Add check_state_kwargs, check_generator_kwargs in optimize

This is a more comprehensive ways on tweaking the check of arguments.

Author: goerz

src/optimize.jl

@@ -31,9 +31,8 @@ result = optimize(problem; method=Krotov)
 
 If `check` is true (default), the `initial_state` and `generator` of each
 trajectory is checked with [`check_state`](@ref) and [`check_generator`](@ref).
-Any other keyword argument temporarily overrides the corresponding keyword
-argument in [`problem`](@ref ControlProblem). These arguments are available to
-the optimizer, see each optimization package's documentation for details.
+Additional keyword arguments can be passed to these two `check` routines by
+passing a named-tuple as `check_state_kwargs` and `check_generator_kwargs`.
 
 The `callback` can be given as a function to be called after each iteration in
 order to analyze the progress of the optimization or to modify the state of
@@ -54,7 +53,8 @@ the method-specific [`make_print_iters`](@ref) to print the progress of the
 optimization after each iteration. This automatic callback runs after any
 manually given `callback`.
 
-All remaining keyword argument are method-specific.
+All remaining keyword argument are method-specific and temporarily overrides
+the corresponding keyword argument in [`problem`](@ref ControlProblem).
 To obtain the documentation for which options a particular method uses, run,
 e.g.,
 
@@ -75,10 +75,8 @@ function optimize(
     check = get(problem.kwargs, :check, true),
     print_iters = get(problem.kwargs, :print_iters, true),
     callback = get(problem.kwargs, :callback, nothing),
-    for_expval = true, # undocumented
-    for_pwc = true,  # undocumented
-    for_time_continuous = false,  # undocumented
-    for_parameterization = false, # undocumented
+    check_state_kwargs = get(problem.kwargs, :check_state_kwargs, (;)),
+    check_generator_kwargs = get(problem.kwargs, :check_generator_kwargs, (;)),
     kwargs...
 )
 
@@ -113,20 +111,16 @@ function optimize(
     )
 
     if check
-        # TODO: checks will have to be method-dependent, and then we may not
-        # need all the `for_...` keyword arguments
+        # TODO: checks maybe should be method-dependent
         for (i, traj) in enumerate(problem.trajectories)
-            if !check_state(traj.initial_state)
+            if !check_state(traj.initial_state; check_state_kwargs...)
                 error("The `initial_state` of trajectory $i is not valid")
             end
             if !check_generator(
                 traj.generator;
+                check_generator_kwargs...,
                 state = traj.initial_state,
                 tlist = problem.tlist,
-                for_expval,
-                for_pwc,
-                for_time_continuous,
-                for_parameterization,
             )
                 error("The `generator` of trajectory $i is not valid")
             end

test/test_optimize_or_load.jl

@@ -5,6 +5,44 @@ using QuantumControl: @optimize_or_load, load_optimization, save_optimization, o
 using QuantumControlTestUtils.DummyOptimization:
     dummy_control_problem, DummyOptimizationResult
 
+
+@testset "check_state_kwargs and check_generator_kwargs" begin
+
+    problem = dummy_control_problem()
+    outdir = mktempdir()
+    outfile = joinpath(outdir, "optimization_check_kwargs.jld2")
+
+    # Smoke test: extra kwargs are accepted
+    captured = IOCapture.capture(passthrough = false) do
+        @optimize_or_load(
+            outfile,
+            problem;
+            method = :dummymethod,
+            force = true,
+            check_state_kwargs = (; atol = 1e-10),
+            check_generator_kwargs = (; atol = 1e-10),
+        )
+    end
+    @test captured.value isa DummyOptimizationResult
+    @test captured.value.converged
+
+    # Test with teeth: for_time_continuous=true fails check_generator for the
+    # pwc dummy controls (array-based controls cannot be evaluated at a
+    # continuous time t)
+    captured2 = IOCapture.capture(rethrow = Union{}, passthrough = false) do
+        @optimize_or_load(
+            outfile,
+            problem;
+            method = :dummymethod,
+            force = true,
+            check_generator_kwargs = (; for_time_continuous = true),
+        )
+    end
+    @test captured2.value isa Exception
+
+end
+
+
 @testset "metadata" begin
 
     problem = dummy_control_problem()