欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页  >  IT编程

java 读写Parquet格式的数据的示例代码

程序员文章站 2024-03-02 13:01:46
本文介绍了java 读写parquet格式的数据,分享给大家,具体如下: import java.io.bufferedreader; import jav...

本文介绍了java 读写parquet格式的数据,分享给大家,具体如下:

import java.io.bufferedreader;
import java.io.file;
import java.io.filereader;
import java.io.ioexception;
import java.util.random;

import org.apache.hadoop.conf.configuration;
import org.apache.hadoop.fs.path;
import org.apache.log4j.logger;
import org.apache.parquet.example.data.group;
import org.apache.parquet.example.data.groupfactory;
import org.apache.parquet.example.data.simple.simplegroupfactory;
import org.apache.parquet.hadoop.parquetreader;
import org.apache.parquet.hadoop.parquetreader.builder;
import org.apache.parquet.hadoop.parquetwriter;
import org.apache.parquet.hadoop.example.groupreadsupport;
import org.apache.parquet.hadoop.example.groupwritesupport;
import org.apache.parquet.schema.messagetype;
import org.apache.parquet.schema.messagetypeparser;

public class readparquet {
  static logger logger=logger.getlogger(readparquet.class);
  public static void main(string[] args) throws exception {
    
//    parquetwriter("test\\parquet-out2","input.txt");
    parquetreaderv2("test\\parquet-out2");
  }
  
  
  static void parquetreaderv2(string inpath) throws exception{
    groupreadsupport readsupport = new groupreadsupport();
    builder<group> reader= parquetreader.builder(readsupport, new path(inpath));
    parquetreader<group> build=reader.build();
    group line=null;
    while((line=build.read())!=null){
      group time= line.getgroup("time", 0);
        //通过下标和字段名称都可以获取
        /*system.out.println(line.getstring(0, 0)+"\t"+
        line.getstring(1, 0)+"\t"+
        time.getinteger(0, 0)+"\t"+
        time.getstring(1, 0)+"\t");*/
        system.out.println(line.getstring("city", 0)+"\t"+
        line.getstring("ip", 0)+"\t"+
        time.getinteger("ttl", 0)+"\t"+
        time.getstring("ttl2", 0)+"\t");
        //system.out.println(line.tostring());
    }
    system.out.println("读取结束");
  } 
  //新版本中new parquetreader()所有构造方法好像都弃用了,用上面的builder去构造对象
  static void parquetreader(string inpath) throws exception{
    groupreadsupport readsupport = new groupreadsupport();
    parquetreader<group> reader = new parquetreader<group>(new path(inpath),readsupport);
    group line=null;
    while((line=reader.read())!=null){
     system.out.println(line.tostring());
    }
    system.out.println("读取结束");
    
  }
  /**
   * 
   * @param outpath  输出parquet格式
   * @param inpath 输入普通文本文件
   * @throws ioexception
   */
  static void parquetwriter(string outpath,string inpath) throws ioexception{
    messagetype schema = messagetypeparser.parsemessagetype("message pair {\n" +
        " required binary city (utf8);\n" +
        " required binary ip (utf8);\n" +
        " repeated group time {\n"+
          " required int32 ttl;\n"+
           " required binary ttl2;\n"+
        "}\n"+
       "}");
    groupfactory factory = new simplegroupfactory(schema);
    path path = new path(outpath);
    configuration configuration = new configuration();
    groupwritesupport writesupport = new groupwritesupport();
    writesupport.setschema(schema,configuration);
    parquetwriter<group> writer = new parquetwriter<group>(path,configuration,writesupport);
    //把本地文件读取进去,用来生成parquet格式文件
    bufferedreader br =new bufferedreader(new filereader(new file(inpath)));
    string line="";
    random r=new random();
    while((line=br.readline())!=null){
      string[] strs=line.split("\\s+");
      if(strs.length==2) {
        group group = factory.newgroup()
            .append("city",strs[0])
            .append("ip",strs[1]);
        group tmpg =group.addgroup("time");
        tmpg.append("ttl", r.nextint(9)+1);
        tmpg.append("ttl2", r.nextint(9)+"_a");
        writer.write(group);
      }
    }
    system.out.println("write end");
    writer.close();
  }
}

说下schema(写parquet格式数据需要schema,读取的话"自动识别"了schema)

/*
 * 每一个字段有三个属性:重复数、数据类型和字段名,重复数可以是以下三种:
 *     required(出现1次)
 *     repeated(出现0次或多次) 
 *     optional(出现0次或1次)
 * 每一个字段的数据类型可以分成两种:
 *     group(复杂类型)
 *     primitive(基本类型)
 * 数据类型有
 * int64, int32, boolean, binary, float, double, int96, fixed_len_byte_array
 */

这个repeated和required 不光是次数上的区别,序列化后生成的数据类型也不同,比如repeqted修饰 ttl2 打印出来为 wrappedarray([7,7_a]) 而 required修饰 ttl2 打印出来为 [7,7_a]  除了用messagetypeparser.parsemessagetype类生成messagetype 还可以用下面方法

(注意这里有个坑--spark里会有这个问题--ttl2这里 as(originaltype.utf8) 和 required binary city (utf8)作用一样,加上utf8,在读取的时候可以转为stringtype,不加的话会报错 [b cannot be cast to java.lang.string  )

/*messagetype schema = messagetypeparser.parsemessagetype("message pair {\n" +
        " required binary city (utf8);\n" +
        " required binary ip (utf8);\n" +
        "repeated group time {\n"+
        "required int32 ttl;\n"+
        "required binary ttl2;\n"+
        "}\n"+
        "}");*/
    
//import org.apache.parquet.schema.types;
messagetype schema = types.buildmessage() 
      .required(primitivetypename.binary).as(originaltype.utf8).named("city") 
      .required(primitivetypename.binary).as(originaltype.utf8).named("ip") 
      .repeatedgroup().required(primitivetypename.int32).named("ttl")
              .required(primitivetypename.binary).as(originaltype.utf8).named("ttl2")
              .named("time")
     .named("pair"); 

解决 [b cannot be cast to java.lang.string 异常:

1.要么生成parquet文件的时候加个utf8
2.要么读取的时候再提供一个同样的schema类指定该字段类型,比如下面:

java 读写Parquet格式的数据的示例代码

maven依赖(我用的1.7)

<dependency>
  <groupid>org.apache.parquet</groupid>
  <artifactid>parquet-hadoop</artifactid>
  <version>1.7.0</version>
</dependency>

以上就是本文的全部内容,希望对大家的学习有所帮助,也希望大家多多支持。