mapreduce实现全局排序
直接附代码,说明都在源码里了。
package com.hadoop.totalsort; import java.io.IOException; import java.util.ArrayList; 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.NullWritable; import org.apache.hadoop.io.SequenceFile; import org.apache.hadoop.io.Text; import org.apache.hadoop.mapred.FileInputFormat; import org.apache.hadoop.mapred.FileSplit; import org.apache.hadoop.mapred.InputSplit; import org.apache.hadoop.mapred.JobConf; import org.apache.hadoop.mapred.LineRecordReader; import org.apache.hadoop.mapred.RecordReader; import org.apache.hadoop.mapred.Reporter; import org.apache.hadoop.util.IndexedSortable; import org.apache.hadoop.util.QuickSort; public class SamplerInputFormat extends FileInputFormat<Text, Text> { static final String PARTITION_FILENAME = "_partition.lst"; static final String SAMPLE_SIZE = "terasort.partitions.sample"; private static JobConf lastConf = null; private static InputSplit[] lastResult = null; static class TextSampler implements IndexedSortable { public ArrayList<Text> records = new ArrayList<Text>(); public int compare(int arg0, int arg1) { Text right = records.get(arg0); Text left = records.get(arg1); return right.compareTo(left); } public void swap(int arg0, int arg1) { Text right = records.get(arg0); Text left = records.get(arg1); records.set(arg0, left); records.set(arg1, right); } public void addKey(Text key) { records.add(new Text(key)); } //将采集出来的key数据排序 public Text[] createPartitions(int numPartitions) { int numRecords = records.size(); if (numPartitions > numRecords) { throw new IllegalArgumentException("Requested more partitions than input keys (" + numPartitions + " > " + numRecords + ")"); } new QuickSort().sort(this, 0, records.size()); float stepSize = numRecords / (float) numPartitions; //采集的时候应该是采了100条记录,从10个分片查找的,此处再取numPartitions-1条 Text[] result = new Text[numPartitions - 1]; for (int i = 1; i < numPartitions; ++i) { result[i - 1] = records.get(Math.round(stepSize * i)); } return result; } } public static void writePartitionFile(JobConf conf, Path partFile) throws IOException { //前段代码从分片中采集数据,通过sampler.addKey存入TextSampler中的records数组 SamplerInputFormat inputFormat = new SamplerInputFormat(); TextSampler sampler = new TextSampler(); Text key = new Text(); Text value = new Text(); int partitions = conf.getNumReduceTasks(); // Reducer任务的个数 long sampleSize = conf.getLong(SAMPLE_SIZE, 100); // 采集数据-键值对的个数 InputSplit[] splits = inputFormat.getSplits(conf, conf.getNumMapTasks());// 获得数据分片 int samples = Math.min(10, splits.length);// 采集分片的个数 ,采集10个分片 long recordsPerSample = sampleSize / samples;// 每个分片采集的键值对个数 int sampleStep = splits.length / samples; // 采集分片的步长 ,总的分片个数/要采集的分片个数 long records = 0; for (int i = 0; i < samples; i++) { //1...10分片数 RecordReader<Text, Text> reader = inputFormat.getRecordReader(splits[sampleStep * i], conf, null); while (reader.next(key, value)) { sampler.addKey(key); //将key值增加到sampler的records数组 records += 1; if ((i + 1) * recordsPerSample <= records) { //目的是均匀采集各分片的条数,比如采集到第5个分片,那么记录条数应该小于5个分片应该的条数 break; } } } FileSystem outFs = partFile.getFileSystem(conf); if (outFs.exists(partFile)) { outFs.delete(partFile, false); } SequenceFile.Writer writer = SequenceFile.createWriter(outFs, conf, partFile, Text.class, NullWritable.class); NullWritable nullValue = NullWritable.get(); for (Text split : sampler.createPartitions(partitions)) { //调用createPartitions方法,排序采集出来的数据,并取partitions条 writer.append(split, nullValue); } writer.close(); } static class TeraRecordReader implements RecordReader<Text, Text> { private LineRecordReader in; private LongWritable junk = new LongWritable(); private Text line = new Text(); private static int KEY_LENGTH = 10; public TeraRecordReader(Configuration job, FileSplit split) throws IOException { in = new LineRecordReader(job, split); } public void close() throws IOException { in.close(); } public Text createKey() { // TODO Auto-generated method stub return new Text(); } public Text createValue() { return new Text(); } public long getPos() throws IOException { // TODO Auto-generated method stub return in.getPos(); } public float getProgress() throws IOException { // TODO Auto-generated method stub return in.getProgress(); } public boolean next(Text arg0, Text arg1) throws IOException { if (in.next(junk, line)) { //调用父类方法,将value值赋给key // if (line.getLength() < KEY_LENGTH) { arg0.set(line); arg1.clear(); // } else { // byte[] bytes = line.getBytes(); // 默认知道读取要比较值的前10个字节 作为key // // 后面的字节作为value; // arg0.set(bytes, 0, KEY_LENGTH); // arg1.set(bytes, KEY_LENGTH, line.getLength() - KEY_LENGTH); // } return true; } else { return false; } } } @Override public InputSplit[] getSplits(JobConf conf, int splits) throws IOException { if (conf == lastConf) { return lastResult; } lastConf = conf; lastResult = super.getSplits(lastConf, splits); return lastResult; } public org.apache.hadoop.mapred.RecordReader<Text, Text> getRecordReader(InputSplit arg0, JobConf arg1, Reporter arg2) throws IOException { return new TeraRecordReader(arg1, (FileSplit) arg0); } }
转载自:http://www.open-open.com/lib/view/open1381329062408.html
mapreduce实现全局排序的更多相关文章
- MapReduce TotalOrderPartitioner 全局排序
我们知道Mapreduce框架在feed数据给reducer之前会对map output key排序,这种排序机制保证了每一个reducer局部有序,hadoop 默认的partitioner是Has ...
- 三种方法实现Hadoop(MapReduce)全局排序(1)
我们可能会有些需求要求MapReduce的输出全局有序,这里说的有序是指Key全局有序.但是我们知道,MapReduce默认只是保证同一个分区内的Key是有序的,但是不保证全局有序.基于此,本文提供三 ...
- Mapreduce的排序(全局排序、分区加排序、Combiner优化)
一.MR排序的分类 1.部分排序:MR会根据自己输出记录的KV对数据进行排序,保证输出到每一个文件内存都是经过排序的: 2.全局排序: 3.辅助排序:再第一次排序后经过分区再排序一次: 4.二次排序: ...
- 大数据mapreduce全局排序top-N之python实现
a.txt.b.txt文件如下: a.txt hadoop hadoop hadoop hadoop hadoop hadoop hadoop hadoop hadoop hadoop hadoop ...
- MapReduce怎么优雅地实现全局排序
思考 想到全局排序,是否第一想到的是,从map端收集数据,shuffle到reduce来,设置一个reduce,再对reduce中的数据排序,显然这样和单机器并没有什么区别,要知道mapreduce框 ...
- Hadoop对文本文件的快速全局排序
一.背景 Hadoop中实现了用于全局排序的InputSampler类和TotalOrderPartitioner类,调用示例是org.apache.hadoop.examples.Sort. 但是当 ...
- 一起学Hadoop——TotalOrderPartitioner类实现全局排序
Hadoop排序,从大的范围来说有两种排序,一种是按照key排序,一种是按照value排序.如果按照value排序,只需在map函数中将key和value对调,然后在reduce函数中在对调回去.从小 ...
- MapReduce分区和排序
一.排序 排序: 需求:根据用户每月使用的流量按照使用的流量多少排序 接口-->WritableCompareable 排序操作在hadoop中属于默认的行为.默认按照字典殊勋排序. 排序的分类 ...
- Hadoop学习笔记—11.MapReduce中的排序和分组
一.写在之前的 1.1 回顾Map阶段四大步骤 首先,我们回顾一下在MapReduce中,排序和分组在哪里被执行: 从上图中可以清楚地看出,在Step1.4也就是第四步中,需要对不同分区中的数据进行排 ...
随机推荐
- oc 基础知识
一.枚举 结构体 typedef enum{ sexMan, sexWoman }Sex; tydedef struct{ int year; int month; int da ...
- Linux Terminal命令
Linux Terminal命令 1.Ctrl + a 回到命令行の「行首/head」. 2.Ctrl + e 回到命令行の「行尾/tail」, ctrl + end. 3.Ctrl + w 後向/b ...
- 306573704 Char型和string型字符串比较整理(转)
1.赋值 char赋值: char ch1[] = "give me"; char ch2[] = "a cup"; strcpy(ch1,ch2); cout ...
- free -m
free -m total used free shared buffers cached Mem: 7760 1572 6187 0 9 ...
- ajax、json一些整理(3)
写上面那些都是因为对ajax不熟悉 从w3c抄写JS原生ajax的东西补充一些基础 XMLHttpRequest 是 AJAX 的基础. XMLHttpRequest 对象 所有现代浏览器均支持 XM ...
- 配置 Struts2 Hello World
http://javaweb.group.iteye.com/group/wiki/1505-struts2-under-helloworld---how-to-make-the-first-of-t ...
- Angular ng-repeat
<tr ng-repeat="(key,item) in tableData"> <td class="check hidden-xs"> ...
- C++ dll调用-动态(显式)
C++ dll调用-动态(显式) 废话不说上代码, dll 头文件 j_test.h #pragma once extern "C"_declspec(dllexport) voi ...
- 纯JavaScript实现一些小功能
题目链接:http://wenku.baidu.com/link?url=7Gbarr5q9X6h1QFRVAsHmfPp1xXagG209mvrJqBogseb4WLeRqbVKwxQieoh8SL ...
- strstr函数与strcmp函数
1.strstr函数主要完成在一个字串中寻找另外一个字串 函数实现工程如下:摘自http://baike.baidu.com/link?url=RwrzOxs0w68j02J2uQs5u1A56bEN ...