Berkeley DB 使用SecondKey给数据排序

项目中用到了nosql数据库BDB，借此机会研究了一下它的用法。它的官方示例和文档比较丰富，感觉比较容易学习

在开发过程中出现了一个需求，要把数据根据插入时间遍历，个人认为通过第二主键（SecondKey）比较容易实现。以下是我的基本实现过程

1.在ValueBean中加入insertTime属性

public class ValueBean{
	
	private String insertTime;
	private String hostName;
	private byte[] value;
	

	public String getHostName() {
		return hostName;
	}
	public void setHostName(String hostName) {
		this.hostName = hostName;
	}
	public String getInsertTime() {
		return insertTime;
	}
	public void setInsertTime(String insertTime) {
		this.insertTime = insertTime;
	}
	public byte[] getValue() {
		return value;
	}
	public void setValue(byte[] value) {
		this.value = value;
	}


}

其中的hostName属性在主从同步和生成插入时间时用到，value属性就是key-value中的值

2.TupleBinding类

public class ValueBeanBinding extends TupleBinding<ValueBean> {

	@Override
	public ValueBean entryToObject(TupleInput input) {
        String time = input.readString();
        String name = input.readString();
        byte[] value = new byte[input.getBufferLength()-input.getBufferOffset()];//获得value长度
        input.read(value);
        
        ValueBean data = new ValueBean();
        data.setInsertTime(time);
        data.setHostName(name);
        data.setValue(value);
        
        return data;
	}

	@Override
	public void objectToEntry(ValueBean object, TupleOutput output) {
		ValueBean value = object;
		output.writeString(value.getInsertTime());
		output.writeString(value.getHostName());
		output.write(value.getValue());
	}
	
}

此类用于将ValueBean和DatabaseEntry进行转换，两个方法中的属性读写顺序要统一。

3.SecondaryKeyCreator，第二主键生成器

public class SecondKeyCreator implements SecondaryKeyCreator{
	private TupleBinding<ValueBean> theBinding;
	
	
	SecondKeyCreator(TupleBinding<ValueBean> theBinding) {
		this.theBinding = theBinding;
	}

	@Override
	public boolean createSecondaryKey(SecondaryDatabase secondary,
			DatabaseEntry key, DatabaseEntry data, DatabaseEntry result) {
		ValueBean v =
             (ValueBean) theBinding.entryToObject(data);
		String time=v.getInsertTime();

		result.setData(time.getBytes());

		return true;
	}
	
	


}

指定insertTime属性作为第二主键

在插入一个新数据时生成insertTime十分关键，尤其在高并发和互为主从同步时极易出现“第二主键重复”的错误，造成数据无法插入，我了使用当前时间毫秒数+AtomicInteger自增+hostName的asc码之和，保证insertTime的前后大小顺序。

System.currentTimeMillis()*1000000+(add_num.getAndIncrement()%1000)*1000 + host_key

4.创建第二数据库，用于存储secondkey

SecondaryConfig mySecConfig = new SecondaryConfig();
        mySecConfig.setAllowCreate(true);
        mySecConfig.setSortedDuplicates(false);
TupleBinding<ValueBean> tb =new ValueBeanBinding();
SecondKeyCreator  keyCreator = new  SecondKeyCreator(tb);
        mySecConfig.setKeyCreator(keyCreator);
        mySecConfig.setTransactional(envConfig.getTransactional());
        String secDbName = "mySecondaryDatabase";
        
        mySecDb = myEnv.openSecondaryDatabase(null, secDbName,
        		storeDb, mySecConfig);

到此，便可以使用SecondaryCursor的getNext()和getPrev()前后遍历了，getSearchKey()可以找到你想要的位置

=========================================================================================================

今天发了一个问题，cursor在未关闭之前会把所指的记录锁住。在修改被锁数据的时候会报错。

com.sleepycat.je.LockTimeoutException: (JE 4.1.10) Lock expired. Locker 25268096 7_New I/O server worker #1-2_Txn: waited for lock on database=testdb LockAddr:15430449 node=21 type=WRITE grant=WAIT_NEW timeoutMillis=1000 startTime=1323757856515 endTime=1323757857515
Owners: [<LockInfo locker="8385974 -1_New I/O server worker #1-1_ThreadLocker" type="READ"/>]
Waiters: []

所以在打开指针时要指定锁模式

问题代码：

mySecCursor = sdb.openCursor(null, null);

修改为：

CursorConfig config = new CursorConfig();
	    config.setReadUncommitted(true);
		mySecCursor = sdb.openCursor(null, config);