better seed handling

Author: BatyLeo

src/Utils/data_sample.jl

@@ -4,6 +4,7 @@ $TYPEDEF
 Data sample data structure.
 Its main purpose is to store datasets generated by the benchmarks.
 It has 3 main (optional) fields: features `x`, cost parameters `θ`, and solution `y`.
+Currently, all three are restricted to `AbstractArray` or `nothing`.
 Additionally, it has a `context` field (solver and scenario-generation context, spread into the
 maximizer as `maximizer(θ; sample.context...)`) and an `extra` field (non-solver data, never passed
 to the maximizer).
@@ -131,11 +132,11 @@ function Base.show(io::IO, d::DataSample)
     end
     if !isnothing(d.θ)
         θ_str = sprint(show, d.θ; context=io_limited)
-        push!(fields, "θ_true=$θ_str")
+        push!(fields, "θ=$θ_str")
     end
     if !isnothing(d.y)
         y_str = sprint(show, d.y; context=io_limited)
-        push!(fields, "y_true=$y_str")
+        push!(fields, "y=$y_str")
     end
     for (key, value) in pairs(d.context)
         value_str = sprint(show, value; context=io_limited)
@@ -166,7 +167,7 @@ Transform the features in the dataset.
 function StatsBase.transform(t, dataset::AbstractVector{<:DataSample})
     return map(dataset) do d
         (; context, extra, x, θ, y) = d
-        DataSample(StatsBase.transform(t, x), θ, y, context, extra)
+        DataSample(; x=StatsBase.transform(t, x), θ, y, context..., extra)
     end
 end
 

src/Utils/interface.jl

@@ -83,7 +83,6 @@ function generate_dataset(
     rng=MersenneTwister(seed),
     kwargs...,
 )
-    Random.seed!(rng, seed)
     return [
         begin
             sample = generate_sample(bench, rng; kwargs...)
@@ -422,7 +421,6 @@ function generate_dataset(
     rng=MersenneTwister(seed),
     kwargs...,
 )
-    Random.seed!(rng, seed)
     return reduce(
         vcat,
         (
@@ -500,7 +498,6 @@ function generate_dataset(
     rng=MersenneTwister(seed),
     kwargs...,
 )
-    Random.seed!(rng, seed)
     return reduce(
         vcat, (generate_sample(saa, rng; target_policy, kwargs...) for _ in 1:nb_instances)
     )
@@ -561,6 +558,20 @@ meaning (whether uncertainty is independent of decisions).
 """
 abstract type AbstractDynamicBenchmark{exogenous} <: AbstractStochasticBenchmark{exogenous} end
 
+"""
+$TYPEDSIGNATURES
+
+Intercepts accidental calls to `generate_sample` on dynamic benchmarks and throws a
+descriptive error pointing at the correct entry point.
+"""
+function generate_sample(bench::AbstractDynamicBenchmark, rng; kwargs...)
+    return error(
+        "`generate_sample` is not supported for dynamic benchmarks ($(typeof(bench))). " *
+        "Use `generate_environments` and " *
+        "`generate_dataset(bench, environments; target_policy=...)` instead.",
+    )
+end
+
 "Alias for [`AbstractDynamicBenchmark`](@ref)`{true}`. Uncertainty is independent of decisions."
 const ExogenousDynamicBenchmark = AbstractDynamicBenchmark{true}
 
@@ -591,7 +602,6 @@ function generate_environments(
     rng=MersenneTwister(seed),
     kwargs...,
 )
-    Random.seed!(rng, seed)
     return [generate_environment(bench, rng; kwargs...) for _ in 1:n]
 end
 
@@ -612,14 +622,8 @@ to obtain standard baseline callables (e.g. the anticipative solver).
 - `rng`: random number generator.
 """
 function generate_dataset(
-    bench::ExogenousDynamicBenchmark,
-    environments::AbstractVector;
-    target_policy,
-    seed=nothing,
-    rng=MersenneTwister(seed),
-    kwargs...,
+    bench::ExogenousDynamicBenchmark, environments::AbstractVector; target_policy, kwargs...
 )
-    Random.seed!(rng, seed)
     return reduce(vcat, (target_policy(env) for env in environments))
 end
 

test/dynamic_assortment.jl

@@ -321,14 +321,14 @@ end
 
     # vector-of-environments overload
     dataset = generate_dataset(b, envs; target_policy=target_policy)
-    @test dataset isa Vector{D} where D <: DataSample
+    @test dataset isa Vector{D} where {D<:DataSample}
     @test !isempty(dataset)
     @test all(!isnothing(s.x) for s in dataset)
     @test all(!isnothing(s.y) for s in dataset)
 
     # count-based wrapper
     dataset2 = generate_dataset(b, 3; seed=7, target_policy=target_policy)
-    @test dataset2 isa Vector{D} where D <: DataSample
+    @test dataset2 isa Vector{D} where {D<:DataSample}
     @test !isempty(dataset2)
 end
 

test/dynamic_vsp.jl

@@ -63,14 +63,14 @@ end
 
     # vector-of-environments overload
     dataset = generate_dataset(b, envs; target_policy=target_policy)
-    @test dataset isa Vector{D} where D <: DataSample
+    @test dataset isa Vector{D} where {D<:DataSample}
     @test !isempty(dataset)
     @test all(!isnothing(s.x) for s in dataset)
     @test all(!isnothing(s.y) for s in dataset)
 
     # count-based wrapper
     dataset2 = generate_dataset(b, 3; seed=1, target_policy=target_policy)
-    @test dataset2 isa Vector{D} where D <: DataSample
+    @test dataset2 isa Vector{D} where {D<:DataSample}
     @test !isempty(dataset2)
 
     # seed keyword is forwarded: same seed → same dataset

test/utils.jl

@@ -54,8 +54,8 @@ end
     show(io, sample)
     s = String(take!(io))
     @test occursin("DataSample(", s)
-    @test occursin("θ_true", s)
-    @test occursin("y_true", s)
+    @test occursin("θ", s)
+    @test occursin("y", s)
     @test occursin("instance=\"this is an instance\"", s)
 
     @test propertynames(sample) == (:x, :θ, :y, :context, :extra, :instance)