LearningSolver: Keep original H5 file unmodified

CHANGELOG.md

@@ -5,6 +5,17 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.4.0] - Unreleased
+
+### Added
+
+- Add ML strategies for user cuts
+- Add ML strategies for lazy constraints
+
+### Changed
+
+- LearningSolver.solve no longer generates HDF5 files; use a collector instead.
+
 ## [0.3.0] - 2023-06-08
 
 This is a complete rewrite of the original prototype package, with an entirely new API, focused on performance, scalability and flexibility.
@@ -31,4 +42,4 @@ This is a complete rewrite of the original prototype package, with an entirely n
 
 ## [0.1.0] - 2020-11-23
 
-- Initial public release
+- Initial public release

docs/tutorials/getting-started-gurobipy.ipynb

@@ -695,7 +695,7 @@
    "source": [
     "## Accessing the solution\n",
     "\n",
-    "In the example above, we used `LearningSolver.solve` together with data files to solve both the training and the test instances. The optimal solutions were saved to HDF5 files in the train/test folders, and could be retrieved by reading theses files, but that is not very convenient. In the following example, we show how to build and solve a Pyomo model entirely in-memory, using our trained solver."
+    "In the example above, we used `LearningSolver.solve` together with data files to solve both the training and the test instances. In the following example, we show how to build and solve a Pyomo model entirely in-memory, using our trained solver."
    ]
   },
   {

docs/tutorials/getting-started-jump.ipynb

@@ -592,7 +592,7 @@
    "source": [
     "## Accessing the solution\n",
     "\n",
-    "In the example above, we used `LearningSolver.solve` together with data files to solve both the training and the test instances. The optimal solutions were saved to HDF5 files in the train/test folders, and could be retrieved by reading theses files, but that is not very convenient. In the following example, we show how to build and solve a JuMP model entirely in-memory, using our trained solver."
+    "In the example above, we used `LearningSolver.solve` together with data files to solve both the training and the test instances. In the following example, we show how to build and solve a JuMP model entirely in-memory, using our trained solver."
    ]
   },
   {

docs/tutorials/getting-started-pyomo.ipynb

@@ -712,7 +712,7 @@
    "source": [
     "## Accessing the solution\n",
     "\n",
-    "In the example above, we used `LearningSolver.solve` together with data files to solve both the training and the test instances. The optimal solutions were saved to HDF5 files in the train/test folders, and could be retrieved by reading theses files, but that is not very convenient. In the following example, we show how to build and solve a Pyomo model entirely in-memory, using our trained solver."
+    "In the example above, we used `LearningSolver.solve` together with data files to solve both the training and the test instances. In the following example, we show how to build and solve a Pyomo model entirely in-memory, using our trained solver."
    ]
   },
   {

miplearn/solvers/learning.py

@@ -8,6 +8,7 @@ from typing import List, Any, Union, Dict, Callable, Optional
 from miplearn.h5 import H5File
 from miplearn.io import _to_h5_filename
 from miplearn.solvers.abstract import AbstractModel
+import shutil
 
 
 class LearningSolver:
@@ -25,15 +26,20 @@ class LearningSolver:
         model: Union[str, AbstractModel],
         build_model: Optional[Callable] = None,
     ) -> Dict[str, Any]:
+        h5_filename, mode = NamedTemporaryFile().name, "w"
         if isinstance(model, str):
-            h5_filename = _to_h5_filename(model)
             assert build_model is not None
+            old_h5_filename = _to_h5_filename(model)
             model = build_model(model)
             assert isinstance(model, AbstractModel)
-        else:
-            h5_filename = NamedTemporaryFile().name
+
+            # If the instance has an associate H5 file, we make a temporary copy of it,
+            # then work on that copy. We keep the original file unmodified
+            if exists(old_h5_filename):
+                shutil.copy(old_h5_filename, h5_filename)
+                mode = "r+"
+
         stats: Dict[str, Any] = {}
-        mode = "r+" if exists(h5_filename) else "w"
         with H5File(h5_filename, mode) as h5:
             model.extract_after_load(h5)
             if not self.skip_lp: