Hive Warehouse Connector APIs in Azure HDInsight

This article lists all the APIs supported by Hive warehouse connector. All the examples shown below are run using spark-shell and hive warehouse connector session.

How to create Hive warehouse connector session:

import com.hortonworks.hwc.HiveWarehouseSession
val hive = HiveWarehouseSession.session(spark).build()

Prerequisite

Complete the Hive Warehouse Connector setup steps.

Supported APIs

  • Set the database:

    hive.setDatabase("<database-name>")
    
  • List all databases:

    hive.showDatabases()
    
  • List all tables in the current database

    hive.showTables()
    
  • Describe a table

    // Describes the table <table-name> in the current database
    hive.describeTable("<table-name>")
    
    // Describes the table <table-name> in <database-name>
    hive.describeTable("<database-name>.<table-name>")
    
  • Drop a database

    // ifExists and cascade are boolean variables
    hive.dropDatabase("<database-name>", ifExists, cascade)
    
  • Drop a table in the current database

    // ifExists and purge are boolean variables
    hive.dropTable("<table-name>", ifExists, purge)
    
  • Create a database

    // ifNotExists is boolean variable
    hive.createDatabase("<database-name>", ifNotExists)
    
  • Create a table in current database

    // Returns a builder to create table
    val createTableBuilder = hive.createTable("<table-name>")
    

    Builder for create-table supports only the below operations:

    // Create only if table does not exists already
    createTableBuilder = createTableBuilder.ifNotExists()
    
    // Add columns
    createTableBuilder = createTableBuilder.column("<column-name>", "<datatype>")
    
    // Add partition column
    createTableBuilder = createTableBuilder.partition("<partition-column-name>", "<datatype>")
    
    // Add table properties
    createTableBuilder = createTableBuilder.prop("<key>", "<value>")
    
    // Creates a bucketed table,
    // Parameters are numOfBuckets (integer) followed by column names for bucketing
    createTableBuilder = createTableBuilder.clusterBy(numOfBuckets, "<column1>", .... , "<columnN>")
    
    // Creates the table
    createTableBuilder.create()
    

    Note

    This API creates an ORC formatted table at default location. For other features/options or to create table using hive queries, use executeUpdate API.

  • Read a table

    // Returns a Dataset<Row> that contains data of <table-name> in the current database
    hive.table("<table-name>")
    
  • Execute DDL commands on HiveServer2

    // Executes the <hive-query> against HiveServer2
    // Returns true or false if the query succeeded or failed respectively
    hive.executeUpdate("<hive-query>")
    
    // Executes the <hive-query> against HiveServer2
    // Throws exception, if propagateException is true and query threw excpetion in HiveServer2
    // Returns true or false if the query succeeded or failed respectively
    hive.executeUpdate("<hive-query>", propagateException) // propagate exception is boolean value
    
  • Execute Hive query and load result in Dataset

    • Executing query via LLAP daemons. [Recommended]

      // <hive-query> should be a hive query 
      hive.executeQuery("<hive-query>")
      
    • Executing query through HiveServer2 via JDBC.

      Set spark.datasource.hive.warehouse.smartExecution to false in spark configs before starting spark session to use this API

      hive.execute("<hive-query>")
      
  • Close Hive warehouse connector session

    // Closes all the open connections and
    // release resources/locks from HiveServer2
    hive.close()
    
  • Execute Hive Merge query

    This API creates a Hive merge query of below format

    MERGE INTO <current-db>.<target-table> AS <targetAlias> USING <source expression/table> AS <sourceAlias>
    ON <onExpr>
    WHEN MATCHED [AND <updateExpr>] THEN UPDATE SET <nameValuePair1> ... <nameValuePairN>
    WHEN MATCHED [AND <deleteExpr>] THEN DELETE
    WHEN NOT MATCHED [AND <insertExpr>] THEN INSERT VALUES <value1> ... <valueN>
    
    val mergeBuilder = hive.mergeBuilder() // Returns a builder for merge query
    

    Builder supports the following operations:

    mergeBuilder.mergeInto("<taget-table>", "<targetAlias>")
    
    mergeBuilder.using("<source-expression/table>", "<sourceAlias>")
    
    mergeBuilder.on("<onExpr>")
    
    mergeBuilder.whenMatchedThenUpdate("<updateExpr>", "<nameValuePair1>", ... , "<nameValuePairN>")
    
    mergeBuilder.whenMatchedThenDelete("<deleteExpr>")
    
    mergeBuilder.whenNotMatchedInsert("<insertExpr>", "<value1>", ... , "<valueN>");
    
    // Executes the merge query
    mergeBuilder.merge()
    
  • Write a Dataset to Hive Table in batch

    df.write.format("com.hortonworks.spark.sql.hive.llap.HiveWarehouseConnector")
       .option("table", tableName)
       .mode(SaveMode.Type)
       .save()
    
    • TableName should be of form <db>.<table> or <table>. If no database name is provided, the table will searched/created in the current database

    • SaveMode types are:

      • Append: Appends the dataset to the given table

      • Overwrite: Overwrites the data in the given table with dataset

      • Ignore: Skips write if table already exists, no error thrown

      • ErrorIfExists: Throws error if table already exists

  • Write a Dataset to Hive Table using HiveStreaming

    df.write.format("com.hortonworks.spark.sql.hive.llap.HiveStreamingDataSource")
       .option("database", databaseName)
       .option("table", tableName)
       .option("metastoreUri", "<HMS_URI>")
    // .option("metastoreKrbPrincipal", principal), add if executing in ESP cluster
       .save()
    
     // To write to static partition
     df.write.format("com.hortonworks.spark.sql.hive.llap.HiveStreamingDataSource")
       .option("database", databaseName)
       .option("table", tableName)
       .option("partition", partition)
       .option("metastoreUri", "<HMS URI>")
    // .option("metastoreKrbPrincipal", principal), add if executing in ESP cluster
       .save()
    

    Note

    Stream writes always append data.

  • Writing a spark stream to a Hive Table

    stream.writeStream
        .format("com.hortonworks.spark.sql.hive.llap.streaming.HiveStreamingDataSource")
        .option("metastoreUri", "<HMS_URI>")
        .option("database", databaseName)
        .option("table", tableName)
      //.option("partition", partition) , add if inserting data in partition
      //.option("metastoreKrbPrincipal", principal), add if executing in ESP cluster
        .start()