * HashMap是常用的Java集合之一,是基于哈希表的Map接口的实现。与HashTable主要区别为不支持同步和允许null作为key和value。 17 * HashMap非线程安全,即任一时刻可以有多个线程同时写HashMap,可能会导致数据的不一致。 18 * 如果需要满足线程安全,可以用 Collections的synchronizedMap方法使HashMap具有线程安全的能力,或者使用ConcurrentHashMap。 19 * 在JDK1.6中,HashMap采用数组+链表实现,即使用链表处理冲突,同一hash值的链表都存储在一个链表里。 20 * 但是当位于一个数组中的元素较多,即hash值相等的元素较多时,通过key值依次查找的效率较低。 21 * 而JDK1.8中,HashMap采用数组+链表+红黑树实现,当链表长度超过阈值8时,将链表转换为红黑树,这样大大减少了查找时间。 22 * 原本Map.Entry接口的实现类Entry改名为了Node。转化为红黑树时改用另一种实现TreeNode。 23 */ 24 public class HashMap<K, V> extends AbstractMap<K, V> 25 implements Map<K, V>, Cloneable, Serializable { 26 27 private static final long serialVersionUID = 362498820763181265L; 28 29 30 /** 31 * 默认的初始容量(容量为HashMap中槽的数目)是16,且实际容量必须是2的整数次幂。 32 */ 33 static final int DEFAULT_INITIAL_CAPACITY = 1 << 4; // aka 16 34 35 /** 36 * 最大容量(必须是2的幂且小于2的30次方,传入容量过大将被这个值替换) 37 */ 38 static final int MAXIMUM_CAPACITY = 1 << 30; 39 40 /** 41 * 默认装填因子0.75,如果当前键值对个数 >= HashMap最大容量*装填因子,进行rehash操作 42 */ 43 static final float DEFAULT_LOAD_FACTOR = 0.75f; 44 45 /** 46 * JDK1.8 新加,Entry链表最大长度,当桶中节点数目大于该长度时,将链表转成红黑树存储; 47 */ 48 static final int TREEIFY_THRESHOLD = 8; 49 50 /** 51 * JDK1.8 新加,当桶中节点数小于该长度,将红黑树转为链表存储; 52 */ 53 static final int UNTREEIFY_THRESHOLD = 6; 54 55 /** 56 * 桶可能被转化为树形结构的最小容量。当哈希表的大小超过这个阈值,才会把链式结构转化成树型结构,否则仅采取扩容来尝试减少冲突。 57 * 应该至少4*TREEIFY_THRESHOLD来避免扩容和树形结构化之间的冲突。 58 */ 59 static final int MIN_TREEIFY_CAPACITY = 64; 60 61 /** 62 * JDK1.6用Entry描述键值对,JDK1.8中用Node代替Entry 63 */ 64 static class Node<K, V> implements Map.Entry<K, V> { 65 // hash存储key的hashCode 66 final int hash; 67 // final:一个键值对的key不可改变 68 final K key; 69 V value; 70 //指向下个节点的引用 71 Node<K, V> next; 72 73 //构造函数 74 Node(int hash, K key, V value, Node<K, V> next) { 75 this.hash = hash; 76 this.key = key; 77 this.value = value; 78 this.next = next; 79 } 80 81 public final K getKey() { 82 return key; 83 } 84 85 public final V getValue() { 86 return value; 87 } 88 89 public final String toString() { 90 return key + "=" + value; 91 } 92 93 public final int hashCode() { 94 return Objects.hashCode(key) ^ Objects.hashCode(value); 95 } 96 97 public final V setValue(V newValue) { 98 V oldValue = value; 99 value = newValue; 100 return oldValue; 101 } 102 103 public final boolean equals(Object o) { 104 if (o == this) 105 return true; 106 if (o instanceof Map.Entry) { 107 Map.Entry<?, ?> e = (Map.Entry<?, ?>) o; 108 if (Objects.equals(key, e.getKey()) && 109 Objects.equals(value, e.getValue())) 110 return true; 111 } 112 return false; 113 } 114 } 115 116 /* ---------------- Static utilities -------------- */ 117 118 /** 119 * HashMap中键值对的存储形式为链表节点,hashCode相同的节点(位于同一个桶)用链表组织 120 * hash方法分为三步: 121 * 1.取key的hashCode 122 * 2.key的hashCode高16位异或低16位 123 * 3.将第一步和第二步得到的结果进行取模运算。 124 */ 125 static final int hash(Object key) { 126 int h; 127 //计算key的hashCode, h = Objects.hashCode(key) 128 //h >>> 16表示对h无符号右移16位,高位补0,然后h与h >>> 16按位异或 129 return (key == null) ? 0 : (h = key.hashCode()) ^ (h >>> 16); 130 } 131 132 /** 133 * 如果参数x实现了Comparable接口,返回参数x的类名,否则返回null 134 */ 135 static Class<?> comparableClassFor(Object x) { 136 if (x instanceof Comparable) { 137 Class<?> c; 138 Type[] ts, as; 139 Type t; 140 ParameterizedType p; 141 if ((c = x.getClass()) == String.class) // bypass checks 142 return c; 143 if ((ts = c.getGenericInterfaces()) != null) { 144 for (int i = 0; i < ts.length; ++i) { 145 if (((t = ts[i]) instanceof ParameterizedType) && 146 ((p = (ParameterizedType) t).getRawType() == 147 Comparable.class) && 148 (as = p.getActualTypeArguments()) != null && 149 as.length == 1 && as[0] == c) // type arg is c 150 return c; 151 } 152 } 153 } 154 return null; 155 } 156 157 /** 158 * 如果x的类型为kc,则返回k.compareTo(x),否则返回0 159 */ 160 @SuppressWarnings({"rawtypes", "unchecked"}) // for cast to Comparable 161 static int compareComparables(Class<?> kc, Object k, Object x) { 162 return (x == null || x.getClass() != kc ? 0 : 163 ((Comparable) k).compareTo(x)); 164 } 165 166 /** 167 * 结果为>=cap的最小2的自然数幂 168 */ 169 static final int tableSizeFor(int cap) { 170 //先移位再或运算,最终保证返回值是2的整数幂 171 int n = cap - 1; 172 n |= n >>> 1; 173 n |= n >>> 2; 174 n |= n >>> 4; 175 n |= n >>> 8; 176 n |= n >>> 16; 177 return (n < 0) ? 1 : (n >= MAXIMUM_CAPACITY) ? MAXIMUM_CAPACITY : n + 1; 178 } 179 180 /* ---------------- Fields -------------- */ 181 182 /** 183 * 哈希桶数组,分配的时候,table的长度总是2的幂 184 */ 185 transient Node<K, V>[] table; 186 187 /** 188 * HashMap将数据转换成set的另一种存储形式,这个变量主要用于迭代功能 189 */ 190 transient Set<Map.Entry<K, V>> entrySet; 191 192 /** 193 * 实际存储的数量,则HashMap的size()方法,实际返回的就是这个值,isEmpty()也是判断该值是否为0 194 */ 195 transient int size; 196 197 /** 198 * hashmap结构被改变的次数,fail-fast机制 199 */ 200 transient int modCount; 201 202 /** 203 * HashMap的扩容阈值,在HashMap中存储的Node键值对超过这个数量时,自动扩容容量为原来的二倍 204 * 205 * @serial 206 */ 207 int threshold; 208 209 /** 210 * HashMap的负加载因子,可计算出当前table长度下的扩容阈值:threshold = loadFactor * table.length 211 * 212 * @serial 213 */ 214 final float loadFactor; 215 216 /* ---------------- Public operations -------------- */ 217 218 /** 219 * 使用指定的初始化容量initial capacity 和加载因子load factor构造一个空HashMap 220 * 221 * @param initialCapacity 初始化容量 222 * @param loadFactor 加载因子 223 * @throws IllegalArgumentException 如果指定的初始化容量为负数或者加载因子为非正数 224 */ 225 public HashMap(int initialCapacity, float loadFactor) { 226 if (initialCapacity < 0) 227 throw new IllegalArgumentException("Illegal initial capacity: " + 228 initialCapacity); 229 if (initialCapacity > MAXIMUM_CAPACITY) 230 initialCapacity = MAXIMUM_CAPACITY; 231 if (loadFactor <= 0 || Float.isNaN(loadFactor)) 232 throw new IllegalArgumentException("Illegal load factor: " + 233 loadFactor); 234 this.loadFactor = loadFactor; 235 this.threshold = tableSizeFor(initialCapacity); 236 } 237 238 /** 239 * 使用指定的初始化容量initial capacity和默认加载因子DEFAULT_LOAD_FACTOR(0.75)构造一个空HashMap 240 * 241 * @param initialCapacity 初始化容量 242 * @throws IllegalArgumentException 如果指定的初始化容量为负数 243 */ 244 public HashMap(int initialCapacity) { 245 this(initialCapacity, DEFAULT_LOAD_FACTOR); 246 } 247 248 /** 249 * 使用指定的初始化容量(16)和默认加载因子DEFAULT_LOAD_FACTOR(0.75)构造一个空HashMap 250 */ 251 public HashMap() { 252 this.loadFactor = DEFAULT_LOAD_FACTOR; // all other fields defaulted 253 } 254 255 /** 256 * 使用指定Map m构造新的HashMap。使用指定的初始化容量(16)和默认加载因子DEFAULT_LOAD_FACTOR(0.75) 257 * 258 * @param m 指定的map 259 * @throws NullPointerException 如果指定的map是null 260 */ 261 public HashMap(Map<? extends K, ? extends V> m) { 262 this.loadFactor = DEFAULT_LOAD_FACTOR; 263 putMapEntries(m, false); 264 } 265 266 /** 267 * Map.putAll and Map constructor的实现需要的方法 268 * 将m的键值对插入本map中 269 * 270 * @param m 指定的map 271 * @param evict 初始化map时使用false,否则使用true 272 */ 273 final void putMapEntries(Map<? extends K, ? extends V> m, boolean evict) { 274 int s = m.size(); 275 //如果参数map不为空 276 if (s > 0) { 277 // 判断table是否已经初始化 278 if (table == null) { // pre-size 279 // 未初始化,s为m的实际元素个数 280 float ft = ((float) s / loadFactor) + 1.0F; 281 int t = ((ft < (float) MAXIMUM_CAPACITY) ? 282 (int) ft : MAXIMUM_CAPACITY); 283 // 计算得到的t大于阈值,则初始化阈值 284 if (t > threshold) 285 //根据容量初始化临界值 286 threshold = tableSizeFor(t); 287 // 已初始化,并且m元素个数大于阈值,进行扩容处理 288 } else if (s > threshold) 289 //扩容处理 290 resize(); 291 // 将m中的所有元素添加至HashMap中 292 for (Map.Entry<? extends K, ? extends V> e : m.entrySet()) { 293 K key = e.getKey(); 294 V value = e.getValue(); 295 putVal(hash(key), key, value, false, evict); 296 } 297 } 298 } 299 300 /** 301 * 返回map中键值对映射的个数 302 * 303 * @return map中键值对映射的个数 304 */ 305 public int size() { 306 return size; 307 } 308 309 /** 310 * 如果map中没有键值对映射,返回true 311 * 312 * @return 如果map中没有键值对映射,返回true 313 */ 314 public boolean isEmpty() { 315 return size == 0; 316 } 317 318 /** 319 * 返回指定的key映射的value,如果value为null,则返回null 320 * get可以分为三个步骤: 321 * 1.通过hash(Object key)方法计算key的哈希值hash。 322 * 2.通过getNode( int hash, Object key)方法获取node。 323 * 3.如果node为null,返回null,否则返回node.value。 324 * 325 * @see #put(Object, Object) 326 */ 327 public V get(Object key) { 328 Node<K, V> e; 329 //根据key及其hash值查询node节点,如果存在,则返回该节点的value值 330 return (e = getNode(hash(key), key)) == null ? null : e.value; 331 } 332 333 /** 334 * 根据key的哈希值和key获取对应的节点 335 * getNode可分为以下几个步骤: 336 * 1.如果哈希表为空,或key对应的桶为空,返回null 337 * 2.如果桶中的第一个节点就和指定参数hash和key匹配上了,返回这个节点。 338 * 3.如果桶中的第一个节点没有匹配上,而且有后续节点 339 * 3.1如果当前的桶采用红黑树,则调用红黑树的get方法去获取节点 340 * 3.2如果当前的桶不采用红黑树,即桶中节点结构为链式结构,遍历链表,直到key匹配 341 * 4.找到节点返回null,否则返回null。 342 * 343 * @param hash 指定参数key的哈希值 344 * @param key 指定参数key 345 * @return 返回node,如果没有则返回null 346 */ 347 final Node<K, V> getNode(int hash, Object key) { 348 Node<K, V>[] tab; 349 Node<K, V> first, e; 350 int n; 351 K k; 352 //如果哈希表不为空,而且key对应的桶上不为空 353 if ((tab = table) != null && (n = tab.length) > 0 && 354 (first = tab[(n - 1) & hash]) != null) { 355 //如果桶中的第一个节点就和指定参数hash和key匹配上了 356 if (first.hash == hash && // always check first node 357 ((k = first.key) == key || (key != null && key.equals(k)))) 358 //返回桶中的第一个节点 359 return first; 360 //如果桶中的第一个节点没有匹配上,而且有后续节点 361 if ((e = first.next) != null) { 362 //如果当前的桶采用红黑树,则调用红黑树的get方法去获取节点 363 if (first instanceof TreeNode) 364 return ((TreeNode<K, V>) first).getTreeNode(hash, key); 365 //如果当前的桶不采用红黑树,即桶中节点结构为链式结构 366 do { 367 //遍历链表,直到key匹配 368 if (e.hash == hash && 369 ((k = e.key) == key || (key != null && key.equals(k)))) 370 return e; 371 } while ((e = e.next) != null); 372 } 373 } 374 //如果哈希表为空,或者没有找到节点,返回null 375 return null; 376 } 377 378 /** 379 * 如果map中含有key为指定参数key的键值对,返回true 380 * 381 * @param key 指定参数key 382 * @return 如果map中含有key为指定参数key的键值对,返回true 383 * key. 384 */ 385 public boolean containsKey(Object key) { 386 return getNode(hash(key), key) != null; 387 } 388 389 /** 390 * 将指定参数key和指定参数value插入map中,如果key已经存在,那就替换key对应的value 391 * put(K key, V value)可以分为三个步骤: 392 * 1.通过hash(Object key)方法计算key的哈希值。 393 * 2.通过putVal(hash(key), key, value, false, true)方法实现功能。 394 * 3.返回putVal方法返回的结果。 395 * 396 * @param key 指定key 397 * @param value 指定value 398 * @return 如果value被替换,则返回旧的value,否则返回null。当然,可能key对应的value就是null 399 */ 400 public V put(K key, V value) { 401 // 倒数第二个参数false:表示允许旧值替换 402 // 最后一个参数true:表示HashMap不处于创建模式 403 return putVal(hash(key), key, value, false, true); 404 } 405 406 /** 407 * Map.put和其他相关方法的实现需要的方法 408 * putVal方法可以分为下面的几个步骤: 409 * 1.如果哈希表为空,调用resize()创建一个哈希表。 410 * 2.如果指定参数hash在表中没有对应的桶,即为没有碰撞,直接将键值对插入到哈希表中即可。 411 * 3.如果有碰撞,遍历桶,找到key映射的节点 412 * 3.1桶中的第一个节点就匹配了,将桶中的第一个节点记录起来。 413 * 3.2如果桶中的第一个节点没有匹配,且桶中结构为红黑树,则调用红黑树对应的方法插入键值对。 414 * 3.3如果不是红黑树,那么就肯定是链表。遍历链表,如果找到了key映射的节点,就记录这个节点,退出循环。如果没有找到,在链表尾部插入节点。插入后,如果链的长度大于TREEIFY_THRESHOLD这个临界值,则使用treeifyBin方法把链表转为红黑树。 415 * 4.如果找到了key映射的节点,且节点不为null 416 * 4.1记录节点的vlaue。 417 * 4.2如果参数onlyIfAbsent为false,或者oldValue为null,替换value,否则不替换。 418 * 4.3返回记录下来的节点的value。 419 * 5.如果没有找到key映射的节点(2、3步中讲了,这种情况会插入到hashMap中),插入节点后size会加1,这时要检查size是否大于临界值threshold,如果大于会使用resize方法进行扩容。 420 * 421 * @param hash 指定参数key的哈希值 422 * @param key 指定参数key 423 * @param value 指定参数value 424 * @param onlyIfAbsent 如果为true,即使指定参数key在map中已经存在,也不会替换value 425 * @param evict 如果为false,数组table在创建模式中 426 * @return 如果value被替换,则返回旧的value,否则返回null。当然,可能key对应的value就是null。 427 */ 428 final V putVal(int hash, K key, V value, boolean onlyIfAbsent, 429 boolean evict) { 430 Node<K, V>[] tab; 431 Node<K, V> p; 432 int n, i; 433 //如果哈希表为空,调用resize()创建一个哈希表,并用变量n记录哈希表长度 434 if ((tab = table) == null || (n = tab.length) == 0) 435 n = (tab = resize()).length; 436 /** 437 * 如果指定参数hash在表中没有对应的桶,即为没有碰撞 438 * Hash函数,(n - 1) & hash 计算key将被放置的槽位 439 * (n - 1) & hash 本质上是hash % n,位运算更快 440 */ 441 if ((p = tab[i = (n - 1) & hash]) == null) 442 //直接将键值对插入到map中即可 443 tab[i] = newNode(hash, key, value, null); 444 else {// 桶中已经存在元素 445 Node<K, V> e; 446 K k; 447 // 比较桶中第一个元素(数组中的结点)的hash值相等,key相等 448 if (p.hash == hash && 449 ((k = p.key) == key || (key != null && key.equals(k)))) 450 // 将第一个元素赋值给e,用e来记录 451 e = p; 452 // 当前桶中无该键值对,且桶是红黑树结构,按照红黑树结构插入 453 else if (p instanceof TreeNode) 454 e = ((TreeNode<K, V>) p).putTreeVal(this, tab, hash, key, value); 455 // 当前桶中无该键值对,且桶是链表结构,按照链表结构插入到尾部 456 else { 457 for (int binCount = 0; ; ++binCount) { 458 // 遍历到链表尾部 459 if ((e = p.next) == null) { 460 p.next = newNode(hash, key, value, null); 461 // 检查链表长度是否达到阈值,达到将该槽位节点组织形式转为红黑树 462 if (binCount >= TREEIFY_THRESHOLD - 1) // -1 for 1st 463 treeifyBin(tab, hash); 464 break; 465 } 466 // 链表节点的<key, value>与put操作<key, value>相同时,不做重复操作,跳出循环 467 if (e.hash == hash && 468 ((k = e.key) == key || (key != null && key.equals(k)))) 469 break; 470 p = e; 471 } 472 } 473 // 找到或新建一个key和hashCode与插入元素相等的键值对,进行put操作 474 if (e != null) { // existing mapping for key 475 // 记录e的value 476 V oldValue = e.value; 477 /** 478 * onlyIfAbsent为false或旧值为null时,允许替换旧值 479 * 否则无需替换 480 */ 481 if (!onlyIfAbsent || oldValue == null) 482 e.value = value; 483 // 访问后回调 484 afterNodeAccess(e); 485 // 返回旧值 486 return oldValue; 487 } 488 } 489 // 更新结构化修改信息 490 ++modCount; 491 // 键值对数目超过阈值时,进行rehash 492 if (++size > threshold) 493 resize(); 494 // 插入后回调 495 afterNodeInsertion(evict); 496 return null; 497 } 498 499 /** 500 * 对table进行初始化或者扩容。 501 * 如果table为null,则对table进行初始化 502 * 如果对table扩容,因为每次扩容都是翻倍,与原来计算(n-1)&hash的结果相比,节点要么就在原来的位置,要么就被分配到“原位置+旧容量”这个位置 503 * resize的步骤总结为: 504 * 1.计算扩容后的容量,临界值。 505 * 2.将hashMap的临界值修改为扩容后的临界值 506 * 3.根据扩容后的容量新建数组,然后将hashMap的table的引用指向新数组。 507 * 4.将旧数组的元素复制到table中。 508 * 509 * @return the table 510 */ 511 final Node<K, V>[] resize() { 512 //新建oldTab数组保存扩容前的数组table 513 Node<K, V>[] oldTab = table; 514 //获取原来数组的长度 515 int oldCap = (oldTab == null) ? 0 : oldTab.length; 516 //原来数组扩容的临界值 517 int oldThr = threshold; 518 int newCap, newThr = 0; 519 //如果扩容前的容量 > 0 520 if (oldCap > 0) { 521 //如果原来的数组长度大于最大值(2^30) 522 if (oldCap >= MAXIMUM_CAPACITY) { 523 //扩容临界值提高到正无穷 524 threshold = Integer.MAX_VALUE; 525 //无法进行扩容,返回原来的数组 526 return oldTab; 527 //如果现在容量的两倍小于MAXIMUM_CAPACITY且现在的容量大于DEFAULT_INITIAL_CAPACITY 528 } else if ((newCap = oldCap << 1) < MAXIMUM_CAPACITY && 529 oldCap >= DEFAULT_INITIAL_CAPACITY) 530 //临界值变为原来的2倍 531 newThr = oldThr << 1; 532 } else if (oldThr > 0) //如果旧容量 <= 0,而且旧临界值 > 0 533 //数组的新容量设置为老数组扩容的临界值 534 newCap = oldThr; 535 else { //如果旧容量 <= 0,且旧临界值 <= 0,新容量扩充为默认初始化容量,新临界值为DEFAULT_LOAD_FACTOR * DEFAULT_INITIAL_CAPACITY 536 newCap = DEFAULT_INITIAL_CAPACITY;//新数组初始容量设置为默认值 537 newThr = (int) (DEFAULT_LOAD_FACTOR * DEFAULT_INITIAL_CAPACITY);//计算默认容量下的阈值 538 } 539 // 计算新的resize上限 540 if (newThr == 0) {//在当上面的条件判断中,只有oldThr > 0成立时,newThr == 0 541 //ft为临时临界值,下面会确定这个临界值是否合法,如果合法,那就是真正的临界值 542 float ft = (float) newCap * loadFactor; 543 //当新容量< MAXIMUM_CAPACITY且ft < (float)MAXIMUM_CAPACITY,新的临界值为ft,否则为Integer.MAX_VALUE 544 newThr = (newCap < MAXIMUM_CAPACITY && ft < (float) MAXIMUM_CAPACITY ? 545 (int) ft : Integer.MAX_VALUE); 546 } 547 //将扩容后hashMap的临界值设置为newThr 548 threshold = newThr; 549 //创建新的table,初始化容量为newCap 550 @SuppressWarnings({"rawtypes", "unchecked"}) 551 Node<K, V>[] newTab = (Node<K, V>[]) new Node[newCap]; 552 //修改hashMap的table为新建的newTab 553 table = newTab; 554 //如果旧table不为空,将旧table中的元素复制到新的table中 555 if (oldTab != null) { 556 //遍历旧哈希表的每个桶,将旧哈希表中的桶复制到新的哈希表中 557 for (int j = 0; j < oldCap; ++j) { 558 Node<K, V> e; 559 //如果旧桶不为null,使用e记录旧桶 560 if ((e = oldTab[j]) != null) { 561 //将旧桶置为null 562 oldTab[j] = null; 563 //如果旧桶中只有一个node 564 if (e.next == null) 565 //将e也就是oldTab[j]放入newTab中e.hash & (newCap - 1)的位置 566 newTab[e.hash & (newCap - 1)] = e; 567 //如果旧桶中的结构为红黑树 568 else if (e instanceof TreeNode) 569 //将树中的node分离 570 ((TreeNode<K, V>) e).split(this, newTab, j, oldCap); 571 else { //如果旧桶中的结构为链表,链表重排,jdk1.8做的一系列优化 572 Node<K, V> loHead = null, loTail = null; 573 Node<K, V> hiHead = null, hiTail = null; 574 Node<K, V> next; 575 //遍历整个链表中的节点 576 do { 577 next = e.next; 578 // 原索引 579 if ((e.hash & oldCap) == 0) { 580 if (loTail == null) 581 loHead = e; 582 else 583 loTail.next = e; 584 loTail = e; 585 } else {// 原索引+oldCap 586 if (hiTail == null) 587 hiHead = e; 588 else 589 hiTail.next = e; 590 hiTail = e; 591 } 592 } while ((e = next) != null); 593 // 原索引放到bucket里 594 if (loTail != null) { 595 loTail.next = null; 596 newTab[j] = loHead; 597 } 598 // 原索引+oldCap放到bucket里 599 if (hiTail != null) { 600 hiTail.next = null; 601 newTab[j + oldCap] = hiHead; 602 } 603 } 604 } 605 } 606 } 607 return newTab; 608 } 609 610 /** 611 * 将链表转化为红黑树 612 */ 613 final void treeifyBin(Node<K, V>[] tab, int hash) { 614 int n, index; 615 Node<K, V> e; 616 //如果桶数组table为空,或者桶数组table的长度小于MIN_TREEIFY_CAPACITY,不符合转化为红黑树的条件 617 if (tab == null || (n = tab.length) < MIN_TREEIFY_CAPACITY) 618 //扩容 619 resize(); 620 //如果符合转化为红黑树的条件,而且hash对应的桶不为null 621 else if ((e = tab[index = (n - 1) & hash]) != null) { 622 // 红黑树的头、尾节点 623 TreeNode<K, V> hd = null, tl = null; 624 //遍历链表 625 do { 626 //替换链表node为树node,建立双向链表 627 TreeNode<K, V> p = replacementTreeNode(e, null); 628 // 确定树头节点 629 if (tl == null) 630 hd = p; 631 else { 632 p.prev = tl; 633 tl.next = p; 634 } 635 tl = p; 636 } while ((e = e.next) != null); 637 //遍历链表插入每个节点到红黑树 638 if ((tab[index] = hd) != null) 639 hd.treeify(tab); 640 } 641 } 642 643 /** 644 * 将参数map中的所有键值对映射插入到hashMap中,如果有碰撞,则覆盖value。 645 * 646 * @param m 参数map 647 * @throws NullPointerException 如果map为null 648 */ 649 public void putAll(Map<? extends K, ? extends V> m) { 650 putMapEntries(m, true); 651 } 652 653 /** 654 * 删除hashMap中key映射的node 655 * remove方法的实现可以分为三个步骤: 656 * 1.通过 hash(Object key)方法计算key的哈希值。 657 * 2.通过 removeNode 方法实现功能。 658 * 3.返回被删除的node的value。 659 * 660 * @param key 参数key 661 * @return 如果没有映射到node,返回null,否则返回对应的value 662 */ 663 public V remove(Object key) { 664 Node<K, V> e; 665 //根据key来删除node。removeNode方法的具体实现在下面 666 return (e = removeNode(hash(key), key, null, false, true)) == null ? 667 null : e.value; 668 } 669 670 /** 671 * Map.remove和相关方法的实现需要的方法 672 * removeNode方法的步骤总结为: 673 * 1.如果数组table为空或key映射到的桶为空,返回null。 674 * 2.如果key映射到的桶上第一个node的就是要删除的node,记录下来。 675 * 3.如果桶内不止一个node,且桶内的结构为红黑树,记录key映射到的node。 676 * 4.桶内的结构不为红黑树,那么桶内的结构就肯定为链表,遍历链表,找到key映射到的node,记录下来。 677 * 5.如果被记录下来的node不为null,删除node,size-1被删除。 678 * 6.返回被删除的node。 679 * 680 * @param hash key的哈希值 681 * @param key key的哈希值 682 * @param value 如果 matchValue 为true,则value也作为确定被删除的node的条件之一,否则忽略 683 * @param matchValue 如果为true,则value也作为确定被删除的node的条件之一 684 * @param movable 如果为false,删除node时不会删除其他node 685 * @return 返回被删除的node,如果没有node被删除,则返回null(针对红黑树的删除方法) 686 */ 687 final Node<K, V> removeNode(int hash, Object key, Object value, 688 boolean matchValue, boolean movable) { 689 Node<K, V>[] tab; 690 Node<K, V> p; 691 int n, index; 692 //如果数组table不为空且key映射到的桶不为空 693 if ((tab = table) != null && (n = tab.length) > 0 && 694 (p = tab[index = (n - 1) & hash]) != null) { 695 Node<K, V> node = null, e; 696 K k; 697 V v; 698 //如果桶上第一个node的就是要删除的node 699 if (p.hash == hash && 700 ((k = p.key) == key || (key != null && key.equals(k)))) 701 //记录桶上第一个node 702 node = p; 703 else if ((e = p.next) != null) {//如果桶内不止一个node 704 //如果桶内的结构为红黑树 705 if (p instanceof TreeNode) 706 //记录key映射到的node 707 node = ((TreeNode<K, V>) p).getTreeNode(hash, key); 708 else {//如果桶内的结构为链表 709 do {//遍历链表,找到key映射到的node 710 if (e.hash == hash && 711 ((k = e.key) == key || 712 (key != null && key.equals(k)))) { 713 //记录key映射到的node 714 node = e; 715 break; 716 } 717 p = e; 718 } while ((e = e.next) != null); 719 } 720 } 721 //如果得到的node不为null且(matchValue为false||node.value和参数value匹配) 722 if (node != null && (!matchValue || (v = node.value) == value || 723 (value != null && value.equals(v)))) { 724 //如果桶内的结构为红黑树 725 if (node instanceof TreeNode) 726 //使用红黑树的删除方法删除node 727 ((TreeNode<K, V>) node).removeTreeNode(this, tab, movable); 728 else if (node == p)//如果桶的第一个node的就是要删除的node 729 //删除node 730 tab[index] = node.next; 731 else//如果桶内的结构为链表,使用链表删除元素的方式删除node 732 p.next = node.next; 733 ++modCount;//结构性修改次数+1 734 --size;//哈希表大小-1 735 afterNodeRemoval(node); 736 return node;//返回被删除的node 737 } 738 } 739 return null;//如果数组table为空或key映射到的桶为空,返回null。 740 } 741 742 /** 743 * 删除map中所有的键值对 744 */ 745 public void clear() { 746 Node<K, V>[] tab; 747 modCount++; 748 if ((tab = table) != null && size > 0) { 749 size = 0; 750 for (int i = 0; i < tab.length; ++i) 751 tab[i] = null; 752 } 753 } 754 755 /** 756 * 如果hashMap中的键值对有一对或多对的value为参数value,返回true 757 * 758 * @param value 参数value 759 * @return 如果hashMap中的键值对有一对或多对的value为参数value,返回true 760 */ 761 public boolean containsValue(Object value) { 762 Node<K, V>[] tab; 763 V v; 764 if ((tab = table) != null && size > 0) { 765 //遍历数组table 766 for (int i = 0; i < tab.length; ++i) { 767 //遍历桶中的node 768 for (Node<K, V> e = tab[i]; e != null; e = e.next) { 769 if ((v = e.value) == value || 770 (value != null && value.equals(v))) 771 return true; 772 } 773 } 774 } 775 return false; 776 } 777 778 /** 779 * 返回hashMap中所有key的视图。 780 * 改变hashMap会影响到set,反之亦然。 781 * 如果当迭代器迭代set时,hashMap被修改(除非是迭代器自己的remove()方法),迭代器的结果是不确定的。 782 * set支持元素的删除,通过Iterator.remove、Set.remove、removeAll、retainAll、clear操作删除hashMap中对应的键值对。 783 * 不支持add和addAll方法。 784 * 785 * @return 返回hashMap中所有key的set视图 786 */ 787 public Set<K> keySet() { 788 Set<K> ks = keySet; 789 if (ks == null) { 790 ks = new KeySet(); 791 keySet = ks; 792 } 793 return ks; 794 } 795 796 /** 797 * 内部类KeySet 798 */ 799 final class KeySet extends AbstractSet<K> { 800 public final int size() { 801 return size; 802 } 803 804 public final void clear() { 805 HashMap.this.clear(); 806 } 807 808 public final Iterator<K> iterator() { 809 return new KeyIterator(); 810 } 811 812 public final boolean contains(Object o) { 813 return containsKey(o); 814 } 815 816 public final boolean remove(Object key) { 817 return removeNode(hash(key), key, null, false, true) != null; 818 } 819 820 public final Spliterator<K> spliterator() { 821 return new KeySpliterator<>(HashMap.this, 0, -1, 0, 0); 822 } 823 824 public final void forEach(Consumer<? super K> action) { 825 Node<K, V>[] tab; 826 if (action == null) 827 throw new NullPointerException(); 828 if (size > 0 && (tab = table) != null) { 829 int mc = modCount; 830 for (int i = 0; i < tab.length; ++i) { 831 for (Node<K, V> e = tab[i]; e != null; e = e.next) 832 action.accept(e.key); 833 } 834 if (modCount != mc) 835 throw new ConcurrentModificationException(); 836 } 837 } 838 } 839 840 /** 841 * 返回hashMap中所有value的collection视图 842 * 改变hashMap会改变collection,反之亦然。 843 * 如果当迭代器迭代collection时,hashMap被修改(除非是迭代器自己的remove()方法),迭代器的结果是不确定的。 844 * collection支持元素的删除,通过Iterator.remove、Collection.remove、removeAll、retainAll、clear操作删除hashMap中对应的键值对。 845 * 不支持add和addAll方法。 846 * 847 * @return 返回hashMap中所有key的collection视图 848 */ 849 public Collection<V> values() { 850 Collection<V> vs = values; 851 if (vs == null) { 852 vs = new Values(); 853 values = vs; 854 } 855 return vs; 856 } 857 858 /** 859 * 内部类Values 860 */ 861 final class Values extends AbstractCollection<V> { 862 public final int size() { 863 return size; 864 } 865 866 public final void clear() { 867 HashMap.this.clear(); 868 } 869 870 public final Iterator<V> iterator() { 871 return new ValueIterator(); 872 } 873 874 public final boolean contains(Object o) { 875 return containsValue(o); 876 } 877 878 public final Spliterator<V> spliterator() { 879 return new ValueSpliterator<>(HashMap.this, 0, -1, 0, 0); 880 } 881 882 public final void forEach(Consumer<? super V> action) { 883 Node<K, V>[] tab; 884 if (action == null) 885 throw new NullPointerException(); 886 if (size > 0 && (tab = table) != null) { 887 int mc = modCount; 888 for (int i = 0; i < tab.length; ++i) { 889 for (Node<K, V> e = tab[i]; e != null; e = e.next) 890 action.accept(e.value); 891 } 892 if (modCount != mc) 893 throw new ConcurrentModificationException(); 894 } 895 } 896 } 897 898 /** 899 * 返回hashMap中所有键值对的set视图 900 * 改变hashMap会影响到set,反之亦然。 901 * 如果当迭代器迭代set时,hashMap被修改(除非是迭代器自己的remove()方法),迭代器的结果是不确定的。 902 * set支持元素的删除,通过Iterator.remove、Set.remove、removeAll、retainAll、clear操作删除hashMap中对应的键值对。 903 * 不支持add和addAll方法。 904 * 905 * @return 返回hashMap中所有键值对的set视图 906 */ 907 public Set<Map.Entry<K, V>> entrySet() { 908 Set<Map.Entry<K, V>> es; 909 return (es = entrySet) == null ? (entrySet = new EntrySet()) : es; 910 } 911 912 /** 913 * 内部类EntrySet 914 */ 915 final class EntrySet extends AbstractSet<Map.Entry<K, V>> { 916 public final int size() { 917 return size; 918 } 919 920 public final void clear() { 921 HashMap.this.clear(); 922 } 923 924 public final Iterator<Map.Entry<K, V>> iterator() { 925 return new EntryIterator(); 926 } 927 928 public final boolean contains(Object o) { 929 if (!(o instanceof Map.Entry)) 930 return false; 931 Map.Entry<?, ?> e = (Map.Entry<?, ?>) o; 932 Object key = e.getKey(); 933 Node<K, V> candidate = getNode(hash(key), key); 934 return candidate != null && candidate.equals(e); 935 } 936 937 public final boolean remove(Object o) { 938 if (o instanceof Map.Entry) { 939 Map.Entry<?, ?> e = (Map.Entry<?, ?>) o; 940 Object key = e.getKey(); 941 Object value = e.getValue(); 942 return removeNode(hash(key), key, value, true, true) != null; 943 } 944 return false; 945 } 946 947 public final Spliterator<Map.Entry<K, V>> spliterator() { 948 return new EntrySpliterator<>(HashMap.this, 0, -1, 0, 0); 949 } 950 951 public final void forEach(Consumer<? super Map.Entry<K, V>> action) { 952 Node<K, V>[] tab; 953 if (action == null) 954 throw new NullPointerException(); 955 if (size > 0 && (tab = table) != null) { 956 int mc = modCount; 957 for (int i = 0; i < tab.length; ++i) { 958 for (Node<K, V> e = tab[i]; e != null; e = e.next) 959 action.accept(e); 960 } 961 if (modCount != mc) 962 throw new ConcurrentModificationException(); 963 } 964 } 965 } 966 967 // JDK8重写的方法 968 969 /** 970 * 通过key映射到对应node,如果没映射到则返回默认值defaultValue 971 * 972 * @param key 973 * @param defaultValue 974 * @return key映射到对应的node,如果没映射到则返回默认值defaultValue 975 */ 976 @Override 977 public V getOrDefault(Object key, V defaultValue) { 978 Node<K, V> e; 979 return (e = getNode(hash(key), key)) == null ? defaultValue : e.value; 980 } 981 982 /** 983 * 在hashMap中插入参数key和value组成的键值对,如果key在hashMap中已经存在,不替换value 984 * 985 * @param key 986 * @param value 987 * @return 如果key在hashMap中不存在,返回旧value 988 */ 989 @Override 990 public V putIfAbsent(K key, V value) { 991 return putVal(hash(key), key, value, true, true); 992 } 993 994 /** 995 * 删除hashMap中key为参数key,value为参数value的键值对。如果桶中结构为树,则级联删除 996 * 997 * @param key 998 * @param value 999 * @return 删除成功,返回true 1000 */ 1001 @Override 1002 public boolean remove(Object key, Object value) { 1003 return removeNode(hash(key), key, value, true, true) != null; 1004 } 1005 1006 /** 1007 * 使用newValue替换key和oldValue映射到的键值对中的value 1008 * 1009 * @param key 1010 * @param oldValue 1011 * @param newValue 1012 * @return 替换成功,返回true 1013 */ 1014 @Override 1015 public boolean replace(K key, V oldValue, V newValue) { 1016 Node<K, V> e; 1017 V v; 1018 if ((e = getNode(hash(key), key)) != null && 1019 ((v = e.value) == oldValue || (v != null && v.equals(oldValue)))) { 1020 e.value = newValue; 1021 afterNodeAccess(e); 1022 return true; 1023 } 1024 return false; 1025 } 1026 1027 /** 1028 * 使用参数value替换key映射到的键值对中的value 1029 * 1030 * @param key 1031 * @param value 1032 * @return 替换成功,返回true 1033 */ 1034 @Override 1035 public V replace(K key, V value) { 1036 Node<K, V> e; 1037 if ((e = getNode(hash(key), key)) != null) { 1038 V oldValue = e.value; 1039 e.value = value; 1040 afterNodeAccess(e); 1041 return oldValue; 1042 } 1043 return null; 1044 } 1045 1046 @Override 1047 public V computeIfAbsent(K key, 1048 Function<? super K, ? extends V> mappingFunction) { 1049 if (mappingFunction == null) 1050 throw new NullPointerException(); 1051 int hash = hash(key); 1052 Node<K, V>[] tab; 1053 Node<K, V> first; 1054 int n, i; 1055 int binCount = 0; 1056 TreeNode<K, V> t = null; 1057 Node<K, V> old = null; 1058 if (size > threshold || (tab = table) == null || 1059 (n = tab.length) == 0) 1060 n = (tab = resize()).length; 1061 if ((first = tab[i = (n - 1) & hash]) != null) { 1062 if (first instanceof TreeNode) 1063 old = (t = (TreeNode<K, V>) first).getTreeNode(hash, key); 1064 else { 1065 Node<K, V> e = first; 1066 K k; 1067 do { 1068 if (e.hash == hash && 1069 ((k = e.key) == key || (key != null && key.equals(k)))) { 1070 old = e; 1071 break; 1072 } 1073 ++binCount; 1074 } while ((e = e.next) != null); 1075 } 1076 V oldValue; 1077 if (old != null && (oldValue = old.value) != null) { 1078 afterNodeAccess(old); 1079 return oldValue; 1080 } 1081 } 1082 V v = mappingFunction.apply(key); 1083 if (v == null) { 1084 return null; 1085 } else if (old != null) { 1086 old.value = v; 1087 afterNodeAccess(old); 1088 return v; 1089 } else if (t != null) 1090 t.putTreeVal(this, tab, hash, key, v); 1091 else { 1092 tab[i] = newNode(hash, key, v, first); 1093 if (binCount >= TREEIFY_THRESHOLD - 1) 1094 treeifyBin(tab, hash); 1095 } 1096 ++modCount; 1097 ++size; 1098 afterNodeInsertion(true); 1099 return v; 1100 } 1101 1102 public V computeIfPresent(K key, 1103 BiFunction<? super K, ? super V, ? extends V> remappingFunction) { 1104 if (remappingFunction == null) 1105 throw new NullPointerException(); 1106 Node<K, V> e; 1107 V oldValue; 1108 int hash = hash(key); 1109 if ((e = getNode(hash, key)) != null && 1110 (oldValue = e.value) != null) { 1111 V v = remappingFunction.apply(key, oldValue); 1112 if (v != null) { 1113 e.value = v; 1114 afterNodeAccess(e); 1115 return v; 1116 } else 1117 removeNode(hash, key, null, false, true); 1118 } 1119 return null; 1120 } 1121 1122 @Override 1123 public V compute(K key, 1124 BiFunction<? super K, ? super V, ? extends V> remappingFunction) { 1125 if (remappingFunction == null) 1126 throw new NullPointerException(); 1127 int hash = hash(key); 1128 Node<K, V>[] tab; 1129 Node<K, V> first; 1130 int n, i; 1131 int binCount = 0; 1132 TreeNode<K, V> t = null; 1133 Node<K, V> old = null; 1134 if (size > threshold || (tab = table) == null || 1135 (n = tab.length) == 0) 1136 n = (tab = resize()).length; 1137 if ((first = tab[i = (n - 1) & hash]) != null) { 1138 if (first instanceof TreeNode) 1139 old = (t = (TreeNode<K, V>) first).getTreeNode(hash, key); 1140 else { 1141 Node<K, V> e = first; 1142 K k; 1143 do { 1144 if (e.hash == hash && 1145 ((k = e.key) == key || (key != null && key.equals(k)))) { 1146 old = e; 1147 break; 1148 } 1149 ++binCount; 1150 } while ((e = e.next) != null); 1151 } 1152 } 1153 V oldValue = (old == null) ? null : old.value; 1154 V v = remappingFunction.apply(key, oldValue); 1155 if (old != null) { 1156 if (v != null) { 1157 old.value = v; 1158 afterNodeAccess(old); 1159 } else 1160 removeNode(hash, key, null, false, true); 1161 } else if (v != null) { 1162 if (t != null) 1163 t.putTreeVal(this, tab, hash, key, v); 1164 else { 1165 tab[i] = newNode(hash, key, v, first); 1166 if (binCount >= TREEIFY_THRESHOLD - 1) 1167 treeifyBin(tab, hash); 1168 } 1169 ++modCount; 1170 ++size; 1171 afterNodeInsertion(true); 1172 } 1173 return v; 1174 } 1175 1176 @Override 1177 public V merge(K key, V value, 1178 BiFunction<? super V, ? super V, ? extends V> remappingFunction) { 1179 if (value == null) 1180 throw new NullPointerException(); 1181 if (remappingFunction == null) 1182 throw new NullPointerException(); 1183 int hash = hash(key); 1184 Node<K, V>[] tab; 1185 Node<K, V> first; 1186 int n, i; 1187 int binCount = 0; 1188 TreeNode<K, V> t = null; 1189 Node<K, V> old = null; 1190 if (size > threshold || (tab = table) == null || 1191 (n = tab.length) == 0) 1192 n = (tab = resize()).length; 1193 if ((first = tab[i = (n - 1) & hash]) != null) { 1194 if (first instanceof TreeNode) 1195 old = (t = (TreeNode<K, V>) first).getTreeNode(hash, key); 1196 else { 1197 Node<K, V> e = first; 1198 K k; 1199 do { 1200 if (e.hash == hash && 1201 ((k = e.key) == key || (key != null && key.equals(k)))) { 1202 old = e; 1203 break; 1204 } 1205 ++binCount; 1206 } while ((e = e.next) != null); 1207 } 1208 } 1209 if (old != null) { 1210 V v; 1211 if (old.value != null) 1212 v = remappingFunction.apply(old.value, value); 1213 else 1214 v = value; 1215 if (v != null) { 1216 old.value = v; 1217 afterNodeAccess(old); 1218 } else 1219 removeNode(hash, key, null, false, true); 1220 return v; 1221 } 1222 if (value != null) { 1223 if (t != null) 1224 t.putTreeVal(this, tab, hash, key, value); 1225 else { 1226 tab[i] = newNode(hash, key, value, first); 1227 if (binCount >= TREEIFY_THRESHOLD - 1) 1228 treeifyBin(tab, hash); 1229 } 1230 ++modCount; 1231 ++size; 1232 afterNodeInsertion(true); 1233 } 1234 return value; 1235 } 1236 1237 @Override 1238 public void forEach(BiConsumer<? super K, ? super V> action) { 1239 Node<K, V>[] tab; 1240 if (action == null) 1241 throw new NullPointerException(); 1242 if (size > 0 && (tab = table) != null) { 1243 int mc = modCount; 1244 for (int i = 0; i < tab.length; ++i) { 1245 for (Node<K, V> e = tab[i]; e != null; e = e.next) 1246 action.accept(e.key, e.value); 1247 } 1248 if (modCount != mc) 1249 throw new ConcurrentModificationException(); 1250 } 1251 } 1252 1253 @Override 1254 public void replaceAll(BiFunction<? super K, ? super V, ? extends V> function) { 1255 Node<K, V>[] tab; 1256 if (function == null) 1257 throw new NullPointerException(); 1258 if (size > 0 && (tab = table) != null) { 1259 int mc = modCount; 1260 for (int i = 0; i < tab.length; ++i) { 1261 for (Node<K, V> e = tab[i]; e != null; e = e.next) { 1262 e.value = function.apply(e.key, e.value); 1263 } 1264 } 1265 if (modCount != mc) 1266 throw new ConcurrentModificationException(); 1267 } 1268 } 1269 1270 /* ------------------------------------------------------------ */ 1271 // 克隆和序列化 1272 1273 /** 1274 * 浅拷贝。 1275 * clone方法虽然生成了新的HashMap对象,新的HashMap中的table数组虽然也是新生成的,但是数组中的元素还是引用以前的HashMap中的元素。 1276 * 这就导致在对HashMap中的元素进行修改的时候,即对数组中元素进行修改,会导致原对象和clone对象都发生改变,但进行新增或删除就不会影响对方,因为这相当于是对数组做出的改变,clone对象新生成了一个数组。 1277 * 1278 * @return hashMap的浅拷贝 1279 */ 1280 @SuppressWarnings("unchecked") 1281 @Override 1282 public Object clone() { 1283 HashMap<K, V> result; 1284 try { 1285 result = (HashMap<K, V>) super.clone(); 1286 } catch (CloneNotSupportedException e) { 1287 // this shouldn‘t happen, since we are Cloneable 1288 throw new InternalError(e); 1289 } 1290 result.reinitialize(); 1291 result.putMapEntries(this, false); 1292 return result; 1293 } 1294 1295 // These methods are also used when serializing HashSets 1296 final float loadFactor() { 1297 return loadFactor; 1298 } 1299 1300 final int capacity() { 1301 return (table != null) ? table.length : 1302 (threshold > 0) ? threshold : 1303 DEFAULT_INITIAL_CAPACITY; 1304 } 1305 1306 /** 1307 * 序列化hashMap到ObjectOutputStream中 1308 * 将hashMap的总容量capacity、实际容量size、键值对映射写入到ObjectOutputStream中。键值对映射序列化时是无序的。 1309 * 1310 * @serialData The <i>capacity</i> of the HashMap (the length of the 1311 * bucket array) is emitted (int), followed by the 1312 * <i>size</i> (an int, the number of key-value 1313 * mappings), followed by the key (Object) and value (Object) 1314 * for each key-value mapping. The key-value mappings are 1315 * emitted in no particular order. 1316 */ 1317 private void writeObject(java.io.ObjectOutputStream s) 1318 throws IOException { 1319 int buckets = capacity(); 1320 // Write out the threshold, loadfactor, and any hidden stuff 1321 s.defaultWriteObject(); 1322 //写入总容量 1323 s.writeInt(buckets); 1324 //写入实际容量 1325 s.writeInt(size); 1326 //写入键值对 1327 internalWriteEntries(s); 1328 } 1329 1330 /** 1331 * 到ObjectOutputStream中读取hashMap 1332 * 将hashMap的总容量capacity、实际容量size、键值对映射读取出来 1333 */ 1334 private void readObject(java.io.ObjectInputStream s) 1335 throws IOException, ClassNotFoundException { 1336 // 将hashMap的总容量capacity、实际容量size、键值对映射读取出来 1337 s.defaultReadObject(); 1338 //重置hashMap 1339 reinitialize(); 1340 //如果加载因子不合法,抛出异常 1341 if (loadFactor <= 0 || Float.isNaN(loadFactor)) 1342 throw new InvalidObjectException("Illegal load factor: " + 1343 loadFactor); 1344 s.readInt(); //读出桶的数量,忽略 1345 int mappings = s.readInt(); //读出实际容量size 1346 //如果读出的实际容量size小于0,抛出异常 1347 if (mappings < 0) 1348 throw new InvalidObjectException("Illegal mappings count: " + 1349 mappings); 1350 else if (mappings > 0) { // (if zero, use defaults) 1351 // Size the table using given load factor only if within 1352 // range of 0.25...4.0 1353 //调整hashMap大小 1354 float lf = Math.min(Math.max(0.25f, loadFactor), 4.0f); // 加载因子 1355 float fc = (float) mappings / lf + 1.0f; //初步得到的总容量,后续还会处理 1356 //处理初步得到的容量,确认最终的总容量 1357 int cap = ((fc < DEFAULT_INITIAL_CAPACITY) ? 1358 DEFAULT_INITIAL_CAPACITY : 1359 (fc >= MAXIMUM_CAPACITY) ? 1360 MAXIMUM_CAPACITY : 1361 tableSizeFor((int) fc)); 1362 //计算临界值,得到初步的临界值 1363 float ft = (float) cap * lf; 1364 //得到最终的临界值 1365 threshold = ((cap < MAXIMUM_CAPACITY && ft < MAXIMUM_CAPACITY) ? 1366 (int) ft : Integer.MAX_VALUE); 1367 1368 // Check Map.Entry[].class since it‘s the nearest public type to 1369 // what we‘re actually creating. 1370 SharedSecrets.getJavaOISAccess().checkArray(s, Map.Entry[].class, cap); 1371 //新建桶数组table 1372 @SuppressWarnings({"rawtypes", "unchecked"}) 1373 Node<K, V>[] tab = (Node<K, V>[]) new Node[cap]; 1374 table = tab; 1375 1376 // 读出key和value,并组成键值对插入hashMap中 1377 for (int i = 0; i < mappings; i++) { 1378 @SuppressWarnings("unchecked") 1379 K key = (K) s.readObject(); 1380 @SuppressWarnings("unchecked") 1381 V value = (V) s.readObject(); 1382 putVal(hash(key), key, value, false, false); 1383 } 1384 } 1385 } 1386 1387 /* ------------------------------------------------------------ */ 1388 // iterators 1389 1390 abstract class HashIterator { 1391 Node<K, V> next; // next entry to return 1392 Node<K, V> current; // current entry 1393 int expectedModCount; // for fast-fail 1394 int index; // current slot 1395 1396 HashIterator() { 1397 expectedModCount = modCount; 1398 Node<K, V>[] t = table; 1399 current = next = null; 1400 index = 0; 1401 if (t != null && size > 0) { // advance to first entry 1402 do { 1403 } while (index < t.length && (next = t[index++]) == null); 1404 } 1405 } 1406 1407 public final boolean hasNext() { 1408 return next != null; 1409 } 1410 1411 final Node<K, V> nextNode() { 1412 Node<K, V>[] t; 1413 Node<K, V> e = next; 1414 if (modCount != expectedModCount) 1415 throw new ConcurrentModificationException(); 1416 if (e == null) 1417 throw new NoSuchElementException(); 1418 if ((next = (current = e).next) == null && (t = table) != null) { 1419 do { 1420 } while (index < t.length && (next = t[index++]) == null); 1421 } 1422 return e; 1423 } 1424 1425 public final void remove() { 1426 Node<K, V> p = current; 1427 if (p == null) 1428 throw new IllegalStateException(); 1429 if (modCount != expectedModCount) 1430 throw new ConcurrentModificationException(); 1431 current = null; 1432 K key = p.key; 1433 removeNode(hash(key), key, null, false, false); 1434 expectedModCount = modCount; 1435 } 1436 } 1437 1438 final class KeyIterator extends HashIterator 1439 implements Iterator<K> { 1440 public final K next() { 1441 return nextNode().key; 1442 } 1443 } 1444 1445 final class ValueIterator extends HashIterator 1446 implements Iterator<V> { 1447 public final V next() { 1448 return nextNode().value; 1449 } 1450 } 1451 1452 final class EntryIterator extends HashIterator 1453 implements Iterator<Map.Entry<K, V>> { 1454 public final Map.Entry<K, V> next() { 1455 return nextNode(); 1456 } 1457 } 1458 1459 /* ------------------------------------------------------------ */ 1460 // spliterators 1461 1462 static class HashMapSpliterator<K, V> { 1463 final HashMap<K, V> map; 1464 Node<K, V> current; //记录当前的节点 1465 int index; //当前节点的下标 1466 int fence; //堆大小 1467 int est; //估计大小 1468 int expectedModCount; // for comodification checks 1469 1470 HashMapSpliterator(HashMap<K, V> m, int origin, 1471 int fence, int est, 1472 int expectedModCount) { 1473 this.map = m; 1474 this.index = origin; 1475 this.fence = fence; 1476 this.est = est; 1477 this.expectedModCount = expectedModCount; 1478 } 1479 1480 final int getFence() { // initialize fence and size on first use 1481 int hi; 1482 if ((hi = fence) < 0) { 1483 HashMap<K, V> m = map; 1484 est = m.size; 1485 expectedModCount = m.modCount; 1486 Node<K, V>[] tab = m.table; 1487 hi = fence = (tab == null) ? 0 : tab.length; 1488 } 1489 return hi; 1490 } 1491 1492 public final long estimateSize() { 1493 getFence(); // force init 1494 return (long) est; 1495 } 1496 } 1497 1498 static final class KeySpliterator<K, V> 1499 extends HashMapSpliterator<K, V> 1500 implements Spliterator<K> { 1501 KeySpliterator(HashMap<K, V> m, int origin, int fence, int est, 1502 int expectedModCount) { 1503 super(m, origin, fence, est, expectedModCount); 1504 } 1505 1506 public KeySpliterator<K, V> trySplit() { 1507 int hi = getFence(), lo = index, mid = (lo + hi) >>> 1; 1508 return (lo >= mid || current != null) ? null : 1509 new KeySpliterator<>(map, lo, index = mid, est >>>= 1, 1510 expectedModCount); 1511 } 1512 1513 public void forEachRemaining(Consumer<? super K> action) { 1514 int i, hi, mc; 1515 if (action == null) 1516 throw new NullPointerException(); 1517 HashMap<K, V> m = map; 1518 Node<K, V>[] tab = m.table; 1519 if ((hi = fence) < 0) { 1520 mc = expectedModCount = m.modCount; 1521 hi = fence = (tab == null) ? 0 : tab.length; 1522 } else 1523 mc = expectedModCount; 1524 if (tab != null && tab.length >= hi && 1525 (i = index) >= 0 && (i < (index = hi) || current != null)) { 1526 Node<K, V> p = current; 1527 current = null; 1528 do { 1529 if (p == null) 1530 p = tab[i++]; 1531 else { 1532 action.accept(p.key); 1533 p = p.next; 1534 } 1535 } while (p != null || i < hi); 1536 if (m.modCount != mc) 1537 throw new ConcurrentModificationException(); 1538 } 1539 } 1540 1541 public boolean tryAdvance(Consumer<? super K> action) { 1542 int hi; 1543 if (action == null) 1544 throw new NullPointerException(); 1545 Node<K, V>[] tab = map.table; 1546 if (tab != null && tab.length >= (hi = getFence()) && index >= 0) { 1547 while (current != null || index < hi) { 1548 if (current == null) 1549 current = tab[index++]; 1550 else { 1551 K k = current.key; 1552 current = current.next; 1553 action.accept(k); 1554 if (map.modCount != expectedModCount) 1555 throw new ConcurrentModificationException(); 1556 return true; 1557 } 1558 } 1559 } 1560 return false; 1561 } 1562 1563 public int characteristics() { 1564 return (fence < 0 || est == map.size ? Spliterator.SIZED : 0) | 1565 Spliterator.DISTINCT; 1566 } 1567 } 1568 1569 static final class ValueSpliterator<K, V> 1570 extends HashMapSpliterator<K, V> 1571 implements Spliterator<V> { 1572 ValueSpliterator(HashMap<K, V> m, int origin, int fence, int est, 1573 int expectedModCount) { 1574 super(m, origin, fence, est, expectedModCount); 1575 } 1576 1577 public ValueSpliterator<K, V> trySplit() { 1578 int hi = getFence(), lo = index, mid = (lo + hi) >>> 1; 1579 return (lo >= mid || current != null) ? null : 1580 new ValueSpliterator<>(map, lo, index = mid, est >>>= 1, 1581 expectedModCount); 1582 } 1583 1584 public void forEachRemaining(Consumer<? super V> action) { 1585 int i, hi, mc; 1586 if (action == null) 1587 throw new NullPointerException(); 1588 HashMap<K, V> m = map; 1589 Node<K, V>[] tab = m.table; 1590 if ((hi = fence) < 0) { 1591 mc = expectedModCount = m.modCount; 1592 hi = fence = (tab == null) ? 0 : tab.length; 1593 } else 1594 mc = expectedModCount; 1595 if (tab != null && tab.length >= hi && 1596 (i = index) >= 0 && (i < (index = hi) || current != null)) { 1597 Node<K, V> p = current; 1598 current = null; 1599 do { 1600 if (p == null) 1601 p = tab[i++]; 1602 else { 1603 action.accept(p.value); 1604 p = p.next; 1605 } 1606 } while (p != null || i < hi); 1607 if (m.modCount != mc) 1608 throw new ConcurrentModificationException(); 1609 } 1610 } 1611 1612 public boolean tryAdvance(Consumer<? super V> action) { 1613 int hi; 1614 if (action == null) 1615 throw new NullPointerException(); 1616 Node<K, V>[] tab = map.table; 1617 if (tab != null && tab.length >= (hi = getFence()) && index >= 0) { 1618 while (current != null || index < hi) { 1619 if (current == null) 1620 current = tab[index++]; 1621 else { 1622 V v = current.value; 1623 current = current.next; 1624 action.accept(v); 1625 if (map.modCount != expectedModCount) 1626 throw new ConcurrentModificationException(); 1627 return true; 1628 } 1629 } 1630 } 1631 return false; 1632 } 1633 1634 public int characteristics() { 1635 return (fence < 0 || est == map.size ? Spliterator.SIZED : 0); 1636 } 1637 } 1638 1639 static final class EntrySpliterator<K, V> 1640 extends HashMapSpliterator<K, V> 1641 implements Spliterator<Map.Entry<K, V>> { 1642 EntrySpliterator(HashMap<K, V> m, int origin, int fence, int est, 1643 int expectedModCount) { 1644 super(m, origin, fence, est, expectedModCount); 1645 } 1646 1647 public EntrySpliterator<K, V> trySplit() { 1648 int hi = getFence(), lo = index, mid = (lo + hi) >>> 1; 1649 return (lo >= mid || current != null) ? null : 1650 new EntrySpliterator<>(map, lo, index = mid, est >>>= 1, 1651 expectedModCount); 1652 } 1653 1654 public void forEachRemaining(Consumer<? super Map.Entry<K, V>> action) { 1655 int i, hi, mc; 1656 if (action == null) 1657 throw new NullPointerException(); 1658 HashMap<K, V> m = map; 1659 Node<K, V>[] tab = m.table; 1660 if ((hi = fence) < 0) { 1661 mc = expectedModCount = m.modCount; 1662 hi = fence = (tab == null) ? 0 : tab.length; 1663 } else 1664 mc = expectedModCount; 1665 if (tab != null && tab.length >= hi && 1666 (i = index) >= 0 && (i < (index = hi) || current != null)) { 1667 Node<K, V> p = current; 1668 current = null; 1669 do { 1670 if (p == null) 1671 p = tab[i++]; 1672 else { 1673 action.accept(p); 1674 p = p.next; 1675 } 1676 } while (p != null || i < hi); 1677 if (m.modCount != mc) 1678 throw new ConcurrentModificationException(); 1679 } 1680 } 1681 1682 public boolean tryAdvance(Consumer<? super Map.Entry<K, V>> action) { 1683 int hi; 1684 if (action == null) 1685 throw new NullPointerException(); 1686 Node<K, V>[] tab = map.table; 1687 if (tab != null && tab.length >= (hi = getFence()) && index >= 0) { 1688 while (current != null || index < hi) { 1689 if (current == null) 1690 current = tab[index++]; 1691 else { 1692 Node<K, V> e = current; 1693 current = current.next; 1694 action.accept(e); 1695 if (map.modCount != expectedModCount) 1696 throw new ConcurrentModificationException(); 1697 return true; 1698 } 1699 } 1700 } 1701 return false; 1702 } 1703 1704 public int characteristics() { 1705 return (fence < 0 || est == map.size ? Spliterator.SIZED : 0) | 1706 Spliterator.DISTINCT; 1707 } 1708 } 1709 1710 /* ------------------------------------------------------------ */ 1711 // LinkedHashMap support 1712 1713 1714 /* 1715 * The following package-protected methods are designed to be 1716 * overridden by LinkedHashMap, but not by any other subclass. 1717 * Nearly all other internal methods are also package-protected 1718 * but are declared final, so can be used by LinkedHashMap, view 1719 * classes, and HashSet. 1720 */ 1721 1722 // 创建一个链表结点 1723 Node<K, V> newNode(int hash, K key, V value, Node<K, V> next) { 1724 return new Node<>(hash, key, value, next); 1725 } 1726 1727 // 替换一个链表节点 1728 Node<K, V> replacementNode(Node<K, V> p, Node<K, V> next) { 1729 return new Node<>(p.hash, p.key, p.value, next); 1730 } 1731 1732 // 创建一个红黑树节点 1733 TreeNode<K, V> newTreeNode(int hash, K key, V value, Node<K, V> next) { 1734 return new TreeNode<>(hash, key, value, next); 1735 } 1736 1737 // 替换一个红黑树节点 1738 TreeNode<K, V> replacementTreeNode(Node<K, V> p, Node<K, V> next) { 1739 return new TreeNode<>(p.hash, p.key, p.value, next); 1740 } 1741 1742 /** 1743 * Reset to initial default state. Called by clone and readObject. 1744 */ 1745 void reinitialize() { 1746 table = null; 1747 entrySet = null; 1748 keySet = null; 1749 values = null; 1750 modCount = 0; 1751 threshold = 0; 1752 size = 0; 1753 } 1754 1755 // Callbacks to allow LinkedHashMap post-actions 1756 void afterNodeAccess(Node<K, V> p) { 1757 } 1758 1759 void afterNodeInsertion(boolean evict) { 1760 } 1761 1762 void afterNodeRemoval(Node<K, V> p) { 1763 } 1764 1765 // 写入hashMap键值对到ObjectOutputStream中 1766 void internalWriteEntries(java.io.ObjectOutputStream s) throws IOException { 1767 Node<K, V>[] tab; 1768 if (size > 0 && (tab = table) != null) { 1769 for (int i = 0; i < tab.length; ++i) { 1770 for (Node<K, V> e = tab[i]; e != null; e = e.next) { 1771 s.writeObject(e.key); 1772 s.writeObject(e.value); 1773 } 1774 } 1775 } 1776 } 1777 1778 /* ------------------------------------------------------------ */ 1779 // Tree bins 1780 1781 /** 1782 * JDK1.8新增,用来支持桶的红黑树结构实现 1783 * 性质1. 节点是红色或黑色。 1784 * 性质2. 根是黑色。 1785 * 性质3. 所有叶子都是黑色(叶子是NIL节点)。 1786 * 性质4. 每个红色节点必须有两个黑色的子节点。(从每个叶子到根的所有路径上不能有两个连续的红色节点。) 1787 * 性质5. 从任一节点到其每个叶子的所有简单路径都包含相同数目的黑色节点。 1788 */ 1789 1790 static final class TreeNode<K, V> extends LinkedHashMap.Entry<K, V> { 1791 TreeNode<K, V> parent; //节点的父亲 1792 TreeNode<K, V> left; //节点的左孩子 1793 TreeNode<K, V> right; //节点的右孩子 1794 TreeNode<K, V> prev; //节点的前一个节点 1795 boolean red; //true表示红节点,false表示黑节点 1796 1797 TreeNode(int hash, K key, V val, Node<K, V> next) { 1798 super(hash, key, val, next); 1799 } 1800 1801 /** 1802 * 获取红黑树的根 1803 */ 1804 final TreeNode<K, V> root() { 1805 for (TreeNode<K, V> r = this, p; ; ) { 1806 if ((p = r.parent) == null) 1807 return r; 1808 r = p; 1809 } 1810 } 1811 1812 /** 1813 * 确保root是桶中的第一个元素 ,将root移到中中的第一个 1814 */ 1815 static <K, V> void moveRootToFront(Node<K, V>[] tab, TreeNode<K, V> root) { 1816 int n; 1817 if (root != null && tab != null && (n = tab.length) > 0) { 1818 int index = (n - 1) & root.hash; 1819 TreeNode<K, V> first = (TreeNode<K, V>) tab[index]; 1820 if (root != first) { 1821 Node<K, V> rn; 1822 tab[index] = root; 1823 TreeNode<K, V> rp = root.prev; 1824 if ((rn = root.next) != null) 1825 ((TreeNode<K, V>) rn).prev = rp; 1826 if (rp != null) 1827 rp.next = rn; 1828 if (first != null) 1829 first.prev = root; 1830 root.next = first; 1831 root.prev = null; 1832 } 1833 assert checkInvariants(root); 1834 } 1835 } 1836 1837 /** 1838 * 查找hash为h,key为k的节点 1839 */ 1840 final TreeNode<K, V> find(int h, Object k, Class<?> kc) { 1841 TreeNode<K, V> p = this; 1842 do { 1843 int ph, dir; 1844 K pk; 1845 TreeNode<K, V> pl = p.left, pr = p.right, q; 1846 if ((ph = p.hash) > h) 1847 p = pl; 1848 else if (ph < h) 1849 p = pr; 1850 else if ((pk = p.key) == k || (k != null && k.equals(pk))) 1851 return p; 1852 else if (pl == null) 1853 p = pr; 1854 else if (pr == null) 1855 p = pl; 1856 else if ((kc != null || 1857 (kc = comparableClassFor(k)) != null) && 1858 (dir = compareComparables(kc, k, pk)) != 0) 1859 p = (dir < 0) ? pl : pr; 1860 else if ((q = pr.find(h, k, kc)) != null) 1861 return q; 1862 else 1863 p = pl; 1864 } while (p != null); 1865 return null; 1866 } 1867 1868 /** 1869 * 获取树节点,通过根节点查找 1870 */ 1871 final TreeNode<K, V> getTreeNode(int h, Object k) { 1872 return ((parent != null) ? root() : this).find(h, k, null); 1873 } 1874 1875 /** 1876 * 比较2个对象的大小 1877 */ 1878 static int tieBreakOrder(Object a, Object b) { 1879 int d; 1880 if (a == null || b == null || 1881 (d = a.getClass().getName(). 1882 compareTo(b.getClass().getName())) == 0) 1883 d = (System.identityHashCode(a) <= System.identityHashCode(b) ? 1884 -1 : 1); 1885 return d; 1886 } 1887 1888 /** 1889 * 将链表转为二叉树 1890 * 1891 * @return root of tree 1892 */ 1893 final void treeify(Node<K, V>[] tab) { 1894 TreeNode<K, V> root = null; 1895 for (TreeNode<K, V> x = this, next; x != null; x = next) { 1896 next = (TreeNode<K, V>) x.next; 1897 x.left = x.right = null; 1898 if (root == null) { 1899 x.parent = null; 1900 x.red = false; 1901 root = x; 1902 } else { 1903 K k = x.key; 1904 int h = x.hash; 1905 Class<?> kc = null; 1906 for (TreeNode<K, V> p = root; ; ) { 1907 int dir, ph; 1908 K pk = p.key; 1909 if ((ph = p.hash) > h) 1910 dir = -1; 1911 else if (ph < h) 1912 dir = 1; 1913 else if ((kc == null && 1914 (kc = comparableClassFor(k)) == null) || 1915 (dir = compareComparables(kc, k, pk)) == 0) 1916 dir = tieBreakOrder(k, pk); 1917 1918 TreeNode<K, V> xp = p; 1919 if ((p = (dir <= 0) ? p.left : p.right) == null) { 1920 x.parent = xp; 1921 if (dir <= 0) 1922 xp.left = x; 1923 else 1924 xp.right = x; 1925 root = balanceInsertion(root, x); 1926 break; 1927 } 1928 } 1929 } 1930 } 1931 moveRootToFront(tab, root); 1932 } 1933 1934 /** 1935 * 将二叉树转为链表 1936 */ 1937 final Node<K, V> untreeify(HashMap<K, V> map) { 1938 Node<K, V> hd = null, tl = null; 1939 for (Node<K, V> q = this; q != null; q = q.next) { 1940 Node<K, V> p = map.replacementNode(q, null); 1941 if (tl == null) 1942 hd = p; 1943 else 1944 tl.next = p; 1945 tl = p; 1946 } 1947 return hd; 1948 } 1949 1950 /** 1951 * 添加一个键值对 1952 */ 1953 final TreeNode<K, V> putTreeVal(HashMap<K, V> map, Node<K, V>[] tab, 1954 int h, K k, V v) { 1955 Class<?> kc = null; 1956 boolean searched = false; 1957 TreeNode<K, V> root = (parent != null) ? root() : this; 1958 for (TreeNode<K, V> p = root; ; ) { 1959 int dir, ph; 1960 K pk; 1961 if ((ph = p.hash) > h) 1962 dir = -1; 1963 else if (ph < h) 1964 dir = 1; 1965 else if ((pk = p.key) == k || (k != null && k.equals(pk))) 1966 return p; 1967 else if ((kc == null && 1968 (kc = comparableClassFor(k)) == null) || 1969 (dir = compareComparables(kc, k, pk)) == 0) { 1970 if (!searched) { 1971 TreeNode<K, V> q, ch; 1972 searched = true; 1973 if (((ch = p.left) != null && 1974 (q = ch.find(h, k, kc)) != null) || 1975 ((ch = p.right) != null && 1976 (q = ch.find(h, k, kc)) != null)) 1977 return q; 1978 } 1979 dir = tieBreakOrder(k, pk); 1980 } 1981 1982 TreeNode<K, V> xp = p; 1983 if ((p = (dir <= 0) ? p.left : p.right) == null) { 1984 Node<K, V> xpn = xp.next; 1985 TreeNode<K, V> x = map.newTreeNode(h, k, v, xpn); 1986 if (dir <= 0) 1987 xp.left = x; 1988 else 1989 xp.right = x; 1990 xp.next = x; 1991 x.parent = x.prev = xp; 1992 if (xpn != null) 1993 ((TreeNode<K, V>) xpn).prev = x; 1994 moveRootToFront(tab, balanceInsertion(root, x)); 1995 return null; 1996 } 1997 } 1998 } 1999 2000 /** 2001 * Removes the given node, that must be present before this call. 2002 * This is messier than typical red-black deletion code because we 2003 * cannot swap the contents of an interior node with a leaf 2004 * successor that is pinned by "next" pointers that are accessible 2005 * independently during traversal. So instead we swap the tree 2006 * linkages. If the current tree appears to have too few nodes, 2007 * the bin is converted back to a plain bin. (The test triggers 2008 * somewhere between 2 and 6 nodes, depending on tree structure). 2009 */ 2010 final void removeTreeNode(HashMap<K, V> map, Node<K, V>[] tab, 2011 boolean movable) { 2012 int n; 2013 if (tab == null || (n = tab.length) == 0) 2014 return; 2015 int index = (n - 1) & hash; 2016 TreeNode<K, V> first = (TreeNode<K, V>) tab[index], root = first, rl; 2017 TreeNode<K, V> succ = (TreeNode<K, V>) next, pred = prev; 2018 if (pred == null) 2019 tab[index] = first = succ; 2020 else 2021 pred.next = succ; 2022 if (succ != null) 2023 succ.prev = pred; 2024 if (first == null) 2025 return; 2026 if (root.parent != null) 2027 root = root.root(); 2028 if (root == null || root.right == null || 2029 (rl = root.left) == null || rl.left == null) { 2030 tab[index] = first.untreeify(map); // too small 2031 return; 2032 } 2033 TreeNode<K, V> p = this, pl = left, pr = right, replacement; 2034 if (pl != null && pr != null) { 2035 TreeNode<K, V> s = pr, sl; 2036 while ((sl = s.left) != null) // find successor 2037 s = sl; 2038 boolean c = s.red; 2039 s.red = p.red; 2040 p.red = c; // swap colors 2041 TreeNode<K, V> sr = s.right; 2042 TreeNode<K, V> pp = p.parent; 2043 if (s == pr) { // p was s‘s direct parent 2044 p.parent = s; 2045 s.right = p; 2046 } else { 2047 TreeNode<K, V> sp = s.parent; 2048 if ((p.parent = sp) != null) { 2049 if (s == sp.left) 2050 sp.left = p; 2051 else 2052 sp.right = p; 2053 } 2054 if ((s.right = pr) != null) 2055 pr.parent = s; 2056 } 2057 p.left = null; 2058 if ((p.right = sr) != null) 2059 sr.parent = p; 2060 if ((s.left = pl) != null) 2061 pl.parent = s; 2062 if ((s.parent = pp) == null) 2063 root = s; 2064 else if (p == pp.left) 2065 pp.left = s; 2066 else 2067 pp.right = s; 2068 if (sr != null) 2069 replacement = sr; 2070 else 2071 replacement = p; 2072 } else if (pl != null) 2073 replacement = pl; 2074 else if (pr != null) 2075 replacement = pr; 2076 else 2077 replacement = p; 2078 if (replacement != p) { 2079 TreeNode<K, V> pp = replacement.parent = p.parent; 2080 if (pp == null) 2081 root = replacement; 2082 else if (p == pp.left) 2083 pp.left = replacement; 2084 else 2085 pp.right = replacement; 2086 p.left = p.right = p.parent = null; 2087 } 2088 2089 TreeNode<K, V> r = p.red ? root : balanceDeletion(root, replacement); 2090 2091 if (replacement == p) { // detach 2092 TreeNode<K, V> pp = p.parent; 2093 p.parent = null; 2094 if (pp != null) { 2095 if (p == pp.left) 2096 pp.left = null; 2097 else if (p == pp.right) 2098 pp.right = null; 2099 } 2100 } 2101 if (movable) 2102 moveRootToFront(tab, r); 2103 } 2104 2105 /** 2106 * 将结点太多的桶分割 2107 * 2108 * @param map the map 2109 * @param tab the table for recording bin heads 2110 * @param index the index of the table being split 2111 * @param bit the bit of hash to split on 2112 */ 2113 final void split(HashMap<K, V> map, Node<K, V>[] tab, int index, int bit) { 2114 TreeNode<K, V> b = this; 2115 // Relink into lo and hi lists, preserving order 2116 TreeNode<K, V> loHead = null, loTail = null; 2117 TreeNode<K, V> hiHead = null, hiTail = null; 2118 int lc = 0, hc = 0; 2119 for (TreeNode<K, V> e = b, next; e != null; e = next) { 2120 next = (TreeNode<K, V>) e.next; 2121 e.next = null; 2122 if ((e.hash & bit) == 0) { 2123 if ((e.prev = loTail) == null) 2124 loHead = e; 2125 else 2126 loTail.next = e; 2127 loTail = e; 2128 ++lc; 2129 } else { 2130 if ((e.prev = hiTail) == null) 2131 hiHead = e; 2132 else 2133 hiTail.next = e; 2134 hiTail = e; 2135 ++hc; 2136 } 2137 } 2138 2139 if (loHead != null) { 2140 if (lc <= UNTREEIFY_THRESHOLD) 2141 tab[index] = loHead.untreeify(map); 2142 else { 2143 tab[index] = loHead; 2144 if (hiHead != null) // (else is already treeified) 2145 loHead.treeify(tab); 2146 } 2147 } 2148 if (hiHead != null) { 2149 if (hc <= UNTREEIFY_THRESHOLD) 2150 tab[index + bit] = hiHead.untreeify(map); 2151 else { 2152 tab[index + bit] = hiHead; 2153 if (loHead != null) 2154 hiHead.treeify(tab); 2155 } 2156 } 2157 } 2158 2159 /* ------------------------------------------------------------ */ 2160 // 红黑树方法,都是从CLR中修改的 2161 2162 /** 2163 * 左旋转 2164 * 2165 * @param root 2166 * @param p 2167 * @param <K> 2168 * @param <V> 2169 * @return 2170 */ 2171 static <K, V> TreeNode<K, V> rotateLeft(TreeNode<K, V> root, 2172 TreeNode<K, V> p) { 2173 TreeNode<K, V> r, pp, rl; 2174 if (p != null && (r = p.right) != null) { 2175 if ((rl = p.right = r.left) != null) 2176 rl.parent = p; 2177 if ((pp = r.parent = p.parent) == null) 2178 (root = r).red = false; 2179 else if (pp.left == p) 2180 pp.left = r; 2181 else 2182 pp.right = r; 2183 r.left = p; 2184 p.parent = r; 2185 } 2186 return root; 2187 } 2188 2189 /** 2190 * 右旋转 2191 * 2192 * @param root 2193 * @param p 2194 * @param <K> 2195 * @param <V> 2196 * @return 2197 */ 2198 static <K, V> TreeNode<K, V> rotateRight(TreeNode<K, V> root, 2199 TreeNode<K, V> p) { 2200 TreeNode<K, V> l, pp, lr; 2201 if (p != null && (l = p.left) != null) { 2202 if ((lr = p.left = l.right) != null) 2203 lr.parent = p; 2204 if ((pp = l.parent = p.parent) == null) 2205 (root = l).red = false; 2206 else if (pp.right == p) 2207 pp.right = l; 2208 else 2209 pp.left = l; 2210 l.right = p; 2211 p.parent = l; 2212 } 2213 return root; 2214 } 2215 2216 /** 2217 * 保证插入后平衡 2218 * 2219 * @param root 2220 * @param x 2221 * @param <K> 2222 * @param <V> 2223 * @return 2224 */ 2225 static <K, V> TreeNode<K, V> balanceInsertion(TreeNode<K, V> root, 2226 TreeNode<K, V> x) { 2227 x.red = true; 2228 for (TreeNode<K, V> xp, xpp, xppl, xppr; ; ) { 2229 if ((xp = x.parent) == null) { 2230 x.red = false; 2231 return x; 2232 } else if (!xp.red || (xpp = xp.parent) == null) 2233 return root; 2234 if (xp == (xppl = xpp.left)) { 2235 if ((xppr = xpp.right) != null && xppr.red) { 2236 xppr.red = false; 2237 xp.red = false; 2238 xpp.red = true; 2239 x = xpp; 2240 } else { 2241 if (x == xp.right) { 2242 root = rotateLeft(root, x = xp); 2243 xpp = (xp = x.parent) == null ? null : xp.parent; 2244 } 2245 if (xp != null) { 2246 xp.red = false; 2247 if (xpp != null) { 2248 xpp.red = true; 2249 root = rotateRight(root, xpp); 2250 } 2251 } 2252 } 2253 } else { 2254 if (xppl != null && xppl.red) { 2255 xppl.red = false; 2256 xp.red = false; 2257 xpp.red = true; 2258 x = xpp; 2259 } else { 2260 if (x == xp.left) { 2261 root = rotateRight(root, x = xp); 2262 xpp = (xp = x.parent) == null ? null : xp.parent; 2263 } 2264 if (xp != null) { 2265 xp.red = false; 2266 if (xpp != null) { 2267 xpp.red = true; 2268 root = rotateLeft(root, xpp); 2269 } 2270 } 2271 } 2272 } 2273 } 2274 } 2275 2276 /** 2277 * 删除后调整平衡 2278 * 2279 * @param root 2280 * @param x 2281 * @param <K> 2282 * @param <V> 2283 * @return 2284 */ 2285 static <K, V> TreeNode<K, V> balanceDeletion(TreeNode<K, V> root, 2286 TreeNode<K, V> x) { 2287 for (TreeNode<K, V> xp, xpl, xpr; ; ) { 2288 if (x == null || x == root) 2289 return root; 2290 else if ((xp = x.parent) == null) { 2291 x.red = false; 2292 return x; 2293 } else if (x.red) { 2294 x.red = false; 2295 return root; 2296 } else if ((xpl = xp.left) == x) { 2297 if ((xpr = xp.right) != null && xpr.red) { 2298 xpr.red = false; 2299 xp.red = true; 2300 root = rotateLeft(root, xp); 2301 xpr = (xp = x.parent) == null ? null : xp.right; 2302 } 2303 if (xpr == null) 2304 x = xp; 2305 else { 2306 TreeNode<K, V> sl = xpr.left, sr = xpr.right; 2307 if ((sr == null || !sr.red) && 2308 (sl == null || !sl.red)) { 2309 xpr.red = true; 2310 x = xp; 2311 } else { 2312 if (sr == null || !sr.red) { 2313 if (sl != null) 2314 sl.red = false; 2315 xpr.red = true; 2316 root = rotateRight(root, xpr); 2317 xpr = (xp = x.parent) == null ? 2318 null : xp.right; 2319 } 2320 if (xpr != null) { 2321 xpr.red = (xp == null) ? false : xp.red; 2322 if ((sr = xpr.right) != null) 2323 sr.red = false; 2324 } 2325 if (xp != null) { 2326 xp.red = false; 2327 root = rotateLeft(root, xp); 2328 } 2329 x = root; 2330 } 2331 } 2332 } else { // symmetric 2333 if (xpl != null && xpl.red) { 2334 xpl.red = false; 2335 xp.red = true; 2336 root = rotateRight(root, xp); 2337 xpl = (xp = x.parent) == null ? null : xp.left; 2338 } 2339 if (xpl == null) 2340 x = xp; 2341 else { 2342 TreeNode<K, V> sl = xpl.left, sr = xpl.right; 2343 if ((sl == null || !sl.red) && 2344 (sr == null || !sr.red)) { 2345 xpl.red = true; 2346 x = xp; 2347 } else { 2348 if (sl == null || !sl.red) { 2349 if (sr != null) 2350 sr.red = false; 2351 xpl.red = true; 2352 root = rotateLeft(root, xpl); 2353 xpl = (xp = x.parent) == null ? 2354 null : xp.left; 2355 } 2356 if (xpl != null) { 2357 xpl.red = (xp == null) ? false : xp.red; 2358 if ((sl = xpl.left) != null) 2359 sl.red = false; 2360 } 2361 if (xp != null) { 2362 xp.red = false; 2363 root = rotateRight(root, xp); 2364 } 2365 x = root; 2366 } 2367 } 2368 } 2369 } 2370 } 2371 2372 /** 2373 * 检测是否符合红黑树 2374 */ 2375 static <K, V> boolean checkInvariants(TreeNode<K, V> t) { 2376 TreeNode<K, V> tp = t.parent, tl = t.left, tr = t.right, 2377 tb = t.prev, tn = (TreeNode<K, V>) t.next; 2378 if (tb != null && tb.next != t) 2379 return false; 2380 if (tn != null && tn.prev != t) 2381 return false; 2382 if (tp != null && t != tp.left && t != tp.right) 2383 return false; 2384 if (tl != null && (tl.parent != t || tl.hash > t.hash)) 2385 return false; 2386 if (tr != null && (tr.parent != t || tr.hash < t.hash)) 2387 return false; 2388 if (t.red && tl != null && tl.red && tr != null && tr.red) 2389 return false; 2390 if (tl != null && !checkInvariants(tl)) 2391 return false; 2392 if (tr != null && !checkInvariants(tr)) 2393 return false; 2394 return true; 2395 } 2396 } 2397 2398 }