Building WENO advection with immersed grid fails due to grid size

[tomchor]

The following snippet fails:

using Oceananigans
grid_base = RectilinearGrid(size=(8, 8, 8), extent = (1,1,1), halo=(4,4,4))
bathymetry(x, y) = -0.5
grid = ImmersedBoundaryGrid(grid_base, GridFittedBottom(bathymetry))
WENO(grid=grid)

The error is

julia> WENO(grid=grid)
ERROR: ArgumentError: The dimensions of the immersed boundary (16, 16) do not match the grid size (18, 18)
Stacktrace:
 [1] validate_ib_size(grid::RectilinearGrid{Float64, Periodic, Periodic, Bounded, Float64, Float64, Float64, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, CPU}, ib::GridFittedBottom{OffsetArrays.OffsetMatrix{Float64, Matrix{Float64}}, Oceananigans.ImmersedBoundaries.CenterImmersedCondition})
   @ Oceananigans.ImmersedBoundaries ~/.julia/packages/Oceananigans/9znTg/src/ImmersedBoundaries/grid_fitted_immersed_boundaries.jl:92
 [2] ImmersedBoundaryGrid(grid::RectilinearGrid{Float64, Periodic, Periodic, Bounded, Float64, Float64, Float64, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, CPU}, ib::GridFittedBottom{OffsetArrays.OffsetMatrix{Float64, Matrix{Float64}}, Oceananigans.ImmersedBoundaries.CenterImmersedCondition})
   @ Oceananigans.ImmersedBoundaries ~/.julia/packages/Oceananigans/9znTg/src/ImmersedBoundaries/grid_fitted_immersed_boundaries.jl:85
 [3] with_halo(halo::Tuple{Int64, Int64, Int64}, ibg::ImmersedBoundaryGrid{Float64, Periodic, Periodic, Bounded, RectilinearGrid{Float64, Periodic, Periodic, Bounded, Float64, Float64, Float64, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, CPU}, GridFittedBottom{OffsetArrays.OffsetMatrix{Float64, Matrix{Float64}}, Oceananigans.ImmersedBoundaries.CenterImmersedCondition}, Nothing, CPU})
   @ Oceananigans.ImmersedBoundaries ~/.julia/packages/Oceananigans/9znTg/src/ImmersedBoundaries/ImmersedBoundaries.jl:136
 [4] #compute_reconstruction_coefficients#26
   @ ~/.julia/packages/Oceananigans/9znTg/src/Advection/reconstruction_coefficients.jl:224 [inlined]
 [5] WENO(FT::DataType; order::Int64, grid::ImmersedBoundaryGrid{Float64, Periodic, Periodic, Bounded, RectilinearGrid{Float64, Periodic, Periodic, Bounded, Float64, Float64, Float64, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, OffsetArrays.OffsetVector{Float64, StepRangeLen{Float64, Base.TwicePrecision{Float64}, Base.TwicePrecision{Float64}, Int64}}, CPU}, GridFittedBottom{OffsetArrays.OffsetMatrix{Float64, Matrix{Float64}}, Oceananigans.ImmersedBoundaries.CenterImmersedCondition}, Nothing, CPU}, zweno::Bool, bounds::Nothing)
   @ Oceananigans.Advection ~/.julia/packages/Oceananigans/9znTg/src/Advection/weno_reconstruction.jl:121
 [6] top-level scope
   @ REPL[5]:1

In order to work we need to call WENO as

WENO(grid=grid_base)

Is this by design or is it a bug?

[simone-silvestri]

It is not by design, because this is not a specific WENO problem. It's a problem with using with_halo on immersed boundaries (the bottom field does not have the same size as the new grid when trying to enlarge the halos). I thought it had been fixed already.

I ll fix it in a new PR

[tomchor]

Thanks @simone-silvestri! Should we add a test?

[tomchor]

This seems to not be an issue anymore:

julia> using Oceananigans
[ Info: Precompiling Oceananigans [9e8cae18-63c1-5223-a75c-80ca9d6e9a09]

julia> grid_base = RectilinearGrid(size=(8, 8, 8), extent = (1,1,1), halo=(4,4,4))
8×8×8 RectilinearGrid{Float64, Periodic, Periodic, Bounded} on CPU with 4×4×4 halo
├── Periodic x ∈ [0.0, 1.0)  regularly spaced with Δx=0.125
├── Periodic y ∈ [0.0, 1.0)  regularly spaced with Δy=0.125
└── Bounded  z ∈ [-1.0, 0.0] regularly spaced with Δz=0.125

julia> bathymetry(x, y) = -0.5
bathymetry (generic function with 1 method)

julia> grid = ImmersedBoundaryGrid(grid_base, GridFittedBottom(bathymetry))
8×8×8 ImmersedBoundaryGrid{Float64, Periodic, Periodic, Bounded} on CPU with 4×4×4 halo:
├── immersed_boundary: GridFittedBottom(mean(z)=-0.5, min(z)=-0.5, max(z)=-0.5)
├── underlying_grid: 8×8×8 RectilinearGrid{Float64, Periodic, Periodic, Bounded} on CPU with 4×4×4 halo
├── Periodic x ∈ [0.0, 1.0)  regularly spaced with Δx=0.125
├── Periodic y ∈ [0.0, 1.0)  regularly spaced with Δy=0.125
└── Bounded  z ∈ [-1.0, 0.0] regularly spaced with Δz=0.125

julia> WENO(grid=grid)
WENO reconstruction order 5 
 Smoothness formulation: 
    └── Z-weno 
 Boundary scheme: 
    └── WENO reconstruction order 3
 Symmetric scheme: 
    └── Centered reconstruction order 4
 Directions:
    ├── X regular 
    ├── Y regular 
    └── Z regular

So I'm closing this issue.