SparkSession — The Entry Point to Spark SQL

SparkSession is the entry point to Spark SQL. It is the very first object you have to create to start developing Spark SQL applications using the fully-typed Dataset (and untyped DataFrame) data abstractions.

SparkSession has merged SQLContext and HiveContext in one object as of Spark 2.0.0.

You use the SparkSession.builder method to create an instance of SparkSession.

import org.apache.spark.sql.SparkSession
val spark: SparkSession = SparkSession.builder
  .appName("My Spark Application")  // optional and will be autogenerated if not specified
  .master("local[*]")               // avoid hardcoding the deployment environment
  .enableHiveSupport()              // self-explanatory, isn't it?
  .config("spark.sql.warehouse.dir", "target/spark-warehouse")

And stop the current SparkSession using stop method.


You can have multiple SparkSessions in a single Spark application.

Internally, SparkSession requires a SparkContext and an optional SharedState (that represents the shared state across SparkSession instances).

Table 1. SparkSession’s Class and Instance Methods
Method Description


"Opens" a builder to get or create a SparkSession instance


Returns the current version of Spark.


Use import spark.implicits._ to import the implicits conversions and create Datasets from (almost arbitrary) Scala objects.


Creates an empty Dataset[T].


Creates a Dataset[Long].


Executes a SQL query (and returns a DataFrame).


Access to user-defined functions (UDFs).


Creates a DataFrame from a table.


Access to the catalog of the entities of structured queries


Access to DataFrameReader to read a DataFrame from external files and storage systems.


Access to the current runtime configuration.


Access to DataStreamReader to read streaming datasets.


Access to StreamingQueryManager to manage structured streaming queries.


Creates a new SparkSession.


Stops the SparkSession.


Use spark.sql.warehouse.dir Spark property to change the location of Hive’s hive.metastore.warehouse.dir property, i.e. the location of the Hive local/embedded metastore database (using Derby).

Refer to SharedState in this document to learn about (the low-level details of) Spark SQL support for Apache Hive.

See also the official Hive Metastore Administration document.

Table 2. SparkSession’s (Lazily-Initialized) Attributes (in alphabetical order)
Name Description sessionState

Creating SparkSession Instance


Creating SessionState — instantiateSessionState Internal Method

instantiateSessionState(className: String, sparkSession: SparkSession): SessionState
instantiateSessionState is used exclusively when SparkSession is requested for the current SessionState.

Creating SparkSession Using Builder Pattern — builder Method

builder(): Builder

builder creates a new Builder that you use to build a fully-configured SparkSession using a fluent API.

import org.apache.spark.sql.SparkSession
val builder = SparkSession.builder
Read about Fluent interface design pattern in Wikipedia, the free encyclopedia.

Accessing Version of Spark — version Method

version: String

version returns the version of Apache Spark in use.

Internally, version uses spark.SPARK_VERSION value that is the version property in properties file on CLASSPATH.

Implicit Conversions — implicits object

The implicits object is a helper class with the Scala implicit methods (aka conversions) to convert Scala objects to Datasets, DataFrames and Columns. It also defines Encoders for Scala’s "primitive" types, e.g. Int, Double, String, and their products and collections.


Import the implicits by import spark.implicits._.

val spark = SparkSession.builder.getOrCreate()
import spark.implicits._

implicits object offers support for creating Dataset from RDD of any type (for which an encoder exists in scope), or case classes or tuples, and Seq.

implicits object also offers conversions from Scala’s Symbol or $ to Column.

It also offers conversions from RDD or Seq of Product types (e.g. case classes or tuples) to DataFrame. It has direct conversions from RDD of Int, Long and String to DataFrame with a single column name _1.

It is only possible to call toDF methods on RDD objects of Int, Long, and String "primitive" types.

Creating Empty Dataset — emptyDataset method

emptyDataset[T: Encoder]: Dataset[T]

emptyDataset creates an empty Dataset (assuming that future records being of type T).

scala> val strings = spark.emptyDataset[String]
strings: org.apache.spark.sql.Dataset[String] = [value: string]

scala> strings.printSchema
 |-- value: string (nullable = true)

emptyDataset creates a LocalRelation logical query plan.

Creating Dataset from Local Collections and RDDs — createDataset methods

createDataset[T : Encoder](data: Seq[T]): Dataset[T]
createDataset[T : Encoder](data: RDD[T]): Dataset[T]

createDataset is an experimental API to create a Dataset from a local Scala collection, i.e. Seq[T], Java’s List[T], or a distributed RDD[T].

scala> val one = spark.createDataset(Seq(1))
one: org.apache.spark.sql.Dataset[Int] = [value: int]

|    1|

createDataset creates a LocalRelation logical query plan (for the input data collection) or LogicalRDD (for the input RDD[T]).


You’d be better off using Scala implicits and toDS method instead (that does this conversion automatically for you).

val spark: SparkSession = ...
import spark.implicits._

scala> val one = Seq(1).toDS
one: org.apache.spark.sql.Dataset[Int] = [value: int]

Internally, createDataset first looks up the implicit expression encoder in scope to access the AttributeReferences (of the schema).

Only unresolved expression encoders are currently supported.

The expression encoder is then used to map elements (of the input Seq[T]) into a collection of InternalRows. With the references and rows, createDataset returns a Dataset with a LocalRelation logical query plan.

Creating Dataset With Single Long Column — range methods

range(end: Long): Dataset[java.lang.Long]
range(start: Long, end: Long): Dataset[java.lang.Long]
range(start: Long, end: Long, step: Long): Dataset[java.lang.Long]
range(start: Long, end: Long, step: Long, numPartitions: Int): Dataset[java.lang.Long]

range family of methods create a Dataset of Long numbers.

scala> spark.range(start = 0, end = 4, step = 2, numPartitions = 5).show
| id|
|  0|
|  2|
The three first variants (that do not specify numPartitions explicitly) use SparkContext.defaultParallelism for the number of partitions numPartitions.

Internally, range creates a new Dataset[Long] with Range logical plan and Encoders.LONG encoder.

Creating Empty DataFrame —  emptyDataFrame method

emptyDataFrame: DataFrame

emptyDataFrame creates an empty DataFrame (with no rows and columns).

It calls createDataFrame with an empty RDD[Row] and an empty schema StructType(Nil).

Creating DataFrames from RDDs with Explicit Schema — createDataFrame method

createDataFrame(rowRDD: RDD[Row], schema: StructType): DataFrame

createDataFrame creates a DataFrame using RDD[Row] and the input schema. It is assumed that the rows in rowRDD all match the schema.

Executing SQL Queries — sql method

sql(sqlText: String): DataFrame

sql executes the sqlText SQL statement and creates a DataFrame.


sql is imported in spark-shell so you can execute SQL statements as if sql were a part of the environment.

scala> spark.version
res0: String = 2.2.0-SNAPSHOT

scala> :imports
 1) import spark.implicits._       (72 terms, 43 are implicit)
 2) import spark.sql               (1 terms)
scala> sql("SHOW TABLES")
res0: org.apache.spark.sql.DataFrame = [tableName: string, isTemporary: boolean]

scala> sql("DROP TABLE IF EXISTS testData")
res1: org.apache.spark.sql.DataFrame = []

// Let's create a table to SHOW it
spark.range(10).write.option("path", "/tmp/test").saveAsTable("testData")

scala> sql("SHOW TABLES").show
| testdata|      false|

Internally, sql requests the current ParserInterface to execute a SQL query that gives a LogicalPlan.

sql uses SessionState to access the current ParserInterface.

sql then creates a DataFrame using the current SparkSession (itself) and the LogicalPlan.


spark-sql is the main SQL environment in Spark to work with pure SQL statements (where you do not have to use Scala to execute them).

spark-sql> show databases;
Time taken: 0.028 seconds, Fetched 1 row(s)

Accessing UDF Registration Interface — udf Attribute

udf: UDFRegistration

udf attribute gives access to UDFRegistration that allows registering user-defined functions for SQL-based query expressions.

val spark: SparkSession = ...
spark.udf.register("myUpper", (s: String) => s.toUpperCase)

val strs = ('a' to 'c').map(_.toString).toDS

scala> sql("SELECT *, myUpper(value) UPPER FROM strs").show
|    a|    A|
|    b|    B|
|    c|    C|

Internally, it is an alias for SessionState.udf.

Creating DataFrames from Tables — table method

table(tableName: String): DataFrame

table creates a DataFrame from records in the tableName table (if exists).

val df = spark.table("mytable")

Accessing Metastore — catalog Attribute

catalog: Catalog

catalog attribute is a (lazy) interface to the current metastore, i.e. data catalog (of relational entities like databases, tables, functions, table columns, and temporary views).

All methods in Catalog return Datasets.
|              name|database|description|tableType|isTemporary|
|my_permanent_table| default|       null|  MANAGED|      false|
|              strs|    null|       null|TEMPORARY|       true|

Internally, catalog creates a CatalogImpl (referencing the current SparkSession).

Accessing DataFrameReader — read method

read: DataFrameReader

read method returns a DataFrameReader that is used to read data from external storage systems and load it into a DataFrame.

val spark: SparkSession = // create instance
val dfReader: DataFrameReader =

Runtime Configuration — conf attribute

conf: RuntimeConfig

conf returns the current runtime configuration (as RuntimeConfig) that wraps SQLConf.


readStream method

readStream: DataStreamReader

readStream returns a new DataStreamReader.

streams Attribute

streams: StreamingQueryManager

streams attribute gives access to StreamingQueryManager (through SessionState).

val spark: SparkSession = ...

streamingQueryManager Attribute

streamingQueryManager is…​

listenerManager Attribute

listenerManager is…​


ExecutionListenerManager is…​

functionRegistry Attribute

functionRegistry is…​

experimentalMethods Attribute

experimental: ExperimentalMethods

experimentalMethods is an extension point with ExperimentalMethods that is a per-session collection of extra strategies and Rule[LogicalPlan]s.

experimental is used in SparkPlanner and SparkOptimizer. Hive and Structured Streaming use it for their own extra strategies and optimization rules.

newSession method

newSession(): SparkSession

newSession creates (starts) a new SparkSession (with the current SparkContext and SharedState).

scala> println(sc.version)

scala> val newSession = spark.newSession
newSession: org.apache.spark.sql.SparkSession = org.apache.spark.sql.SparkSession@122f58a

sharedState Attribute

sharedState is the current SharedState. It is created lazily when first accessed.


SharedState is an internal class that holds the shared state across active SQL sessions (as SparkSession instances) by sharing CacheManager, SQLListener, and ExternalCatalog.


Enable INFO logging level for org.apache.spark.sql.internal.SharedState logger to see what happens inside.

Add the following line to conf/

Refer to Logging.

SharedState requires a SparkContext when created. It also adds hive-site.xml to Hadoop’s Configuration in the current SparkContext if found on CLASSPATH.

hive-site.xml is an optional Hive configuration file when working with Hive in Spark.

The fully-qualified class name is org.apache.spark.sql.internal.SharedState.

SharedState is created lazily, i.e. when first accessed after SparkSession is created. It can happen when a new session is created or when the shared services are accessed. It is created with a SparkContext.

When created, SharedState sets hive.metastore.warehouse.dir to spark.sql.warehouse.dir if hive.metastore.warehouse.dir is not set or spark.sql.warehouse.dir is set. Otherwise, when hive.metastore.warehouse.dir is set and spark.sql.warehouse.dir is not, spark.sql.warehouse.dir gets set to hive.metastore.warehouse.dir. You should see the following INFO message in the logs:

INFO spark.sql.warehouse.dir is not set, but hive.metastore.warehouse.dir is set. Setting spark.sql.warehouse.dir to the value of hive.metastore.warehouse.dir ('[hiveWarehouseDir]').

You should see the following INFO message in the logs:

INFO SharedState: Warehouse path is '[warehousePath]'.

Stopping SparkSession — stop Method

stop(): Unit

stop stops the SparkSession, i.e. stops the underlying SparkContext.

baseRelationToDataFrame Method


results matching ""

    No results matching ""