Fix tests and deprecation

Author: BatyLeo

src/DecisionFocusedLearningBenchmarks.jl

@@ -73,7 +73,7 @@ export generate_scenario
 export generate_baseline_policies
 export generate_statistical_model
 export generate_maximizer
-export generate_anticipative_solution
+export generate_anticipative_solver, generate_parametric_anticipative_solver
 export is_exogenous, is_endogenous
 
 export objective_value

src/DynamicVehicleScheduling/DynamicVehicleScheduling.jl

@@ -116,13 +116,13 @@ end
 $TYPEDSIGNATURES
 
 Return the anticipative solver for the dynamic vehicle scheduling benchmark.
-The callable takes a scenario and solver kwargs (including `instance`) and returns a
-training trajectory as a `Vector{DataSample}`.
+The callable takes an environment and solver kwargs and returns a training trajectory
+as a `Vector{DataSample}`. Set `reset_env=true` (default) to reset the environment
+before solving, or `reset_env=false` to plan from the current state.
 """
 function Utils.generate_anticipative_solver(::DynamicVehicleSchedulingBenchmark)
-    return (scenario; instance, kwargs...) -> begin
-        env = DVSPEnv(instance, scenario)
-        _, trajectory = anticipative_solver(env; reset_env=false, kwargs...)
+    return (env; reset_env=true, kwargs...) -> begin
+        _, trajectory = anticipative_solver(env; reset_env, kwargs...)
         return trajectory
     end
 end

src/Utils/Utils.jl

@@ -32,7 +32,7 @@ export generate_statistical_model, generate_maximizer
 export generate_scenario
 export generate_environment, generate_environments
 export generate_baseline_policies
-export generate_anticipative_solution
+export generate_anticipative_solver, generate_parametric_anticipative_solver
 
 export plot_data, compute_gap
 export grid_graph, get_path, path_to_matrix

src/Utils/interface.jl

@@ -262,18 +262,15 @@ spread from `sample.context`:
 function generate_scenario end
 
 """
-    generate_anticipative_solver(::AbstractStochasticBenchmark{true}) -> callable
+    generate_anticipative_solver(::AbstractBenchmark) -> callable
 
-Return a callable that computes the anticipative solution for a given scenario.
-The instance and other solver-relevant fields are spread from the sample context.
+Return a callable that computes the anticipative solution.
 
 - For [`AbstractStochasticBenchmark`](@ref): returns `(scenario; context...) -> y`.
 - For [`AbstractDynamicBenchmark`](@ref): returns
-  `(scenario; context...) -> Vector{DataSample}` — a full training trajectory.
-
-    solver = generate_anticipative_solver(bench)
-    y          = solver(scenario; sample.context...)  # stochastic
-    trajectory = solver(scenario; sample.context...)  # dynamic
+  `(env; reset_env=true, kwargs...) -> Vector{DataSample}`, a full training trajectory.
+  `reset_env=true` resets the env before solving (initial dataset building);
+  `reset_env=false` starts from the current env state.
 """
 function generate_anticipative_solver end
 
@@ -287,16 +284,6 @@ parametric anticipative subproblem:
 """
 function generate_parametric_anticipative_solver end
 
-"""
-    generate_anticipative_solution(::AbstractStochasticBenchmark, instance, scenario; kwargs...)
-
-!!! warning "Deprecated"
-    Use [`generate_anticipative_solver`](@ref) instead, which returns a callable
-    `(scenario; kwargs...) -> y` consistent with the [`generate_maximizer`](@ref)
-    convention.
-"""
-function generate_anticipative_solution end
-
 """
 $TYPEDSIGNATURES
 

test/dynamic_vsp.jl

@@ -25,8 +25,7 @@
     @test mean(r_lazy) <= mean(r_greedy)
 
     env = environments[1]
-    scenario = env.scenario
-    v, y = generate_anticipative_solution(b, env, scenario; nb_epochs=2, reset_env=true)
+    y = generate_anticipative_solver(b)(env; nb_epochs=2)
 
     maximizer = generate_maximizer(b)
 
@@ -44,11 +43,10 @@
     @test size(x, 1) == 2
     @test size(x2, 1) == 27
 
-    anticipative_value, solution = generate_anticipative_solution(b, env; reset_env=true)
+    solution = generate_anticipative_solver(b)(env)
     reset!(env; reset_rng=true)
     cost = sum(step!(env, sample.y) for sample in solution)
     cost2 = sum(sample.reward for sample in solution)
-    @test isapprox(cost, anticipative_value; atol=1e-5)
     @test isapprox(cost, cost2; atol=1e-5)
 end
 

test/dynamic_vsp_plots.jl

@@ -12,7 +12,7 @@
     @test fig1 isa Plots.Plot
 
     scenario = env.scenario
-    v, y = generate_anticipative_solution(b, env, scenario; nb_epochs=3, reset_env=true)
+    y = generate_anticipative_solver(b)(env; nb_epochs=3)
 
     fig2 = DVSP.plot_epochs(y)
     @test fig2 isa Plots.Plot