Fix problem of LargeNeighborhoodSearchRCPSP solver :

- problem was happening when trying to call the solver in an instance without addditional constraint and the notebook was failing (silently)
- the bug is now fixed and a unit test is added

discrete_optimization/generic_rcpsp_tools/large_neighborhood_search_scheduling.py

@@ -287,15 +287,20 @@ def build_constraint_handler(rcpsp_problem: ANY_RCPSP, graph, **kwargs):
             graph=graph,
             fraction_subproblem=kwargs.get("fraction_subproblem", 0.05),
         )
-        n3 = NeighborConstraintBreaks(
-            problem=rcpsp_problem,
-            graph=graph,
-            fraction_subproblem=kwargs.get("fraction_subproblem", 0.05),
-            other_constraint_handler=n1,
-        )
-        n_mix = NeighborBuilderMix(
-            list_neighbor=[n1, n2, n3], weight_neighbor=[0.2, 0.5, 0.3]
-        )
+        if rcpsp_problem.includes_special_constraint():
+            n3 = NeighborConstraintBreaks(
+                problem=rcpsp_problem,
+                graph=graph,
+                fraction_subproblem=kwargs.get("fraction_subproblem", 0.05),
+                other_constraint_handler=n1,
+            )
+            n_mix = NeighborBuilderMix(
+                list_neighbor=[n1, n2, n3], weight_neighbor=[0.2, 0.5, 0.3]
+            )
+        else:
+            n_mix = NeighborBuilderMix(
+                list_neighbor=[n1, n2], weight_neighbor=[0.5, 0.5]
+            )
         basic_constraint_builder = BasicConstraintBuilder(
             neighbor_builder=n_mix,
             preemptive=kwargs.get("preemptive", False),

discrete_optimization/generic_rcpsp_tools/neighbor_tools_rcpsp.py

@@ -1099,7 +1099,7 @@ def __init__(
     ):
         if (
             not hasattr(problem, "do_special_constraints")
-            or not problem.do_special_constraints
+            or not problem.includes_special_constraint()
         ):
             raise ValueError(
                 "NeighborConstraintBreaks is meant for problems with special constraints"

examples/rcpsp/rcpsp_lns_example.py

@@ -0,0 +1,37 @@
+#  Copyright (c) 2024 AIRBUS and its affiliates.
+#  This source code is licensed under the MIT license found in the
+#  LICENSE file in the root directory of this source tree.
+import logging
+
+from discrete_optimization.generic_rcpsp_tools.large_neighborhood_search_scheduling import (
+    LargeNeighborhoodSearchScheduling,
+)
+from discrete_optimization.generic_tools.cp_tools import ParametersCP
+from discrete_optimization.rcpsp.rcpsp_model import RCPSPModel
+from discrete_optimization.rcpsp.rcpsp_parser import get_data_available, parse_file
+
+logging.basicConfig(level=logging.INFO)
+
+
+def example_lns_solver():
+    files_available = get_data_available()
+    file = [f for f in files_available if "j1201_1.sm" in f][0]
+    rcpsp_problem: RCPSPModel = parse_file(file)
+    solver = LargeNeighborhoodSearchScheduling(problem=rcpsp_problem)
+    parameters_cp = ParametersCP.default()
+    parameters_cp.time_limit_iter0 = 5
+    parameters_cp.time_limit = 2
+    results = solver.solve(
+        nb_iteration_lns=100,
+        skip_first_iteration=False,
+        stop_first_iteration_if_optimal=False,
+        parameters_cp=parameters_cp,
+        nb_iteration_no_improvement=200,
+        max_time_seconds=100,
+    )
+    sol, fit = results.get_best_solution_fit()
+    assert rcpsp_problem.satisfy(sol)
+
+
+if __name__ == "__main__":
+    example_lns_solver()

notebooks/RCPSP tutorials/RCPSP-6 Large Neighbourhood Search .ipynb

@@ -225,14 +225,20 @@
    "outputs": [],
    "source": [
     "# Different constraint handler methods\n",
+    "from discrete_optimization.generic_rcpsp_tools.neighbor_builder import (\n",
+    "    ObjectiveSubproblem,\n",
+    "    ParamsConstraintBuilder,\n",
+    "    mix_lot,\n",
+    ")\n",
+    "\n",
     "constraint_handler = mix_lot(\n",
     "    rcpsp_model=rcpsp_problem,\n",
     "    nb_cut_parts=[4, 5, 6],\n",
     "    fraction_subproblems=[0.3],\n",
     "    params_list=[\n",
     "        ParamsConstraintBuilder(\n",
     "            minus_delta_primary=60,\n",
-    "            plus_delta_primary=60,152f1299af4f3f60662a07d1591f3938d973a6c4\n",
+    "            plus_delta_primary=60,\n",
     "            minus_delta_secondary=20,\n",
     "            plus_delta_secondary=20,\n",
     "            constraint_max_time_to_current_solution=True,\n",
@@ -306,15 +312,23 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "%matplotlib notebook\n",
     "import matplotlib.pyplot as plt\n",
     "\n",
+    "%matplotlib inline\n",
     "fig, ax = plt.subplots(1)\n",
     "ax.plot([x[1] for x in results.list_solution_fits], marker=\"o\")\n",
     "ax.set_ylabel(\"- makespan\")\n",
     "ax.set_xlabel(\"# solution found\")\n",
     "plt.show()"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "9bed3315-16ef-47a4-acce-f84d27f17046",
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {
@@ -333,7 +347,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.8.11"
+   "version": "3.9.18"
   },
   "toc": {
    "base_numbering": 1,

tests/rcpsp/solver/test_large_neighborhood_search_solver.py

@@ -0,0 +1,35 @@
+#  Copyright (c) 2024 AIRBUS and its affiliates.
+#  This source code is licensed under the MIT license found in the
+#  LICENSE file in the root directory of this source tree.
+import pytest
+
+from discrete_optimization.generic_rcpsp_tools.large_neighborhood_search_scheduling import (
+    LargeNeighborhoodSearchScheduling,
+)
+from discrete_optimization.generic_tools.cp_tools import ParametersCP
+from discrete_optimization.rcpsp.rcpsp_model import RCPSPModel
+from discrete_optimization.rcpsp.rcpsp_parser import get_data_available, parse_file
+
+
+@pytest.mark.parametrize(
+    "file_name",
+    ["j1201_1.sm", "j1010_9.mm"],
+)
+def test_lns_solver(file_name):
+    files_available = get_data_available()
+    file = [f for f in files_available if file_name in f][0]
+    rcpsp_problem: RCPSPModel = parse_file(file)
+    solver = LargeNeighborhoodSearchScheduling(problem=rcpsp_problem)
+    parameters_cp = ParametersCP.default()
+    parameters_cp.time_limit_iter0 = 5
+    parameters_cp.time_limit = 2
+    results = solver.solve(
+        nb_iteration_lns=100,
+        skip_first_iteration=False,
+        stop_first_iteration_if_optimal=False,
+        parameters_cp=parameters_cp,
+        nb_iteration_no_improvement=50,
+        max_time_seconds=20,
+    )
+    sol, fit = results.get_best_solution_fit()
+    assert rcpsp_problem.satisfy(sol)