Add Solver.resolve() and Model.apply_result() for iterative re-solve workflows

doc/index.rst

@@ -116,6 +116,7 @@ This package is published under MIT license.
    piecewise-linear-constraints
    piecewise-linear-constraints-tutorial
    manipulating-models
+   iterative-resolving
    testing-framework
    transport-tutorial
    infeasible-model

doc/iterative-resolving.nblink

@@ -0,0 +1,3 @@
+{
+    "path": "../examples/iterative-resolving.ipynb"
+}

doc/release_notes.rst

@@ -4,6 +4,8 @@ Release Notes
 Upcoming Version
 ----------------
 
+* Add ``Solver.resolve()`` method for re-solving a modified native solver model without rebuilding the linopy model. Implemented for HiGHS, Gurobi, and Mosek.
+* Add ``Model.apply_result()`` to map a solver ``Result`` back onto model variables and constraints, enabling iterative solve workflows.
 * Harmonize coordinate alignment for operations with subset/superset objects:
   - Multiplication and division fill missing coords with 0 (variable doesn't participate)
   - Addition and subtraction of constants fill missing coords with 0 (identity element) and pin result to LHS coords

examples/iterative-resolving.ipynb

@@ -0,0 +1,302 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "d21785bf",
+   "metadata": {},
+   "source": [
+    "# Iterative Re-solving\n",
+    "\n",
+    "In many optimization workflows you need to solve a model, tweak parameters on the native solver object, and re-solve — without rebuilding the entire linopy model. This is common in:\n",
+    "\n",
+    "- Sensitivity analysis (varying objective coefficients or bounds)\n",
+    "- Decomposition algorithms (Benders, column generation)\n",
+    "- Rolling-horizon / receding-horizon schemes\n",
+    "\n",
+    "Linopy supports this via two methods:\n",
+    "\n",
+    "- **`Solver.resolve()`** — re-solves an existing native solver model and returns a `Result`\n",
+    "- **`Model.apply_result()`** — maps a `Result` back onto the linopy model's variables and constraints\n",
+    "\n",
+    "Currently `resolve()` is implemented for the **HiGHS**, **Gurobi**, and **Mosek** solvers."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "f7f604f5",
+   "metadata": {},
+   "source": [
+    "## Setup\n",
+    "\n",
+    "We start with a simple two-variable LP."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "122ee88e",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2026-03-16T08:48:03.207441Z",
+     "iopub.status.busy": "2026-03-16T08:48:03.207185Z",
+     "iopub.status.idle": "2026-03-16T08:48:03.649145Z",
+     "shell.execute_reply": "2026-03-16T08:48:03.648818Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "\n",
+    "from linopy import Model\n",
+    "from linopy.solvers import Highs"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "7c05110c",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2026-03-16T08:48:03.650662Z",
+     "iopub.status.busy": "2026-03-16T08:48:03.650440Z",
+     "iopub.status.idle": "2026-03-16T08:48:03.744258Z",
+     "shell.execute_reply": "2026-03-16T08:48:03.743824Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "m = Model()\n",
+    "x = m.add_variables(lower=0, upper=10, name=\"x\")\n",
+    "y = m.add_variables(lower=0, upper=10, name=\"y\")\n",
+    "m.add_constraints(x + y >= 8, name=\"demand\")\n",
+    "m.add_constraints(x + y <= 15, name=\"capacity\")\n",
+    "m.objective = 2 * x + 3 * y\n",
+    "m"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "574be9bd",
+   "metadata": {},
+   "source": [
+    "## Initial Solve\n",
+    "\n",
+    "Solve the model as usual. After solving, linopy stores the native solver object in `m.solver_model`."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "548b7a4d",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2026-03-16T08:48:03.745807Z",
+     "iopub.status.busy": "2026-03-16T08:48:03.745531Z",
+     "iopub.status.idle": "2026-03-16T08:48:03.787345Z",
+     "shell.execute_reply": "2026-03-16T08:48:03.786876Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "m.solve(solver_name=\"highs\")\n",
+    "print(f\"Objective: {m.objective.value}\")\n",
+    "print(f\"x = {m.solution['x'].values.item():.2f}\")\n",
+    "print(f\"y = {m.solution['y'].values.item():.2f}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "214cbe09",
+   "metadata": {},
+   "source": [
+    "## Modify the Native Solver Model\n",
+    "\n",
+    "Access the native HiGHS object and change the objective coefficients. Here we make `x` much more expensive, so the solver should prefer `y`."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "b662db35",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2026-03-16T08:48:03.789208Z",
+     "iopub.status.busy": "2026-03-16T08:48:03.788835Z",
+     "iopub.status.idle": "2026-03-16T08:48:03.801031Z",
+     "shell.execute_reply": "2026-03-16T08:48:03.800672Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "h = m.solver_model\n",
+    "n_cols = h.getNumCol()\n",
+    "new_costs = np.array([10.0, 1.0], dtype=float)\n",
+    "h.changeColsCost(n_cols, np.arange(n_cols, dtype=np.int32), new_costs)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "9e55e60b",
+   "metadata": {},
+   "source": [
+    "## Re-solve and Apply\n",
+    "\n",
+    "Create a solver instance, call `resolve()`, then apply the result back to the linopy model."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "949ce62e",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2026-03-16T08:48:03.802549Z",
+     "iopub.status.busy": "2026-03-16T08:48:03.802443Z",
+     "iopub.status.idle": "2026-03-16T08:48:03.818907Z",
+     "shell.execute_reply": "2026-03-16T08:48:03.818509Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "solver = Highs()\n",
+    "result = solver.resolve(h, sense=m.sense)\n",
+    "m.apply_result(result, solver_name=\"highs\")\n",
+    "\n",
+    "print(f\"Objective: {m.objective.value}\")\n",
+    "print(f\"x = {m.solution['x'].values.item():.2f}\")\n",
+    "print(f\"y = {m.solution['y'].values.item():.2f}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "5b2eab9a",
+   "metadata": {},
+   "source": [
+    "Since `x` is now 10× more expensive than `y`, the solver allocates as much as possible to `y`."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "7fd50c4d",
+   "metadata": {},
+   "source": [
+    "## Iterating Over Parameter Sweeps\n",
+    "\n",
+    "The `resolve` / `apply_result` pattern is efficient for parameter sweeps because it avoids rebuilding the model from scratch each time."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "bb3ac524",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2026-03-16T08:48:03.820485Z",
+     "iopub.status.busy": "2026-03-16T08:48:03.820382Z",
+     "iopub.status.idle": "2026-03-16T08:48:03.862574Z",
+     "shell.execute_reply": "2026-03-16T08:48:03.862143Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "\n",
+    "cost_x_values = [1.0, 2.0, 5.0, 10.0, 20.0]\n",
+    "results = []\n",
+    "\n",
+    "for cost_x in cost_x_values:\n",
+    "    h.changeColsCost(\n",
+    "        n_cols, np.arange(n_cols, dtype=np.int32), np.array([cost_x, 3.0], dtype=float)\n",
+    "    )\n",
+    "    result = solver.resolve(h, sense=m.sense)\n",
+    "    m.apply_result(result, solver_name=\"highs\")\n",
+    "    results.append(\n",
+    "        {\n",
+    "            \"cost_x\": cost_x,\n",
+    "            \"objective\": m.objective.value,\n",
+    "            \"x\": m.solution[\"x\"].values.item(),\n",
+    "            \"y\": m.solution[\"y\"].values.item(),\n",
+    "        }\n",
+    "    )\n",
+    "\n",
+    "pd.DataFrame(results)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "6f84f855",
+   "metadata": {},
+   "source": [
+    "## Handling Infeasible Results\n",
+    "\n",
+    "When a re-solve yields an infeasible or otherwise non-optimal result, `apply_result` returns the status without setting variable solutions. You can also pass `solution=None` in a `Result` object."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "acc2e8bb",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2026-03-16T08:48:03.864315Z",
+     "iopub.status.busy": "2026-03-16T08:48:03.864083Z",
+     "iopub.status.idle": "2026-03-16T08:48:03.879731Z",
+     "shell.execute_reply": "2026-03-16T08:48:03.879245Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from linopy.constants import Result, Status, TerminationCondition\n",
+    "\n",
+    "status = Status.from_termination_condition(TerminationCondition.infeasible)\n",
+    "result = Result(status=status, solution=None)\n",
+    "\n",
+    "s, tc = m.apply_result(result)\n",
+    "print(f\"Status: {s}, Termination: {tc}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "898f4824",
+   "metadata": {},
+   "source": [
+    "## API Reference\n",
+    "\n",
+    "**`Solver.resolve(solver_model, sense=\"min\")`**\n",
+    "\n",
+    "Re-solves a native solver object and returns a `Result`. Implemented for `Highs`, `Gurobi`, and `Mosek`. Raises `NotImplementedError` for other solvers.\n",
+    "\n",
+    "**`Model.apply_result(result, solver_name=None)`**\n",
+    "\n",
+    "Applies a `Result` to the model:\n",
+    "1. Resets existing solutions\n",
+    "2. Sets status and termination condition\n",
+    "3. Maps primal values back to variables\n",
+    "4. Maps dual values back to constraints (if available)\n",
+    "\n",
+    "Returns a `(status, termination_condition)` tuple."
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}