Hadoop MapReduce 学习笔记(十二) MapReduce实现类似SQL的order by/排序3 改进及改正

古城热线

　　本博客属创文章,转载请注明出处:http://guoyunsky.iyunv.com/blog/1235954
　　
　　本博客已迁移到本人独立博客:  http://www.yun5u.com/articles/hadoop-mapreduce-sql-order-by-sort-improve-fix.html
　　请先阅读:             
　　           1.Hadoop MapReduce 学习笔记(一) 序言和准备  
　　            2.Hadoop MapReduce 学习笔记(二) 序言和准备 2  
　　3.Hadoop MapReduce 学习笔记(八) MapReduce实现类似SQL的order by/排序
　　4.Hadoop MapReduce 学习笔记(九) MapReduce实现类似SQL的order by/排序 正确写法
　　5.Hadoop MapReduce 学习笔记(十) MapReduce实现类似SQL的order by/排序2 对多个字段排序
　　6.Hadoop MapReduce 学习笔记(十一) MapReduce实现类似SQL的order by/排序3 改进
　　          下一篇:
　　上一篇博客 Hadoop MapReduce 学习笔记(十一) MapReduce实现类似SQL的order by/排序3 改进 获得的结果并不是正确的结果,折腾了一小时没找到原因.于是参考hadoop/examples下面的SecondarySort.照搬里面的一些做法才纠正.这里先标记一下,待日后了解原理后再找出答案.

package com.guoyun.hadoop.mapreduce.study;

import java.io.IOException;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.RawComparator;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.io.WritableComparator;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.Mapper;
import org.apache.hadoop.mapreduce.Partitioner;
import org.apache.hadoop.mapreduce.Reducer;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
/**
* 通过MapReduce实现类似SELECT ＊ FROM TABLE ORDER BY COL1 ASC，COL2 DESC功能
* 也就是对多个字段的排序
* 相比 ＠OrderByMultiMapReduceTest,主要引入了Partitioner和GroupingComparator,提升性能
* 由于生成的数据frameworkName比较固定(具体请查看 @MyMapReduceMultiColumnTest 如何生成的数据)
* 所以这里获取map输出key的frameworkName属性,交给Partitioner和GroupingComparator来确定相同
* frameworkName的数据输出到相同的Reduce上,尽可能减少Reduce之前的清洗和排序工作,提升性能.
* 具体Partitioner和GroupingComparator的用法请查看Hadoop说明.
* 这里只是我目前对Partitioner和GroupingComparator的理解,刻意安排的输入数据.一切还需要验证中,待有机会
* 查看map和reduce源码后再来求证.
* 本类相比 @OrderByMultiMapReduceImproveTest 纠正了结果不正确的错误
*
* 注：
* 查看结果可以发现，其实这也是一个group by的实现
*/
public class OrderByMultiMapReduceImproveFixTest extends
OrderByMultiMapReduceTest {
public static final Logger log=LoggerFactory.getLogger(OrderByMultiMapReduceImproveFixTest.class);
public OrderByMultiMapReduceImproveFixTest(long dataLength, String inputPath,
String outputPath) throws Exception {
super(dataLength, inputPath, outputPath);
// TODO Auto-generated constructor stub
}
public OrderByMultiMapReduceImproveFixTest(long dataLength) throws Exception {
super(dataLength);
// TODO Auto-generated constructor stub
}
public OrderByMultiMapReduceImproveFixTest(String inputPath, String outputPath) {
super(inputPath, outputPath);
// TODO Auto-generated constructor stub
}
public OrderByMultiMapReduceImproveFixTest(String outputPath)
throws Exception {
super(outputPath);
// TODO Auto-generated constructor stub
}
/**
* 继承OrderMultiColumnWritable，新增WritableComparator，并注入到WritableComparator中
* 增加本类就可以解决OrderByMultiMapReduceImproveTest输出结果不一致的错误,具体原因还待探索
*/
public static class OrderMultiColumnFixWritable extends OrderMultiColumnWritable{
/**
* 增加这个WritableComparator就可以解决OrderByMultiMapReduceImproveTest
* 原理还不清楚,待探索
*/
public static class MyComparator extends WritableComparator {
public MyComparator() {
super(OrderMultiColumnWritable.class);
}
public int compare(byte[] b1, int s1, int l1,
byte[] b2, int s2, int l2) {
return compareBytes(b1, s1, l1, b2, s2, l2);
}
}
static {                                       
// register this comparator
WritableComparator.define(OrderMultiColumnWritable.class, new MyComparator());
}
}
/**
* map,get the source datas,and generate a (key,value) pair as (MultiWritable,NullWritable)
*/
public static class MyMapper extends Mapper<LongWritable,Text,OrderMultiColumnWritable,LongWritable>{
private OrderMultiColumnWritable writeKey=new OrderMultiColumnWritable();
private LongWritable writeValue=new LongWritable(0);
@Override
protected void map(LongWritable key, Text value, Context context)
throws IOException, InterruptedException {
log.debug("begin to map");
String[] splits=null;
try {
splits=value.toString().split("\\t");
if(splits!=null&&splits.length==2){
writeKey.set(splits[0],Long.parseLong(splits[1].trim()));
writeValue.set(writeKey.getNumber());
}
} catch (NumberFormatException e) {
log.error("map error:"+e.getMessage());
}
context.write(writeKey, writeValue);
}
}
/**
* reduce,only use to output the result
*/
public static class MyReducer
extends Reducer<OrderMultiColumnWritable,LongWritable,Text,LongWritable>{
private Text writeKey=new Text();
@Override
protected void reduce(OrderMultiColumnWritable key,
Iterable<LongWritable> values,Context context) throws IOException,
InterruptedException {
writeKey.set(key.getFrameworkName());
for(LongWritable value:values){
context.write(writeKey, value);
}
}
}
/**
* partitioner
*/
public static class MyPartitioner extends Partitioner<OrderMultiColumnWritable,LongWritable>{
@Override
public int getPartition(OrderMultiColumnWritable key, LongWritable value,
int numbers) {
return (int)Math.abs(key.getFrameworkName().hashCode()%numbers);
}
}

/**
* GroupingComparator
*/
public static class MyGroupingComparator implements RawComparator<OrderMultiColumnWritable>{
@Override
public int compare(OrderMultiColumnWritable o1,
OrderMultiColumnWritable o2) {
return o1.getFrameworkName().compareTo(o2.getFrameworkName());
}
@Override
public int compare(byte[] b1, int s1, int l1, byte[] b2, int s2,
int l2) {
return WritableComparator.compareBytes(b1,s1,l1,b2,s2,l2);
}
}
public static void main(String[] args){
MyMapReduceTest mapReduceTest=null;
Configuration conf=null;
Job job=null;
FileSystem fs=null;
Path inputPath=null;
Path outputPath=null;
long begin=0;
String input="testDatas/mapreduce/MRInput_Multi_OrderBy";
String output="testDatas/mapreduce/MROutput_Multi_OrderBy_Improve_Fix";

try {
// 直接使用MRInput_Single_OrderBy的输入数据,不重新生成数据,以便比对结果是否正确
// 和MROutput_Multi_OrderBy输出结果进行比对
mapReduceTest=new OrderByMultiMapReduceImproveFixTest(input,output);
inputPath=new Path(mapReduceTest.getInputPath());
outputPath=new Path(mapReduceTest.getOutputPath());
conf=new Configuration();
job=new Job(conf,"OrderBy");
fs=FileSystem.getLocal(conf);
if(fs.exists(outputPath)){
if(!fs.delete(outputPath,true)){
System.err.println("Delete output file:"+mapReduceTest.getOutputPath()+" failed!");
return;
}
}

job.setJarByClass(OrderByMultiMapReduceImproveFixTest.class);
job.setMapOutputKeyClass(OrderMultiColumnWritable.class);
job.setMapOutputValueClass(LongWritable.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(LongWritable.class);
job.setMapperClass(MyMapper.class);
job.setReducerClass(MyReducer.class);
job.setPartitionerClass(MyPartitioner.class);
job.setGroupingComparatorClass(MyGroupingComparator.class);
job.setNumReduceTasks(2);
FileInputFormat.addInputPath(job, inputPath);
FileOutputFormat.setOutputPath(job, outputPath);
begin=System.currentTimeMillis();
job.waitForCompletion(true);
System.out.println("===================================================");
if(mapReduceTest.isGenerateDatas()){
System.out.println("The maxValue is:"+mapReduceTest.getMaxValue());
System.out.println("The minValue is:"+mapReduceTest.getMinValue());
}
System.out.println("Spend time:"+(System.currentTimeMillis()-begin));
// Spend time:1270
} catch (Exception e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
}
}

 
 
更多文章、感悟、分享、勾搭，请用微信扫描: