Specialize `SubcellLimiterIDP` for certain equations

[patrickersing]

I would like to apply the SubcellLimiterIDP for shallow water equations in TrixiSW. In order to do this I need to modify the blending coefficient alpha computation to set alpha = 1 to have pure FV near dry states and possibly introduce a special version of the idp_local_twosided!.

I think the easiest way to do this would be to just add an additional argument equations to this function

Trixi.jl/src/solvers/dgsem_tree/subcell_limiters_2d.jl

Lines 33 to 35 in a5a6e27

	function (limiter::SubcellLimiterIDP)(u::AbstractArray{<:Any, 4}, semi, dg::DGSEM, t,
	dt;
	kwargs...)

and then use equations::ShallowWaterEquations for dispatch, such that we can create a specialized version.

@bennibolm @amrueda Does this sound good or can someone see a better way to achieve this?

[amrueda]

I think that would be the easiest way to deal with dry states in TrixiSW.

Alternatively, if one wants to minimize the amount of repeated code, one could define a set of limiters to be called when initializing SubcellLimiterIDP (for instance, limiter_set = [idp_local_twosided!, idp_positivity!] or similar), and replace

Trixi.jl/src/solvers/dgsem_tree/subcell_limiters_2d.jl

Lines 40 to 50 in a5a6e27

	if limiter.local_twosided
	@trixi_timeit timer() "local twosided" idp_local_twosided!(alpha, limiter,
	u, t, dt, semi)
	end
	if limiter.positivity
	@trixi_timeit timer() "positivity" idp_positivity!(alpha, limiter, u, dt, semi)
	end
	if limiter.local_onesided
	@trixi_timeit timer() "local onesided" idp_local_onesided!(alpha, limiter,
	u, t, dt, semi)
	end

with a loop over limiter_set. The TrixiSW-specific limiters could be defined in TrixiSW. This requires a bit more implementation effort, so I leave it up to you. 🙂

What do you think, @bennibolm?

[amrueda]

BTW, you will probably need to add the argument equations to the calc_bounds* routines as well in order to account for the bottom topography.

[bennibolm]

I agree with @amrueda . It's probably easier (and completely fine) to just add equations and duplicate some code.
Cooler would be the other approach. I'm happy to help either way.

[patrickersing]

Thanks for the feedback! The idea to have a limiter_set sounds like a nice feature. It would make it convenient to add new limiters without modifying the struct. However, as it is I think the current structure of SubcellLimiterIDP doesn't really allow to add custom limiters, because all information is stored directly in the struct, so this might require quite some changes. I think in that case we would require each limiter to have their own type such that we could define new limiters in TrixiSW and pass them when initializing SubcellLimiterIDP.

If you anticipate other cases where you require custom limiters we could look into this. But if for now it is only my case that is the exception, I would be fine with just dispatching on the equations and accept the small amount of code duplication.

[amrueda]

I would be fine with just dispatching on the equations and accept the small amount of code duplication.

Sounds good!