test: fix formatting

MilesCranmer · MilesCranmer · commit ed347c3ec713 · 2025-11-07T02:11:43.000Z
diff --git a/test/integration/ad/test_derivatives.jl b/test/integration/ad/test_derivatives.jl
@@ -17,7 +17,8 @@
                 sin(
                     custom_cos(
                         sin(1.2926733 - 1.6606787) / sin(
-                            ((0.14577048 * x1) + ((0.111149654 + x1) - -0.8298334)) - -1.2071426
+                            ((0.14577048 * x1) + ((0.111149654 + x1) - -0.8298334)) -
+                            -1.2071426,
                         ),
                     ) * (custom_cos(x3 - 2.3201916) + ((x1 - (x1 * x2)) / x2)),
                 ) / (0.14854191 - ((custom_cos(x2) * -1.6047639) - 0.023943262))
@@ -64,14 +65,16 @@
             @test array_test(predicted_output, true_output)
 
             true_grad = gradient(
-                (x1, x2, x3) -> sum(equation.(x1, x2, x3)), [X[i, :] for i in 1:nfeatures]...
+                (x1, x2, x3) -> sum(equation.(x1, x2, x3)),
+                [X[i, :] for i in 1:nfeatures]...,
             )
             # Convert tuple of vectors to matrix:
             true_grad = reduce(hcat, true_grad)'
             predicted_grad = eval_grad_tree_array(tree, X, options; variable=true)[2]
             predicted_grad2 =
                 reduce(
-                    hcat, [eval_diff_tree_array(tree, X, options, i)[2] for i in 1:nfeatures]
+                    hcat,
+                    [eval_diff_tree_array(tree, X, options, i)[2] for i in 1:nfeatures],
                 )'
 
             # Print largest difference between predicted_grad, true_grad:
diff --git a/test/integration/ad/test_tree_construction.jl b/test/integration/ad/test_tree_construction.jl
@@ -5,9 +5,9 @@
     using SymbolicRegression: eval_loss, eval_cost, Dataset
     using ForwardDiff
     include(joinpath(@__DIR__, "..", "..", "test_params.jl"))
-    
+
     x1 = 2.0
-    
+
     function make_options_maker(binop, unaop; kw...)
         @nospecialize binop unaop kw
         return Options(;
@@ -19,37 +19,41 @@
             kw...,
         )
     end
-    
+
     # Initialize functions in Base....
     for unaop in
         [cos, exp, safe_log, safe_log2, safe_log10, safe_sqrt, relu, gamma, safe_acosh],
         binop in [sub]
-    
+
         let
             make_options = Fix{1}(Fix{2}(make_options_maker, unaop), binop)
             options = make_options()
             @extend_operators options
-    
+
             # for unaop in
             f_true = (x,) -> binop(abs(3.0 * unaop(x))^2.0, -1.2)
-    
+
             # binop at outside:
             const_tree = Node(
-                5, (^)(Node(2, Node(; val=3.0) * Node(1, Node("x1"))), 2.0), Node(; val=-1.2)
+                5,
+                (^)(Node(2, Node(; val=3.0) * Node(1, Node("x1"))), 2.0),
+                Node(; val=-1.2),
             )
             const_tree_bad = Node(
-                5, (^)(Node(2, Node(; val=3.0) * Node(1, Node("x1"))), 2.1), Node(; val=-1.3)
+                5,
+                (^)(Node(2, Node(; val=3.0) * Node(1, Node("x1"))), 2.1),
+                Node(; val=-1.3),
             )
             n = count_nodes(const_tree)
-    
+
             true_result = f_true(x1)
-    
+
             result = eval(Meta.parse(string_tree(const_tree, options)))
-    
+
             # Test Basics
             @test n == 9
             @test result == true_result
-    
+
             types_to_test = [Float32, Float64, BigFloat]
             if unaop == cos
                 # Other unary operators produce numbers too large
@@ -62,10 +66,10 @@
                 else
                     zero_tolerance = 1e-6
                 end
-    
+
                 tree = convert(Node{T}, const_tree)
                 tree_bad = convert(Node{T}, const_tree_bad)
-    
+
                 Random.seed!(0)
                 N = 100
                 if unaop in [safe_log, safe_log2, safe_log10, safe_acosh, safe_sqrt]
@@ -77,40 +81,41 @@
                 if unaop == safe_acosh
                     X = X .+ T(1.0)
                 end
-    
+
                 y = T.(f_true.(X[1, :]))
                 dataset = Dataset(X, y)
                 test_y, complete = eval_tree_array(tree, X, make_options())
                 test_y2, complete2 = differentiable_eval_tree_array(tree, X, make_options())
-    
+
                 # Test Evaluation
                 @test complete == true
                 @test all(abs.(test_y .- y) / N .< zero_tolerance)
                 @test complete2 == true
                 @test all(abs.(test_y2 .- y) / N .< zero_tolerance)
-    
+
                 # Test loss:
                 @test abs(eval_loss(tree, dataset, make_options())) < zero_tolerance
                 @test eval_loss(tree, dataset, make_options()) ==
                     eval_cost(dataset, tree, make_options())[2]
-    
+
                 #Test Scoring
                 @test abs(eval_cost(dataset, tree, make_options(; parsimony=0.0))[1]) <
                     zero_tolerance
                 @test eval_cost(dataset, tree, make_options(; parsimony=1.0))[1] > 1.0
                 @test eval_cost(dataset, tree, make_options())[1] <
                     eval_cost(dataset, tree_bad, make_options())[1]
-    
+
                 dataset_with_larger_baseline = deepcopy(dataset)
                 dataset_with_larger_baseline.baseline_loss = one(T) * 10
                 @test eval_cost(dataset_with_larger_baseline, tree_bad, make_options())[1] <
                     eval_cost(dataset, tree_bad, make_options())[1]
-    
+
                 # Test gradients:
                 df_true = x -> ForwardDiff.derivative(f_true, x)
                 dy = T.(df_true.(X[1, :]))
                 test_dy = ForwardDiff.gradient(
-                    _x -> sum(differentiable_eval_tree_array(tree, _x, make_options())[1]), X
+                    _x -> sum(differentiable_eval_tree_array(tree, _x, make_options())[1]),
+                    X,
                 )
                 test_dy = test_dy[1, 1:end]
                 @test all(abs.(test_dy .- dy) / N .< zero_tolerance)
diff --git a/test/integration/ext/json3_recorder/test_recorder.jl b/test/integration/ext/json3_recorder/test_recorder.jl
@@ -3,12 +3,12 @@
     using SymbolicRegression.UtilsModule: recursive_merge
     using JSON3
     include(joinpath(@__DIR__, "..", "..", "..", "test_params.jl"))
-    
+
     base_dir = mktempdir()
     recorder_file = joinpath(base_dir, "pysr_recorder.json")
     X = 2 .* randn(Float32, 2, 1000)
     y = 3 * cos.(X[2, :]) + X[1, :] .^ 2 .- 2
-    
+
     options = SymbolicRegression.Options(;
         binary_operators=(+, *, /, -),
         unary_operators=(cos,),
@@ -19,24 +19,24 @@
         maxsize=20,
         complexity_of_operators=[cos => 2],
     )
-    
+
     hall_of_fame = equation_search(
         X, y; niterations=5, options=options, parallelism=:multithreading
     )
-    
+
     data = open(options.recorder_file, "r") do io
         JSON3.read(io; allow_inf=true)
     end
-    
+
     @test haskey(data, :options)
     @test haskey(data, :out1_pop1)
     @test haskey(data, :out1_pop2)
     @test haskey(data, :mutations)
-    
+
     # Test that "Options" is part of the string in `data.options`:
     @test contains(data.options, "Options")
     @test length(data.mutations) > 1000
-    
+
     # Check whether 10 random elements have the right properties:
     for (i, key) in enumerate(keys(data.mutations))
         @test haskey(data.mutations[key], :events)
@@ -48,6 +48,6 @@
             break
         end
     end
-    
+
     @test_throws ErrorException recursive_merge()
 end
diff --git a/test/integration/ext/loopvectorization/test_nan_detection.jl b/test/integration/ext/loopvectorization/test_nan_detection.jl
@@ -2,13 +2,15 @@
     println("Testing NaN detection.")
     using SymbolicRegression
     using LoopVectorization
-    
+
     for T in [Float16, Float32, Float64], turbo in [true, false]
         T == Float16 && turbo && continue
         local options, tree, X
-    
+
         options = Options(;
-            binary_operators=(+, *, /, -, ^), unary_operators=(cos, sin, exp, sqrt), turbo=turbo
+            binary_operators=(+, *, /, -, ^),
+            unary_operators=(cos, sin, exp, sqrt),
+            turbo=turbo,
         )
         @extend_operators options
         # Creating a NaN via computation.
@@ -17,27 +19,27 @@
         X = fill(T(100), 1, 10)
         output, flag = eval_tree_array(tree, X, options)
         @test !flag
-    
+
         # Creating a NaN/Inf via division by constant zero.
         tree = cos(Node(T; feature=1) / zero(T))
         tree = convert(Node{T}, tree)
         output, flag = eval_tree_array(tree, X, options)
         @test !flag
-    
+
         # Creating a NaN via sqrt(-1):
         tree = safe_sqrt(Node(T; feature=1) - 1)
         tree = convert(Node{T}, tree)
         X = fill(T(0), 1, 10)
         output, flag = eval_tree_array(tree, X, options)
         @test !flag
-    
+
         # Creating a NaN via pow(-1, 0.5):
         tree = (^)(Node(T; feature=1) - 1, 0.5)
         tree = convert(Node{T}, tree)
         X = fill(T(0), 1, 10)
         output, flag = eval_tree_array(tree, X, options)
         @test !flag
-    
+
         # Having a NaN/Inf constants:
         tree = cos(Node(T; feature=1) + T(Inf))
         tree = convert(Node{T}, tree)
@@ -48,6 +50,6 @@
         output, flag = eval_tree_array(tree, X, options)
         @test !flag
     end
-    
+
     println("Passed.")
 end
diff --git a/test/integration/ext/loopvectorization/test_turbo_nan.jl b/test/integration/ext/loopvectorization/test_turbo_nan.jl
@@ -1,9 +1,9 @@
 @testitem "Test turbo NaN with LoopVectorization" begin
     using SymbolicRegression
     using LoopVectorization
-    
+
     bad_op(x::T) where {T} = (x >= 0) ? x : T(0)
-    
+
     options = Options(;
         unary_operators=(sin, exp, sqrt, bad_op),
         binary_operators=(+, *),
@@ -14,19 +14,19 @@
         # ^ Leave as deprecated param, just to test.
         parsimony=0.01,
     )
-    
+
     tree = Node(3, Node(1, Node(; val=-π / 2)))
-    
+
     # Should still be safe against domain errors:
     try
         tree([0.0]', options)
         @test true
     catch e
         @test false
     end
-    
+
     tree = Node(3, Node(1, Node(; feature=1)))
-    
+
     try
         tree([-π / 2]', options)
         @test true
diff --git a/test/integration/ext/symbolicutils/test_simplification.jl b/test/integration/ext/symbolicutils/test_simplification.jl
@@ -63,8 +63,9 @@
         ((x2 + x2) * ((-0.5982493 / pow_abs2(x1, x2)) / -0.54734415)) + (
             sin(
                 custom_cos2(
-                    sin(1.2926733 - 1.6606787) /
-                    sin(((0.14577048 * x1) + ((0.111149654 + x1) - -0.8298334)) - -1.2071426),
+                    sin(1.2926733 - 1.6606787) / sin(
+                        ((0.14577048 * x1) + ((0.111149654 + x1) - -0.8298334)) - -1.2071426
+                    ),
                 ) * (custom_cos2(x3 - 2.3201916) + ((x1 - (x1 * x2)) / x2)),
             ) / (0.14854191 - ((custom_cos2(x2) * -1.6047639) - 0.023943262))
         )
diff --git a/test/runtests.jl b/test/runtests.jl
@@ -34,7 +34,9 @@ ENV["SYMBOLIC_REGRESSION_TEST"] = "true"
         end
 
         # Run all test items in the integration directory
-        @run_package_tests(filter = ti -> startswith(ti.filename, integration_dir), verbose = true)
+        @run_package_tests(
+            filter = ti -> startswith(ti.filename, integration_dir), verbose = true
+        )
     else
         # Unit tests
         test_dir = joinpath(@__DIR__, TEST_GROUP)
