Fix proper convert for data_definition to input_data. (#1461)

src/evidently/future/metrics/classification.py

@@ -8,6 +8,7 @@
 from typing import Type
 from typing import TypeVar
 
+from evidently.base_metric import InputData
 from evidently.base_metric import Metric
 from evidently.future.metric_types import BoundTest
 from evidently.future.metric_types import ByLabelCalculation
@@ -19,6 +20,7 @@
 from evidently.future.metric_types import TMetricResult
 from evidently.future.metrics._legacy import LegacyMetricCalculation
 from evidently.future.report import Context
+from evidently.future.report import _default_input_data_generator
 from evidently.future.tests import Reference
 from evidently.future.tests import eq
 from evidently.future.tests import gt
@@ -44,9 +46,6 @@ class ClassificationQualityByLabel(ByLabelMetric):
     probas_threshold: Optional[float] = None
     k: Optional[int] = None
 
-    # def _default_tests_with_reference(self, context: Context) -> List[BoundTest]:
-    #     return [eq(Reference(relative=0.2)).bind_single(self.get_fingerprint())]
-
 
 class ClassificationQualityBase(SingleValueMetric):
     probas_threshold: Optional[float] = None
@@ -69,6 +68,11 @@ def _get_dummy_value(
 TSingleValueMetric = TypeVar("TSingleValueMetric", bound=ClassificationQuality)
 
 
+def _gen_classification_input_data(context: "Context") -> InputData:
+    default_input_data = _default_input_data_generator(context)
+    return default_input_data
+
+
 class LegacyClassificationQualityByClass(
     ByLabelCalculation[TByLabelMetric],
     LegacyMetricCalculation[

src/evidently/future/metrics/regression.py

@@ -16,6 +16,7 @@
 from evidently.future.metric_types import TSingleValueMetric
 from evidently.future.metrics._legacy import LegacyMetricCalculation
 from evidently.future.report import Context
+from evidently.future.report import _default_input_data_generator
 from evidently.future.tests import Reference
 from evidently.future.tests import eq
 from evidently.future.tests import gt
@@ -41,6 +42,23 @@
 }
 
 
+def _gen_regression_input_data(context: "Context") -> InputData:
+    default_data = _default_input_data_generator(context)
+    regression = context.data_definition.get_regression("default")
+    if regression is None:
+        raise ValueError("No default regression in data definition")
+    default_data.column_mapping.target = regression.target
+    default_data.column_mapping.prediction = regression.prediction
+
+    definition = create_data_definition(
+        default_data.reference_data,
+        default_data.current_data,
+        default_data.column_mapping,
+    )
+    default_data.data_definition = definition
+    return default_data
+
+
 class LegacyRegressionMeanStdMetric(
     MeanStdCalculation[TMeanStdMetric],
     LegacyMetricCalculation[MeanStdValue, TMeanStdMetric, RegressionQualityMetricResults, RegressionQualityMetric],
@@ -50,30 +68,17 @@ class LegacyRegressionMeanStdMetric(
     def legacy_metric(self) -> RegressionQualityMetric:
         return RegressionQualityMetric()
 
-    def _gen_input_data(self, context: "Context") -> InputData:
-        default_data = super()._gen_input_data(context)
-        regression = context.data_definition.get_regression("default")
-        if regression is None:
-            raise ValueError("No default regression in data definition")
-        default_data.column_mapping.target = regression.target
-        default_data.column_mapping.prediction = regression.prediction
-
-        definition = create_data_definition(
-            default_data.reference_data,
-            default_data.current_data,
-            default_data.column_mapping,
-        )
-        default_data.data_definition = definition
-        return default_data
-
     def get_additional_widgets(self, context: "Context") -> List[BaseWidgetInfo]:
         result = []
         for field, metric in ADDITIONAL_WIDGET_MAPPING.items():
             if hasattr(self.metric, field) and getattr(self.metric, field):
-                _, widgets = context.get_legacy_metric(metric, self._gen_input_data)
+                _, widgets = context.get_legacy_metric(metric, _gen_regression_input_data)
                 result += widgets
         return result
 
+    def _gen_input_data(self, context: "Context") -> InputData:
+        return _gen_regression_input_data(context)
+
 
 class LegacyRegressionSingleValueMetric(
     SingleValueCalculation[TSingleValueMetric],
@@ -84,30 +89,17 @@ class LegacyRegressionSingleValueMetric(
     def legacy_metric(self) -> RegressionQualityMetric:
         return RegressionQualityMetric()
 
-    def _gen_input_data(self, context: "Context") -> InputData:
-        default_data = super()._gen_input_data(context)
-        regression = context.data_definition.get_regression("default")
-        if regression is None:
-            raise ValueError("No default regression in data definition")
-        default_data.column_mapping.target = regression.target
-        default_data.column_mapping.prediction = regression.prediction
-
-        definition = create_data_definition(
-            default_data.reference_data,
-            default_data.current_data,
-            default_data.column_mapping,
-        )
-        default_data.data_definition = definition
-        return default_data
-
     def get_additional_widgets(self, context: "Context") -> List[BaseWidgetInfo]:
         result = []
         for field, metric in ADDITIONAL_WIDGET_MAPPING.items():
             if hasattr(self.metric, field) and getattr(self.metric, field):
-                _, widgets = context.get_legacy_metric(metric, self._gen_input_data)
+                _, widgets = context.get_legacy_metric(metric, _gen_regression_input_data)
                 result += widgets
         return result
 
+    def _gen_input_data(self, context: "Context") -> InputData:
+        return _gen_regression_input_data(context)
+
 
 class MeanError(MeanStdMetric):
     error_plot: bool = True
@@ -269,20 +261,7 @@ def legacy_metric(self) -> RegressionDummyMetric:
         return RegressionDummyMetric()
 
     def _gen_input_data(self, context: "Context") -> InputData:
-        default_data = super()._gen_input_data(context)
-        regression = context.data_definition.get_regression("default")
-        if regression is None:
-            raise ValueError("No default regression in data definition")
-        default_data.column_mapping.target = regression.target
-        default_data.column_mapping.prediction = regression.prediction
-
-        definition = create_data_definition(
-            default_data.reference_data,
-            default_data.current_data,
-            default_data.column_mapping,
-        )
-        default_data.data_definition = definition
-        return default_data
+        return _gen_regression_input_data(context)
 
 
 class LegacyRegressionDummyValueMetric(
@@ -295,20 +274,7 @@ def legacy_metric(self) -> RegressionDummyMetric:
         return RegressionDummyMetric()
 
     def _gen_input_data(self, context: "Context") -> InputData:
-        default_data = super()._gen_input_data(context)
-        regression = context.data_definition.get_regression("default")
-        if regression is None:
-            raise ValueError("No default regression in data definition")
-        default_data.column_mapping.target = regression.target
-        default_data.column_mapping.prediction = regression.prediction
-
-        definition = create_data_definition(
-            default_data.reference_data,
-            default_data.current_data,
-            default_data.column_mapping,
-        )
-        default_data.data_definition = definition
-        return default_data
+        return _gen_regression_input_data(context)
 
 
 class DummyMAE(SingleValueMetric):

src/evidently/future/presets/classification.py

@@ -24,6 +24,7 @@
 from evidently.future.metrics.classification import DummyF1Score
 from evidently.future.metrics.classification import DummyPrecision
 from evidently.future.metrics.classification import DummyRecall
+from evidently.future.metrics.classification import _gen_classification_input_data
 from evidently.future.report import Context
 from evidently.metrics import ClassificationConfusionMatrix
 from evidently.metrics import ClassificationDummyMetric
@@ -93,15 +94,25 @@ def generate_metrics(self, context: "Context") -> List[Metric]:
         return metrics
 
     def render(self, context: "Context", results: Dict[MetricId, MetricResult]) -> List[BaseWidgetInfo]:
-        _, render = context.get_legacy_metric(ClassificationQualityMetric(probas_threshold=self._probas_threshold))
+        _, render = context.get_legacy_metric(
+            ClassificationQualityMetric(probas_threshold=self._probas_threshold),
+            _gen_classification_input_data,
+        )
         if self._conf_matrix:
-            render += context.get_legacy_metric(ClassificationConfusionMatrix(probas_threshold=self._probas_threshold))[
-                1
-            ]
+            render += context.get_legacy_metric(
+                ClassificationConfusionMatrix(probas_threshold=self._probas_threshold),
+                _gen_classification_input_data,
+            )[1]
         if self._pr_curve:
-            render += context.get_legacy_metric(ClassificationPRCurve(probas_threshold=self._probas_threshold))[1]
+            render += context.get_legacy_metric(
+                ClassificationPRCurve(probas_threshold=self._probas_threshold),
+                _gen_classification_input_data,
+            )[1]
         if self._pr_table:
-            render += context.get_legacy_metric(ClassificationPRTable(probas_threshold=self._probas_threshold))[1]
+            render += context.get_legacy_metric(
+                ClassificationPRTable(probas_threshold=self._probas_threshold),
+                _gen_classification_input_data,
+            )[1]
         return render
 
 
@@ -119,7 +130,10 @@ def generate_metrics(self, context: "Context") -> List[Metric]:
         ]
 
     def render(self, context: "Context", results: Dict[MetricId, MetricResult]):
-        render = context.get_legacy_metric(ClassificationQualityByClass(self._probas_threshold, self._k))[1]
+        render = context.get_legacy_metric(
+            ClassificationQualityByClass(self._probas_threshold, self._k),
+            _gen_classification_input_data,
+        )[1]
         widget = render
         widget[0].params["counters"][0]["label"] = "Classification Quality by Label"
         return widget
@@ -138,7 +152,10 @@ def generate_metrics(self, context: "Context") -> List[Metric]:
         ]
 
     def render(self, context: "Context", results: Dict[MetricId, MetricResult]) -> List[BaseWidgetInfo]:
-        _, widgets = context.get_legacy_metric(ClassificationDummyMetric(self._probas_threshold, self._k))
+        _, widgets = context.get_legacy_metric(
+            ClassificationDummyMetric(self._probas_threshold, self._k),
+            _gen_classification_input_data,
+        )
         return widgets
 
 

src/evidently/future/presets/dataset_stats.py

@@ -28,6 +28,7 @@
 from evidently.future.metrics.dataset_statistics import EmptyColumnsCount
 from evidently.future.metrics.dataset_statistics import EmptyRowsCount
 from evidently.future.report import Context
+from evidently.future.report import _default_input_data_generator
 from evidently.metric_results import Label
 from evidently.metrics import DatasetSummaryMetric
 from evidently.model.widget import BaseWidgetInfo
@@ -224,7 +225,7 @@ def generate_metrics(self, context: Context) -> List[Metric]:
 
     def render(self, context: Context, results: Dict[MetricId, MetricResult]) -> List[BaseWidgetInfo]:
         metric = DatasetSummaryMetric()
-        _, render = context.get_legacy_metric(metric)
+        _, render = context.get_legacy_metric(metric, _default_input_data_generator)
         return render
 
 

src/evidently/future/presets/drift.py

@@ -12,6 +12,7 @@
 from evidently.future.metrics import ValueDrift
 from evidently.future.metrics.column_statistics import DriftedColumnsCount
 from evidently.future.report import Context
+from evidently.future.report import _default_input_data_generator
 from evidently.metrics import DataDriftTable
 from evidently.metrics import DatasetDriftMetric
 from evidently.metrics.data_drift.embedding_drift_methods import DriftMethod
@@ -110,7 +111,8 @@ def render(self, context: Context, results: Dict[MetricId, MetricResult]) -> Lis
                 num_stattest_threshold=self.num_stattest_threshold,
                 text_stattest_threshold=self.text_stattest_threshold,
                 per_column_stattest_threshold=self.per_column_stattest_threshold,
-            )
+            ),
+            _default_input_data_generator,
         )[1]
         table = context.get_legacy_metric(
             DataDriftTable(
@@ -125,7 +127,8 @@ def render(self, context: Context, results: Dict[MetricId, MetricResult]) -> Lis
                 num_stattest_threshold=self.num_stattest_threshold,
                 text_stattest_threshold=self.text_stattest_threshold,
                 per_column_stattest_threshold=self.per_column_stattest_threshold,
-            )
+            ),
+            _default_input_data_generator,
         )[1]
         return dataset_drift + table
 

src/evidently/future/presets/regression.py

@@ -14,6 +14,7 @@
 from evidently.future.metrics import DummyRMSE
 from evidently.future.metrics import MeanError
 from evidently.future.metrics import R2Score
+from evidently.future.metrics.regression import _gen_regression_input_data
 from evidently.future.report import Context
 from evidently.metrics import RegressionDummyMetric
 from evidently.metrics import RegressionErrorDistribution
@@ -45,13 +46,25 @@ def generate_metrics(self, context: Context) -> List[Metric]:
         ]
 
     def render(self, context: Context, results: Dict[MetricId, MetricResult]) -> List[BaseWidgetInfo]:
-        widgets = context.get_legacy_metric(RegressionQualityMetric())[1]
+        widgets = context.get_legacy_metric(
+            RegressionQualityMetric(),
+            _gen_regression_input_data,
+        )[1]
         if self._pred_actual_plot:
-            widgets += context.get_legacy_metric(RegressionPredictedVsActualPlot())[1]
+            widgets += context.get_legacy_metric(
+                RegressionPredictedVsActualPlot(),
+                _gen_regression_input_data,
+            )[1]
         if self._error_plot:
-            widgets += context.get_legacy_metric(RegressionErrorPlot())[1]
+            widgets += context.get_legacy_metric(
+                RegressionErrorPlot(),
+                _gen_regression_input_data,
+            )[1]
         if self._error_distr:
-            widgets += context.get_legacy_metric(RegressionErrorDistribution())[1]
+            widgets += context.get_legacy_metric(
+                RegressionErrorDistribution(),
+                _gen_regression_input_data,
+            )[1]
         return widgets
 
 
@@ -64,7 +77,10 @@ def generate_metrics(self, context: Context) -> List[Metric]:
         ]
 
     def render(self, context: Context, results: Dict[MetricId, MetricResult]) -> List[BaseWidgetInfo]:
-        return context.get_legacy_metric(RegressionDummyMetric())[1]
+        return context.get_legacy_metric(
+            RegressionDummyMetric(),
+            _gen_regression_input_data,
+        )[1]
 
 
 class RegressionPreset(MetricContainer):

src/evidently/future/report.py

@@ -140,15 +140,15 @@ def get_reference_metric_result(self, metric: Metric) -> MetricResult:
     def get_legacy_metric(
         self,
         metric: LegacyMetric[T],
-        input_data_generator: Optional[Callable[["Context"], InputData]] = None,
+        input_data_generator: Optional[Callable[["Context"], InputData]],
     ) -> Tuple[T, List[BaseWidgetInfo]]:
         if input_data_generator is None:
             input_data_generator = _default_input_data_generator
         input_data = input_data_generator(self)
         dependencies = _discover_dependencies(metric)
         for _, obj in dependencies:
             if isinstance(obj, LegacyMetric):
-                (result, render) = self.get_legacy_metric(obj)
+                (result, render) = self.get_legacy_metric(obj, input_data_generator)
                 object.__setattr__(obj, "get_result", lambda: result)
             else:
                 raise ValueError(f"unexpected type {type(obj)}")