摘要：對(duì)象池類的成員應(yīng)該都是靜態(tài)的。事實(shí)上，由于對(duì)象池技術(shù)將對(duì)象限制在一定的數(shù)量，也有效地減少了應(yīng)用程序內(nèi)存上的開(kāi)銷。對(duì)生成時(shí)開(kāi)銷不大的對(duì)象進(jìn)行池化，反而可能會(huì)出現(xiàn)維護(hù)對(duì)象池的開(kāi)銷大于生成新對(duì)象的開(kāi)銷，從而使性能降低的情況。

在學(xué)習(xí) React 事件系統(tǒng)的時(shí)候，在事件分發(fā)的 dispatch方法發(fā)現(xiàn)了調(diào)用了一個(gè) pooledClass 方法，一時(shí)半會(huì)沒(méi)看明白這個(gè)方法的用意。

我們先看一下是怎么用的：

// step1
function TopLevelCallbackBookKeeping(topLevelType, nativeEvent) {
  this.topLevelType = topLevelType;
  this.nativeEvent = nativeEvent;
  this.ancestors = [];
}
Object.assign(TopLevelCallbackBookKeeping.prototype, {
  destructor: function() {
    this.topLevelType = null;
    this.nativeEvent = null;
    this.ancestors.length = 0;
  },
});
PooledClass.addPoolingTo(
  TopLevelCallbackBookKeeping,
  PooledClass.twoArgumentPooler
);

// step2
var bookKeeping = TopLevelCallbackBookKeeping.getPooled(
  topLevelType,
  nativeEvent
);
// bookKeeping 是 TopLevelCallbackBookKeeping 的實(shí)例
try {
  // Event queue being processed in the same cycle allows
  // `preventDefault`.
  ReactUpdates.batchedUpdates(handleTopLevelImpl, bookKeeping);
} finally {
  //釋放
  TopLevelCallbackBookKeeping.release(bookKeeping);
}

那么這里為什么不直接 new 一個(gè) TopLevelCallbackBookKeeping, 而要通過(guò)這個(gè) PooledClass 來(lái)返回 TopLevelCallbackBookKeeping 的實(shí)例呢

單例模式是限制了一個(gè)類只能有一個(gè)實(shí)例，對(duì)象池模式則是限制一個(gè)類實(shí)例的個(gè)數(shù)。對(duì)象池類就像是一個(gè)對(duì)象管理員，它以Static列表（也就是裝對(duì)象的池子）的形式存存儲(chǔ)某個(gè)實(shí)例數(shù)受限的類的實(shí)例，每一個(gè)實(shí)例還要加一個(gè)標(biāo)記，標(biāo)記該實(shí)例是否被占用。當(dāng)類初始化的時(shí)候，這個(gè)對(duì)象池就被初始化了，實(shí)例就被創(chuàng)建出來(lái)。然后，用戶可以向這個(gè)類索取實(shí)例，如果池中所有的實(shí)例都已經(jīng)被占用了，那么拋出異常。用戶用完以后，還要把實(shí)例“還”回來(lái)，即釋放占用。對(duì)象池類的成員應(yīng)該都是靜態(tài)的。用戶也不應(yīng)該能訪問(wèn)池子里裝著的對(duì)象的構(gòu)造函數(shù)，以防用戶繞開(kāi)對(duì)象池創(chuàng)建實(shí)例。書(shū)上說(shuō)這個(gè)模式會(huì)用在數(shù)據(jù)庫(kù)連接的管理上。比如，每個(gè)用戶的連接數(shù)是有限的，這樣每個(gè)連接就是一個(gè)池子里的一個(gè)對(duì)象，“連接池”類就可以控制連接數(shù)了。
如果說(shuō)每次觸發(fā) dispatch 的時(shí)候都用 new TopLevelCallbackBookKeeping 來(lái) new 一個(gè)對(duì)象，那么當(dāng)觸發(fā)很多次 dispatch 的時(shí)候，就會(huì)導(dǎo)致生成多個(gè)對(duì)象無(wú)法銷毀（多個(gè)bookKeeping的引用次數(shù)一直為1），導(dǎo)致內(nèi)存溢出。

對(duì)象池技術(shù)基本原理的核心有兩點(diǎn)：緩存和共享，即對(duì)于那些被頻繁使用的對(duì)象，在使用完后，不立即將它們釋放，而是將它們緩存起來(lái)，以供后續(xù)的應(yīng)用程序重復(fù)使用，從而減少創(chuàng)建對(duì)象和釋放對(duì)象的次數(shù)，進(jìn)而改善應(yīng)用程序的性能。事實(shí)上，由于對(duì)象池技術(shù)將對(duì)象限制在一定的數(shù)量，也有效地減少了應(yīng)用程序內(nèi)存上的開(kāi)銷。

將用過(guò)的對(duì)象保存起來(lái)，等下一次需要這種對(duì)象的時(shí)候，再拿出來(lái)重復(fù)使用，從而在一定程度上減少頻繁創(chuàng)建對(duì)象所造成的開(kāi)銷。React 的 pooledClass.js 就是一個(gè)例子：

var invariant = require("invariant");

/**
 * Static poolers. Several custom versions for each potential number of
 * arguments. A completely generic pooler is easy to implement, but would
 * require accessing the `arguments` object. In each of these, `this` refers to
 * the Class itself, not an instance. If any others are needed, simply add them
 * here, or in their own files.
 */
var oneArgumentPooler = function(copyFieldsFrom) {
  var Klass = this;
  if (Klass.instancePool.length) {
    var instance = Klass.instancePool.pop();
    Klass.call(instance, copyFieldsFrom);
    return instance;
  } else {
    return new Klass(copyFieldsFrom);
  }
};

...
var standardReleaser = function(instance) {
  var Klass = this;
  invariant(
    instance instanceof Klass,
    "Trying to release an instance into a pool of a different type."
  );
  instance.destructor();
  if (Klass.instancePool.length < Klass.poolSize) {
    Klass.instancePool.push(instance);
  }
};

var DEFAULT_POOL_SIZE = 10;
var DEFAULT_POOLER = oneArgumentPooler;

/**
 * Augments `CopyConstructor` to be a poolable class, augmenting only the class
 * itself (statically) not adding any prototypical fields. Any CopyConstructor
 * you give this may have a `poolSize` property, and will look for a
 * prototypical `destructor` on instances (optional).
 *
 * @param {Function} CopyConstructor Constructor that can be used to reset.
 * @param {Function} pooler Customizable pooler.
 */
var addPoolingTo = function(CopyConstructor, pooler) {
  var NewKlass = CopyConstructor;
  NewKlass.instancePool = [];
  NewKlass.getPooled = pooler || DEFAULT_POOLER;
  if (!NewKlass.poolSize) {
    NewKlass.poolSize = DEFAULT_POOL_SIZE;
  }
  NewKlass.release = standardReleaser;
  return NewKlass;
};

var PooledClass = {
  addPoolingTo: addPoolingTo,
  oneArgumentPooler: oneArgumentPooler,
  twoArgumentPooler: twoArgumentPooler,
  threeArgumentPooler: threeArgumentPooler,
  fourArgumentPooler: fourArgumentPooler,
  fiveArgumentPooler: fiveArgumentPooler,
};

module.exports = PooledClass;

具體分為三步

addPoolingTo 添加對(duì)象到池子

調(diào)用的時(shí)候發(fā)現(xiàn)是否有緩存，有緩存就pop()出來(lái)用， 沒(méi)有緩存就新增一個(gè)

使用完成之后，釋放對(duì)象，緩存進(jìn)去

說(shuō)的再簡(jiǎn)單一點(diǎn)就是

創(chuàng)建對(duì)象 addPoolingTo()

借取對(duì)象 getPooled()

歸還對(duì)象 release()

并非所有對(duì)象都適合拿來(lái)池化――因?yàn)榫S護(hù)對(duì)象池也要造成一定開(kāi)銷。對(duì)生成時(shí)開(kāi)銷不大的對(duì)象進(jìn)行池化，反而可能會(huì)出現(xiàn)“維護(hù)對(duì)象池的開(kāi)銷”大于“生成新對(duì)象的開(kāi)銷”，從而使性能降低的情況。但是對(duì)于生成時(shí)開(kāi)銷可觀的對(duì)象，池化技術(shù)就是提高性能的有效策略了。