feat(ComparisonReport)!: Make reports_ a dict instead of a list

skore/src/skore/_sklearn/_comparison/feature_importance_accessor.py

@@ -46,9 +46,7 @@ def coefficients(self) -> FeatureImportanceDisplay:
         """
         similar_reports = defaultdict(list)
 
-        for report, name in zip(
-            self._parent.reports_, self._parent.report_names_, strict=False
-        ):
+        for name, report in self._parent.reports_.items():
             report = cast(CrossValidationReport | EstimatorReport, report)
             feature_names = (
                 report.feature_importance.coefficients().frame().index.tolist()

skore/src/skore/_sklearn/_comparison/metrics_accessor.py

@@ -242,7 +242,7 @@ def _compute_metric_scores(
 
             generator = parallel(
                 joblib.delayed(getattr(report.metrics, report_metric_name))(**kwargs)
-                for report in self._parent.reports_
+                for report in self._parent.reports_.values()
             )
             individual_results = []
             for result in generator:
@@ -256,15 +256,15 @@ def _compute_metric_scores(
             if self._parent._reports_type == "EstimatorReport":
                 results = _combine_estimator_results(
                     individual_results,
-                    estimator_names=self._parent.report_names_,
+                    estimator_names=self._parent.reports_.keys(),
                     indicator_favorability=metric_kwargs.get(
                         "indicator_favorability", False
                     ),
                 )
             else:  # "CrossValidationReport"
                 results = _combine_cross_validation_results(
                     individual_results,
-                    estimator_names=self._parent.report_names_,
+                    estimator_names=self._parent.reports_.keys(),
                     indicator_favorability=metric_kwargs.get(
                         "indicator_favorability", False
                     ),
@@ -325,10 +325,10 @@ def timings(
             timings: pd.DataFrame = pd.concat(
                 [
                     pd.Series(report.metrics.timings())
-                    for report in self._parent.reports_
+                    for report in self._parent.reports_.values()
                 ],
                 axis=1,
-                keys=self._parent.report_names_,
+                keys=self._parent.reports_.keys(),
             )
             timings.index = timings.index.str.replace("_", " ").str.capitalize()
 
@@ -342,19 +342,19 @@ def timings(
         else:  # "CrossValidationReport"
             results = [
                 report.metrics.timings(aggregate=None)
-                for report in self._parent.reports_
+                for report in self._parent.reports_.values()
             ]
 
             # Put dataframes in the right shape
             for i, result in enumerate(results):
                 result.index.name = "Metric"
                 result.columns = pd.MultiIndex.from_product(
-                    [[self._parent.report_names_[i]], result.columns]
+                    [[list(self._parent.reports_.keys())[i]], result.columns]
                 )
 
             timings = _combine_cross_validation_results(
                 results,
-                self._parent.report_names_,
+                self._parent.reports_.keys(),
                 indicator_favorability=False,
                 aggregate=aggregate,
             )
@@ -1218,9 +1218,7 @@ def _get_display(
             y_pred: list[YPlotData] = []
 
             if self._parent._reports_type == "EstimatorReport":
-                for report, report_name in zip(
-                    self._parent.reports_, self._parent.report_names_, strict=False
-                ):
+                for report_name, report in self._parent.reports_.items():
                     report_X, report_y, _ = (
                         report.metrics._get_X_y_and_data_source_hash(
                             data_source=data_source,
@@ -1264,16 +1262,16 @@ def _get_display(
                     y_true=y_true,
                     y_pred=y_pred,
                     report_type="comparison-estimator",
-                    estimators=[report.estimator_ for report in self._parent.reports_],
+                    estimators=[
+                        report.estimator_ for report in self._parent.reports_.values()
+                    ],
                     ml_task=self._parent._ml_task,
                     data_source=data_source,
                     **display_kwargs,
                 )
 
             else:
-                for report, report_name in zip(
-                    self._parent.reports_, self._parent.report_names_, strict=False
-                ):
+                for report_name, report in self._parent.reports_.items():
                     for split, estimator_report in enumerate(report.estimator_reports_):
                         report_X, report_y, _ = (
                             estimator_report.metrics._get_X_y_and_data_source_hash(
@@ -1321,7 +1319,7 @@ def _get_display(
                     report_type="comparison-cross-validation",
                     estimators=[
                         estimator_report.estimator_
-                        for report in self._parent.reports_
+                        for report in self._parent.reports_.values()
                         for estimator_report in report.estimator_reports_
                     ],
                     ml_task=self._parent._ml_task,

skore/src/skore/_sklearn/_comparison/report.py

@@ -20,7 +20,7 @@
     from skore._sklearn._comparison.feature_importance_accessor import (
         _FeatureImportanceAccessor,
     )
-    from skore._sklearn._estimator.metrics_accessor import _MetricsAccessor
+    from skore._sklearn._comparison.metrics_accessor import _MetricsAccessor
 
     ReportType = Literal["EstimatorReport", "CrossValidationReport"]
 
@@ -57,15 +57,9 @@ class ComparisonReport(_BaseReport, DirNamesMixin):
 
     Attributes
     ----------
-    reports_ : list of :class:`~skore.EstimatorReport` or list of \
-        :class:`~skore.CrossValidationReport`
+    reports_ : dict mapping names to reports
         The compared reports.
 
-    report_names_ : list of str
-        The names of the compared estimators. If the names are not customized (i.e. the
-        class names are used), a de-duplication process is used to make sure that the
-        names are distinct.
-
     See Also
     --------
     skore.EstimatorReport
@@ -87,13 +81,13 @@ class names are used), a de-duplication process is used to make sure that the
     >>> estimator_2 = LogisticRegression(C=2)  # Different regularization
     >>> estimator_report_2 = EstimatorReport(estimator_2, **split_data)
     >>> report = ComparisonReport([estimator_report_1, estimator_report_2])
-    >>> report.report_names_
-    ['LogisticRegression_1', 'LogisticRegression_2']
+    >>> report.reports_
+    {'LogisticRegression_1': ..., 'LogisticRegression_2': ...}
     >>> report = ComparisonReport(
     ...     {"model1": estimator_report_1, "model2": estimator_report_2}
     ... )
-    >>> report.report_names_
-    ['model1', 'model2']
+    >>> report.reports_
+    {'model1': ..., 'model2': ...}
 
     >>> from sklearn.datasets import make_classification
     >>> from sklearn.linear_model import LogisticRegression
@@ -122,8 +116,7 @@ def _validate_reports(
         | list[CrossValidationReport]
         | dict[str, CrossValidationReport],
     ) -> tuple[
-        list[EstimatorReport] | list[CrossValidationReport],
-        list[str],
+        dict[str, EstimatorReport] | dict[str, CrossValidationReport],
         ReportType,
         PositiveLabel,
     ]:
@@ -136,11 +129,8 @@ def _validate_reports(
 
         Returns
         -------
-        list of EstimatorReport or list of CrossValidationReport
+        dict
             The validated reports.
-        list of str
-            The report names, either taken from dict keys or computed from the estimator
-            class names.
         {"EstimatorReport", "CrossValidationReport"}
             The inferred type of the reports that will be compared.
         int, float, bool, str or None
@@ -222,7 +212,12 @@ class names.
         else:
             deduped_report_names = report_names
 
-        return reports_list, deduped_report_names, reports_type, pos_label
+        reports_dict = cast(
+            dict[str, EstimatorReport] | dict[str, CrossValidationReport],
+            dict(zip(deduped_report_names, reports_list, strict=True)),
+        )
+
+        return reports_dict, reports_type, pos_label
 
     def __init__(
         self,
@@ -244,7 +239,7 @@ def __init__(
         - all estimators have non-empty X_test and y_test,
         - all estimators have the same X_test and y_test.
         """
-        self.reports_, self.report_names_, self._reports_type, self._pos_label = (
+        self.reports_, self._reports_type, self._pos_label = (
             ComparisonReport._validate_reports(reports)
         )
 
@@ -256,7 +251,7 @@ def __init__(
             low=np.iinfo(np.int64).min, high=np.iinfo(np.int64).max
         )
         self._cache: dict[tuple[Any, ...], Any] = {}
-        self._ml_task = self.reports_[0]._ml_task
+        self._ml_task = list(self.reports_.values())[0]._ml_task  # type: ignore
 
     def clear_cache(self) -> None:
         """Clear the cache.
@@ -279,7 +274,7 @@ def clear_cache(self) -> None:
         >>> report._cache
         {}
         """
-        for report in self.reports_:
+        for report in self.reports_.values():
             report.clear_cache()
         self._cache = {}
 
@@ -332,7 +327,7 @@ def cache_predictions(
         total_estimators = len(self.reports_)
         progress.update(main_task, total=total_estimators)
 
-        for report in self.reports_:
+        for report in self.reports_.values():
             # Share the parent's progress bar with child report
             report._progress_info = {"current_progress": progress}
             report.cache_predictions(response_methods=response_methods, n_jobs=n_jobs)
@@ -418,7 +413,7 @@ def get_predictions(
                 X=X,
                 pos_label=pos_label,
             )
-            for report in self.reports_
+            for report in self.reports_.values()
         ]
 
     @property

skore/src/skore/_utils/_accessor.py

@@ -180,7 +180,7 @@ def check(accessor: Any) -> bool:
 
         parent = accessor._parent
         parent_estimators = []
-        for parent_report in parent.reports_:
+        for parent_report in parent.reports_.values():
             if parent._reports_type == "CrossValidationReport":
                 parent_report = cast(CrossValidationReport, parent_report)
                 parent_estimators.append(parent_report.estimator_reports_[0].estimator_)
@@ -203,7 +203,8 @@ def _check_any_sub_report_has_metric(metric: str) -> Callable[[Any], bool]:
 
     def check(accessor: Any) -> bool:
         return any(
-            hasattr(report.metrics, metric) for report in accessor._parent.reports_
+            hasattr(report.metrics, metric)
+            for report in accessor._parent.reports_.values()
         )
 
     return check

skore/src/skore/_utils/_progress_bar.py

@@ -70,7 +70,7 @@ def wrapper(*args: Any, **kwargs: Any) -> T:
             # assigning progress to child reports
             reports_to_cleanup: list[Any] = []
             if hasattr(self_obj, "reports_"):
-                for report in self_obj.reports_:
+                for report in self_obj.reports_.values():
                     if hasattr(report, "_progress_info"):
                         report._progress_info = {"current_progress": progress}
                         reports_to_cleanup.append(report)

skore/tests/unit/displays/precision_recall_curve/test_comparison_cross_validation.py

@@ -28,13 +28,13 @@ def test_binary_classification(
 
     pos_label = 1
     n_reports = len(report.reports_)
-    n_splits = len(report.reports_[0].estimator_reports_)
+    n_splits = len(list(report.reports_.values())[0].estimator_reports_)
 
     display.plot()
     assert isinstance(display.lines_, list)
     assert len(display.lines_) == n_reports * n_splits
     default_colors = sample_mpl_colormap(pyplot.cm.tab10, 10)
-    for i, estimator_name in enumerate(report.report_names_):
+    for i, estimator_name in enumerate(report.reports_.keys()):
         precision_recall_mpl = display.lines_[i * n_splits]
         assert isinstance(precision_recall_mpl, Line2D)
 
@@ -76,15 +76,15 @@ def test_multiclass_classification(
 
     labels = display.precision_recall["label"].cat.categories
     n_reports = len(report.reports_)
-    n_splits = len(report.reports_[0].estimator_reports_)
+    n_splits = len(list(report.reports_.values())[0].estimator_reports_)
 
     display.plot()
     assert isinstance(display.lines_, list)
     assert len(display.lines_) == n_reports * len(labels) * n_splits
 
     default_colors = sample_mpl_colormap(pyplot.cm.tab10, 10)
     for i, ((estimator_idx, estimator_name), label) in enumerate(
-        product(enumerate(report.report_names_), labels)
+        product(enumerate(report.reports_.keys()), labels)
     ):
         precision_recall_mpl = display.lines_[i * n_splits]
         assert isinstance(precision_recall_mpl, Line2D)
@@ -219,7 +219,7 @@ def test_binary_classification_constructor(forest_binary_classification_data):
         assert df.query("estimator_name == 'estimator_2'")[
             "split"
         ].unique().tolist() == list(range(cv + 1))
-        assert df["estimator_name"].unique().tolist() == report.report_names_
+        assert df["estimator_name"].unique().tolist() == list(report.reports_.keys())
         assert df["label"].unique() == 1
 
     assert len(display.average_precision) == cv + (cv + 1)
@@ -245,7 +245,7 @@ def test_multiclass_classification_constructor(forest_multiclass_classification_
         assert df.query("estimator_name == 'estimator_2'")[
             "split"
         ].unique().tolist() == list(range(cv + 1))
-        assert df["estimator_name"].unique().tolist() == report.report_names_
+        assert df["estimator_name"].unique().tolist() == list(report.reports_.keys())
         np.testing.assert_array_equal(df["label"].unique(), classes)
 
     assert len(display.average_precision) == len(classes) * cv + len(classes) * (cv + 1)

skore/tests/unit/displays/precision_recall_curve/test_comparison_estimator.py

@@ -43,7 +43,7 @@ def test_binary_classification(pyplot, logistic_binary_classification_with_train
     display.plot()
     expected_colors = sample_mpl_colormap(pyplot.cm.tab10, 10)
     for idx, (estimator_name, line) in enumerate(
-        zip(report.report_names_, display.lines_, strict=False)
+        zip(report.reports_.keys(), display.lines_, strict=False)
     ):
         assert isinstance(line, mpl.lines.Line2D)
         average_precision = display.average_precision.query(
@@ -96,14 +96,14 @@ def test_multiclass_classification(
     assert isinstance(display, PrecisionRecallCurveDisplay)
     check_display_data(display)
 
-    class_labels = report.reports_[0].estimator_.classes_
+    class_labels = list(report.reports_.values())[0].estimator_.classes_
 
     display.plot()
     assert isinstance(display.lines_, list)
     assert len(display.lines_) == len(class_labels) * 2
     default_colors = sample_mpl_colormap(pyplot.cm.tab10, 10)
     for idx, (estimator_name, expected_color) in enumerate(
-        zip(report.report_names_, default_colors, strict=False)
+        zip(report.reports_.keys(), default_colors, strict=False)
     ):
         for class_label_idx, class_label in enumerate(class_labels):
             roc_curve_mpl = display.lines_[idx * len(class_labels) + class_label_idx]
@@ -350,7 +350,7 @@ def test_binary_classification_constructor(
     index_columns = ["estimator_name", "split", "label"]
     for df in [display.precision_recall, display.average_precision]:
         assert all(col in df.columns for col in index_columns)
-        assert df["estimator_name"].unique().tolist() == report.report_names_
+        assert df["estimator_name"].unique().tolist() == list(report.reports_.keys())
         assert df["split"].isnull().all()
         assert df["label"].unique() == 1
 
@@ -378,7 +378,7 @@ def test_multiclass_classification_constructor(
     index_columns = ["estimator_name", "split", "label"]
     for df in [display.precision_recall, display.average_precision]:
         assert all(col in df.columns for col in index_columns)
-        assert df["estimator_name"].unique().tolist() == report.report_names_
+        assert df["estimator_name"].unique().tolist() == list(report.reports_.keys())
         assert df["split"].isnull().all()
         np.testing.assert_array_equal(df["label"].unique(), np.unique(y_train))
 

skore/tests/unit/displays/prediction_error/test_comparison_cross_validation.py

@@ -142,7 +142,7 @@ def test_constructor(linear_regression_data):
     assert df.query("estimator_name == 'estimator_2'")[
         "split"
     ].unique().tolist() == list(range(cv + 1))
-    assert df["estimator_name"].unique().tolist() == report.report_names_
+    assert df["estimator_name"].unique().tolist() == list(report.reports_.keys())
 
 
 def test_frame(comparison_cross_validation_reports_regression):

skore/tests/unit/displays/prediction_error/test_comparison_estimator.py

@@ -275,5 +275,5 @@ def test_constructor(linear_regression_with_train_test):
     index_columns = ["estimator_name", "split"]
     df = display._prediction_error
     assert all(col in df.columns for col in index_columns)
-    assert df["estimator_name"].unique().tolist() == report.report_names_
+    assert df["estimator_name"].unique().tolist() == list(report.reports_.keys())
     assert df["split"].isnull().all()