Predictions: allow predicting probabilities for classless data

Before, combo box for selection of class probabilities was only shown if the
data on the input had a dicrete class. Now, the combo is shown as soon as
there is one predictor which works with discrete class vars. If class is
missing from the file, only model probabilities options are shown there.

Author: markotoplak

Orange/widgets/evaluate/owpredictions.py

@@ -266,10 +266,16 @@ def _set_target_combos(self):
             self.target_class = self.TARGET_AVERAGE
         else:
             self.shown_probs = self.NO_PROBS
+            model = prob_combo.model()
+            for v in (self.DATA_PROBS, self.BOTH_PROBS):
+                item = model.item(v)
+                item.setFlags(item.flags() & ~Qt.ItemIsEnabled)
 
     def _update_control_visibility(self):
         for widget in self._prob_controls:
-            widget.setVisible(self.is_discrete_class)
+            widget.setVisible(any((slot.predictor.domain.class_var is not None and
+                                  slot.predictor.domain.class_var.is_discrete)
+                                  for slot in self.predictors))
 
         for widget in self._target_controls:
             widget.setVisible(self.is_discrete_class and self.show_scores)

Orange/widgets/evaluate/tests/test_owpredictions.py

@@ -37,6 +37,7 @@ class TestOWPredictions(WidgetTest):
     def setUp(self):
         self.widget = self.create_widget(OWPredictions)  # type: OWPredictions
         self.iris = Table("iris")
+        self.iris_classless = self.iris.transform(Domain(self.iris.domain.attributes, []))
         self.housing = Table("housing")
 
     def test_rowCount_from_model(self):
@@ -718,16 +719,32 @@ def test_update_delegates_continuous(self):
 
         widget.data = Table.from_list(Domain([], ContinuousVariable("c")), [])
 
+        # only regression
+        all_predictors = widget.predictors
+        widget.predictors = [widget.predictors[-1]]
         widget._update_control_visibility()
         self.assertTrue(widget.controls.shown_probs.isHidden())
         self.assertTrue(widget.controls.target_class.isHidden())
 
+        # regression and classification
+        widget.predictors = all_predictors
+        widget._update_control_visibility()
+        self.assertFalse(widget.controls.shown_probs.isHidden())
+        self.assertTrue(widget.controls.target_class.isHidden())
+
         widget._set_class_values()
         self.assertEqual(widget.class_values, list("abcde"))
 
         widget._set_target_combos()
         self.assertEqual(widget.shown_probs, widget.NO_PROBS)
 
+        def is_enabled(prob_item):
+            return widget.controls.shown_probs.model().item(prob_item).flags() & Qt.ItemIsEnabled
+        self.assertTrue(is_enabled(widget.NO_PROBS))
+        self.assertTrue(is_enabled(widget.MODEL_PROBS))
+        self.assertFalse(is_enabled(widget.DATA_PROBS))
+        self.assertFalse(is_enabled(widget.BOTH_PROBS))
+
         widget._update_prediction_delegate()
         for delegate in widget._delegates:
             self.assertEqual(list(delegate.shown_probabilities), [])
@@ -857,6 +874,37 @@ def test_output_regression(self):
             out.metas,
             np.hstack([pred.results.predicted.T for pred in widget.predictors]))
 
+    def test_classless(self):
+        widget = self.widget
+        iris012 = self.iris
+        purge = Remove(class_flags=Remove.RemoveUnusedValues)
+        iris01 = purge(iris012[:100])
+        iris12 = purge(iris012[50:])
+
+        bayes01 = NaiveBayesLearner()(iris01)
+        bayes12 = NaiveBayesLearner()(iris12)
+        bayes012 = NaiveBayesLearner()(iris012)
+
+        self.send_signal(widget.Inputs.data, self.iris_classless)
+        self.send_signal(widget.Inputs.predictors, bayes01, 0)
+        self.send_signal(widget.Inputs.predictors, bayes12, 1)
+        self.send_signal(widget.Inputs.predictors, bayes012, 2)
+
+        for i, pred in enumerate(widget.predictors):
+            p = pred.results.unmapped_probabilities
+            p[0] = 10 + 100 * i + np.arange(p.shape[1])
+            pred.results.unmapped_predicted[:] = i
+
+        widget.shown_probs = widget.NO_PROBS
+        widget._commit_predictions()
+        out = self.get_output(widget.Outputs.predictions)
+        self.assertEqual(list(out.metas[0]), [0, 1, 2])
+
+        widget.shown_probs = widget.MODEL_PROBS
+        widget._commit_predictions()
+        out = self.get_output(widget.Outputs.predictions)
+        self.assertEqual(list(out.metas[0]), [0, 10, 11, 1, 110, 111, 2, 210, 211, 212])
+
     @patch("Orange.widgets.evaluate.owpredictions.usable_scorers",
            Mock(return_value=[_Scorer]))
     def test_change_target(self):