Also return errorsmatrix in convergence_test

NEWS.md

@@ -5,16 +5,23 @@ Trixi.jl follows the interpretation of
 used in the Julia ecosystem. Notable changes will be documented in this file
 for human readability.
 
+## Changes when updating to v0.15 from v0.14.x
+
+#### Changed
+
+- `convergence_test` now returns both the mean convergence rates and the full errors matrix ([#2753]).
+
 ## Changes when updating to v0.14 from v0.13.x
 
 #### Changed
 
-- van Leer's limiters changed to snake case: `vanLeer` => `vanleer`
-- A couple `struct`s have been made completely immutable, or only a couple fields thereof. Most notably, `save_solution.condition.save_initial_solution` where `save_solution isa SavesolutionCallback` can no longer be directly changed. Instead, the `@reset` macro from [Accessors.jl](https://github.com/JuliaObjects/Accessors.jl) is used in the elixirs.
+- van Leer's limiters changed to snake case: `vanLeer` => `vanleer` ([#2744])
+- A couple `struct`s have been made completely immutable, or only a couple fields thereof ([#2640]). Most notably, `save_solution.condition.save_initial_solution` where `save_solution isa SavesolutionCallback` can no longer be directly changed. Instead, the `@reset` macro from [Accessors.jl](https://github.com/JuliaObjects/Accessors.jl) is used in the elixirs.
 
 ## Changes in the v0.13 lifecycle
 
 #### Added
+
 - Initial 3D support for subcell limiting with `P4estMesh` was added ([#2582] and [#2647]).
   In the new version, IDP positivity limiting for conservative variables (using
   the keyword `positivity_variables_cons` in `SubcellLimiterIDP()`) and nonlinear
@@ -25,7 +32,7 @@ for human readability.
   When the trait is set to `Trixi.True()`, a single method has to be defined, that computes and returns the tuple
   `flux_cons(u_ll, u_rr) + 0.5f0 * flux_noncons(u_ll, u_rr)` and
   `flux_cons(u_ll, u_rr) + 0.5f0 * flux_noncons(u_rr, u_ll)`.
-- Added support for `source_terms_parabolic`, which allows users to specify gradient-dependent source terms when solving parabolic equations ([#2721]). 
+- Added support for `source_terms_parabolic`, which allows users to specify gradient-dependent source terms when solving parabolic equations ([#2721]).
 
 ## Changes when updating to v0.13 from v0.12.x
 

src/auxiliary/special_elixirs.jl

@@ -36,7 +36,7 @@ function convergence_test(mod::Module, elixir::AbstractString, iterations,
 
         include_refined(mod, elixir, initial_resolution, iter; kwargs)
 
-        # @invokelatest is required for interactive use 
+        # @invokelatest is required for interactive use
         # due to world age issues on Julia 1.12 (and newer)
         l2_error, linf_error = @invokelatest mod.analysis_callback(@invokelatest mod.sol)
 
@@ -49,7 +49,7 @@ function convergence_test(mod::Module, elixir::AbstractString, iterations,
     end
 
     # Use raw error values to compute EOC
-    # @invokelatest is required for interactive use 
+    # @invokelatest is required for interactive use
     # due to world age issues on Julia 1.12 (and newer)
     return analyze_convergence(errors, iterations, (@invokelatest mod.semi))
 end
@@ -123,7 +123,7 @@ function analyze_convergence(errors, iterations,
         println("-"^100)
     end
 
-    return eoc_mean_values
+    return eoc_mean_values, errorsmatrix
 end
 
 function convergence_test(elixir::AbstractString, iterations, RealT = Float64;

src/semidiscretization/semidiscretization_coupled.jl

@@ -778,6 +778,7 @@ function analyze_convergence(errors_coupled, iterations,
     end
 
     eoc_mean_values = Vector{Dict{Symbol, Any}}(undef, nsystems(semi_coupled))
+    errorsmatrix = Vector{Dict{Symbol, Matrix{Float64}}}(undef, nsystems(semi_coupled))
     for i in eachsystem(semi_coupled)
         # Use visual cues to separate output from multiple systems
         println()
@@ -790,9 +791,10 @@ function analyze_convergence(errors_coupled, iterations,
         _, equations, _, _ = mesh_equations_solver_cache(semi)
         variablenames = varnames(cons2cons, equations)
 
-        eoc_mean_values[i] = analyze_convergence(errors[i], iterations, variablenames)
+        eoc_mean_values[i], errorsmatrix[i] = analyze_convergence(errors[i], iterations,
+                                                                  variablenames)
     end
 
-    return eoc_mean_values
+    return eoc_mean_values, errorsmatrix
 end
 end # @muladd

test/test_dgmulti_1d.jl

@@ -66,9 +66,9 @@ end
 
 @trixi_testset "elixir_euler_flux_diff.jl (convergence)" begin
     using Trixi: convergence_test
-    mean_convergence = convergence_test(@__MODULE__,
-                                        joinpath(EXAMPLES_DIR,
-                                                 "elixir_euler_flux_diff.jl"), 3)
+    mean_convergence, _ = convergence_test(@__MODULE__,
+                                           joinpath(EXAMPLES_DIR,
+                                                    "elixir_euler_flux_diff.jl"), 3)
     @test isapprox(mean_convergence[:l2],
                    [4.1558759698638434, 3.977911306037128, 4.041421206468769],
                    rtol = 0.05)

test/test_dgmulti_2d.jl

@@ -262,9 +262,9 @@ end
 
 @trixi_testset "elixir_euler_weakform.jl (convergence)" begin
     using Trixi: convergence_test
-    mean_convergence = convergence_test(@__MODULE__,
-                                        joinpath(EXAMPLES_DIR,
-                                                 "elixir_euler_weakform.jl"), 2)
+    mean_convergence, _ = convergence_test(@__MODULE__,
+                                           joinpath(EXAMPLES_DIR,
+                                                    "elixir_euler_weakform.jl"), 2)
     @test isapprox(mean_convergence[:l2],
                    [
                        4.243843382379403,

test/test_special_elixirs.jl

@@ -21,28 +21,28 @@ EXAMPLES_DIR = examples_dir()
 
 @testset "Convergence test" begin
     @timed_testset "tree_2d_dgsem" begin
-        mean_convergence = convergence_test(@__MODULE__,
-                                            joinpath(EXAMPLES_DIR, "tree_2d_dgsem",
-                                                     "elixir_advection_extended.jl"),
-                                            3, initial_refinement_level = 2)
+        mean_convergence, _ = convergence_test(@__MODULE__,
+                                               joinpath(EXAMPLES_DIR, "tree_2d_dgsem",
+                                                        "elixir_advection_extended.jl"),
+                                               3, initial_refinement_level = 2)
         @test isapprox(mean_convergence[:l2], [4.0], rtol = 0.05)
     end
 
     @timed_testset "structured_2d_dgsem" begin
-        mean_convergence = convergence_test(@__MODULE__,
-                                            joinpath(EXAMPLES_DIR,
-                                                     "structured_2d_dgsem",
-                                                     "elixir_advection_extended.jl"),
-                                            3, cells_per_dimension = (5, 9))
+        mean_convergence, _ = convergence_test(@__MODULE__,
+                                               joinpath(EXAMPLES_DIR,
+                                                        "structured_2d_dgsem",
+                                                        "elixir_advection_extended.jl"),
+                                               3, cells_per_dimension = (5, 9))
         @test isapprox(mean_convergence[:l2], [4.0], rtol = 0.05)
     end
 
     @timed_testset "structured_2d_dgsem coupled" begin
-        mean_convergence = convergence_test(@__MODULE__,
-                                            joinpath(EXAMPLES_DIR,
-                                                     "structured_2d_dgsem",
-                                                     "elixir_advection_coupled.jl"),
-                                            3)
+        mean_convergence, _ = convergence_test(@__MODULE__,
+                                               joinpath(EXAMPLES_DIR,
+                                                        "structured_2d_dgsem",
+                                                        "elixir_advection_coupled.jl"),
+                                               3)
         @test isapprox(mean_convergence[1][:l2], [4.0], rtol = 0.05)
         @test isapprox(mean_convergence[2][:l2], [4.0], rtol = 0.05)
     end
@@ -52,37 +52,37 @@ EXAMPLES_DIR = examples_dir()
         no_refine = @cfunction((p4est, which_tree, quadrant)->Cint(0), Cint,
                                (Ptr{Trixi.p4est_t}, Ptr{Trixi.p4est_topidx_t},
                                 Ptr{Trixi.p4est_quadrant_t}))
-        mean_convergence = convergence_test(@__MODULE__,
-                                            joinpath(EXAMPLES_DIR, "p4est_2d_dgsem",
-                                                     "elixir_euler_source_terms_nonconforming_unstructured_flag.jl"),
-                                            2, refine_fn_c = no_refine)
+        mean_convergence, _ = convergence_test(@__MODULE__,
+                                               joinpath(EXAMPLES_DIR, "p4est_2d_dgsem",
+                                                        "elixir_euler_source_terms_nonconforming_unstructured_flag.jl"),
+                                               2, refine_fn_c = no_refine)
         @test isapprox(mean_convergence[:linf], [3.2, 3.2, 4.0, 3.7], rtol = 0.05)
     end
 
     @timed_testset "structured_3d_dgsem" begin
-        mean_convergence = convergence_test(@__MODULE__,
-                                            joinpath(EXAMPLES_DIR,
-                                                     "structured_3d_dgsem",
-                                                     "elixir_advection_basic.jl"),
-                                            2, cells_per_dimension = (7, 4, 5))
+        mean_convergence, _ = convergence_test(@__MODULE__,
+                                               joinpath(EXAMPLES_DIR,
+                                                        "structured_3d_dgsem",
+                                                        "elixir_advection_basic.jl"),
+                                               2, cells_per_dimension = (7, 4, 5))
         @test isapprox(mean_convergence[:l2], [4.0], rtol = 0.05)
     end
 
     @timed_testset "p4est_3d_dgsem" begin
-        mean_convergence = convergence_test(@__MODULE__,
-                                            joinpath(EXAMPLES_DIR, "p4est_3d_dgsem",
-                                                     "elixir_advection_unstructured_curved.jl"),
-                                            2, initial_refinement_level = 0)
+        mean_convergence, _ = convergence_test(@__MODULE__,
+                                               joinpath(EXAMPLES_DIR, "p4est_3d_dgsem",
+                                                        "elixir_advection_unstructured_curved.jl"),
+                                               2, initial_refinement_level = 0)
         @test isapprox(mean_convergence[:l2], [2.7], rtol = 0.05)
     end
 
     @timed_testset "paper_self_gravitating_gas_dynamics" begin
-        mean_convergence = convergence_test(@__MODULE__,
-                                            joinpath(EXAMPLES_DIR,
-                                                     "paper_self_gravitating_gas_dynamics",
-                                                     "elixir_eulergravity_convergence.jl"),
-                                            2, tspan = (0.0, 0.25),
-                                            initial_refinement_level = 1)
+        mean_convergence, _ = convergence_test(@__MODULE__,
+                                               joinpath(EXAMPLES_DIR,
+                                                        "paper_self_gravitating_gas_dynamics",
+                                                        "elixir_eulergravity_convergence.jl"),
+                                               2, tspan = (0.0, 0.25),
+                                               initial_refinement_level = 1)
         @test isapprox(mean_convergence[:l2], 4 * ones(4), atol = 0.4)
     end
 end