This repository contains three example applications that are discussed in the blog post "Building Stateful Streaming Applications with Apache Flink". The post presents Apache Flink's APIs for stateful stream processing. Please have a look at a previous blog post to learn more about "Stateful Stream Processing with Apache Flink".

You can fork and clone this repository to run the example applications. All three applications process a stream of taxi ride events that is ingested from a gzipped CSV file, which can be downloaded from Google Drive:

Taxi Ride Data Set

Modify the applications and experiment with Flink's APIs to learn more about Flink and stateful stream processing.

Happy Hacking!

• WorkingTimeMontior: This event-driven application is based on Flink's DataStream API and a ProcessFunction. The application ingests a stream of taxi ride events and notifies taxi drivers when they have to stop working according to work time regulations. A driver's shift may last for at most 12 hours, meaning that rides may only be started within a 12 hour period that begins with the first ride of a shift. After a 12 hour shift, a driver needs to take a break of at least 8 hours. The application notifies drivers until when the last ride of their shift may be started and when they violate the regulation.

• AreasTotalPerHour: This streaming analytics application is based on Flink's SQL support. The application runs a SQL query on a stream of taxi ride events. The query computes the average total amount paid per drop off-location and hour. We discretize the coordinates of the drop-off location into cells of 250x250 meters.

• AreasTotalPerHourOfDay: This streaming analytics application is based on Flink's SQL support. The application runs a SQL query on a stream of taxi ride events. The query computes the average total amount paid per drop off-location and hour of day. We discretize the coordinates of the drop-off location into cells of 250x250 meters.