[FIX] Predictions: Handle discrete target with no values

Orange/base.py

@@ -259,13 +259,13 @@ def backmap_value(self, value, mapped_probs, n_values, backmappers):
 
         value = backmapper(value)
         nans = np.isnan(value)
-        if not np.any(nans):
+        if not np.any(nans) or n_values[0] < 2:
             return value
         if mapped_probs is not None:
             value[nans] = np.argmax(mapped_probs[nans], axis=1)
         else:
-            value[nans] = np.RandomState(0).choice(
-                backmapper(np.arange(0, n_values[0] - 1))
+            value[nans] = np.random.RandomState(0).choice(
+                backmapper(np.arange(0, n_values[0] - 1)),
                 (np.sum(nans), ))
         return value
 

Orange/widgets/evaluate/owconfusionmatrix.py

@@ -118,6 +118,7 @@ class Outputs:
     class Error(widget.OWWidget.Error):
         no_regression = Msg("Confusion Matrix cannot show regression results.")
         invalid_values = Msg("Evaluation Results input contains invalid values")
+        empty_input = widget.Msg("Empty result on input. Nothing to display.")
 
     def __init__(self):
         super().__init__()
@@ -245,11 +246,14 @@ def set_results(self, results):
         if results is not None and results.data is not None:
             data = results.data[results.row_indices]
 
+        self.Error.no_regression.clear()
+        self.Error.empty_input.clear()
         if data is not None and not data.domain.has_discrete_class:
             self.Error.no_regression()
             data = results = None
-        else:
-            self.Error.no_regression.clear()
+        elif results is not None and not results.actual.size:
+            self.Error.empty_input()
+            data = results = None
 
         nan_values = False
         if results is not None:

Orange/widgets/evaluate/owrocanalysis.py

@@ -303,10 +303,6 @@ class OWROCAnalysis(widget.OWWidget):
     class Inputs:
         evaluation_results = Input("Evaluation Results", Orange.evaluation.Results)
 
-    class Warning(widget.OWWidget.Warning):
-        empty_results = widget.Msg(
-            "Empty results on input. There is nothing to display.")
-
     target_index = settings.Setting(0)
     selected_classifiers = []
 

Orange/widgets/evaluate/tests/test_owpredictions.py

@@ -7,8 +7,12 @@
 from Orange.data.io import TabReader
 from Orange.widgets.tests.base import WidgetTest
 from Orange.widgets.evaluate.owpredictions import OWPredictions
+from Orange.widgets.evaluate.owcalibrationplot import OWCalibrationPlot
+from Orange.widgets.evaluate.owconfusionmatrix import OWConfusionMatrix
+from Orange.widgets.evaluate.owliftcurve import OWLiftCurve
+from Orange.widgets.evaluate.owrocanalysis import OWROCAnalysis
 
-from Orange.data import Table, Domain
+from Orange.data import Table, Domain, DiscreteVariable
 from Orange.modelling import ConstantLearner, TreeLearner
 from Orange.evaluation import Results
 from Orange.widgets.tests.utils import excepthook_catch
@@ -49,6 +53,38 @@ def test_nan_target_input(self):
         evres = self.get_output(self.widget.Outputs.evaluation_results)
         self.assertEqual(len(evres.data), 0)
 
+    def test_no_values_target(self):
+        train = Table("titanic")
+        model = ConstantLearner()(train)
+        self.send_signal(self.widget.Inputs.predictors, model)
+        domain = Domain([DiscreteVariable("status", values=["first", "third"]),
+                         DiscreteVariable("age", values=["adult", "child"]),
+                         DiscreteVariable("sex", values=["female", "male"])],
+                        [DiscreteVariable("survived", values=[])])
+        test = Table(domain, np.array([[0, 0, 1], [0, 1, 0], [1, 0, 0]]),
+                     np.full((3, 1), np.nan))
+        self.send_signal(self.widget.Inputs.data, test)
+        pred = self.get_output(self.widget.Outputs.predictions)
+        self.assertEqual(len(pred), len(test))
+
+        results = self.get_output(self.widget.Outputs.evaluation_results)
+
+        cm_widget = self.create_widget(OWConfusionMatrix)
+        self.send_signal(cm_widget.Inputs.evaluation_results, results,
+                         widget=cm_widget)
+
+        ra_widget = self.create_widget(OWROCAnalysis)
+        self.send_signal(ra_widget.Inputs.evaluation_results, results,
+                         widget=ra_widget)
+
+        lc_widget = self.create_widget(OWLiftCurve)
+        self.send_signal(lc_widget.Inputs.evaluation_results, results,
+                         widget=lc_widget)
+
+        cp_widget = self.create_widget(OWCalibrationPlot)
+        self.send_signal(cp_widget.Inputs.evaluation_results, results,
+                         widget=cp_widget)
+
     def test_mismatching_targets(self):
         warning = self.widget.Warning
 

Orange/widgets/evaluate/utils.py

@@ -32,6 +32,9 @@ def anynan(a):
     elif not results.data.domain.has_discrete_class:
         error_group.invalid_results(
             "Discrete outcome variable is required")
+    elif not results.actual.size:
+        error_group.invalid_results(
+            "Empty result on input. Nothing to display.")
     elif check_nan and (anynan(results.actual) or
                         anynan(results.predicted) or
                         (results.probabilities is not None and