test passing kwargs in comparison[cv] case; fix bugs

Author: auguste-probabl

skore/src/skore/sklearn/_plot/metrics/roc_curve.py

@@ -153,7 +153,7 @@ def _plot_single_estimator(
         estimator_name: str,
         roc_curve_kwargs: list[dict[str, Any]],
         plot_chance_level: bool = True,
-        chance_level_kwargs: Optional[dict[str, Any]] = None,
+        chance_level_kwargs: Optional[dict[str, Any]],
     ) -> tuple[Axes, list[Line2D], Union[str, None]]:
         """Plot ROC curve for a single estimator.
 
@@ -272,7 +272,7 @@ def _plot_cross_validated_estimator(
         estimator_name: str,
         roc_curve_kwargs: list[dict[str, Any]],
         plot_chance_level: bool = True,
-        chance_level_kwargs: Optional[dict[str, Any]] = None,
+        chance_level_kwargs: Optional[dict[str, Any]],
     ) -> tuple[Axes, list[Line2D], Union[str, None]]:
         """Plot ROC curve for a cross-validated estimator.
 
@@ -398,7 +398,7 @@ def _plot_comparison_estimator(
         estimator_names: list[str],
         roc_curve_kwargs: list[dict[str, Any]],
         plot_chance_level: bool = True,
-        chance_level_kwargs: Optional[dict[str, Any]] = None,
+        chance_level_kwargs: Optional[dict[str, Any]],
     ) -> tuple[Axes, list[Line2D], Union[str, None]]:
         """Plot ROC curve of several estimators.
 
@@ -518,7 +518,7 @@ def _plot_comparison_cross_validation(
         estimator_names: list[str],
         roc_curve_kwargs: list[dict[str, Any]],
         plot_chance_level: bool = True,
-        chance_level_kwargs: Optional[dict[str, Any]] = None,
+        chance_level_kwargs: Optional[dict[str, Any]],
     ) -> tuple[Axes, list[Line2D], Union[str, None]]:
         """Plot ROC curve of several cross-validations.
 
@@ -568,11 +568,11 @@ def _plot_comparison_cross_validation(
                     "roc_auc"
                 ]
 
+                line_kwargs["color"] = colors[report_idx]
+                line_kwargs["alpha"] = 0.6
                 line_kwargs_validated = _validate_style_kwargs(
                     line_kwargs, roc_curve_kwargs[report_idx]
                 )
-                line_kwargs_validated["color"] = colors[report_idx]
-                line_kwargs_validated["alpha"] = 0.6
 
                 for split_index, segment in roc_curve.groupby("split_index"):
                     if split_index == 0:
@@ -620,6 +620,7 @@ def _plot_comparison_cross_validation(
                 colormaps.get_cmap("tab10"),
                 10 if len(estimator_names) < 10 else len(estimator_names),
             )
+            idx = 0
 
             for est_idx, estimator_name in enumerate(estimator_names):
                 est_color = colors[est_idx]
@@ -633,12 +634,6 @@ def _plot_comparison_cross_validation(
                         f"label == {label} & estimator_name == '{estimator_name}'"
                     )["roc_auc"]
 
-                    line_kwargs_validated = _validate_style_kwargs(
-                        line_kwargs, roc_curve_kwargs[est_idx]
-                    )
-                    line_kwargs_validated["color"] = est_color
-                    line_kwargs_validated["alpha"] = 0.6
-
                     for split_index, segment in roc_curve.groupby("split_index"):
                         if split_index == 0:
                             label_kwargs = {
@@ -651,13 +646,21 @@ def _plot_comparison_cross_validation(
                         else:
                             label_kwargs = {}
 
+                        line_kwargs["color"] = est_color
+                        line_kwargs["alpha"] = 0.6
+                        line_kwargs_validated = _validate_style_kwargs(
+                            line_kwargs, roc_curve_kwargs[idx]
+                        )
+
                         (line,) = self.ax_[label_idx].plot(
                             segment["fpr"],
                             segment["tpr"],
                             **(line_kwargs_validated | label_kwargs),
                         )
                         lines.append(line)
 
+                        idx = idx + 1
+
                     info_pos_label = f"\n(Positive label: {label})"
                     _set_axis_labels(self.ax_[label_idx], info_pos_label)
 
@@ -784,6 +787,8 @@ def plot(
                 self._plot_comparison_cross_validation(
                     estimator_names=self.roc_auc["estimator_name"].unique(),
                     roc_curve_kwargs=roc_curve_kwargs,
+                    plot_chance_level=plot_chance_level,
+                    chance_level_kwargs=chance_level_kwargs,
                 )
             )
         else:

skore/tests/unit/sklearn/plot/roc_curve/test_comparison_cross_validation.py

@@ -137,3 +137,97 @@ def test_multiclass_classification(pyplot, multiclass_classification_report):
         assert ax.get_adjustable() == "box"
         assert ax.get_aspect() in ("equal", 1.0)
         assert ax.get_xlim() == ax.get_ylim() == (-0.01, 1.01)
+
+
+def test_binary_classification_wrong_kwargs(pyplot, binary_classification_report):
+    """Check that we raise a proper error message when passing an inappropriate
+    value for the `roc_curve_kwargs` argument."""
+    report = binary_classification_report
+    display = report.metrics.roc()
+    err_msg = (
+        "You intend to plot multiple curves. We expect `roc_curve_kwargs` to be a "
+        "list of dictionaries with the same length as the number of curves. "
+        "Got 2 instead of 10."
+    )
+    with pytest.raises(ValueError, match=err_msg):
+        display.plot(roc_curve_kwargs=[{}, {}])
+
+
+@pytest.mark.parametrize("roc_curve_kwargs", [[{"color": "red"}] * 10])
+def test_binary_classification_kwargs(
+    pyplot, binary_classification_report, roc_curve_kwargs
+):
+    """Check that we can pass keyword arguments to the ROC curve plot."""
+    report = binary_classification_report
+    display = report.metrics.roc()
+    display.plot(
+        roc_curve_kwargs=roc_curve_kwargs, chance_level_kwargs={"color": "blue"}
+    )
+    assert display.lines_[0].get_color() == "red"
+    assert display.chance_level_.get_color() == "blue"
+
+    # check the `.style` display setter
+    display.plot()  # default style
+    assert display.lines_[0].get_color() == (
+        np.float64(0.12156862745098039),
+        np.float64(0.4666666666666667),
+        np.float64(0.7058823529411765),
+        np.float64(1.0),
+    )
+    assert display.chance_level_.get_color() == "k"
+
+    display.set_style(
+        roc_curve_kwargs=roc_curve_kwargs, chance_level_kwargs={"color": "blue"}
+    )
+    display.plot()
+    assert display.lines_[0].get_color() == "red"
+    assert display.chance_level_.get_color() == "blue"
+
+    # overwrite the style that was set above
+    display.plot(
+        roc_curve_kwargs=[{"color": "#1f77b4"}] * 10,
+        chance_level_kwargs={"color": "red"},
+    )
+    assert display.lines_[0].get_color() == "#1f77b4"
+    assert display.chance_level_.get_color() == "red"
+
+
+def test_multiclass_classification_wrong_kwargs(
+    pyplot, multiclass_classification_report
+):
+    """Check that we raise a proper error message when passing an inappropriate
+    value for the `roc_curve_kwargs` argument."""
+    report = multiclass_classification_report
+    display = report.metrics.roc()
+    err_msg = "You intend to plot multiple curves."
+    with pytest.raises(ValueError, match=err_msg):
+        display.plot(roc_curve_kwargs=[{}, {}])
+
+    with pytest.raises(ValueError, match=err_msg):
+        display.plot(roc_curve_kwargs={})
+
+
+def test_multiclass_classification_kwargs(pyplot, multiclass_classification_report):
+    """Check that we can pass keyword arguments to the ROC curve plot for
+    multiclass classification."""
+    report = multiclass_classification_report
+    display = report.metrics.roc()
+    display.plot(
+        roc_curve_kwargs=(
+            [{"color": "red"}] * 10
+            + [{"color": "blue"}] * 10
+            + [{"color": "green"}] * 10
+        ),
+        chance_level_kwargs={"color": "blue"},
+    )
+    assert display.lines_[0].get_color() == "red"
+    assert display.lines_[10].get_color() == "blue"
+    assert display.lines_[20].get_color() == "green"
+    assert display.chance_level_[0].get_color() == "blue"
+
+    display.plot(plot_chance_level=False)
+    assert display.chance_level_ is None
+
+    display.plot(despine=False)
+    assert display.ax_[0].spines["top"].get_visible()
+    assert display.ax_[0].spines["right"].get_visible()