Add test for code export

docs/source/releases.rst

@@ -15,6 +15,7 @@ Maintenance:
    - doc build and deploy
    - pre-commit check
    - check for changelog
+ - Small changes to avoid FutureWarnings and/or DeprecationWarnings.
 
 Bugfixes:
  - #137: raise an output if ``output_directory`` is non-empty.

gama/postprocessing/ensemble.py

@@ -137,7 +137,7 @@ def __init__(
         self,
         metric,
         y: pd.DataFrame,
-        evaluation_library: EvaluationLibrary = None,
+        evaluation_library: EvaluationLibrary,
         shrink_on_pickle=True,
         downsample_to: Optional[int] = 10_000,
         use_top_n_only: Optional[int] = 200,
@@ -166,11 +166,7 @@ def __init__(
                 "metric must be specified as string or `gama.ea.metrics.Metric`."
             )
 
-        if evaluation_library is None:
-            raise ValueError(
-                "`evaluation_library` is None but must be EvaluationLibrary."
-            )
-        elif not isinstance(evaluation_library, EvaluationLibrary):
+        if not isinstance(evaluation_library, EvaluationLibrary):
             raise TypeError(
                 "`evaluation_library` must be of type "
                 "gama.utilities.evaluation_library.EvaluationLibrary."

tests/unit/test_auto_ensemble.py

@@ -1,6 +1,17 @@
 from sklearn.svm import LinearSVC
 from sklearn.datasets import load_iris
-from gama.postprocessing.ensemble import fit_and_weight
+from sklearn.pipeline import Pipeline
+from sklearn.ensemble import VotingClassifier
+
+from gama.genetic_programming.compilers.scikitlearn import compile_individual
+from gama.genetic_programming.components.individual import Individual
+from gama.postprocessing.ensemble import (
+    EnsemblePostProcessing,
+    fit_and_weight,
+    EnsembleClassifier,
+)
+from gama.utilities.evaluation_library import Evaluation, EvaluationLibrary
+from gama.utilities.metrics import Metric
 
 
 def test_fit_and_weight():
@@ -14,3 +25,32 @@ def test_fit_and_weight():
     assert 1 == w
     _, w = fit_and_weight((bad_estimator, x, y, 1))
     assert 0 == w
+
+
+def test_code_export_produces_working_code(GNB, ForestPipeline):
+    x, y = load_iris(return_X_y=True, as_frame=True)
+
+    ensemble = EnsemblePostProcessing()
+
+    ensemble._ensemble = EnsembleClassifier(
+        Metric("neg_log_loss"),
+        y,
+        evaluation_library=EvaluationLibrary(n=None),
+    )
+    gnb = GNB
+    gnb._to_pipeline = compile_individual
+    fp = ForestPipeline
+    fp._to_pipeline = compile_individual
+    ensemble._ensemble._models = {
+        "a": (Evaluation(gnb), 1),
+        "b": (Evaluation(fp), 2),
+    }
+    ensemble._ensemble._metric = Metric("neg_log_loss")
+
+    code = ensemble.to_code()
+    local = {}
+    exec(code, {}, local)
+    exported_ensemble = local["ensemble"]  # should be defined in exported code
+    assert isinstance(exported_ensemble, VotingClassifier)
+    exported_ensemble.fit(x, y)
+    assert 0.9 < exported_ensemble.score(x, y)