Spark的那些事(四) java操作kudu全示例(含sparksql)

上文提到，使用kudu等列式存储将数据以update模式写入kudu.
下面说一下java操作kudu的相关demo。java操作kudu在git上有相关demo，而spark操作kudu并没有。cloudera官网的操作中只提到了scala版本。本文列举java操作kudu的全示例，仅供入门参考。(痛苦的是sparksql查询kudu的java实现，官方没有示例，google也不好用)

1）pom依赖

  <!-- https://mvnrepository.com/artifact/org.apache.kudu/kudu-client -->
        <dependency>
            <groupId>org.apache.kudu</groupId>
            <artifactId>kudu-client</artifactId>
            <version>1.5.0-cdh5.13.1</version>
            <scope>test</scope>
        </dependency>

        <!-- https://mvnrepository.com/artifact/org.apache.kudu/kudu-client-tools -->
        <dependency>
            <groupId>org.apache.kudu</groupId>
            <artifactId>kudu-client-tools</artifactId>
            <version>1.5.0-cdh5.13.1</version>
        </dependency>

        <!-- https://mvnrepository.com/artifact/org.apache.kudu/kudu-spark2 -->
        <dependency>
            <groupId>org.apache.kudu</groupId>
            <artifactId>kudu-spark2_2.11</artifactId>
            <version>1.6.0</version>
        </dependency>

本文用的是cloudera版本，添加：

 <repositories>
        <repository>
            <id>cloudera</id>
            <url>https://repository.cloudera.com/artifactory/cloudera-repos/</url>
        </repository>
    </repositories>

2）功能列表：
使用kuduClient创建表；
使用kuduClient添加数据；
使用kuduClient更新数据；
使用kuduClient查询数据；
使用kuduClient删除表；
使用sparksql查询数据；
使用spark---kuduContext判断表存在

ps：sparksql查询数据在cloudera官网只有scala版本。google也难找到java版的具体写法。查看源码，实际上通过format来指定包路径，制定的路径下包含实现了sparksql的DefaultSource即可。如spark.kudu包中存在DefaultSource类便可以被sparksql识别。
举一反三，其他的库也可以通过此方式访问。同时要扩展集成一个可以供sparksql查询的库也可以通过此方式实现。

package org.apache.kudu.spark.kudu
@org.apache.yetus.audience.InterfaceStability.Unstableclass DefaultSource() extends scala.AnyRef with org.apache.spark.sql.sources.RelationProvider with org.apache.spark.sql.sources.CreatableRelationProvider with org.apache.spark.sql.sources.SchemaRelationProvider {
  val TABLE_KEY : java.lang.String = { /* compiled code */ }
  val KUDU_MASTER : java.lang.String = { /* compiled code */ }
  val OPERATION : java.lang.String = { /* compiled code */ }
  val FAULT_TOLERANT_SCANNER : java.lang.String = { /* compiled code */ }
  val SCAN_LOCALITY : java.lang.String = { /* compiled code */ }  def defaultMasterAddrs : scala.Predef.String = { /* compiled code */ }
  override def createRelation(sqlContext : org.apache.spark.sql.SQLContext, parameters : scala.Predef.Map[scala.Predef.String, scala.Predef.String]) : org.apache.spark.sql.sources.BaseRelation = { /* compiled code */ }
  override def createRelation(sqlContext : org.apache.spark.sql.SQLContext, mode : org.apache.spark.sql.SaveMode, parameters : scala.Predef.Map[scala.Predef.String, scala.Predef.String], data : org.apache.spark.sql.DataFrame) : org.apache.spark.sql.sources.BaseRelation = { /* compiled code */ }
  override def createRelation(sqlContext : org.apache.spark.sql.SQLContext, parameters : scala.Predef.Map[scala.Predef.String, scala.Predef.String], schema : org.apache.spark.sql.types.StructType) : org.apache.spark.sql.sources.BaseRelation = { /* compiled code */ }
}

3）代码示例：

import jdk.nashorn.internal.ir.annotations.Ignore;import org.apache.kudu.ColumnSchema;import org.apache.kudu.Schema;import org.apache.kudu.Type;import org.apache.kudu.client.*;import org.apache.kudu.spark.kudu.KuduContext;import org.apache.spark.SparkConf;import org.apache.spark.SparkContext;import org.apache.spark.sql.Dataset;import org.apache.spark.sql.SparkSession;import org.apache.spark.sql.types.DataTypes;import org.apache.spark.sql.types.StructField;import org.apache.spark.sql.types.StructType;import org.junit.Test;import java.util.ArrayList;import java.util.Arrays;import java.util.List;/**
 * @ClassName: KuduUtil
 * @Description:用于操作kudu的示例代码
 * @author jason.li
 * @date 2018年1月11日 下午3:45:06
 */@Ignorepublic class KuduUtil {    private static final String KUDU_MASTER = "10.1.0.20:7051";    private static String tableName = "TestKudu";    @Test
    public void kuduCreateTableTest(){
        KuduClient client = new KuduClient.KuduClientBuilder(KUDU_MASTER).build();        try {
            List<ColumnSchema> columns = new ArrayList(2);
            columns.add(new ColumnSchema.ColumnSchemaBuilder("key", Type.STRING)
                    .key(true)
                    .build());
            columns.add(new ColumnSchema.ColumnSchemaBuilder("value", Type.STRING)
                    .build());
            List<String> rangeKeys = new ArrayList<>();
            rangeKeys.add("key");
            Schema schema = new Schema(columns);
            client.createTable(tableName, schema,                    new CreateTableOptions().setRangePartitionColumns(rangeKeys));
        } catch (Exception e) {
            e.printStackTrace();
        }finally {            try {
                client.shutdown();
            } catch (Exception e) {
                e.printStackTrace();
            }
        }
    }    @Test
    public void kuduSaveTest(){
        KuduClient client = new KuduClient.KuduClientBuilder(KUDU_MASTER).build();        try{
            KuduTable table = client.openTable(tableName);
            KuduSession session = client.newSession();
            System.out.println("-------start--------"+System.currentTimeMillis());            for (int i = 30000; i < 31000; i++) {
                Insert insert = table.newInsert();
                PartialRow row = insert.getRow();
                row.addString(0, i+"");
                row.addString(1, "aaa");
                OperationResponse operationResponse =  session.apply(insert);
            }
            System.out.println("-------end--------"+System.currentTimeMillis());
        } catch (Exception e) {
            e.printStackTrace();
        }finally {            try {
                client.shutdown();
            } catch (Exception e) {
                e.printStackTrace();
            }
        }
    }    @Test
    public void kuduUpdateTest(){

        KuduClient client = new KuduClient.KuduClientBuilder(KUDU_MASTER).build();        try {
        KuduTable table = client.openTable(tableName);
            KuduSession session = client.newSession();
                Update update = table.newUpdate();
                PartialRow row = update.getRow();
                row.addString("key", 4+"");
                row.addString("value", "value " + 10);
            OperationResponse operationResponse =  session.apply(update);

           System.out.print(operationResponse.getRowError());

        } catch (Exception e) {
            e.printStackTrace();
        }finally {            try {
                client.shutdown();
            } catch (Exception e) {
                e.printStackTrace();
            }
        }

    }    @Test
    public void kuduSearchTest(){
        KuduClient client = new KuduClient.KuduClientBuilder(KUDU_MASTER).build();        try {
            KuduTable table = client.openTable(tableName);
        List<String> projectColumns = new ArrayList<>(1);
        projectColumns.add("value");
        KuduScanner scanner = client.newScannerBuilder(table)
                .setProjectedColumnNames(projectColumns)
                .build();        while (scanner.hasMoreRows()) {
            RowResultIterator results = scanner.nextRows();            while (results.hasNext()) {
                RowResult result = results.next();
                System.out.println(result.getString(0));
            }
        }
        } catch (Exception e) {
            e.printStackTrace();
        }finally {            try {
                client.shutdown();
            } catch (Exception e) {
                e.printStackTrace();
            }
        }
    }    @Test
    public void kuduDelTabletest(){
        KuduClient client = new KuduClient.KuduClientBuilder(KUDU_MASTER).build();        try {
            client.deleteTable(tableName);
        } catch (Exception e) {
            e.printStackTrace();
        } finally {            try {
                client.shutdown();
            } catch (Exception e) {
                e.printStackTrace();
            }
        }
    }    @Test
    public void searchBysparkSql(){
        SparkSession sparkSession = getSparkSession();
        List<StructField> fields = Arrays.asList(
                DataTypes.createStructField("key", DataTypes.StringType, true),
                DataTypes.createStructField("value", DataTypes.StringType, true));
        StructType schema = DataTypes.createStructType(fields);
        Dataset ds =  sparkSession.read().format("org.apache.kudu.spark.kudu").
                schema(schema).option("kudu.master","10.1.0.20:7051").option("kudu.table","TestKudu").load();
        ds.registerTempTable("abc");
        sparkSession.sql("select * from abc").show();
    }    @Test
    public void checkTableExistByKuduContext(){
        SparkSession sparkSession = getSparkSession();
        KuduContext context = new KuduContext("10.1.0.20:7051",sparkSession.sparkContext());
        System.out.println(tableName +" is exist = "context.tableExists(tableName));
    }    public SparkSession getSparkSession(){
        SparkConf conf = new SparkConf().setAppName("test")
                .setMaster("local[*]")
                .set("spark.driver.userClassPathFirst", "true");

        conf.set("spark.sql.crossJoin.enabled", "true");
        SparkContext sparkContext = new SparkContext(conf);
        SparkSession sparkSession = SparkSession.builder().sparkContext(sparkContext).getOrCreate();        return sparkSession;
    }
}

作者：假文艺的真码农
链接：https://www.jianshu.com/p/ef1a621fc6ea