java 读写Parquet格式的数据的示例代码
作者：Nucky_yang
本篇文章主要介绍了java 读写Parquet格式的数据的示例代码，小编觉得挺不错的，现在分享给大家，也给大家做个参考。一起跟随小编过来看看吧
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脚本之家。