Inpromve basic coverage

BatyLeo · BatyLeo · commit d32b1e813be1 · 2025-08-21T16:20:45.000+02:00
diff --git a/docs/src/api/0_interface.md b/docs/src/api/0_interface.md
@@ -1,3 +1,7 @@
+```@meta
+CollapsedDocStrings = true
+```
+
 # Interface
 
 ## Public
diff --git a/docs/src/api/argmax.md b/docs/src/api/argmax.md
@@ -1,3 +1,7 @@
+```@meta
+CollapsedDocStrings = true
+```
+
 # Argmax
 
 ## Public
diff --git a/docs/src/api/argmax_2d.md b/docs/src/api/argmax_2d.md
@@ -1,3 +1,7 @@
+```@meta
+CollapsedDocStrings = true
+```
+
 # Argmax2D
 
 ## Public
diff --git a/docs/src/api/dvsp.md b/docs/src/api/dvsp.md
@@ -1,3 +1,7 @@
+```@meta
+CollapsedDocStrings = true
+```
+
 # Dynamic Vehicle Scheduling
 
 ## Public
diff --git a/docs/src/api/dynamic_assorment.md b/docs/src/api/dynamic_assorment.md
@@ -1,3 +1,7 @@
+```@meta
+CollapsedDocStrings = true
+```
+
 # Dynamic Assortment
 
 ## Public
diff --git a/docs/src/api/fixed_shortest_path.md b/docs/src/api/fixed_shortest_path.md
@@ -1,3 +1,7 @@
+```@meta
+CollapsedDocStrings = true
+```
+
 # Fixed-size shortest path
 
 ## Public
diff --git a/docs/src/api/portfolio_optimization.md b/docs/src/api/portfolio_optimization.md
@@ -1,3 +1,7 @@
+```@meta
+CollapsedDocStrings = true
+```
+
 # Subset selection
 
 ## Public
diff --git a/docs/src/api/ranking.md b/docs/src/api/ranking.md
@@ -1,3 +1,7 @@
+```@meta
+CollapsedDocStrings = true
+```
+
 # Ranking
 
 ## Public
diff --git a/docs/src/api/subset_selection.md b/docs/src/api/subset_selection.md
@@ -1,3 +1,7 @@
+```@meta
+CollapsedDocStrings = true
+```
+
 # Subset selection
 
 ## Public
diff --git a/docs/src/api/vsp.md b/docs/src/api/vsp.md
@@ -1,3 +1,7 @@
+```@meta
+CollapsedDocStrings = true
+```
+
 # Stochastic Vehicle Scheduling
 
 ## Public
diff --git a/docs/src/api/warcraft.md b/docs/src/api/warcraft.md
@@ -1,3 +1,7 @@
+```@meta
+CollapsedDocStrings = true
+```
+
 # Warcraft
 
 ## Public
diff --git a/src/DecisionFocusedLearningBenchmarks.jl b/src/DecisionFocusedLearningBenchmarks.jl
@@ -3,22 +3,22 @@ module DecisionFocusedLearningBenchmarks
 using DataDeps
 using Requires: @require
 
-function _euro_neurips_unpack(local_filepath)
-    directory = dirname(local_filepath)
-    unpack(local_filepath)
-    # Move instances and delete the rest
-    for filepath in readdir(
-        joinpath(directory, "euro-neurips-vrp-2022-quickstart-main", "instances"); join=true
-    )
-        if endswith(filepath, ".txt")
-            mv(filepath, joinpath(directory, basename(filepath)))
+function __init__()
+    function _euro_neurips_unpack(local_filepath)
+        directory = dirname(local_filepath)
+        unpack(local_filepath)
+        # Move instances and delete the rest
+        for filepath in readdir(
+            joinpath(directory, "euro-neurips-vrp-2022-quickstart-main", "instances");
+            join=true,
+        )
+            if endswith(filepath, ".txt")
+                mv(filepath, joinpath(directory, basename(filepath)))
+            end
         end
+        rm(joinpath(directory, "euro-neurips-vrp-2022-quickstart-main"); recursive=true)
+        return nothing
     end
-    rm(joinpath(directory, "euro-neurips-vrp-2022-quickstart-main"); recursive=true)
-    return nothing
-end
-
-function __init__()
     # Register the Warcraft dataset
     ENV["DATADEPS_ALWAYS_ACCEPT"] = "true"
     register(
@@ -64,7 +64,7 @@ using .Utils
 export AbstractBenchmark, AbstractStochasticBenchmark, AbstractDynamicBenchmark, DataSample
 export AbstractEnvironment, get_seed, is_terminated, observe, reset!, step!
 
-export Policy, run_policy!
+export Policy, evaluate_policy!
 
 export generate_sample, generate_dataset, generate_environments, generate_environment
 export generate_scenario
diff --git a/src/DynamicAssortment/environment.jl b/src/DynamicAssortment/environment.jl
@@ -43,7 +43,7 @@ function Environment(instance::Instance; seed=0, rng::AbstractRNG=MersenneTwiste
         features=full_features,
         d_features=zeros(2, N),
     )
-    Utils.reset!(env; reset_seed=true)
+    Utils.reset!(env; reset_rng=true)
     return env
 end
 
@@ -58,8 +58,8 @@ prices(b::Environment) = b.instance.prices
 ## Basic operations of environment
 
 # Reset the environment
-function Utils.reset!(env::Environment; reset_seed=false, seed=env.seed)
-    reset_seed && Random.seed!(env.rng, seed)
+function Utils.reset!(env::Environment; reset_rng=false, seed=env.seed)
+    reset_rng && Random.seed!(env.rng, seed)
 
     env.step = 1
 
diff --git a/src/DynamicVehicleScheduling/anticipative_solver.jl b/src/DynamicVehicleScheduling/anticipative_solver.jl
@@ -49,8 +49,9 @@ function anticipative_solver(
     two_dimensional_features=env.instance.two_dimensional_features,
     reset_env=true,
     nb_epochs=typemax(Int),
+    seed=get_seed(env),
 )
-    reset_env && reset!(env; reset_seed=true)
+    reset_env && reset!(env; reset_rng=true, seed)
 
     start_epoch = current_epoch(env)
     end_epoch = min(last_epoch(env), start_epoch + nb_epochs - 1)
diff --git a/src/DynamicVehicleScheduling/environment.jl b/src/DynamicVehicleScheduling/environment.jl
@@ -70,10 +70,10 @@ Utils.is_terminated(env::DVSPEnv) = current_epoch(env) > last_epoch(env)
 $TYPEDSIGNATURES
 
 Reset the environment to its initial state.
-Also reset the seed if `reset_seed` is set to true.
+Also reset the rng to `seed` if `reset_rng` is set to true.
 """
-function Utils.reset!(env::DVSPEnv; seed=get_seed(env), reset_seed=false)
-    if reset_seed
+function Utils.reset!(env::DVSPEnv; seed=get_seed(env), reset_rng=false)
+    if reset_rng
         Random.seed!(env.rng, seed)
     end
     env.scenario = Utils.generate_scenario(env; rng=env.rng)
diff --git a/src/PortfolioOptimization/PortfolioOptimization.jl b/src/PortfolioOptimization/PortfolioOptimization.jl
@@ -105,40 +105,6 @@ end
 """
 $TYPEDSIGNATURES
 
-Generate a dataset of labeled instances for the portfolio optimization problem.
-"""
-function Utils.generate_dataset(
-    bench::PortfolioOptimizationBenchmark,
-    dataset_size::Int=10;
-    seed::Int=0,
-    type::Type=Float32,
-)
-    (; d, p, deg, ν, L, f) = bench
-    rng = MersenneTwister(seed)
-
-    # Features
-    features = [randn(rng, type, p) for _ in 1:dataset_size]
-
-    # True weights
-    B = rand(rng, Bernoulli(0.5), d, p)
-    c̄ = [
-        (0.05 / type(sqrt(p)) .* B * features[i] .+ 0.1^(1 / deg)) .^ deg for
-        i in 1:dataset_size
-    ]
-    costs = [c̄ᵢ .+ L * f .+ 0.01 .* ν .* randn(rng, type, d) for c̄ᵢ in c̄]
-
-    maximizer = Utils.generate_maximizer(bench)
-    solutions = maximizer.(costs)
-
-    return [
-        DataSample(; x, θ_true, y_true) for
-        (x, θ_true, y_true) in zip(features, costs, solutions)
-    ]
-end
-
-"""
-$TYPEDSIGNATURES
-
 Initialize a linear model for `bench` using `Flux`.
 """
 function Utils.generate_statistical_model(bench::PortfolioOptimizationBenchmark)
diff --git a/src/Utils/Utils.jl b/src/Utils/Utils.jl
@@ -21,7 +21,7 @@ include("misc.jl")
 include("model_builders.jl")
 
 export DataSample, Policy
-export run_policy!
+export evaluate_policy!
 export TopKMaximizer
 
 export AbstractEnvironment, get_seed, is_terminated, observe, reset!, step!
diff --git a/src/Utils/environment.jl b/src/Utils/environment.jl
@@ -34,10 +34,10 @@ This function should return a tuple of two elements:
 function observe end
 
 """
-    reset!(env::AbstractEnvironment; reset_seed::Bool, seed=get_seed(env)) --> Nothing
+    reset!(env::AbstractEnvironment; reset_rng::Bool, seed=get_seed(env)) --> Nothing
 
 Reset the environment to its initial state.
-If `reset_seed` is true, the random number generator is reset to the given `seed`.
+If `reset_rng` is true, the random number generator is reset to the given `seed`.
 """
 function reset! end
 
diff --git a/src/Utils/policy.jl b/src/Utils/policy.jl
@@ -31,9 +31,9 @@ $TYPEDSIGNATURES
 Run the policy on the environment and return the total reward and a dataset of observations.
 By default, the environment is reset before running the policy.
 """
-function run_policy!(policy, env::AbstractEnvironment; kwargs...)
+function evaluate_policy!(policy, env::AbstractEnvironment; kwargs...)
     total_reward = 0.0
-    reset!(env; reset_seed=false)
+    reset!(env; reset_rng=false)
     local labeled_dataset
     while !is_terminated(env)
         y = policy(env; kwargs...)
@@ -52,35 +52,49 @@ function run_policy!(policy, env::AbstractEnvironment; kwargs...)
     return total_reward, labeled_dataset
 end
 
-function run_policy!(policy, envs::Vector{<:AbstractEnvironment}; kwargs...)
-    E = length(envs)
-    rewards = zeros(Float64, E)
-    datasets = map(1:E) do e
-        reward, dataset = run_policy!(policy, envs[e]; kwargs...)
-        rewards[e] = reward
-        return dataset
-    end
-    return rewards, vcat(datasets...)
-end
+# function evaluate_policy!(policy, envs::Vector{<:AbstractEnvironment}; kwargs...)
+#     E = length(envs)
+#     rewards = zeros(Float64, E)
+#     datasets = map(1:E) do e
+#         reward, dataset = evaluate_policy!(policy, envs[e]; kwargs...)
+#         rewards[e] = reward
+#         return dataset
+#     end
+#     return rewards, vcat(datasets...)
+# end
+
+"""
+$TYPEDSIGNATURES
 
-function run_policy!(
-    policy, env::AbstractEnvironment, episodes::Int; seed=get_seed(env), kwargs...
+Evaluate the policy on the environment and return the total reward and a dataset of observations.
+By default, the environment is reset before running the policy.
+"""
+function evaluate_policy!(
+    policy, env::AbstractEnvironment, episodes::Int=1; seed=get_seed(env), kwargs...
 )
-    reset!(env; reset_seed=true, seed)
+    reset!(env; reset_rng=true, seed)
     total_reward = 0.0
     datasets = map(1:episodes) do _i
-        reward, dataset = run_policy!(policy, env; kwargs...)
+        reward, dataset = evaluate_policy!(policy, env; kwargs...)
         total_reward += reward
         return dataset
     end
     return total_reward / episodes, vcat(datasets...)
 end
 
-function run_policy!(policy, envs::Vector{<:AbstractEnvironment}, episodes::Int; kwargs...)
+"""
+$TYPEDSIGNATURES
+
+Run the policy on the environments and return the total rewards and a dataset of observations.
+By default, the environments are reset before running the policy.
+"""
+function evaluate_policy!(
+    policy, envs::Vector{<:AbstractEnvironment}, episodes::Int=1; kwargs...
+)
     E = length(envs)
     rewards = zeros(Float64, E)
     datasets = map(1:E) do e
-        reward, dataset = run_policy!(policy, envs[e], episodes; kwargs...)
+        reward, dataset = evaluate_policy!(policy, envs[e], episodes; kwargs...)
         rewards[e] = reward
         return dataset
     end
diff --git a/test/dynamic_vsp.jl b/test/dynamic_vsp.jl
@@ -1,9 +1,14 @@
 @testitem "DVSP" begin
     using DecisionFocusedLearningBenchmarks.DynamicVehicleScheduling
+    using Statistics: mean
 
     b = DynamicVehicleSchedulingBenchmark(; two_dimensional_features=true)
+
+    @test is_exogenous(b)
+    @test !is_endogenous(b)
+
     dataset = generate_dataset(b, 10)
-    environments = generate_environments(b, dataset)
+    environments = generate_environments(b, dataset; seed=0)
 
     env = environments[1]
     get_seed(env)
@@ -12,10 +17,12 @@
     lazy = policies[1]
     greedy = policies[2]
 
-    d = run_policy!(lazy, env, 1; seed=0)[2]
+    d = evaluate_policy!(lazy, env, 1; seed=0)[2]
+
+    r_lazy, d = evaluate_policy!(lazy, environments, 10)
+    r_greedy, d = evaluate_policy!(greedy, environments, 10)
 
-    r, d = run_policy!(lazy, environments, 10)
-    r, d = run_policy!(greedy, environments, 10)
+    @test mean(r_lazy) <= mean(r_greedy)
 
     env = environments[1]
     instance = dataset[1].instance
diff --git a/test/utils.jl b/test/utils.jl
@@ -23,3 +23,35 @@
         @test max(h, w) <= length(path) <= h + w
     end
 end
+
+@testitem "DataSample" begin
+    using DecisionFocusedLearningBenchmarks
+    using StableRNGs
+
+    rng = StableRNG(1234)
+
+    function random_sample()
+        return DataSample(;
+            x=randn(rng, 10, 5),
+            θ_true=rand(rng, 5),
+            y_true=rand(rng, 10),
+            instance="this is an instance",
+        )
+    end
+
+    sample = random_sample()
+    @test sample isa DataSample
+
+    io = IOBuffer()
+    show(io, sample)
+    @test String(take!(io)) ==
+        "DataSample(x=$(sample.x), θ_true=$(sample.θ_true), y_true=$(sample.y_true), instance=$(sample.instance))"
+end
+
+@testitem "Maximizers" begin
+    using DecisionFocusedLearningBenchmarks.Utils: TopKMaximizer
+    top_k = TopKMaximizer(3)
+    @test top_k([1, 2, 3, 4, 5]) == [0, 0, 1, 1, 1]
+    @test top_k([5, 4, 3, 2, 1]) == [1, 1, 1, 0, 0]
+    @test_throws(AssertionError, top_k([1, 2]))
+end
