Fixing Evaluation class and testing

Author: mmerce

bigml/evaluation.py

@@ -26,7 +26,7 @@
 from bigml.basemodel import retrieve_resource, get_resource_dict
 
 CLASSIFICATION_METRICS = [
-    "accuracy", "precision", "recall", "phi" "phi_coefficient",
+    "accuracy", "precision", "recall", "phi", "phi_coefficient",
     "f_measure", "confusion_matrix", "per_class_statistics"]
 
 REGRESSION_METRICS = ["mean_absolute_error", "mean_squared_error", "r_squared"]
@@ -35,6 +35,7 @@
 class ClassificationEval():
     """A class to store the classification metrics """
     def __init__(self, name, per_class_statistics):
+
         self.name = name
         for statistics in per_class_statistics:
             if statistics["class_name"] == name:

bigml/tests/create_evaluation_steps.py

@@ -15,11 +15,14 @@
 # License for the specific language governing permissions and limitations
 # under the License.
 
+import json
+
 from bigml.api import HTTP_CREATED
 from bigml.api import FINISHED, FAULTY
+from bigml.evaluation import Evaluation
 
 from .read_resource_steps import wait_until_status_code_is
-from .world import world, eq_, ok_
+from .world import world, eq_, ok_, res_filename, approx_
 
 def i_create_an_evaluation(step, shared=None):
     """Step: I create an evaluation for the model with the dataset"""
@@ -105,5 +108,18 @@ def the_measured_measure_is_value(step, measure, value):
 
 
 def the_measured_measure_is_greater_value(step, measure, value):
-    """#@step(r'the measured <measure> is greater than <value>"""
+    """Step: the measured <measure> is greater than <value>"""
     ok_(float(world.evaluation['result']['model'][measure]) > float(value))
+
+def i_create_a_local_evaluation(step, filename):
+    """Step: I create an Evaluation from the JSON file"""
+    filename = res_filename(filename)
+    with open(filename) as handler:
+        evaluation = json.load(handler)
+    local_evaluation = Evaluation(evaluation)
+    step.bigml["local_evaluation"] = local_evaluation
+
+def the_local_metric_is_value(step, metric, value):
+    """Step: The metric in the local evaluation is <value> """
+    approx_(getattr(step.bigml["local_evaluation"], metric), value,
+            precision=4)

bigml/tests/test_14_create_evaluations.py

@@ -214,3 +214,29 @@ def test_scenario4(self):
                 self, example["evaluation_wait"])
             evaluation_create.the_measured_measure_is_value(
                 self, example["metric"], example["value"])
+
+    def test_scenario5(self):
+        """
+        Scenario5: Successfully instantiating Evaluation:
+            Given a stored evaluation "<data>" file
+            When I create an Evaluation for the JSON
+            Then the measured "<metric>" is <value>
+        """
+        show_doc(self.test_scenario5)
+        headers = ["data", "metric", "value"]
+        examples = [
+            ['data/classification_evaluation.json', 'phi',
+             0.64837],
+            ['data/classification_evaluation.json', 'accuracy',
+             0.91791],
+            ['data/classification_evaluation.json', 'precision',
+             0.86639],
+            ['data/regression_evaluation.json', 'r_squared',
+             0.9288]]
+        for example in examples:
+            example = dict(zip(headers, example))
+            show_method(self, self.bigml["method"], example)
+            evaluation_create.i_create_a_local_evaluation(
+                self, example["data"])
+            evaluation_create.the_local_metric_is_value(
+                self, example["metric"], example["value"])

bigml/tests/world.py

@@ -179,9 +179,8 @@ def approx_(number_a, number_b, msg=None, precision=5):
     """Wrapper for pytest approx function"""
     epsilon = math.pow(0.1, precision)
     if msg is None:
-        assert number_a == pytest.approx(number_b, abs=epsilon)
-    else:
-        assert number_a == pytest.approx(number_b, abs=epsilon), msg
+        msg = "%s != %s" % (repr(number_a), repr(number_b))
+    assert number_a == pytest.approx(number_b, abs=epsilon), msg
 
 
 class World: