Added JF test + test dep (i guess i forgot?). Also formats on 2.12.

.JuliaFormatter.toml

@@ -1,4 +1,7 @@
 indent = 4
 margin = 92
 normalize_line_endings = "unix"
+v2_stable_multiline_strings = true
+conditional_to_if = false
+pipe_to_function_call = false
 style="blue"

CHANGELOG.md

@@ -7,6 +7,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Added
+
+- Adds `JuliaFormatter` testing which was forgotten (#60).
+
 ### Removed
 
 - `DynamicPPLExt` no longer requires `ForwardDiff` as a triggering library to load.

benchmarks/ParallelMCMCBenchmarks/scripts/compare_pr_benchmarks.jl

@@ -135,7 +135,7 @@ function write_markdown(path, rows; warn_ratio, fail_ratio)
                 " |",
             )
         end
-        println(io)
+        return println(io)
     end
     return path
 end

benchmarks/ParallelMCMCBenchmarks/scripts/new_bench.jl

@@ -88,7 +88,7 @@ function build_raw_deer_problem(
 
     tape = map(1:T) do _
         ξ = copyto!(similar(x0, D), randn(rng, FP, D))
-        ParallelMCMC.MALATapeElement(ξ, FP(rand(rng)))
+        return ParallelMCMC.MALATapeElement(ξ, FP(rand(rng)))
     end
 
     rec = ParallelMCMC._build_mala_deer_rec(

benchmarks/ParallelMCMCBenchmarks/scripts/prof_view.jl

@@ -61,7 +61,7 @@ function build_raw_deer_problem(
 
     tape = map(1:T) do _
         ξ = copyto!(similar(x0, D), randn(rng, FP, D))
-        ParallelMCMC.MALATapeElement(ξ, FP(rand(rng)))
+        return ParallelMCMC.MALATapeElement(ξ, FP(rand(rng)))
     end
 
     rec = ParallelMCMC._build_mala_deer_rec(

benchmarks/ParallelMCMCBenchmarks/scripts/profile_deer_logreg_components.jl

@@ -35,7 +35,7 @@ function make_tape(rng::Random.AbstractRNG, x0::AbstractVector, dim::Int, t_len:
     FP = typeof(epsilon)
     return map(1:t_len) do _
         xi = copyto!(similar(x0, dim), randn(rng, FP, dim))
-        ParallelMCMC.MALATapeElement(xi, FP(rand(rng)))
+        return ParallelMCMC.MALATapeElement(xi, FP(rand(rng)))
     end
 end
 

benchmarks/ParallelMCMCBenchmarks/src/pr_suite.jl

@@ -282,7 +282,7 @@ end
 function write_results(path::AbstractString, results::Dict{String,Any})
     mkpath(dirname(path))
     open(path, "w") do io
-        TOML.print(io, results; sorted=true)
+        return TOML.print(io, results; sorted=true)
     end
     return path
 end
@@ -312,7 +312,7 @@ function write_markdown(path::AbstractString, results::Dict{String,Any})
                 " B |",
             )
         end
-        println(io)
+        return println(io)
     end
     return path
 end

docs/src/90-contributing.md

@@ -23,3 +23,11 @@ Otherwise, say what your proposed solution is and wait for a discussion around i
 
 If your solution involves code, tests, benchmarks, or documentation builds, check the [developer documentation](91-developer.md).
 Otherwise, you can use the GitHub interface directly to create your pull request.
+
+## Pull request checklist
+
+Before submitting a pull request, please make sure:
+
+- You have added tests for your feature or fix, where applicable.
+- You have formatted the code with `JuliaFormatter` using the version specified by the `test` project’s compatibility bounds.
+- All tests pass.

docs/src/assets/make_julia_deer_gif.jl

@@ -370,13 +370,13 @@ function main()
         animation_paths = vcat(frame_paths, fill(last(frame_paths), 8))
         tmp_output = joinpath(frame_dir, "julia_deer_posterior.gif")
         run(`$convert -delay 7 -loop 0 $animation_paths -layers Optimize $tmp_output`)
-        cp(tmp_output, output; force=true)
+        return cp(tmp_output, output; force=true)
     end
 
     final_error = maximum(abs.(last(iterates) .- final_trajectory))
     println("wrote ", output)
     println("last DEER metric: ", @sprintf("%.3g", last(metrics)))
-    println("max error vs sequential taped MALA: ", @sprintf("%.3g", final_error))
+    return println("max error vs sequential taped MALA: ", @sprintf("%.3g", final_error))
 end
 
 main()

ext/DynamicPPLExt.jl

@@ -31,9 +31,7 @@ chain = sample(model, AdaptiveMALASampler(0.3; n_warmup=500), 2_000;
                chain_type=FlexiChains.VNChain, discard_warmup=true, progress=true)
 ```
 """
-function ParallelMCMC.DensityModel(
-    turing_model::DynamicPPL.Model; ad_backend, hvp=nothing
-)
+function ParallelMCMC.DensityModel(turing_model::DynamicPPL.Model; ad_backend, hvp=nothing)
     # Sample in linked/unconstrained space and let DynamicPPL provide the gradient.
     ld = DynamicPPL.LogDensityFunction(
         turing_model,
@@ -126,7 +124,11 @@ for (Ttrans, Tspl, Tstate) in (
             chain_type::Type{SymChain};
             kwargs...,
         )
-            throw(ArgumentError("FlexiChains.SymChain is not supported for DynamicPPL models; please use VNChain instead."))
+            throw(
+                ArgumentError(
+                    "FlexiChains.SymChain is not supported for DynamicPPL models; please use VNChain instead.",
+                ),
+            )
         end
     end
 end
@@ -140,7 +142,7 @@ function ParallelMCMC._construct_flexichain(
 )
     pwss = map(enumerate(eachrow(vals))) do (i, val)
         stats = map(v -> v[i], internals)
-        DynamicPPL.ParamsWithStats(val, model.logdensity.ld, stats)
+        return DynamicPPL.ParamsWithStats(val, model.logdensity.ld, stats)
     end
     return AbstractMCMC.from_samples(VNChain, hcat(pwss))
 end

src/DEER/DEER.jl

@@ -203,7 +203,7 @@ struct ReverseOnGrad <: HVPStrategy end
 _strategy_from(::DI.ForwardOverAnything) = ForwardOnGrad()
 _strategy_from(::DI.HVPMode) = ReverseOnGrad()
 
-function _hvp_strategy(backend::Union{AbstractADType, DI.SecondOrder})
+function _hvp_strategy(backend::Union{AbstractADType,DI.SecondOrder})
     return _strategy_from(DI.hvp_mode(backend))
 end
 
@@ -320,9 +320,8 @@ end
 @inline _rademacher!(z::AbstractArray, rng::AbstractRNG, ::Nothing) = _rademacher!(z, rng)
 
 @inline _rademacher_matrix!(Z::AbstractMatrix, rng::AbstractRNG) = _rademacher!(Z, rng)
-@inline _rademacher_matrix!(Z::AbstractMatrix, rng::AbstractRNG, host) = _rademacher!(
-    Z, rng, host
-)
+@inline _rademacher_matrix!(Z::AbstractMatrix, rng::AbstractRNG, host) =
+    _rademacher!(Z, rng, host)
 
 function jac_diag_via_jvps(rec::TapedRecursion, x::AbstractVector, t::Int)
     D = length(x)

src/interface.jl

@@ -631,7 +631,11 @@ function _construct_flexichain(
             elseif n isa TKey || n isa Symbol
                 to_parameter(n)
             else
-                throw(ArgumentError("param_names must be a collection of Pairs, Symbols, or $TKey, got $(typeof(n))"))
+                throw(
+                    ArgumentError(
+                        "param_names must be a collection of Pairs, Symbols, or $TKey, got $(typeof(n))",
+                    ),
+                )
             end
         end
     end

test/Project.toml

@@ -8,6 +8,7 @@ Enzyme = "7da242da-08ed-463a-9acd-ee780be4f1d9"
 FlexiChains = "4a37a8b9-6e57-4b92-8664-298d46e639f7"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
 JET = "c3a54625-cd67-489e-a8e7-0a5a0ff4e31b"
+JuliaFormatter = "98e50ef6-434e-11e9-1051-2b60c6c9e899"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 LogDensityProblems = "6fdf6af0-433a-55f7-b3ed-c6c6e0b8df7c"
 LogDensityProblemsAD = "996a588d-648d-4e1f-a8f0-a84b347e47b1"
@@ -23,3 +24,6 @@ Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [extras]
 CUDA_Runtime_jll = "76a88914-d11a-5bdc-97e0-2f5a05c973a2"
+
+[compat]
+JuliaFormatter = "2.12"

test/test-AbstractMCMC-Interface.jl

@@ -241,7 +241,7 @@ gradlogp_iface(x) = -x
 
             @test chain isa SymChain
             @test FlexiChains.parameters(chain) == [:param]
-            @test size(chain[:param, stack=true]) == (50, 1, 2)
+            @test size(chain[:param, stack = true]) == (50, 1, 2)
         end
 
         @testset "with vector-valued varnames" begin
@@ -256,15 +256,19 @@ gradlogp_iface(x) = -x
 
             @test chain isa VNChain
             @test FlexiChains.parameters(chain) == [@varname(param)]
-            @test size(chain[@varname(param), stack=true]) == (50, 1, 2)
+            @test size(chain[@varname(param), stack = true]) == (50, 1, 2)
         end
     end
 
     @testset "invalid param_names throws" begin
         model = DensityModel(logp_iface, gradlogp_iface, 2)
         @test_throws "param_names must be a collection" sample(
-            model, MALASampler(0.15), 50;
-            chain_type=SymChain, progress=false, param_names=["mu", "sigma"],
+            model,
+            MALASampler(0.15),
+            50;
+            chain_type=SymChain,
+            progress=false,
+            param_names=["mu", "sigma"],
         )
     end
 

test/test-Adaptive-MALA.jl

@@ -183,12 +183,7 @@ end
     sampler = AdaptiveMALASampler(0.2; n_warmup=50)
 
     chain = sample(
-        MersenneTwister(1),
-        model,
-        sampler,
-        150;
-        chain_type=SymChain,
-        progress=false,
+        MersenneTwister(1), model, sampler, 150; chain_type=SymChain, progress=false
     )
 
     @test chain isa SymChain
@@ -219,12 +214,7 @@ end
     sampler = AdaptiveMALASampler(0.1; n_warmup=n_w)
 
     chain = sample(
-        MersenneTwister(3),
-        model,
-        sampler,
-        n_w + 50;
-        chain_type=VNChain,
-        progress=false,
+        MersenneTwister(3), model, sampler, n_w + 50; chain_type=VNChain, progress=false
     )
 
     # Filter by is_warmup flag to avoid off-by-one from the init transition.

test/test-Code-Quality.jl

@@ -1,7 +1,11 @@
 using Test
-using Aqua, JET
+using Aqua, JET, JuliaFormatter
 using ParallelMCMC
 
+@testset "Blue Formatting" begin
+    @test JuliaFormatter.format(ParallelMCMC; verbose=false, overwrite=false)
+end
+
 @testset "Aqua" begin
     Aqua.test_all(ParallelMCMC)
 end

test/test-DEER-Interface.jl

@@ -212,12 +212,7 @@ end
     sampler = ParallelMALASampler(0.05; T=16, backend=_AD)
 
     chain = sample(
-        MersenneTwister(1),
-        model,
-        sampler,
-        100;
-        chain_type=SymChain,
-        progress=false,
+        MersenneTwister(1), model, sampler, 100; chain_type=SymChain, progress=false
     )
 
     @test chain isa SymChain
@@ -229,7 +224,7 @@ end
     param_names = FlexiChains.parameters(chain)
     @test length(param_names) == 1
     name = only(param_names)
-    @test size(chain[name, stack=true], 3) == 2
+    @test size(chain[name, stack = true], 3) == 2
 end
 
 @testset "ParallelMALASampler bundle_samples fallback path (thinning)" begin
@@ -279,12 +274,7 @@ end
     sampler = ParallelMALASampler(0.1; T=32, damping=0.5, backend=_AD)
 
     chain = sample(
-        MersenneTwister(2025),
-        model,
-        sampler,
-        5_000;
-        chain_type=SymChain,
-        progress=false,
+        MersenneTwister(2025), model, sampler, 5_000; chain_type=SymChain, progress=false
     )
 
     burn = 500

test/test-DEER-Turing-Logistic.jl

@@ -110,7 +110,7 @@ end
     @test chain isa VNChain
     @test FlexiChains.niters(chain) == 400
     @test @varname(β) in FlexiChains.parameters(chain)
-    @test all(isfinite, chain[@varname(β), stack=true])
+    @test all(isfinite, chain[@varname(β), stack = true])
 end
 
 @testset "ParallelMALASampler Turing logistic: posterior sign correct" begin
@@ -136,7 +136,7 @@ end
         progress=false,
     )
 
-    post = chain[@varname(β), stack=true][201:end, :, :]
+    post = chain[@varname(β), stack = true][201:end, :, :]
     β_mean = vec(mean(post; dims=1))
 
     @test sign(β_mean[1]) == sign(_LR_β_true[1])
@@ -165,7 +165,7 @@ end
         progress=false,
         discard_warmup=true,
     )
-    β_mala = vec(mean(mala_chain[:x, stack=true]; dims=1))
+    β_mala = vec(mean(mala_chain[:x, stack = true]; dims=1))
 
     deer_chain = sample(
         MersenneTwister(42),
@@ -183,7 +183,7 @@ end
         chain_type=SymChain,
         progress=false,
     )
-    β_deer = vec(mean(deer_chain[:x, stack=true][201:end, :, :]; dims=1))
+    β_deer = vec(mean(deer_chain[:x, stack = true][201:end, :, :]; dims=1))
 
     @test abs(β_deer[1] - β_mala[1]) < 0.25
     @test abs(β_deer[2] - β_mala[2]) < 0.25

test/test-GPU-AD-HVP.jl

@@ -26,7 +26,6 @@ end
 if !_ADHVP_GPU_AVAILABLE
     @info "GPU AD-HVP test: CUDA not functional — skipping"
 else
-
     #=
     Multivariate Gaussian target with X'X/N perturbation:
       logp(β) = -0.5 (||β||^2 + ||Xβ||^2 / N)
@@ -39,7 +38,7 @@ else
     function _logp_single(β, X)
         Xβ = pmcmc_matmul(X, β)
         N = oftype(zero(eltype(β)), size(X, 1))
-        -oftype(zero(eltype(β)), 0.5) * (sum(abs2, β) + sum(abs2, Xβ) / N)
+        return -oftype(zero(eltype(β)), 0.5) * (sum(abs2, β) + sum(abs2, Xβ) / N)
     end
 
     function _gradlogp_single(β, X)
@@ -54,8 +53,8 @@ else
     function _logp_batch(B, X)
         XB = pmcmc_matmul(X, B)
         N = oftype(zero(eltype(B)), size(X, 1))
-        -oftype(zero(eltype(B)), 0.5) .*
-        (vec(sum(abs2, B; dims=1)) .+ vec(sum(abs2, XB; dims=1)) ./ N)
+        return -oftype(zero(eltype(B)), 0.5) .*
+               (vec(sum(abs2, B; dims=1)) .+ vec(sum(abs2, XB; dims=1)) ./ N)
     end
 
     function _gradlogp_batch(B, X)

test/test-GPU-Performance.jl

@@ -29,7 +29,6 @@ end
 if !_PERF_GPU_AVAILABLE
     @info "GPU performance test: CUDA not functional — skipping"
 else
-
     #=
     Multivariate Gaussian target — well-conditioned, optimal MALA acceptance from
     any start. Lets ε be set analytically so the chain actually moves and DEER

test/test-Jacobian-Estimator.jl

@@ -23,13 +23,13 @@ Base.axes(x::TaggedVector) = axes(x.data)
 Base.getindex(x::TaggedVector, i::Int) = x.data[i]
 Base.setindex!(x::TaggedVector, v, i::Int) = setindex!(x.data, v, i)
 function Base.similar(x::TaggedVector, ::Type{T}, dims::Dims{1}) where {T}
-    TaggedVector(Vector{T}(undef, dims[1]))
+    return TaggedVector(Vector{T}(undef, dims[1]))
 end
 function Base.similar(x::TaggedVector, ::Type{T}, n::Int) where {T}
-    TaggedVector(Vector{T}(undef, n))
+    return TaggedVector(Vector{T}(undef, n))
 end
 function Base.similar(x::TaggedVector, dims::Dims{1})
-    TaggedVector(Vector{eltype(x)}(undef, dims[1]))
+    return TaggedVector(Vector{eltype(x)}(undef, dims[1]))
 end
 Base.similar(x::TaggedVector, n::Int) = TaggedVector(Vector{eltype(x)}(undef, n))
 Base.copy(x::TaggedVector) = TaggedVector(copy(x.data))