Efficient functional reactive programming through incremental behaviors

Abstract

Many types of software are inherently event-driven ranging from web applications to embedded devices and traditionally, such applications are implemented using imperative callbacks. An alternative approach to writing such programs is functional reactive programming (FRP). FRP offers abstractions to make event-driven programming convenient, safe and composable, but they come at a price. FRP behaviors cannot efficiently deal with larger, incrementally constructed values such as a collection of messages or a list of connected devices. Since these situations occur naturally, it hinders the use of FRP. We report on a new FRP primitive: ‘incremental behavior’. We show that the semantics fit within existing FRP semantics and that their API can be used as a foundation for more ad-hoc solutions, such as incremental collections and discrete behaviors. Finally, we present benchmarks that demonstrate the advantages of incremental behaviors in terms of reduced computation time and bandwidth.