Hikari定制连接池容器ConcurrentBag

它是为了实现比 LinkedBlockingQueue 和 LinkedTransferQueue 更高的性能而特别定制的，根据连接池的特殊场景做了一些性能优化

容器类定义

public class ConcurrentBag implements AutoCloseable {
   // 底层存储 CopyOnWriteArrayList，可以无锁安全的读取连接池的信息
   private final CopyOnWriteArrayList sharedList;
   private final boolean weakThreadLocals; //默认是false 
     // 线程隔离的 FastList
   private final ThreadLocal> threadList;
     // 其实就是HikariPool，用来新建PoolEntry
   private final IBagStateListener listener;
     // 等待获取db连接的线程个数
   private final AtomicInteger waiters;
   private volatile boolean closed;

   private final SynchronousQueue handoffQueue;
  //..
}

容器存储对象(条目)的定义

// 容器存放的的 IConcurrentBagEntry 接口实现类 PoolEntry 
public interface IConcurrentBagEntry
   {  
      // 空闲状态
      int STATE_NOT_IN_USE = 0;
      // 连接在使用
      int STATE_IN_USE = 1;
      // remove的时候先标记为被移除
      int STATE_REMOVED = -1;
      // 被预留状态，不可用但是可以移除 
      // 主要在检查线程中，要对连接进行softEvictConnection，确保能从not-in-use转到 reserved
      int STATE_RESERVED = -2;

      boolean compareAndSet(int expectState, int newState);
      void setState(int newState);
      int getState();
   }

它的几种状态扭转

image.png

连接池容器初始化

public ConcurrentBag(final IBagStateListener listener)
{
     // 其实就是HikariPool, 负责创建PoolEntry
   this.listener = listener;
     // false
   this.weakThreadLocals = useWeakThreadLocals();
   this.handoffQueue = new SynchronousQueue<>(true);
   // 等待获取db连接的线程个数
   this.waiters = new **AtomicInteger**();
   //底层存储 CopyOnWriteArrayList
   this.sharedList = new **CopyOnWriteArrayList**<>();
   if (weakThreadLocals) {
      this.threadList = ThreadLocal.withInitial(() -> new ArrayList<>(16));
   }
   else {
            // ThreadLocal, 使用自定义的FastList
      this.threadList = ThreadLocal.withInitial(() -> new **FastList**<>(IConcurrentBagEntry.class, 16));
   }
}

容器操作 crud

1. 从容器获取对象 borrow(long timeout, TimeUnit timeunit)

public T borrow(long timeout, final TimeUnit timeUnit) throws InterruptedException
{
   // Try the thread-local list first
   // **1.** 先从 threadlocal的FastList中取，倒序遍历, 同一个线程归还连接后再获取最新的连接 可避免一些检查
   // 如果能取到就直接返回 这样无需锁
   final List