Hive自定义函数的使用

想要从日志数据中分析一下操作系统、浏览器、版本使用情况,但是hive中的函数不能直接解析useragent,于是可以写一个UDF来解析。useragent用于表示用户的当前操作系统,浏览器版本信息,形如:

Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/31.0.1650.63 Safari/537.36 180.173.196.29

其中解析ua可以用一个开源的工具包,叫做useragentutils.jar来处理,但是不能直接引入这个包,因为Hadoop和hive都不支持直接引用第三方的包,要导入源码。项目结构应该如下图

Hive自定义函数的使用

下面的代码用来打印出操作系统、浏览器版本信息:

import org.apache.hadoop.hive.ql.exec.UDF;
import org.apache.hadoop.io.Text;

import eu.bitwalker.useragentutils.UserAgent;

public class ParseUserAgent_UDF extends UDF{
 public Text evaluate(final Text userAgent){
  StringBuilder builder = new StringBuilder();
  UserAgent ua = new UserAgent(userAgent.toString());
  builder.append(ua.getOperatingSystem()+"\t"+ua.getBrowser()+"\t"+ua.getBrowserVersion());
  return new Text(builder.toString());
 }
}

使用:打成jar包,hive中add jar xx.jar;

create temporary function ua_parse as 'com.xx.ParseUserAgent_UDF';

select ua_parse(ua) from table_name limit 3;

结果:

WINDOWS_7      CHROME21        21.0.1180.89
WINDOWS_7      CHROME33        33.0.1750.146
WINDOWS_7      CHROME21        21.0.1180.89


此种方式只能处理一行,生成一行,无法进行统计分析。

下面使用UDTF(User Defined Table Generating Function),处理一行,生成多列。

import java.util.ArrayList;

import org.apache.hadoop.hive.ql.exec.UDFArgumentException;
import org.apache.hadoop.hive.ql.exec.UDFArgumentLengthException;
import org.apache.hadoop.hive.ql.metadata.HiveException;
import org.apache.hadoop.hive.ql.udf.generic.GenericUDTF;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorFactory;
import org.apache.hadoop.hive.serde2.objectinspector.StructObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.primitive.PrimitiveObjectInspectorFactory;

import eu.bitwalker.useragentutils.UserAgent;

public class ParseUserAgent_UDTF extends GenericUDTF{
 @Override
 public StructObjectInspector initialize(ObjectInspector[] args) throws UDFArgumentException {
  if (args.length != 1) {
   throw new UDFArgumentLengthException("ExplodeMap takes only one argument");
  }
  if (args[0].getCategory() != ObjectInspector.Category.PRIMITIVE) {
   throw new UDFArgumentException("ExplodeMap takes string as a parameter");
  }
  ArrayList<String> fieldNames = new ArrayList<String>();
  ArrayList<ObjectInspector> fieldOIs = new ArrayList<ObjectInspector>();
  fieldNames.add("system");
  fieldOIs.add(PrimitiveObjectInspectorFactory.javaStringObjectInspector);
  fieldNames.add("browser");
  fieldOIs.add(PrimitiveObjectInspectorFactory.javaStringObjectInspector);
  fieldNames.add("version");
  fieldOIs.add(PrimitiveObjectInspectorFactory.javaStringObjectInspector);
  return ObjectInspectorFactory.getStandardStructObjectInspector(fieldNames, fieldOIs);
 }
 @Override
 public void process(Object[] arg){
  try {
   if(arg == null || arg.length == 0)
    return;
   String input = arg[0].toString();
   String result[] = ua_parse(input).split("\t");
   forward(result);
  } catch (Exception e) {
   e.printStackTrace();
  }
 }
 
 @Override
 public void close() throws HiveException {
 
 }
 public String ua_parse(String userAgent){
  StringBuilder builder = new StringBuilder();
  UserAgent ua = new UserAgent(userAgent.toString());
  builder.append(ua.getOperatingSystem()+"\t"+ua.getBrowser()+"\t"+ua.getBrowserVersion());
  return builder.toString();
 }
}

select t.browser,count(*) c from (select ua_parse(ua) as (system,browser,version) from table_name) t group by t.browser order by c desc;

前十名:

CHROME31        987220571
UNKNOWN 708890045
IE8    420021677
IE7    411500373
MOBILE_SAFARI  291920740
IE6    217574865
IE11    179582201
IE9    165160040
CHROME30        158623163
CHROME21        155192489

未识别的还是很多!

基于Hadoop集群的Hive安装

Hive内表和外表的区别

Hadoop + Hive + Map +reduce 集群安装部署

Hive本地独立模式安装

Hive学习之WordCount单词统计

Hive运行架构及配置部署

内容版权声明:除非注明,否则皆为本站原创文章。

转载注明出处:https://www.heiqu.com/d41e9d73a6d63204e367b9b29e90cf8b.html