Functional Reactive Animation (Fran)

 

by Conal Elliott and Paul Hudak

Presentation by Daniel Bugl

Overview

  • Recap: FRP
     
  • Fran?
     
  • Problems with Fran
     
  • Comparison to Elm paper
     
  • Results of papers

Recap: FRP

  • functional: pure functional programming paradigms
     
  • reactive: applied to time-varying values

functional reactive programming

Fran?

  • Fran = functional reactive animation
     
  • a set of data types in the programming language Haskell
     
  • implemented in Hugs

Why Fran?

  • goal: simplify reactive animations
     
  • get the benefits of functional (declarative) programming
     
  • express the "what" of an interactive animation
     
  • automate the "how"
  • possible performance optimizations in the "how"

functional reactive animation

Fran

  • behaviors: time-varying, reactive values
    • most traditional values are behaviors
    • e.g. time-varying images → animations
    • defined via data types
       
  • events: sets of arbitrarily complex conditions
    • e.g. mouse button presses
    • specified as functions of continuous time

functional reactive animation

Data types

  • behaviors
  • events
lbp : Time -> Event (Event ())
  • reactivity = behaviors + events
data Behavior a = Behavior (Time -> a)

Examples

bigger (sin time) circle `over`
bigger (cos time) square
  • animations
  • reactive animations
colorCycle t0 =
  red   `untilB` lbp t0 *=> \t1 ->
  green `untilB` lbp t1 *=> \t2 ->
  colorCycle t2

untilB : Behavior a -> Event (Behavior a)
         -> Behavior a
(*=>) : Event a -> (Time -> b) -> Event b

Problems with Fran

functional reactive animation

  • implicit treatment of time
    • sample input data as quickly as possible
    • call functions to compute events
    • praised for its great expressiveness
       
  • sounds good?
  • problem: unnecessary recomputation

Solution: Elm

comparison of the two papers

  • Fran: continuous resampling of input data
    → pull approach

     
  • Elm: assume discrete events (values only change when events occur)
    → push approach (global event dispatcher)

Results of papers

  • Fran:
    • expressive system to write reactive animations
    • focus on the "what", not the "how" (declarative)
    • however, performance problems with events
       
  • Elm:
    • based on Fran
    • a practical, asynchronous FRP language
    • allowed efficient execution of FRP programs
      (push approach)

Questions?

Global Event Dispatcher

  • Global event dispatcher
    • dispatches all events to all signal sources
       
    • signal sources decide if value changed or not
       
    • if not, return:
noChange v

More Examples

  • events
predicate (time^2 == 5) t0
lbp t0 .|. predicate (time^2 == 5) t0
(.|.) : Event a -> Event a -> Event a
(==>) : Event a -> (a -> b) -> Event b
(*=>) : Event a -> (Time -> b) -> Event b
(-=>) : Event a -> b -> Event b
  • behaviors
at : Behavior a -> Time -> a
Made with Slides.com