This is an example implementation of an event-driven architecture. Its goal is to support the implementation of modules with little to no interconnection, thus allowing for reusability, testability, and readability.
This application implements three modules:
- Manager starts measurements, and display them.
- Sampler generates measurement samples
- Aggregator collects data from Sampler and condition it, before reporting to the Manager
Important point: the modules themselves do not know about one another. They do not include header files of other modules.
Another module, event_dispatch binds the modules into an application.
Other applications (whether it be another product, or a unit test, or a simulation) may implement other connections between the modules, with another event_dispatch implementation.
The simplest module to begin reading is sampler. Observe how it:
- owns its event
- defines a type for the data to join to the event
- has only functions for API
- only includes
event_dispatch.h, no other modules - publishes an event, with joined data, when done
The module manager is of similar complexity, but also takes a config for its init function. This module handles printing messages to the UART output.
Notice the use of anonymous union and struct in the event data type.
The module aggregator publishes two types of event: measurement request, and measurement done. It does not know who consumes these events, as it shouldn't.
Notice how the event data uses anonymous union to save memory.
See event_dispatch_proc and its sub-functions to see the connections between event and action. The sub-functions interpret the data according to the type of event, then call to the relevant functions.
Note how the sampler event SAMPLER_EVENT_DONE connects to two actions:
- add measurement to the aggregator
- display progress via the manager