根据上回对PooledConnection的分析,下面是对一个可重用PooledConnection的实现:
public class PooledConnection implements Connection
{
public interface Pool
{
//引入这个interface, 是因为我们的PooledConnection只需要知道如何返还Connection。
//本着接口最小化原则,我们只定义我们需要的操作。
void closeConnection(Connection conn);
}
private interface ConnectionState
{
//state pattern的interface.
public ConnectionState close() throws SQLException;
//close()方法是唯一引起状态转移的方法。
boolean isClosed();
Connection getOpenConnection() throws SQLException;
}
private static class ClosedConnection implements ConnectionState
{
public final ConnectionState close()
{
return this;
}
//当一个Connection已经closed了的时候,它实际上已经死了。所有对它的操作,除了isClosed()和close(),
只产生异常。所以,一个closed的Connection,
它已经不需要保存那个物理数据库连接和对出身ConnectionPool的连接。而且因为所有的
closed connection的状态都一样,所以可以用singleton来节省内存。
public final Connection getOpenConnection() throws SQLException
{
throw new SQLException("Connection closed");
}
public final boolean isClosed(){return true;}
private ClosedConnection(){}
private static final ConnectionState _instance = new
ClosedConnection();
static ConnectionState instance(Connection conn, Pool pool){return
_instance;}
}
private static class OpenConnection implements ConnectionState
{
private final Pool pool;
private final Connection conn;
public final ConnectionState close()
{
//对一个open connection的关闭,会把原始数据库连接返还到connection
pool. 同时,该连接死亡。
pool.closeConnection(conn);
return ClosedConnection.instance(conn, pool);
}
public final Connection getOpenConnection()
{
return conn;
}
public final boolean isClosed()
{
return false;
}
OpenConnection(Connection conn, Pool pool)
{
this.conn = conn; this.pool = pool;
}
static ConnectionState instance(Connection conn, Pool pool)
{
return new OpenConnection(conn, pool);
}
}
private ConnectionState state;
//用静态的工厂方法,可以隐藏我们的实现类,以后,根据需要,我们可以方便地修改实现类,比如用内部类取代。
//根据要修饰的Connection的状态,初始化PooledConnection
public static Connection decorate(Connection conn, Pool pool) throws
SQLException
{
if(conn.isClosed())
{
return new PooledConnection(ClosedConnection.instance(conn, pool));
}
else
{
return new PooledConnection(OpenConnection.instance(conn, pool));
}
}
private PooledConnection(ConnectionState state)
{
this.state = state;
}
public final boolean isClosed()
{
return state.isClosed();
}
public final void close() throws SQLException
{
state = state.close();
}
private final Connection getOpenConnection() throws SQLException
{
return state.getOpenConnection();
}
/*****然后,做委托****/
public final Statement createStatement() throws SQLException
{
return getOpenConnection().createStatement();
}
public final void rollback()throws SQLException
{
getOpenConnection().rollback();
}
//等等等等
}
好,再来看看ConnectionPoolImpl怎样使用PooledConnection.
public class ConnectionPoolImpl implements ConnectionPool
{
public synchronized Connection getConnection()
{
Connection ret;
如果pool里有Connection
从pool中去掉一个Connection conn;
clients++;
ret = conn;
否则,如果clients<maxClients
生成一个新的连接conn
clients++;
ret = conn;
否则,wait(),直到pool中有空闲Connection
//下面的这个匿名类实际上是个adapter pattern. J
return PooledConnection.decorate(ret, new PooledConnection.Pool
{
public void closeConnection(Connection conn)
{
ConnectionPoolImpl.this.closeConnection(conn);
}
}
}
//其他都和原来一样
}
这样,所有对ConnectionPool的实现,都可以在返回一个物理Connection之前,把它用PooledConnection封装一下。如此,代码得到了重用。ConnectionPool的实现者可以把主要精力放在怎样处理池的各种功能。而不是怎样包装Connection.
世界真美好!
不过。。。。。
万一,李四忘了用PooledConnection包装他的Connection怎么办?编译器不会报错,因为反正都是Connection类型。
“你也太杞人忧天了吧?他忘了揍他不就得了?”哎,保不齐呀!人不是机器,总有犯错的时候,到时候揍他有啥用?还手疼呢。
同学们,今天的家庭作业是:想办法让李四的健忘症不会影响我们的ConnectionPool大业。
上一页
下一页
|
|