Understanding Spark SQL, DataFrames, and Datasets

There was a lot of confusion about the Datasets and DataFrame APIs, so in this article, we will learn about Spark SQL, DataFrames, and Datasets.

Spark SQL

It is a Spark Module for structured data processing, which allows you to write less code to get things done, and underneath the covers, it intelligently performs optimizations. The Spark SQL module consists of two main parts. We will only discuss the first part in this article, which is the representation of the Structure APIs, called DataFrames and Datasets, which define the high-level APIs for working with structured data.

One of the cool features of the Spark SQL module is the ability to execute SQL queries to perform data processing and the result of the queries will be returned as a Dataset or DataFrame. The Spark SQL module makes it easy to read data and write data from and to any of the following formats; CSV, XML, and JSON, and common formats for binary data are Avro, Parquet, and ORC.

DataFrames

A dataframe is a distributed collection of data that is organized into rows, where each row consists of a set of columns, and each column has a name and an associated type. In other words, this distributed collection of data has a structure defined by a schema. You can think you of it as a table in a relational database, but under the hood, it has much richer optimizations.

Like the RDD, the DataFrame offers two type of operations: transformations and actions

Transformations are lazily evaluated, and actions are eagerlyevaluated.

Creating DataFrames

There are several ways to create a DataFrame; one common thing among them is the need to provide a schema, either implicitly or explicitly.

The following code will work perfectly from Spark 2.x with Scala 2.11

Creating DataFrames from RDDs

val rdd = sc.parallelize(1 to 10).map(x => (x, x * x))
val dataframe = spark.createDataFrame(rdd).toDF("key", "sqaure")
dataframe.show()

//Output:

+---+-----+
|key|value|
+---+-----+
|  1|    1|
|  2|    4|
|  3|    9|
|  4|   16|
|  5|   25|
|  6|   36|
|  7|   49|
|  8|   64|
|  9|   81|
| 10|  100|
+---+-----+

Datasets

A Dataset is a strongly typed, immutable collection of data. Similar to a DataFrame, the data in a Dataset is mapped to a defined schema. It is more about type safety and is object-oriented.

There are a few important differences between a DataFrame and a Dataset.

• Each row in a Dataset is represented by a user-defined object so that you can refer to an individual column as a member variable of that object. This provides you with compile-type safety.
• A Dataset has helpers called encoders, which are smart and efficient encoding utilities that convert data inside each user-defined object into a compact binary format. This translates into a reduction of memory usage if and when a Dataset is cached in memory as well as a reduction in the number of bytes that Spark needs to transfer over a network during the shuffling process.

Creating Datasets

There are a few ways to create a Dataset:

• The first way is to transform a DataFrame to a Dataset using the as(Symbol) function of the DataFrame class.
• The second way is to use the SparkSession.createDataset() function to create a Dataset from a local collection of objects.
• The third way is to use the toDS implicit conversion utility.

Let's see different ways of creating Datasets

// create a Dataset using SparkSession.createDataset() and the toDS 
val movies = Seq(Movie("DDLJ", "Awesome", 2018L),  Movie("ADHM", "Nice", 2018L))
val moviesDS = spark.createDataset(localMovies)
moviesDS.show()
val moviesDS1 = localMovies.toDS()
localMoviesDS1.show()

// Encoders are created for case classes
case class Employee(name: String, age: Long)
val caseClassDS = Seq(Employee("Amy", 32)).toDS
caseClassDS.show()

// convert DataFrame to strongly typed Dataset
case class Movie(actor_name:String, movie_title:String, produced_year:Long)
val movies = Seq(("Damon, Matt", "The Bourne Ultimatum", 2007L),
    ("Damon, Matt", "Good Will Hunting", 1997L))
val moviesDF = movies.toDF.as[Movie]

Thank you for reading this article, I hope it was helpful to you.