Hadoop支持的文件格式之Parquet

0x00 文章內容

1. 行存儲與列存儲
2. 編碼實現Parquet格式的讀寫
3. 彩蛋

0x01 行存儲與列存儲

1. Avro與Parquet

a. 請參考文章：Hadoop支持的文件格式之Avro的0x01 行存儲與列存儲

0x02 編碼實現Parquet格式的讀寫

1. 編碼實現讀寫Parquet文件

a. 引入Parquet相關jar包

    <!--添加Parquet依賴-->
    <dependency>
        <groupId>org.apache.parquet</groupId>
        <artifactId>parquet-column</artifactId>
        <version>1.8.1</version>
    </dependency>
    <dependency>
        <groupId>org.apache.parquet</groupId>
        <artifactId>parquet-hadoop</artifactId>
        <version>1.8.1</version>
    </dependency>

b. 完整的寫Parquet文件代碼（寫到HDFS）

package com.shaonaiyi.hadoop.filetype.parquet;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.parquet.column.ParquetProperties;
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.ParquetWriter;
import org.apache.parquet.hadoop.example.GroupWriteSupport;
import org.apache.parquet.hadoop.metadata.CompressionCodecName;
import org.apache.parquet.schema.MessageType;
import org.apache.parquet.schema.MessageTypeParser;

import java.io.IOException;

/**
 * @Author [email protected]
 * @Date 2019/12/18 10:14
 * @Description 編碼實現寫Parquet文件
 */
public class ParquetFileWriter {

    public static void main(String[] args) throws IOException {
        MessageType schema = MessageTypeParser.parseMessageType("message Person {\n" +
                "    required binary name;\n" +
                "    required int32 age;\n" +
                "    required int32 favorite_number;\n" +
                "    required binary favorite_color;\n" +
                "}");

        Configuration configuration = new Configuration();
        Path path = new Path("hdfs://master:9999/user/hadoop-sny/mr/filetype/parquet/data.parquet");
        GroupWriteSupport writeSupport = new GroupWriteSupport();
        GroupWriteSupport.setSchema(schema, configuration);
        ParquetWriter<Group> writer = new ParquetWriter<Group>(path, writeSupport,
                CompressionCodecName.SNAPPY,
                ParquetWriter.DEFAULT_BLOCK_SIZE,
                ParquetWriter.DEFAULT_PAGE_SIZE,
                ParquetWriter.DEFAULT_PAGE_SIZE,
                ParquetWriter.DEFAULT_IS_DICTIONARY_ENABLED,
                ParquetWriter.DEFAULT_IS_VALIDATING_ENABLED,
                ParquetProperties.WriterVersion.PARQUET_1_0, configuration);

        GroupFactory groupFactory = new SimpleGroupFactory(schema);
        Group group = groupFactory.newGroup()
                .append("name", "shaonaiyi")
                .append("age", 18)
                .append("favorite_number", 7)
                .append("favorite_color", "red");

        writer.write(group);

        writer.close();
    }

}

c. 完整的讀Parquet文件代碼（從HDFS讀）

package com.shaonaiyi.hadoop.filetype.parquet;

import org.apache.hadoop.fs.Path;
import org.apache.parquet.example.data.Group;
import org.apache.parquet.hadoop.ParquetReader;
import org.apache.parquet.hadoop.example.GroupReadSupport;

import java.io.IOException;

/**
 * @Author [email protected]
 * @Date 2019/12/18 10:18
 * @Description 編碼實現讀Parquet文件
 */
public class ParquetFileReader {

    public static void main(String[] args) throws IOException {


        Path path = new Path("hdfs://master:9999/user/hadoop-sny/mr/filetype/parquet/parquet-data.parquet");
        GroupReadSupport readSupport = new GroupReadSupport();
        ParquetReader<Group> reader = new ParquetReader<>(path, readSupport);

        Group result = reader.read();
        System.out.println("name：" + result.getString("name", 0).toString());
        System.out.println("age：" + result.getInteger("age", 0));
        System.out.println("favorite_number：" + result.getInteger("favorite_number", 0));
        System.out.println("favorite_color：" + result.getString("favorite_color", 0));
    }

}

2. 查看讀寫Parquet文件結果

a. 寫Parquet文件

b. 讀Parquet文件

3. 編碼實現讀寫Parquet文件（HDFS）

a. 引入Parquet與Avro關聯的jar包

    <dependency>
        <groupId>org.apache.parquet</groupId>
        <artifactId>parquet-avro</artifactId>
        <version>1.8.1</version>
    </dependency>

從上面的代碼我們可以看出，以下面這種方式定義Schema很不友好：

    MessageType schema = MessageTypeParser.parseMessageType("message Person {\n" +
            "    required binary name;\n" +
            "    required int32 age;\n" +
            "    required int32 favorite_number;\n" +
            "    required binary favorite_color;\n" +
            "}");

所以我們可以將Parquet與Avro關聯，直接使用Avro的Schema即可。

b. 完整的寫Parquet文件代碼（HDFS）

package com.shaonaiyi.hadoop.filetype.parquet;

import com.shaonaiyi.hadoop.filetype.avro.Person;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.mapreduce.*;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
import org.apache.hadoop.mapreduce.task.JobContextImpl;
import org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl;
import org.apache.parquet.avro.AvroParquetOutputFormat;

import java.io.IOException;

/**
 * @Author [email protected]
 * @Date 2019/12/18 10:47
 * @Description 編碼實現寫Parquet文件（HDFS）
 */
public class MRAvroParquetFileWriter {

    public static void main(String[] args) throws IOException, IllegalAccessException, InstantiationException, ClassNotFoundException, InterruptedException {
        //1 構建一個job實例
        Configuration hadoopConf = new Configuration();
        Job job = Job.getInstance(hadoopConf);

        //2 設置job的相關屬性
        job.setOutputKeyClass(NullWritable.class);
        job.setOutputValueClass(Person.class);
        job.setOutputFormatClass(AvroParquetOutputFormat.class);
        //AvroJob.setOutputKeySchema(job, Schema.create(Schema.Type.INT));
        AvroParquetOutputFormat.setSchema(job, Person.SCHEMA$);


        //3 設置輸出路徑
        FileOutputFormat.setOutputPath(job, new Path("hdfs://master:9999/user/hadoop-sny/mr/filetype/avro-parquet"));

        //4 構建JobContext
        JobID jobID = new JobID("jobId", 123);
        JobContext jobContext = new JobContextImpl(job.getConfiguration(), jobID);

        //5 構建taskContext
        TaskAttemptID attemptId = new TaskAttemptID("attemptId", 123, TaskType.REDUCE, 0, 0);
        TaskAttemptContext hadoopAttemptContext = new TaskAttemptContextImpl(job.getConfiguration(), attemptId);

        //6 構建OutputFormat實例
        OutputFormat format = job.getOutputFormatClass().newInstance();

        //7 設置OutputCommitter
        OutputCommitter committer = format.getOutputCommitter(hadoopAttemptContext);
        committer.setupJob(jobContext);
        committer.setupTask(hadoopAttemptContext);

        //8 獲取writer寫數據，寫完關閉writer
        RecordWriter<Void, Person> writer = format.getRecordWriter(hadoopAttemptContext);
        Person person = new Person();
        person.setName("shaonaiyi");
        person.setAge(18);
        person.setFavoriteNumber(7);
        person.setFavoriteColor("red");
        writer.write(null, person);
        writer.close(hadoopAttemptContext);

        //9 committer提交job和task
        committer.commitTask(hadoopAttemptContext);
        committer.commitJob(jobContext);
    }

}

c. 完整的讀Parquet文件代碼（HDFS）

package com.shaonaiyi.hadoop.filetype.parquet;

import com.shaonaiyi.hadoop.filetype.avro.Person;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.io.NullWritable;
import org.apache.hadoop.mapreduce.*;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.task.JobContextImpl;
import org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl;
import org.apache.parquet.avro.AvroParquetInputFormat;

import java.io.IOException;
import java.util.List;
import java.util.function.Consumer;
/**
 * @Author [email protected]
 * @Date 2019/12/18 10:52
 * @Description 編碼實現讀Parquet文件（HDFS）
 */
public class MRAvroParquetFileReader {

    public static void main(String[] args) throws IOException, IllegalAccessException, InstantiationException {
        //1 構建一個job實例
        Configuration hadoopConf = new Configuration();

        Job job = Job.getInstance(hadoopConf);

        //2 設置需要讀取的文件全路徑
        FileInputFormat.setInputPaths(job, "hdfs://master:9999/user/hadoop-sny/mr/filetype/avro-parquet");

        //3 獲取讀取文件的格式
        AvroParquetInputFormat inputFormat = AvroParquetInputFormat.class.newInstance();

        AvroParquetInputFormat.setAvroReadSchema(job, Person.SCHEMA$);
        //AvroJob.setInputKeySchema(job, Person.SCHEMA$);

        //4 獲取需要讀取文件的數據塊的分區信息
        //4.1 獲取文件被分成多少數據塊了
        JobID jobID = new JobID("jobId", 123);
        JobContext jobContext = new JobContextImpl(job.getConfiguration(), jobID);

        List<InputSplit> inputSplits = inputFormat.getSplits(jobContext);

        //讀取每一個數據塊的數據
        inputSplits.forEach(new Consumer<InputSplit>() {
            @Override
            public void accept(InputSplit inputSplit) {
                TaskAttemptID attemptId = new TaskAttemptID("jobTrackerId", 123, TaskType.MAP, 0, 0);
                TaskAttemptContext hadoopAttemptContext = new TaskAttemptContextImpl(job.getConfiguration(), attemptId);
                RecordReader<NullWritable, Person> reader = null;
                try {
                    reader = inputFormat.createRecordReader(inputSplit, hadoopAttemptContext);
                    reader.initialize(inputSplit, hadoopAttemptContext);
                    while (reader.nextKeyValue()) {
                        System.out.println(reader.getCurrentKey());
                        Person person = reader.getCurrentValue();
                        System.out.println(person);
                    }
                    reader.close();
                } catch (IOException e) {
                    e.printStackTrace();
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
            }
        });

    }

}

4. 查看讀寫Parquet文件（HDFS）結果

a. 寫Parquet文件（HDFS）

b. 讀Parquet文件（HDFS），Key沒有設置值

0x03 彩蛋

1. 編寫讀寫Parquet文件Demo

package com.shaonaiyi.hadoop.filetype.parquet;

import com.shaonaiyi.hadoop.filetype.avro.Person;
import org.apache.hadoop.fs.Path;
import org.apache.parquet.avro.AvroParquetReader;
import org.apache.parquet.avro.AvroParquetWriter;
import org.apache.parquet.hadoop.ParquetReader;
import org.apache.parquet.hadoop.ParquetWriter;
import org.apache.parquet.hadoop.metadata.CompressionCodecName;

import java.io.IOException;

/**
 * @Author [email protected]
 * @Date 2019/12/18 11:11
 * @Description 編寫讀寫Parquet文件Demo
 */
public class AvroParquetDemo {

    public static void main(String[] args) throws IOException {
        Person person = new Person();
        person.setName("shaonaiyi");
        person.setAge(18);
        person.setFavoriteNumber(7);
        person.setFavoriteColor("red");

        Path path = new Path("hdfs://master:9999/user/hadoop-sny/mr/filetype/avro-parquet2");

        ParquetWriter<Object> writer = AvroParquetWriter.builder(path)
                .withSchema(Person.SCHEMA$)
                .withCompressionCodec(CompressionCodecName.SNAPPY)
                .build();

        writer.write(person);

        writer.close();

        ParquetReader<Object> avroParquetReader = AvroParquetReader.builder(path).build();
        Person record = (Person)avroParquetReader.read();
        System.out.println("name：" + record.getName());
        System.out.println("age：" + record.get("age").toString());
        System.out.println("favorite_number：" + record.get("favorite_number").toString());
        System.out.println("favorite_color：" + record.get("favorite_color"));

    }


}

2. 控制檯可以讀出文件
   
3. HDFS上也有數據了
   

0xFF 總結

1. 在MapReduce作業中如何使用：

    job.setInputFormatClass(AvroParquetInputFormat.class);
    AvroParquetInputFormat.setAvroReadSchema(job, Person.SCHEMA$);
    
    job.setOutputFormatClass(ParquetOutputFormat.class);
    AvroParquetOutputFormat.setSchema(job, Person.SCHEMA$);

2. 文章：網站用戶行爲分析項目之會話切割（二） 中 9. 保存統計結果 時就是以Parquet的格式保存下來的。
3. Hadoop支持的文件格式系列：
   Hadoop支持的文件格式之Text
   Hadoop支持的文件格式之Avro
   Hadoop支持的文件格式之Parquet
   Hadoop支持的文件格式之SequenceFile

---

作者簡介：邵奈一
全棧工程師、市場洞察者、專欄編輯
| 公衆號 | 微信 | 微博 | CSDN | 簡書 |

福利：
邵奈一的技術博客導航
邵奈一 原創不易，如轉載請標明出處。

---