The problems with eager or push updating in Observables.jl are:
- updating multiple observables in sequence fails if they're mapped together and are only valid after all are changed to their new state
- every update traverses the whole computation graph it's connected to, potentially wasting a lot of resources if the result is not used
PullObservables solve these issues by only recomputing their values once requested (pulled). Here's an example:
using PullObservables
x = PullObservable([1, 2, 3])
y = PullObservable([4, 5, 6])
z = PullObservable{Vector{Int}}()
map!(z, x, y) do x, y
println("Computing z")
x .+ y
end
@show z[]
println("Updating x")
x[] = [1, 2, 3, 4]
println("Updating y")
y[] = [4, 5, 6, 7]
@show z[]Computing z
z[] = [5, 7, 9]
Updating x
Updating y
Computing z
z[] = [5, 7, 9, 11]