AndroidX的内置对象池

我们经常会遇到使用对象复用的场景,比如线程池,数据库连接池,Android的消息机制里面的Message的通过静态方法obtain()获取消息,EventBus在的解析监听事件的订阅者对象里的标有@subscribe的方法而引入的SubscriberMethodFinder.FindState对象池FIND_STATE_POOL,Glide在请求加载图像的时候EngineJob内部的Pools.Pool> pool和解码DecodeJob内部的Pools.Pool> pool。因为对象创建的开销过大,为了避免每一次用到某个对象的时候都去new一个新的对象。对象池提供了这样一种机制,当我们需要某个对象的时候,我们希望先尝试从对象池(相当于缓存)中获取,如果有就直接返回,此时对象池应该空一个位子出来(即对象池的元素个数减一),没有则新创建一个对象返回,当对象用完以后,如果对象池还有空余位置则存放入其中同时对象池中元素的个数+1。
如何实现一个对象池呢?软件开发中经常说的一句话是面向接口而不是面向实现编程,因此将对象池抽象成一个接口。由于接口是功能的集合,对象池应该对外提供哪些功能呢?既然是对象池,那么它必然有一个获取对象和回收对象的方法。由于我们希望它是一个通用型的接口,对任意的类型都适用,任意类型都能实现获取和回收,因此需要用到泛型,我们用Pool来表示对象池,用acquire来表示获取对象,用recycle来表示回收对象,于是就有可以得到下面这个接口。

public interface Pool {
  //从对象池中获取对象
  T acquire();
  //将不需要的对象放回对象池,返回值boolen类型表示是否回收成功
  boolean recycle(T t);
}

有了上面的接口,我们就来尝试何实现一个简单的对象池。既然是对象池,必然容量是有限的,同时必须大于0,如何存放对象呢,最简单的方式是用数组。我们获取对象的时候就查询一下这个数组,看看其中有没有可用的元素,有就返回此元素同时将该处位原来的元素置空,不然会出现内存泄漏,没有就返回null。回收对象的时候先遍历一下数组是不是已经填满了,没有填满,就放入数组中,于是有下面的简单实现。

 public class SimplePool implements Pool {
        private final Object[] mElements;
        private final int mMaxPoolSize;
        public SimplePool(int maxPoolSize) {
            if (maxPoolSize <= 0) {
                throw new IllegalArgumentException("The max pool size must be > 0");
            }
            mMaxPoolSize = maxPoolSize;
            mElements = new Object[maxPoolSize];
        }

        @Override
        public T acquire() {
            for (int i = 0; i < mMaxPoolSize; i++) {
                if (mElements[i] != null) {
                    T ele = (T) mElements[i];
                    mElements[i] = null;//置空,避免内存泄漏
                    return ele;
                }
            }
           return null;
        }

        @Override
        public boolean recycle(T t) {
            for (int i = 0; i < mMaxPoolSize; i++) {
                if (mElements[i] == null) {
                    mElements[i] = t;
                    return true;
                }
            }
            return false;
        }
    }

有了这个简单的设计之后,就可以使用了,一般都是将这个类简单的设计在某个工具类中使用,我们不妨定义一个工具类叫PoolUtils,定义要复用的对象类型为HeavyObject,如下所示:

class HeavyObject {
    public HeavyObject (){
      //开销大的操作
     。。。。
    }
}

class PoolUtils {
  private static final Pool sPool = new SimplePool<>(10);
  //获取对象
  public static HeavyObject obtain() {
    HeavyObject obj = sPool.acquire();
    return obj != null ? obj : new HeavyObject(); //没有就创建新的
  }
  //回收对象
  public static boolean recycle(HeavyObject object) {
   return sPool.recycle(object);
  }
}

在回头看看我们实现的SimplePool设计得如何呢?emmm,我们发现aquirerecycle方法总会从从头到尾到数组的遍历。假如我们的缓存数组mElements的大小为n,开始时数组里面其实并没有元素,此时aquire方法的时间复杂度读为O(n)(并且还没有获取到元素),recycle方法的时间复杂度为O(1)(放在下标为0的位置),而当数组mElements中填满元素后,aquire方法的时间复杂度读为O(1)(返回下标为0处的元素),recycle方法的时间复杂度为O(n)(遍历完数组后仍然不能存放)。我们希望aquirerecycle方法无论在mElements是否有数据的情况下都能做到时间复杂度为O(1),那么如何做到这一点呢?
不难通过分析发现,aquirerecycle方法只要记住上次存取的位置就行了,我们将元素连续存放,比如上次是在下标为i处存放了元素(下标从0~i都有数据),那么aquire则从i处获取,recycle方法则在i+1处存放,只要i+1.实际上i+1即为缓存池里面已经存放的元素的个数,于是我们引入一个变量mPoolSize用于记录缓存池里面已经存放的元素个数。对SimplePool做一次修改。

    public class SimplePool implements Pool {
        private final Object[] mElements;
        private final int mMaxPoolSize;
        private int mPoolSize;

        public SimplePool(int maxPoolSize) {
            if (maxPoolSize <= 0) {
                throw new IllegalArgumentException("The max pool size must be > 0");
            }
            mMaxPoolSize = maxPoolSize;
            mElements = new Object[maxPoolSize];
        }

        @Override
        public T acquire() {
            if (mPoolSize > 0) {
                int lastPooledIndex = mPoolSize - 1;
                T t= (T) mElements[lastPooledIndex];
                mElements[lastPooledIndex] = null;//置空,避免内存泄漏
                mPoolSize--;
                return t;
            }
            return null;
        }

        @Override
        public boolean recycle(T t) {
            if (mPoolSize < mMaxPoolSize) {
                mElements[mPoolSize] = t;
                mPoolSize++;
                return true;
            }
            return false;
        }
    }

通过重构,现在的acquirerecycle方法只需要做一次判断,而不用从头开始遍历数组,因此时间复杂度都成了O(1).
看起来一切都很完美了,但是事情好像并没有那么简单,在多线程的情况下,比如某一时刻A线程正在执行aquire操作,B线程正在执行recycle操作很容易出现很容出现数据不一致的的安全性问题。看来我们得考虑一下线程安全的问题了。
为了能给用户提供一个分别线程安全和线程不安全的对象池,我们将线程安全的对象池的实现命名为SynchronizedPool,并给出实现

public class SynchronizedPool implements Pool {
        private final Object mLock = new Object();
        private final Object[] mElements;
        private final int mMaxPoolSize;
        private int mPoolSize;

        public SynchronizedPool(int maxPoolSize) {
            if (maxPoolSize <= 0) {
                throw new IllegalArgumentException("The max pool size must be > 0");
            }
            mMaxPoolSize = maxPoolSize;
            mElements = new Object[maxPoolSize];
        }

        @Override
        public T acquire() {
            synchronized (mLock) {
                if (mPoolSize > 0) {
                    final int lastPooledIndex = mPoolSize - 1;
                    T instance = (T) mElements[lastPooledIndex];
                    mElements[lastPooledIndex] = null;//置空,避免内存泄漏
                    mPoolSize--;
                    return instance;
                }
                return null;
            }
        }

        @Override
        public boolean recycle(T t) {
            synchronized (mLock) {
                if (mPoolSize < mMaxPoolSize) {
                    mElements[mPoolSize] = t;
                    mPoolSize++;
                    return true;
                }
                return false;
            }
        }
    }

于是我们能够提供了一个用户自己选择的对象池创建工厂方法,,让用户自定义是需要线程安全的还是线程不安全的版本:

class PoolFactory {
  public static  Pool createPool(boolean threadSafe, int maxPoolSize) {
            return threadSafe ? new SynchronizedPool(maxPoolSize) : new SimplePool(maxPoolSize);
    }
 }

看到到这里读者会问,这和标题有什么关系?实际上androidX已经提供了一个内置的对象复用池,如下所示

package androidx.core.util;

import androidx.annotation.NonNull;
import androidx.annotation.Nullable;

/**
 * Helper class for creating pools of objects. An example use looks like this:
 * 
 * public class MyPooledClass {
 *
 *     private static final SynchronizedPool sPool =
 *             new SynchronizedPool(10);
 *
 *     public static MyPooledClass obtain() {
 *         MyPooledClass instance = sPool.acquire();
 *         return (instance != null) ? instance : new MyPooledClass();
 *     }
 *
 *     public void recycle() {
 *          // Clear state if needed.
 *          sPool.release(this);
 *     }
 *
 *     . . .
 * }
 * 
* */ public final class Pools { /** * Interface for managing a pool of objects. * * @param The pooled type. */ public interface Pool { /** * @return An instance from the pool if such, null otherwise. */ @Nullable T acquire(); /** * Release an instance to the pool. * * @param instance The instance to release. * @return Whether the instance was put in the pool. * * @throws IllegalStateException If the instance is already in the pool. */ boolean release(@NonNull T instance); } private Pools() { /* do nothing - hiding constructor */ } /** * Simple (non-synchronized) pool of objects. * * @param The pooled type. */ public static class SimplePool implements Pool { private final Object[] mPool; private int mPoolSize; /** * Creates a new instance. * * @param maxPoolSize The max pool size. * * @throws IllegalArgumentException If the max pool size is less than zero. */ public SimplePool(int maxPoolSize) { if (maxPoolSize <= 0) { throw new IllegalArgumentException("The max pool size must be > 0"); } mPool = new Object[maxPoolSize]; } @Override @SuppressWarnings("unchecked") public T acquire() { if (mPoolSize > 0) { final int lastPooledIndex = mPoolSize - 1; T instance = (T) mPool[lastPooledIndex]; mPool[lastPooledIndex] = null;//置空,避免内存泄漏 mPoolSize--; return instance; } return null; } @Override public boolean release(@NonNull T instance) { if (isInPool(instance)) { throw new IllegalStateException("Already in the pool!"); } if (mPoolSize < mPool.length) { mPool[mPoolSize] = instance; mPoolSize++; return true; } return false; } private boolean isInPool(@NonNull T instance) { for (int i = 0; i < mPoolSize; i++) { if (mPool[i] == instance) { return true; } } return false; } } /** * Synchronized) pool of objects. * * @param The pooled type. */ public static class SynchronizedPool extends SimplePool { private final Object mLock = new Object(); /** * Creates a new instance. * * @param maxPoolSize The max pool size. * * @throws IllegalArgumentException If the max pool size is less than zero. */ public SynchronizedPool(int maxPoolSize) { super(maxPoolSize); } @Override public T acquire() { synchronized (mLock) { return super.acquire(); } } @Override public boolean release(@NonNull T element) { synchronized (mLock) { return super.release(element); } } } }

通过对比我们的设计和androidX的设计,发现androidX的SimplePool多了一个isInPool方法,也就是校验是否复用池里面已经有相同的对象了,如果存在就没有必要再放入一个元素,不然也就失去了复用的意义。同时SynchronizedPool的方法的实现是继承了SimplePool,只是在复写acquirerelease方法的时候加了一个锁,这样代码的复用率更高了。
如果读者有看过Glide源码的话,会发现Glide我们前面的PoolFactory上面更进了一步,它提供了将工厂模式和对象池复用的精妙操作,一起来学习一下。
定义一个工厂接口Factory

    /**
     * Creates new instances of the given type.
     * 创建指定类型的新的实例
     *
     * @param  The type of Object that will be created.
     */
    public interface Factory {
        T create();
    }

定义了一个对象被放回对象池时可能进行重置状态操作的接口Resetter

      /**
     * Resets state when objects are returned to the pool.
     * 当对象放回复用池时重置状态
     *
     * @param  The type of Object that will be reset.
     */
    public interface Resetter {
        void reset(@NonNull T object);
    }

定义了处理回收标记的抽象类StateVerifier

    public abstract class StateVerifier {

        /**
         * Throws an exception if we believe our object is recycled and inactive (i.e. is currently in an
         * object pool).
         * 如果我们确信我们的对象已经被回收并且不再处于活跃状态(例如当前在对象池)时抛出异常
         */
        public abstract void throwIfRecycled();

        /** 
         * Sets whether or not our object is recycled. 
         * 标记对象是否已经被回收
         * */
        abstract void setRecycled(boolean isRecycled);
        
    }

定义了需要可回收时的校验接口Poolable

   /**
     * Allows additional verification to catch errors caused by using objects while they are in an
     * object pool.
     * 允许额外的校验来捕获因使用在对象池的对象而产生的异常,
     */
    public interface Poolable {
        @NonNull
        StateVerifier getVerifier();
    }

将上面的整合一下

  private static final class FactoryPool implements Pool {
    private final Factory factory;
    private final Resetter resetter;
    private final Pool pool;

    FactoryPool(@NonNull Pool pool, @NonNull Factory factory, @NonNull Resetter resetter) {
      this.pool = pool;
      this.factory = factory;
      this.resetter = resetter;
    }

    @Override
    public T acquire() {
      T result = pool.acquire();//先从对象池中获取
      if (result == null) {
        result = factory.create();//没获取到则通过工厂方法来创建
        if (Log.isLoggable(TAG, Log.VERBOSE)) {
          Log.v(TAG, "Created new " + result.getClass());
        }
      }
      if (result instanceof Poolable) {
        ((Poolable) result).getVerifier().setRecycled(false /*isRecycled*/);//标记不在对象池中
      }
      return result;
    }

    @Override
    public boolean release(@NonNull T instance) {
      if (instance instanceof Poolable) {
        ((Poolable) instance).getVerifier().setRecycled(true /*isRecycled*/);//标记在复用池中
      }
      resetter.reset(instance);//重置状态的相关操作
      return pool.release(instance);//复用池复用
    }
  }

再对外提供工厂方法

package com.bumptech.glide.util.pool;

import android.util.Log;
import androidx.annotation.NonNull;
import androidx.core.util.Pools.Pool;
import androidx.core.util.Pools.SimplePool;
import androidx.core.util.Pools.SynchronizedPool;
import java.util.ArrayList;
import java.util.List;
/**
 * Provides implementations of {@link Pool} never return {@code null}, log when new instances are
 * created, and that can use the {@link com.bumptech.glide.util.pool.FactoryPools.Poolable}
 * interface to ensure objects aren't used while inside the pool.
 */
public final class FactoryPools {
  private static final String TAG = "FactoryPools";
  private static final int DEFAULT_POOL_SIZE = 20;
  private static final Resetter EMPTY_RESETTER =
      new Resetter() {
        @Override
        public void reset(@NonNull Object object) {
          // Do nothing.
        }
      };

  private FactoryPools() {}

  /**
   * Returns a non-thread safe {@link Pool} that never returns {@code null} from {@link
   * Pool#acquire()} and that contains objects of the type created by the given {@link Factory} with
   * the given maximum size.
   *
   * 

If the pool is empty when {@link Pool#acquire()} is called, the given {@link Factory} will * be used to create a new instance. * 线程不安全 * @param The type of object the pool will contains. */ @NonNull public static Pool simple(int size, @NonNull Factory factory) { return build(new SimplePool(size), factory); } /** * Returns a new thread safe {@link Pool} that never returns {@code null} from {@link * Pool#acquire()} and that contains objects of the type created by the given {@link Factory} with * the given maximum size. * *

If the pool is empty when {@link Pool#acquire()} is called, the given {@link Factory} will * be used to create a new instance. *线程安全 * @param The type of object the pool will contains. */ @NonNull public static Pool threadSafe(int size, @NonNull Factory factory) { return build(new SynchronizedPool(size), factory); } /** * Returns a new {@link Pool} that never returns {@code null} and that contains {@link List Lists} * of a specific generic type with a standard maximum size of 20. * *

If the pool is empty when {@link Pool#acquire()} is called, a new {@link List} will be * created. * * @param The type of object that the {@link List Lists} will contain. */ @NonNull public static Pool> threadSafeList() { return threadSafeList(DEFAULT_POOL_SIZE); } /** * Returns a new thread safe {@link Pool} that never returns {@code null} and that contains {@link * List Lists} of a specific generic type with the given maximum size. * *

If the pool is empty when {@link Pool#acquire()} is called, a new {@link List} will be * created. * * @param The type of object that the {@link List Lists} will contain. */ // Public API. @SuppressWarnings("WeakerAccess") @NonNull public static Pool> threadSafeList(int size) { return build( new SynchronizedPool>(size), new Factory>() { @NonNull @Override public List create() { return new ArrayList<>(); } }, new Resetter>() { @Override public void reset(@NonNull List object) { object.clear(); } }); } @NonNull private static Pool build( @NonNull Pool pool, @NonNull Factory factory) { return build(pool, factory, FactoryPools.emptyResetter()); } @NonNull private static Pool build( @NonNull Pool pool, @NonNull Factory factory, @NonNull Resetter resetter) { return new FactoryPool<>(pool, factory, resetter); } }

相信通过上面这一些列的流程,我们对对象池的设计和实现有了一个大体的了解,同时对设计模式,算法的复杂度,甚至后面的架构设计积累了灵感和素材,值得反复揣摩。

你可能感兴趣的:(AndroidX的内置对象池)