Add parameters to Callback

[glwagner]

Probably related:

I saw the new Simulation API and the new Callback idea. This brought a change: the old way of building a Simulation included a parameters arguments. This parameters was accesible from the progress function. With the new way, we build callbacks and then register them, but there is no way to pass additional parameters to it.

In one of my usages I was hacking a progress function https://github.com/aramirezreyes/RamirezReyes_ShallowWaterInFPlane/blob/c1971c0cb3fc99fbb2dbf3303a8fd60b17a59d5a/scripts/run_oceananigans_example_cpu.jl#L101 to update some arrays on a parameterization. Unfortunately, this depended on the parameters argument. The more generalized idea of the callbacks is very elegant and works nice, but it loses some power from the lack of ability to add additional parameters to the callback function.

Is there a way to recover this ability? (there are other ways of getting it to work but having it in the design would be nice).

Originally posted by @aramirezreyes in #1895 (comment)

[glwagner]

@aramirezreyes one possibility is a parameters feature for Callback; when !isnothing(callback.parameters), the callback function is called with func(simulation, parameters).

[glwagner]

Before this issue is resolved, one solution is to use callable objects:

struct ParameterizedCallback{P}
    parameters :: P
end

(cb::ParameterizedCallback)(simulation) = # function that depends on both `sim` and `cb.parameters`

cb = ParameterizedCallback(parameters) # builds the callable object, capturing `parameters`

simulation.callbacks[:callback] = Callback(cb, IterationInterval(10))

[aramirezreyes]

Thanks!
The solution I came up with is using a function with a default argument.

mycallback(simulation, parameters = my_parameters)
simulation.callbacks[:callback] = Callback(mycallback, IterationInterval(10))

I have not checked if the way in which both capture the object would be equivalent.

[glwagner]

A default argument works great too! You can also capture data as a global variable:

julia> a = 3
3

julia> f(x) = x * a
f (generic function with 4 methods)

julia> f(2)
6