Evaluation model

The core of languages that support Reactive Programming is about how changes are propagated. From the programmer's point of view this should be as much transparent as possible. In other words, a change of a value is automatically propagated to all dependent computations without the user has to manually propagate the changes by himself.

At the language level this means that there should be a mechanism that is triggered when there's an event occurence at an event source. This mechanism will fire a notification about the changes to dependent computations, that will possibly trigger a recomputation, and so on.

The issue of who initiates the propagation of changes is a really important decision in designing a language or a framework that supports RP.

In simple terms the question is whether the source should push new data to its dependents (consumers) or the dependents should pull data from the event source (producer).

In the literature there are two main evaluation models:

• Pull-Based, in which the computation that requires a value has to pull it from the source. The propagation is said to be demand-driven. The pull model has been implemented in Fran, also thanks to the lazy evaluation of Haskell. A main trait of this approach is that the computation requiring the value has the liberty to only pull the new values when if actually needs it. From another point of view, this trait can lead to a significant latency between an occurence of an event and when its reactions are triggered.
• Push-Based, in which when the event source has new data it pushes the data to its dependent computations. The propagation is said to be data-driven. An example of a recent implementation that use this model is Scala.React, that will be introduced later in this thesis. This approach also brings the issue of wasteful recomputation, since every new data that is pushed to the consumer triggers a recomputation.

Tipically, reactive programming languages use either a pull-based or push-based model, but there are also languages that employ both approaches. In latter case there've both the benefits of the push-based model (efficiency and low latency) and those of the pull-based model (flexibility of pulling values based on demand).