Stream Processing with Apache Flink

Chapter 5. The DataStream API (v1.7)

This chapter introduces the basics of Flink’s DataStream API. We show the structure and components of a typical Flink streaming application, discuss Flink’s type systems and the supported data types, and present data and partitioning transformations. Window operators, time-based transformations, stateful operators, and connectors are discussed in the next chapters. After reading this chapter, you will know how to implement a stream processing application with basic functionality. Our code examples use Scala for conciseness, but the Java API is mostly analogous (exceptions or special cases will be pointed out). We also provide complete example applications implemented in Java and Scala in our GitHub repositories.

Hello, Flink!

Let’s start with a simple example to get a first impression of what it is like to write streaming applications with the DataStream API. We will use this example to showcase the basic structure of a Flink program and introduce some important features of the DataStream API. Our example application ingests a stream of temperature measurements from multiple sensors.

First, let’s have a look at the data type we will be using to represent sensor readings:

case class SensorReading(
  id: String,
  timestamp: Long,
  temperature: Double)

The program in Example 5-1 converts the temperatures from Fahrenheit to Celsius and computes the average temperature every 5 seconds for each sensor.

Example 5-1. Compute the average temperature ...