Variable beta?

[mncrowe]

From a technical rather than physical perspective, it might be useful to do variable beta in simulations (e.g. for parameter continuation). I'm guessing this isn't currently possible as L depends on beta and is immutable? Is there currently a way to do this and if not, perhaps it could be considered as an addition?

Matthew

[glwagner]

Huh, isn't L an array? In that case it is mutable, you can simply update L with a new beta:

GeophysicalFlows.jl/src/singlelayerqg.jl

Line 231 in 1dd98e4

L = @. - params.μ - params.ν * grid.Krsq^params.nν + im * params.β * grid.kr * grid.invKrsq

[mncrowe]

ERROR: setfield!: immutable struct of type Equation cannot be changed

The issue is L is buried inside the Equation structure from FourierFlows.jl

[glwagner]

There are two other ways you could achieve this:

with_params utility

function with_params(problem, new_params)
  new_equation = Equation(new_params, problem.grid)
  return FourierFlows.Problem(new_equation, problem.stepper, problem.dt, problem.grid, problem.vars, new_params)
end

That's fast and easy, you re-use the memory allocated for the old problem.

change_params! utility

You could put in a PR to change FourierFlows.Problem to be mutable. There's no reason why it shouldn't be (it's not even a concrete type at the moment):

https://github.com/FourierFlows/FourierFlows.jl/blob/1c412b9708fd5a22a429b43dc51b50b6e5dd7bba/src/problem.jl#L59

Then you could design utility that allows you to change the parameters, eg

function change_params!(problem, new_params)
  new_equation = Equation(new_params, problem.grid)
  problem.equation = new_equation
  problem.params = new_params
  return nothing
end

[mncrowe]

Thanks, with_params is a nice way to do it. I'll try that for now, might do the PR later depending on how much I need to do this.

[glwagner]

You should contribute with_params either way! If its useful to you it will probably be useful to others