Learning Spark, 2nd Edition

Chapter 5. Spark SQL and DataFrames: Interacting with External Data Sources

In the previous chapter, we explored interacting with the built-in data sources in Spark. We also took a closer look at the DataFrame API and its interoperability with Spark SQL. In this chapter, we will focus on how Spark SQL interfaces with external components. Specifically, we discuss how Spark SQL allows you to:

• Use user-defined functions for both Apache Hive and Apache Spark.

• Connect with external data sources such as JDBC and SQL databases, PostgreSQL, MySQL, Tableau, Azure Cosmos DB, and MS SQL Server.

• Work with simple and complex types, higher-order functions, and common relational operators.

We’ll also look at some different options for querying Spark using Spark SQL, such as the Spark SQL shell, Beeline, and Tableau.

Spark SQL and Apache Hive

Spark SQL is a foundational component of Apache Spark that integrates relational processing with Spark’s functional programming API. Its genesis was in previous work on Shark. Shark was originally built on the Hive codebase on top of Apache Spark¹ and became one of the first interactive SQL query engines on Hadoop systems. It demonstrated that it was possible to have the best of both worlds; as fast as an enterprise data warehouse, and scaling as well as Hive/MapReduce.

Spark SQL lets Spark programmers leverage the benefits of faster performance and relational programming (e.g., declarative queries and optimized storage), as well as call complex ...