Merge pull request #42 from biolab/dev

[ENH] use new interface for testing and scoring learners

orangecontrib/survival_analysis/evaluation/scoring.py

@@ -1,7 +1,7 @@
 from lifelines.utils import concordance_index
-from Orange.data import DiscreteVariable, ContinuousVariable
+from Orange.data import DiscreteVariable, ContinuousVariable, Domain
 from Orange.evaluation.scoring import Score
-from orangecontrib.survival_analysis.widgets.data import get_survival_endpoints
+from orangecontrib.survival_analysis.widgets.data import get_survival_endpoints, contains_survival_endpoints
 
 __all__ = ['ConcordanceIndex']
 
@@ -11,8 +11,10 @@ class SurvivalScorer(Score, abstract=True):
         ContinuousVariable,
         DiscreteVariable,
     )
-    is_built_in = False
-    problem_type = 'time_to_event'
+
+    @staticmethod
+    def is_compatible(domain: Domain) -> bool:
+        return contains_survival_endpoints(domain)
 
 
 class ConcordanceIndex(SurvivalScorer):

orangecontrib/survival_analysis/modeling/cox.py

@@ -8,9 +8,9 @@
 from orangecontrib.survival_analysis.widgets.data import (
     contains_survival_endpoints,
     get_survival_endpoints,
+    MISSING_SURVIVAL_DATA,
 )
 
-
 def to_data_frame(table: Table) -> pd.DataFrame:
     columns = table.domain.attributes + table.domain.class_vars
     df = pd.DataFrame({col.name: table.get_column_view(col)[0] for col in columns})
@@ -47,14 +47,14 @@ def __call__(self, data, ret=Model.Value):
 class CoxRegressionLearner(Learner):
     __returns__ = CoxRegressionModel
     supports_multiclass = True
-    learner_adequacy_err_msg = 'Survival variables expected.  Use As Survival Data widget.'
 
     def __init__(self, preprocessors=None, **kwargs):
         self.params = vars()
         super().__init__(preprocessors=preprocessors)
 
-    def check_learner_adequacy(self, domain):
-        return len(domain.class_vars) == 2
+    def incompatibility_reason(self, domain):
+        if not contains_survival_endpoints(domain):
+            return MISSING_SURVIVAL_DATA
 
     def fit_storage(self, data):
         return self.fit(data)
@@ -67,7 +67,7 @@ def _fit_model(self, data):
 
     def fit(self, data):
         if not contains_survival_endpoints(data.domain):
-            raise ValueError(self.learner_adequacy_err_msg)
+            raise ValueError(MISSING_SURVIVAL_DATA)
         time_var, event_var = get_survival_endpoints(data.domain)
 
         df = to_data_frame(data)

orangecontrib/survival_analysis/widgets/data.py

@@ -5,9 +5,15 @@
 from Orange.data import Table, Domain, Variable
 
 
-TIME_VAR = 'time'
-EVENT_VAR = 'event'
-TIME_TO_EVENT_VAR = '_time_to_event_var'
+TIME_VAR: str = 'time'
+EVENT_VAR: str = 'event'
+TIME_TO_EVENT_VAR: str = '_time_to_event_var'
+
+# Error/Warning messages related to survival data tables.
+MISSING_ROWS: str = 'Rows with missing values detected. They will be omitted.'
+MISSING_SURVIVAL_DATA: str = (
+    'No survival data detected. ' 'Use the "As Survival Data" widget or consult the documentation.'
+)
 
 
 def contains_survival_endpoints(domain: Domain):
@@ -35,15 +41,13 @@ def get_survival_endpoints(domain: Domain) -> Tuple[Optional[Variable], Optional
 
 def check_survival_data(f):
     """Check for survival data."""
-    error_msg = 'No survival data detected. Use the "As Survival Data" widget or consult the documentation.'
-    warning_msg = 'Rows with missing values detected. They will be omitted.'
 
     @wraps(f)
     def wrapper(widget, data: Table, *args, **kwargs):
-        widget.Error.add_message('missing_survival_data', UnboundMsg(error_msg))
+        widget.Error.add_message('missing_survival_data', UnboundMsg(MISSING_SURVIVAL_DATA))
         widget.Error.missing_survival_data.clear()
 
-        widget.Warning.add_message('missing_values_detected', UnboundMsg(warning_msg))
+        widget.Warning.add_message('missing_values_detected', UnboundMsg(MISSING_ROWS))
         widget.Warning.missing_values_detected.clear()
 
         if data is not None and isinstance(data, Table):

orangecontrib/survival_analysis/widgets/owcoxregression.py

@@ -123,8 +123,9 @@ def check_data(self):
         self.Error.sparse_not_supported.clear()
         if self.data is not None and self.learner is not None:
             self.Error.data_error.clear()
-            if not self.learner.check_learner_adequacy(self.data.domain):
-                self.Error.data_error(self.learner.learner_adequacy_err_msg)
+            incompatibility_reason = self.learner.incompatibility_reason(self.data.domain)
+            if incompatibility_reason is not None:
+                self.Error.data_error(incompatibility_reason)
             elif not len(self.data):
                 self.Error.data_error("Dataset is empty.")
             elif self.data.X.size == 0: