Hadoop集群上使用JNI,调用资源文件

时间:2021-10-07 21:19:32

hadoop是基于java的数据计算平台,引入第三方库,例如C语言实现的开发包将会大大增强数据分析的效率和能力。 通常在是用一些工具的时候都要用到一些配置文件、资源文件等。接下来,借一个例子来说明hadoop上面如何使用JNI、以及调用资源文件。

首先介绍一下ICTClass,ICTClass是中国科学院开发的一个分词软件(ICTClass官网)。该套软件采用C/C++编写。ICTClass虽然支持java,但是必须使用到的JNI技术。因此,在使用ICTClass之前需要配置好JNI资源以及ICTClass运行所需的一些资源文件。这造成了将ICTClass直接移植到Hadoop之上有点困难。不管怎样,我们先来探讨下怎个使用ICTClass的流程吧。

1. 封装Wrapper类

首先,要使用JNI技术创建ICTClass的Wrapper类ICTCLAS50.java:

package ICTCLAS.I3S.AC;
import java.io.*;
import java.net.URI;
import java.net.URL;
public class ICTCLAS50
{
public native boolean ICTCLAS_Init(byte[] sPath);
public native boolean ICTCLAS_Exit();
public native int ICTCLAS_ImportUserDictFile(byte[] sPath,int eCodeType);
public native int ICTCLAS_SaveTheUsrDic();
public native int ICTCLAS_SetPOSmap(int nPOSmap);
public native boolean ICTCLAS_FileProcess(byte[] sSrcFilename, int eCodeType, int bPOSTagged,byte[] sDestFilename);
public native byte[] ICTCLAS_ParagraphProcess(byte[] sSrc, int eCodeType, int bPOSTagged);
public native byte[] nativeProcAPara(byte[] sSrc, int eCodeType, int bPOStagged);
/* Use static intializer */
static
{

      String absolutePath = new File("").getAbsolutePath();
      System.load(absolutePath+"/ICTClassLib/libICTCLAS50.so");

    }
}

这里要注意,ICTCLAS50这个类必须在pacakage ICTCLAS.I3S.AC下否则将会出错。

2. 打包ICTCLAS以及资源文件

这里要打包两个jar文件。ICTClass.jar中打包了ICTCLAS50类作为即将被调用的API。ICTClassLib.jar打包了ICTCLAS所需的所有资源。

看一下两个jar的清单:

**********************************
ICTClass.jar
ICTClass
|_I3S
|_AC
|_ICTCLAS50.class ___________________________________
ICTClassLib.jar
ICTClassLib
|_libICTCLAS50.so
|_userdict.txt
|_user.lic
|_Configure.xml
|_Data
|_BiWord.big
|_character.idx
|_character.type
|_CoreDict.pdat
|_CoreDict.pos
|_CoreDict.unig
|_FieldDict.pdat
|_FieldDict.pos
|_GranDict.pdat
|_GranDict.pos
|_ICTCLAS30.ctx
|_ICTCLAS_First.map
|_ICTPOS.map
|_nr.ctx
|_nr.fsa
|_nr.role
|_PKU_First.map
|_PKU.map
|_temp
|_UserDict.map
|_UserDict.pdat
|_UserDict.pos

3. 在hadoop上使用JNI和资源

hadoop有很特殊的文件系统,这里笔者针对性介绍一下DistributeCache的机制。Hadoop可以将HDFS的一些文件分发到运行的某台机器的工作目录下,并按照一定的逻辑解压。通过以下API实现:

URI uri = null;
try {
   //这里filepath指的是hadoop文件系统目录
uri = new URI("hdfs://filepath/ICTClassLib.jar#ICTClassLib");
} catch (URISyntaxException e) {
logger.error(e.getMessage(),e);
}
if( uri != null){
DistributedCache.addCacheArchive(uri, jobConf);
}

上面的API将ICTClassLib.jar(当然该文件必须是在hadoop集群上的hdfs文件系统上)分发到tasknode上,并解压到工作目录的link目录下。ICTClassLib.jar包含ICTClass相关的资源文件。在tasknode上,每个task工作目录下的文件是:

Hadoop集群上使用JNI,调用资源文件

从上面可以看出,通过该机制hadoop能够把需要的文件下发到指定的task的工作目录下。

4. Hadoop程序源码(笔者是hadoop菜鸟,限于能力这里涉及到的资源访问使用的都是绝对路径。笔者试过相对路径但是总会抛出错误,无奈选择了这个折衷的办法)。

package com.ict.hadoop;

import java.io.File;
import java.io.IOException;
import java.net.URI;
import java.net.URISyntaxException;
import java.util.ArrayList;
import java.util.HashSet;
import java.util.Iterator;
import java.util.Set; import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.filecache.DistributedCache;
import org.apache.hadoop.fs.FileStatus;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapred.FileInputFormat;
import org.apache.hadoop.mapred.FileOutputFormat;
import org.apache.hadoop.mapred.FileSplit;
import org.apache.hadoop.mapred.InputSplit;
import org.apache.hadoop.mapred.JobClient;
import org.apache.hadoop.mapred.JobConf;
import org.apache.hadoop.mapred.MapReduceBase;
import org.apache.hadoop.mapred.Mapper;
import org.apache.hadoop.mapred.OutputCollector;
import org.apache.hadoop.mapred.Reducer;
import org.apache.hadoop.mapred.Reporter;
import org.apache.hadoop.mapred.TextOutputFormat;
import org.apache.hadoop.mapred.lib.MultipleOutputs;
import org.apache.hadoop.util.GenericOptionsParser;
import org.apache.log4j.Logger; public class WxSegment{
private static Logger logger = Logger.getLogger(WxSegment.class); public static class Map extends MapReduceBase implements Mapper<Text, Text, Text, Text>
{ public ICTCLAS50 testICTCLAS50;
public static Set<String> listAll(String dir) throws IOException {
//获取文件夹内所有文件的操作,可以忽略
} @Override
public void configure(JobConf jobConf){
testICTCLAS50 = new ICTCLAS50();
String absolutePath = new File("").getAbsolutePath()+"/ICTClassLib/";
if(testICTCLAS50.ICTCLAS_Init(absolutePath.getBytes("GB2312")) == false){
System.out.println("Init Fail!");
System.exit(1);
}
super.configure(jobConf); }
public String getSplitFileName(InputSplit inputSplit){
return ((FileSplit)inputSplit).getPath().getName();
} @Override
public void map(Text key, Text value,
OutputCollector<Text,Text> output, Reporter reporter)
throws IOException {
logger.info("AbsolutePath:" + new File("./ICTClassLib").getAbsolutePath()); String lineString = value.toString();
byte nativeBytes[] = testICTCLAS50.ICTCLAS_ParagraphProcess(
lineString.getBytes("GB2312"), 0, 0);
String nativeStr = new String(nativeBytes, 0,
nativeBytes.length, "GB2312");
// System.out.println("The result is :" + nativeStr);
String[] rest = nativeStr.split("\\s+");
for (int i = 0; i < rest.length; i++)
sb.append(rest[i] + " ");
lineString = sb.toString();
logger.info(lineString);
output.collect(key, new Text(lineString)); }
@Override
public void close() throws IOException {
//保存用户字典
testICTCLAS50.ICTCLAS_SaveTheUsrDic();
//释放分词组件资源
testICTCLAS50.ICTCLAS_Exit();
}
} public static class Reduce extends MapReduceBase implements Reducer<Text, Text, Text, NullWritable>{ private MultipleOutputs mos;
private OutputCollector<Text,NullWritable> collector; @Override
public void configure(JobConf conf) {
mos = new MultipleOutputs(conf);
} @Override
public void reduce(Text key, Iterator<Text> values,
OutputCollector<Text,NullWritable> output, Reporter reporter)
throws IOException { System.out.print("This is reduce");
collector = mos.getCollector(key.toString(), reporter);
output.collect( new Text(values.hasNext()?"has values":"no values"),NullWritable.get());
int count = 0;
while(values.hasNext()){
Text tmp = values.next();
collector.collect(tmp,NullWritable.get());
output.collect( tmp,NullWritable.get());
logger.info(tmp);
count ++;
} logger.info("The total count is "+ count); } @Override
public void close() throws IOException {
mos.close();
} }
public static Set<String> listAll(String dir) throws IOException { Configuration conf = new Configuration();
FileSystem fs = FileSystem.get(conf);
FileStatus[] stats = fs.listStatus(new Path(dir));
Set<String> list = new HashSet<String>();
for (int i = 0; i < stats.length; ++i)
{
if ( !stats[i].isDir() )
{
String filename = stats[i].getPath().getName();
filename = filename.replaceAll("[^a-zA-Z0-9]", "");
list.add(filename);
//System.out.println("File: "+ stats[i].getPath().getName());
logger.info("File: "+ filename);
}
else
logger.warn(stats[i].getPath().getName() +" is a directory."); }
fs.close();
return list;
} public static void main(String [] args) throws IOException, URISyntaxException{
Configuration conf = new Configuration();
String[] remainingArgs = new GenericOptionsParser(conf,args).getRemainingArgs();
// new GenericOptionsParser(conf, args).getRemainingArgs(); if (remainingArgs.length != 2) {
System.err.println("Error!");
System.exit(1);
} JobConf jobConf = new JobConf(conf,WxSegment.class); DistributedCache.createSymlink(jobConf);
logger.info("AbsolutePath:" + new File("").getAbsolutePath()); URI uri = null;
try {
uri = new URI("hdfs://filepath/ICTClassLib.jar#ICTClassLib");
} catch (URISyntaxException e) {
logger.error(e.getMessage(),e);
}
if( uri != null
)
logger.info(uri.getPath());
DistributedCache.addCacheArchive(uri, jobConf);
Path in = new Path(remainingArgs[0]);
Path out = new Path(remainingArgs[1]);
FileInputFormat.setInputPaths(jobConf, in);
FileOutputFormat.setOutputPath(jobConf, out); jobConf.setJobName("WxSegment");
jobConf.setMapperClass(Map.class);
jobConf.setReducerClass(Reduce.class);
jobConf.setMapOutputKeyClass(Text.class);
jobConf.setMapOutputValueClass(Text.class);
jobConf.setInputFormat(MsgSplitInputFormat.class); jobConf.setOutputKeyClass(Text.class);
jobConf.setOutputValueClass(NullWritable.class); Set<String> set = listAll(remainingArgs[0]);
for( String filename : set){
MultipleOutputs.addNamedOutput(jobConf,
filename,
TextOutputFormat.class,
Text.class,
NullWritable.class);
} jobConf.setJarByClass(WxSegment.class); JobClient.runJob(jobConf);
}
}

另外一篇值得参考的文章:《如何在Hadoop集群运行JNI程序》