HBase学习笔记-聚合函数

利用HBase的coprocessor特性实现聚合函数，添加coprocessor方式有两种

1、修改hbase-site.xml，添加如下内容

<property>
    <name>hbase.coprocessor.region.classes</name>
    <value>org.apache.hadoop.hbase.coprocessor.AggregateImplementation</value>
</property> 

2、通过调用API，让Table注册该Coprocessor

String coprocessClassName = "org.apache.hadoop.hbase.coprocessor.AggregateImplementation";
 
HBaseAdmin admin = new HBaseAdmin(HBaseConfiguration.create());
 
admin.disableTable(tableName);
 
HTableDescriptor htd = admin.getTableDescriptor(tableName);
 
htd.addCoprocessor(coprocessClassName);
 
admin.modifyTable(tableName, htd);
 
admin.enableTable(tableName);</span>

相关聚合函数调用

/** 统计表行数*/
public static long rowCount(String tableName, String family) {
	AggregationClient ac = new AggregationClient(configuration);  
	Scan scan = new Scan();
	scan.addFamily(Bytes.toBytes(family));
	scan.setFilter(new FirstKeyOnlyFilter());
	long rowCount = 0;
	try {
		rowCount = ac.rowCount(Bytes.toBytes(tableName), new LongColumnInterpreter(), scan);
	} catch (Throwable e) {
		logger.info(e.getMessage(), e);
	}  
	return rowCount;
}
	
/** 求和*/
public static double sum(String tableName, String family, String qualifier) {
	AggregationClient ac = new AggregationClient(configuration);  
	Scan scan = new Scan();
	scan.addColumn(Bytes.toBytes(family), Bytes.toBytes(qualifier));
	double sum = 0;
	try {
		sum = ac.sum(Bytes.toBytes(tableName), new DoubleColumnInterpreter(), scan);
	} catch (Throwable e) {
		logger.info(e.getMessage(), e);
	}  
	return sum;
}
	
/** 求平均值*/
public static double avg(String tableName, String family, String qualifier) {
	AggregationClient ac = new AggregationClient(configuration);  
	Scan scan = new Scan();
	scan.addColumn(Bytes.toBytes(family), Bytes.toBytes(qualifier));
	double avg = 0;
	try {
		avg = ac.avg(Bytes.toBytes(tableName), new DoubleColumnInterpreter(), scan);
	} catch (Throwable e) {
		logger.info(e.getMessage(), e);
	}  
	return avg;
}

注意sum、avg时候里面的列解释器ColumnInterpreter是DoubleColumnInterpreter，而不是HBase里面自带的LongColumnInterpreter。DoubleColumnInterpreter是实现ColumnInterpreter接口的一个子类。

public class DoubleColumnInterpreter implements
		ColumnInterpreter<Double, Double> {

	@Override
	public void write(DataOutput out) throws IOException {
	}

	@Override
	public void readFields(DataInput in) throws IOException {
	}

	@Override
	public Double getValue(byte[] colFamily, byte[] colQualifier, KeyValue kv)
			throws IOException {
		if (kv == null)
			return null;
		return Double.valueOf(new String(kv.getValue()));
	}

	@Override
	public Double add(Double l1, Double l2) {
		if (l1 == null ^ l2 == null) {
			return l1 == null ? l2 : l1;
		} else if (l1 == null) {
			return null;
		}
		return l1 + l2;
	}

	@Override
	public Double getMaxValue() {
		return null;
	}

	@Override
	public Double getMinValue() {
		return null;
	}

	@Override
	public Double multiply(Double o1, Double o2) {
		if (o1 == null ^ o2 == null) {
			return o1 == null ? o2 : o1;
		} else if (o1 == null) {
			return null;
		}
		return o1 * o2;
	}

	@Override
	public Double increment(Double o) {
		return null;
	}

	@Override
	public Double castToReturnType(Double o) {
		return o.doubleValue();
	}

	@Override
	public int compare(Double l1, Double l2) {
		if (l1 == null ^ l2 == null) {
			return l1 == null ? -1 : 1; // either of one is null.
		} else if (l1 == null)
			return 0; // both are null
		return l1.compareTo(l2); // natural ordering.
	}

	@Override
	public double divideForAvg(Double o, Long l) {
		return (o == null || l == null) ? Double.NaN : (o.doubleValue() / l
				.doubleValue());
	}
}

将这个类打成jar包放到HBASE_HOME目录下lib目录下，然后重启hbase，遍可以调用相关聚合函数。