Java API 读取Hive Orc文件

Java API 读取Hive Orc文件

转载：http://lxw1234.com/archives/2015/08/462.htm

Orc是Hive特有的一种列式存储的文件格式，它有着非常高的压缩比和读取效率，因此很快取代了之前的RCFile，成为Hive中非常常用的一种文件格式。

在实际业务场景中，可能需要使用Java API，或者MapReduce读写Orc文件。

本文先介绍使用Java API读取Hive Orc文件。

在Hive中已有一张Orc格式存储的表lxw1234:

该表有四个字段：url、word、freq、weight，类型均为string；

数据只有5条：

 

下面的代码，从表lxw1234对应的HDFS路径中使用API直接读取Orc文件：

 
 
package com.lxw1234.test;
 
import java.util.List;
import java.util.Properties;
 
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hive.ql.io.orc.OrcInputFormat;
import org.apache.hadoop.hive.ql.io.orc.OrcSerde;
import org.apache.hadoop.hive.serde2.objectinspector.StructField;
import org.apache.hadoop.hive.serde2.objectinspector.StructObjectInspector;
import org.apache.hadoop.mapred.FileInputFormat;
import org.apache.hadoop.mapred.InputFormat;
import org.apache.hadoop.mapred.InputSplit;
import org.apache.hadoop.mapred.JobConf;
import org.apache.hadoop.mapred.RecordReader;
import org.apache.hadoop.mapred.Reporter;
 
/**
 * lxw的大数据田地 -- http://lxw1234.com
 * @author lxw.com
 *
 */
public class TestOrcReader {
 
	public static void main(String[] args) throws Exception {		
		JobConf conf = new JobConf();
		Path testFilePath = new Path(args[0]);
		Properties p = new Properties();
		OrcSerde serde = new OrcSerde();
		p.setProperty("columns", "url,word,freq,weight");
		p.setProperty("columns.types", "string:string:string:string");
		serde.initialize(conf, p);
		StructObjectInspector inspector = (StructObjectInspector) serde.getObjectInspector();
		InputFormat in = new OrcInputFormat();
		FileInputFormat.setInputPaths(conf, testFilePath.toString());
		InputSplit[] splits = in.getSplits(conf, 1);
		System.out.println("splits.length==" + splits.length);
		
		conf.set("hive.io.file.readcolumn.ids", "1");
		RecordReader reader = in.getRecordReader(splits[0], conf, Reporter.NULL);
		Object key = reader.createKey();
		Object value = reader.createValue();
		List<? extends StructField> fields = inspector.getAllStructFieldRefs();		
		long offset = reader.getPos();
		while(reader.next(key, value)) {
			Object url = inspector.getStructFieldData(value, fields.get(0));
			Object word = inspector.getStructFieldData(value, fields.get(1));
			Object freq = inspector.getStructFieldData(value, fields.get(2));
			Object weight = inspector.getStructFieldData(value, fields.get(3));
			offset = reader.getPos();
			System.out.println(url + "|" + word + "|" + freq + "|" + weight);
		}
		reader.close();
		
	}
 
}
 

将上面的代码打包成orc.jar；

在Hadoop客户端机器上运行命令：

export HADOOP_CLASSPATH=/usr/local/apache-hive-0.13.1-bin/lib/hive-exec-0.13.1.jar:$HADOOP_CLASSPATH

hadoop jar orc.jar com.lxw1234.test.TestOrcReader /hivedata/warehouse/liuxiaowen.db/lxw1234/

运行结果如下：

 
 
[liuxiaowen@dev tmp]$ hadoop jar orc.jar com.lxw1234.test.TestOrcReader /hivedata/warehouse/liuxiaowen.db/lxw1234/
15/08/18 17:03:37 INFO log.PerfLogger: 
15/08/18 17:03:38 INFO Configuration.deprecation: mapred.input.dir is deprecated. Instead, use mapreduce.input.fileinputformat.inputdir
15/08/18 17:03:38 INFO orc.OrcInputFormat: FooterCacheHitRatio: 0/1
15/08/18 17:03:38 INFO log.PerfLogger: 
splits.length==1
15/08/18 17:03:38 INFO orc.ReaderImpl: Reading ORC rows from hdfs://cdh5/hivedata/warehouse/liuxiaowen.db/lxw1234/000000_0 with {include: null, offset: 0, length: 712}
 
http://cook.mytv365.com/v/20130403/1920802.html|未找到|1|5
 
 
http://cook.mytv365.com/v/20130403/1920802.html|网站|1|3
 
 
http://cook.mytv365.com/v/20130403/1920802.html|广大|1|3
 
 
http://cook.mytv365.com/v/20130403/1920802.html|直播|1|3
 
 
http://cook.mytv365.com/v/20130403/1920802.html|精彩|1|3
 

结果没有问题。