java 读写Parquet格式的数据的示例代码

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:

maven依赖(我用的1.7)

<dependency>

<groupId>org.apache.parquet</groupId>

<artifactId>parquet-hadoop</artifactId>

<version>1.7.0</version>

</dependency>

以上是 java 读写Parquet格式的数据的示例代码 的全部内容, 来源链接: utcz.com/p/214521.html

回到顶部