New KNITRO direct solver interface

doc/OnlineDocs/explanation/experimental/solvers.rst

@@ -54,6 +54,9 @@ with existing interfaces).
    * - HiGHS
      - ``highs``
      - ``highs``
+   * - KNITRO
+     - ``knitro_direct``
+     - ``knitro_direct``
 
 Using the new interfaces through the legacy interface
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

pyomo/contrib/solver/plugins.py

@@ -15,22 +15,28 @@
 from .solvers.gurobi_persistent import GurobiPersistent
 from .solvers.gurobi_direct import GurobiDirect
 from .solvers.highs import Highs
+from .solvers.knitro.direct import KnitroDirectSolver
 
 
 def load():
     SolverFactory.register(
-        name='ipopt', legacy_name='ipopt_v2', doc='The IPOPT NLP solver'
+        name="ipopt", legacy_name="ipopt_v2", doc="The IPOPT NLP solver"
     )(Ipopt, LegacyIpoptSolver)
     SolverFactory.register(
-        name='gurobi_persistent',
-        legacy_name='gurobi_persistent_v2',
-        doc='Persistent interface to Gurobi',
+        name="gurobi_persistent",
+        legacy_name="gurobi_persistent_v2",
+        doc="Persistent interface to Gurobi",
     )(GurobiPersistent)
     SolverFactory.register(
-        name='gurobi_direct',
-        legacy_name='gurobi_direct_v2',
-        doc='Direct (scipy-based) interface to Gurobi',
+        name="gurobi_direct",
+        legacy_name="gurobi_direct_v2",
+        doc="Direct (scipy-based) interface to Gurobi",
     )(GurobiDirect)
     SolverFactory.register(
-        name='highs', legacy_name='highs', doc='Persistent interface to HiGHS'
+        name="highs", legacy_name="highs", doc="Persistent interface to HiGHS"
     )(Highs)
+    SolverFactory.register(
+        name="knitro_direct",
+        legacy_name="knitro_direct",
+        doc="Direct interface to KNITRO solver",
+    )(KnitroDirectSolver)

pyomo/contrib/solver/solvers/knitro/__init__.py

@@ -0,0 +1,10 @@
+#  ___________________________________________________________________________
+#
+#  Pyomo: Python Optimization Modeling Objects
+#  Copyright (c) 2008-2025
+#  National Technology and Engineering Solutions of Sandia, LLC
+#  Under the terms of Contract DE-NA0003525 with National Technology and
+#  Engineering Solutions of Sandia, LLC, the U.S. Government retains certain
+#  rights in this software.
+#  This software is distributed under the 3-clause BSD License.
+#  ___________________________________________________________________________

pyomo/contrib/solver/solvers/knitro/api.py

@@ -0,0 +1,20 @@
+#  ___________________________________________________________________________
+#
+#  Pyomo: Python Optimization Modeling Objects
+#  Copyright (c) 2008-2025
+#  National Technology and Engineering Solutions of Sandia, LLC
+#  Under the terms of Contract DE-NA0003525 with National Technology and
+#  Engineering Solutions of Sandia, LLC, the U.S. Government retains certain
+#  rights in this software.
+#  This software is distributed under the 3-clause BSD License.
+#  ___________________________________________________________________________
+
+from pyomo.common.dependencies import attempt_import
+
+knitro, KNITRO_AVAILABLE = attempt_import("knitro")
+
+
+def get_version() -> str:
+    if not bool(KNITRO_AVAILABLE):
+        return "0.0.0"
+    return knitro.__version__

pyomo/contrib/solver/solvers/knitro/base.py

@@ -0,0 +1,271 @@
+#  ___________________________________________________________________________
+#
+#  Pyomo: Python Optimization Modeling Objects
+#  Copyright (c) 2008-2025
+#  National Technology and Engineering Solutions of Sandia, LLC
+#  Under the terms of Contract DE-NA0003525 with National Technology and
+#  Engineering Solutions of Sandia, LLC, the U.S. Government retains certain
+#  rights in this software.
+#  This software is distributed under the 3-clause BSD License.
+#  ___________________________________________________________________________
+
+from abc import abstractmethod
+from collections.abc import Mapping, Sequence
+from datetime import datetime, timezone
+from io import StringIO
+from typing import Optional
+
+from pyomo.common.collections import ComponentMap
+from pyomo.common.errors import ApplicationError, DeveloperError, PyomoException
+from pyomo.common.numeric_types import value
+from pyomo.common.tee import TeeStream, capture_output
+from pyomo.common.timing import HierarchicalTimer
+from pyomo.contrib.solver.common.base import SolverBase
+from pyomo.contrib.solver.common.results import (
+    Results,
+    SolutionStatus,
+    TerminationCondition,
+)
+from pyomo.contrib.solver.common.util import (
+    IncompatibleModelError,
+    NoDualsError,
+    NoOptimalSolutionError,
+    NoReducedCostsError,
+    NoSolutionError,
+)
+from pyomo.contrib.solver.solvers.knitro.api import knitro
+from pyomo.contrib.solver.solvers.knitro.config import KnitroConfig
+from pyomo.contrib.solver.solvers.knitro.engine import Engine
+from pyomo.contrib.solver.solvers.knitro.package import PackageChecker
+from pyomo.contrib.solver.solvers.knitro.solution import (
+    SolutionLoader,
+    SolutionProvider,
+)
+from pyomo.contrib.solver.solvers.knitro.typing import ItemData, ItemType, ValueType
+from pyomo.contrib.solver.solvers.knitro.utils import KnitroModelData
+from pyomo.core.base.block import BlockData
+from pyomo.core.base.constraint import ConstraintData
+from pyomo.core.base.var import VarData
+from pyomo.core.staleflag import StaleFlagManager
+
+
+class KnitroSolverBase(SolutionProvider, PackageChecker, SolverBase):
+    CONFIG = KnitroConfig()
+    config: KnitroConfig
+
+    _engine: Engine
+    _model_data: KnitroModelData
+    _stream: StringIO
+    _saved_var_values: dict[int, Optional[float]]
+
+    def __init__(self, **kwds) -> None:
+        PackageChecker.__init__(self)
+        SolverBase.__init__(self, **kwds)
+        self._engine = Engine()
+        self._model_data = KnitroModelData()
+        self._stream = StringIO()
+        self._saved_var_values = {}
+
+    def solve(self, model: BlockData, **kwds) -> Results:
+        tick = datetime.now(timezone.utc)
+        self._check_available()
+
+        config = self._build_config(**kwds)
+        timer = config.timer or HierarchicalTimer()
+
+        StaleFlagManager.mark_all_as_stale()
+
+        self._presolve(model, config, timer)
+        self._validate_problem()
+
+        self._stream = StringIO()
+        if config.restore_variable_values_after_solve:
+            self._save_var_values()
+
+        with capture_output(TeeStream(self._stream, *config.tee), capture_fd=False):
+            self._solve(config, timer)
+
+        if config.restore_variable_values_after_solve:
+            self._restore_var_values()
+
+        results = self._postsolve(config, timer)
+
+        tock = datetime.now(timezone.utc)
+
+        results.timing_info.start_timestamp = tick
+        results.timing_info.wall_time = (tock - tick).total_seconds()
+        return results
+
+    def _build_config(self, **kwds) -> KnitroConfig:
+        return self.config(value=kwds, preserve_implicit=True)  # type: ignore
+
+    def _validate_problem(self) -> None:
+        if len(self._model_data.objs) > 1:
+            msg = f"{self.name} does not support multiple objectives."
+            raise IncompatibleModelError(msg)
+
+    def _check_available(self) -> None:
+        avail = self.available()
+        if not avail:
+            msg = f"Solver {self.name} is not available: {avail}."
+            raise ApplicationError(msg)
+
+    def _save_var_values(self) -> None:
+        self._saved_var_values.clear()
+        for var in self._get_vars():
+            self._saved_var_values[id(var)] = value(var.value)
+
+    def _restore_var_values(self) -> None:
+        StaleFlagManager.mark_all_as_stale(delayed=True)
+        for var in self._get_vars():
+            var.set_value(self._saved_var_values[id(var)])
+        StaleFlagManager.mark_all_as_stale()
+
+    @abstractmethod
+    def _presolve(
+        self, model: BlockData, config: KnitroConfig, timer: HierarchicalTimer
+    ) -> None:
+        raise NotImplementedError
+
+    @abstractmethod
+    def _solve(self, config: KnitroConfig, timer: HierarchicalTimer) -> None:
+        raise NotImplementedError
+
+    def _postsolve(self, config: KnitroConfig, timer: HierarchicalTimer) -> Results:
+        status = self._engine.get_status()
+        results = Results()
+        results.solver_name = self.name
+        results.solver_version = self.version()
+        results.solver_log = self._stream.getvalue()
+        results.solver_config = config
+        results.solution_status = self._get_solution_status(status)
+        results.termination_condition = self._get_termination_condition(status)
+        results.incumbent_objective = self._engine.get_obj_value()
+        results.iteration_count = self._engine.get_num_iters()
+        results.timing_info.solve_time = self._engine.get_solve_time()
+        results.timing_info.timer = timer
+
+        if (
+            config.raise_exception_on_nonoptimal_result
+            and results.termination_condition
+            != TerminationCondition.convergenceCriteriaSatisfied
+        ):
+            raise NoOptimalSolutionError()
+
+        results.solution_loader = SolutionLoader(
+            self,
+            has_primals=results.solution_status
+            not in {SolutionStatus.infeasible, SolutionStatus.noSolution},
+            has_reduced_costs=results.solution_status == SolutionStatus.optimal,
+            has_duals=results.solution_status
+            not in {SolutionStatus.infeasible, SolutionStatus.noSolution},
+        )
+        if config.load_solutions:
+            timer.start("load_solutions")
+            results.solution_loader.load_vars()
+            timer.stop("load_solutions")
+
+        return results
+
+    def get_values(
+        self,
+        item_type: type[ItemType],
+        value_type: ValueType,
+        items: Optional[Sequence[ItemType]] = None,
+        *,
+        exists: bool,
+        solution_id: Optional[int] = None,
+    ) -> Mapping[ItemType, float]:
+        error_type = self._get_error_type(item_type, value_type)
+        if not exists:
+            raise error_type()
+        # KNITRO only supports a single solution
+        assert solution_id is None
+        if items is None:
+            items = self._get_items(item_type)
+        x = self._engine.get_values(item_type, value_type, items)
+        if x is None:
+            raise error_type()
+        sign = value_type.sign
+        return ComponentMap([(k, sign * xk) for k, xk in zip(items, x)])
+
+    def get_num_solutions(self) -> int:
+        return self._engine.get_num_solutions()
+
+    def _get_vars(self) -> list[VarData]:
+        return self._model_data.variables
+
+    def _get_items(self, item_type: type[ItemType]) -> Sequence[ItemType]:
+        maps = {
+            VarData: self._model_data.variables,
+            ConstraintData: self._model_data.cons,
+        }
+        return maps[item_type]
+
+    @staticmethod
+    def _get_solution_status(status: int) -> SolutionStatus:
+        if (
+            status == knitro.KN_RC_OPTIMAL
+            or status == knitro.KN_RC_OPTIMAL_OR_SATISFACTORY
+            or status == knitro.KN_RC_NEAR_OPT
+        ):
+            return SolutionStatus.optimal
+        elif status == knitro.KN_RC_FEAS_NO_IMPROVE:
+            return SolutionStatus.feasible
+        elif (
+            status == knitro.KN_RC_INFEASIBLE
+            or status == knitro.KN_RC_INFEAS_CON_BOUNDS
+            or status == knitro.KN_RC_INFEAS_VAR_BOUNDS
+            or status == knitro.KN_RC_INFEAS_NO_IMPROVE
+        ):
+            return SolutionStatus.infeasible
+        else:
+            return SolutionStatus.noSolution
+
+    @staticmethod
+    def _get_termination_condition(status: int) -> TerminationCondition:
+        if (
+            status == knitro.KN_RC_OPTIMAL
+            or status == knitro.KN_RC_OPTIMAL_OR_SATISFACTORY
+            or status == knitro.KN_RC_NEAR_OPT
+        ):
+            return TerminationCondition.convergenceCriteriaSatisfied
+        elif status == knitro.KN_RC_INFEAS_NO_IMPROVE:
+            return TerminationCondition.locallyInfeasible
+        elif (
+            status == knitro.KN_RC_INFEASIBLE
+            or status == knitro.KN_RC_INFEAS_CON_BOUNDS
+            or status == knitro.KN_RC_INFEAS_VAR_BOUNDS
+        ):
+            return TerminationCondition.provenInfeasible
+        elif (
+            status == knitro.KN_RC_UNBOUNDED_OR_INFEAS
+            or status == knitro.KN_RC_UNBOUNDED
+        ):
+            return TerminationCondition.infeasibleOrUnbounded
+        elif (
+            status == knitro.KN_RC_ITER_LIMIT_FEAS
+            or status == knitro.KN_RC_ITER_LIMIT_INFEAS
+        ):
+            return TerminationCondition.iterationLimit
+        elif (
+            status == knitro.KN_RC_TIME_LIMIT_FEAS
+            or status == knitro.KN_RC_TIME_LIMIT_INFEAS
+        ):
+            return TerminationCondition.maxTimeLimit
+        elif status == knitro.KN_RC_USER_TERMINATION:
+            return TerminationCondition.interrupted
+        else:
+            return TerminationCondition.unknown
+
+    @staticmethod
+    def _get_error_type(
+        item_type: type[ItemData], value_type: ValueType
+    ) -> type[PyomoException]:
+        if item_type is VarData and value_type == ValueType.PRIMAL:
+            return NoSolutionError
+        elif item_type is VarData and value_type == ValueType.DUAL:
+            return NoReducedCostsError
+        elif item_type is ConstraintData and value_type == ValueType.DUAL:
+            return NoDualsError
+        raise DeveloperError()