Merge pull request #52 from edx/dylanrhodes/consolidate-variants

Consolidation parameter for extraneous variants

analytics_data_api/utils.py

+from collections import defaultdict
+
 from django.db.models import Q
 from rest_framework.authtoken.models import Token
 
@@ -31,3 +33,53 @@ def set_user_auth_token(user, key):
     Token.objects.create(user=user, key=key)
 
     print "Set API key for user %s to %s" % (user, key)
+
+
+def matching_tuple(answer):
+    """ Return tuple containing values which must match for consolidation. """
+    return (
+        answer.question_text,
+        answer.answer_value_text,
+        answer.answer_value_numeric,
+        answer.problem_display_name,
+        answer.correct,
+    )
+
+
+def consolidate_answers(problem):
+    """ Attempt to consolidate erroneously randomized answers. """
+    answer_sets = defaultdict(list)
+    match_tuple_sets = defaultdict(set)
+
+    for answer in problem:
+        answer.consolidated_variant = False
+
+        answer_sets[answer.value_id].append(answer)
+        match_tuple_sets[answer.value_id].add(matching_tuple(answer))
+
+    # If a part has more than one unique tuple of matching fields, do not consolidate.
+    for _, match_tuple_set in match_tuple_sets.iteritems():
+        if len(match_tuple_set) > 1:
+            return problem
+
+    consolidated_answers = []
+
+    for _, answers in answer_sets.iteritems():
+        consolidated_answer = None
+
+        if len(answers) == 1:
+            consolidated_answers.append(answers[0])
+            continue
+
+        for answer in answers:
+            if not consolidated_answer:
+                consolidated_answer = answer
+
+                consolidated_answer.variant = None
+                consolidated_answer.consolidated_variant = True
+            else:
+                consolidated_answer.count += answer.count
+
+        consolidated_answers.append(consolidated_answer)
+
+    return consolidated_answers

analytics_data_api/v0/serializers.py

@@ -75,6 +75,29 @@ class ProblemResponseAnswerDistributionSerializer(ModelSerializerWithCreatedFiel
         )
 
 
+class ConsolidatedAnswerDistributionSerializer(ProblemResponseAnswerDistributionSerializer):
+    """
+    Serializer for consolidated answer distributions.
+    """
+
+    consolidated_variant = serializers.BooleanField()
+
+    class Meta(ProblemResponseAnswerDistributionSerializer.Meta):
+        fields = ProblemResponseAnswerDistributionSerializer.Meta.fields + ('consolidated_variant',)
+
+    # pylint: disable=super-on-old-class
+    def restore_object(self, attrs, instance=None):
+        """
+        Pops and restores non-model field.
+        """
+
+        consolidated_variant = attrs.pop('consolidated_variant', None)
+        distribution = super(ConsolidatedAnswerDistributionSerializer, self).restore_object(attrs, instance)
+        distribution.consolidated_variant = consolidated_variant
+
+        return distribution
+
+
 class GradeDistributionSerializer(ModelSerializerWithCreatedField):
     """
     Representation of the grade_distribution table without id

analytics_data_api/v0/tests/views/test_problems.py

@@ -20,23 +20,107 @@ class AnswerDistributionTests(TestCaseWithAuthentication):
     @classmethod
     def setUpClass(cls):
         cls.course_id = "org/num/run"
-        cls.module_id = "i4x://org/num/run/problem/RANDOMNUMBER"
-        cls.part_id1 = "i4x-org-num-run-problem-RANDOMNUMBER_2_1"
+        cls.module_id1 = "i4x://org/num/run/problem/RANDOMNUMBER"
+        cls.module_id2 = "i4x://org/num/run/problem/OTHERRANDOM"
+        cls.part_id = "i4x-org-num-run-problem-RANDOMNUMBER_2_1"
+        cls.correct = True
+        cls.value_id1 = '3'
+        cls.value_id2 = '4'
+        cls.answer_value_text = '3'
+        cls.answer_value_numeric = 3.0
+        cls.problem_display_name = 'Test Problem'
+        cls.question_text = 'Question Text'
+
         cls.ad1 = G(
             models.ProblemResponseAnswerDistribution,
             course_id=cls.course_id,
-            module_id=cls.module_id,
-            part_id=cls.part_id1
+            module_id=cls.module_id1,
+            part_id=cls.part_id,
+            correct=cls.correct,
+            value_id=cls.value_id1,
+            answer_value_text=cls.answer_value_text,
+            answer_value_numeric=cls.answer_value_numeric,
+            problem_display_name=cls.problem_display_name,
+            question_text=cls.question_text,
+            variant=123,
+            count=1
+        )
+        cls.ad2 = G(
+            models.ProblemResponseAnswerDistribution,
+            course_id=cls.course_id,
+            module_id=cls.module_id1,
+            part_id=cls.part_id,
+            correct=cls.correct,
+            value_id=cls.value_id1,
+            answer_value_text=cls.answer_value_text,
+            answer_value_numeric=cls.answer_value_numeric,
+            problem_display_name=cls.problem_display_name,
+            question_text=cls.question_text,
+            variant=345,
+            count=2
+        )
+        cls.ad3 = G(
+            models.ProblemResponseAnswerDistribution,
+            course_id=cls.course_id,
+            module_id=cls.module_id1,
+            part_id=cls.part_id,
+        )
+        cls.ad4 = G(
+            models.ProblemResponseAnswerDistribution,
+            course_id=cls.course_id,
+            module_id=cls.module_id2,
+            part_id=cls.part_id,
+            value_id=cls.value_id1,
+            correct=True,
+        )
+        cls.ad5 = G(
+            models.ProblemResponseAnswerDistribution,
+            course_id=cls.course_id,
+            module_id=cls.module_id2,
+            part_id=cls.part_id,
+            value_id=cls.value_id2,
+            correct=True
+        )
+        cls.ad6 = G(
+            models.ProblemResponseAnswerDistribution,
+            course_id=cls.course_id,
+            module_id=cls.module_id2,
+            part_id=cls.part_id,
+            value_id=cls.value_id1,
+            correct=False,
         )
 
-    def test_get(self):
-        response = self.authenticated_get('/api/v0/problems/%s%s' % (self.module_id, self.path))
+    def test_nonconsolidated_get(self):
+        """ Verify that answers which should not be consolidated are not. """
+        response = self.authenticated_get('/api/v0/problems/%s%s' % (self.module_id2, self.path))
         self.assertEquals(response.status_code, 200)
 
-        expected_dict = ProblemResponseAnswerDistributionSerializer(self.ad1).data
-        actual_list = response.data
-        self.assertEquals(len(actual_list), 1)
-        self.assertDictEqual(actual_list[0], expected_dict)
+        expected_data = models.ProblemResponseAnswerDistribution.objects.filter(module_id=self.module_id2)
+        expected_data = [ProblemResponseAnswerDistributionSerializer(answer).data for answer in expected_data]
+
+        for answer in expected_data:
+            answer['consolidated_variant'] = False
+
+        self.assertEqual(response.data, expected_data)
+
+    def test_consolidated_get(self):
+        """ Verify that valid consolidation does occur. """
+        response = self.authenticated_get(
+            '/api/v0/problems/{0}{1}'.format(self.module_id1, self.path))
+        self.assertEquals(response.status_code, 200)
+
+        expected_data = [
+            ProblemResponseAnswerDistributionSerializer(self.ad1).data,
+            ProblemResponseAnswerDistributionSerializer(self.ad3).data,
+        ]
+
+        expected_data[0]['count'] += self.ad2.count
+        expected_data[0]['variant'] = None
+        expected_data[0]['consolidated_variant'] = True
+
+        expected_data[1]['consolidated_variant'] = False
+
+        self.assertEquals(response.data, expected_data)
 
     def test_get_404(self):
         response = self.authenticated_get('/api/v0/problems/%s%s' % ("DOES-NOT-EXIST", self.path))

analytics_data_api/v0/views/problems.py

+"""
+API methods for module level data.
+"""
+
+from itertools import groupby
+
 from rest_framework import generics
 
 from analytics_data_api.v0.models import ProblemResponseAnswerDistribution
-from analytics_data_api.v0.serializers import ProblemResponseAnswerDistributionSerializer
+from analytics_data_api.v0.serializers import ConsolidatedAnswerDistributionSerializer
 from analytics_data_api.v0.models import GradeDistribution
 from analytics_data_api.v0.serializers import GradeDistributionSerializer
 from analytics_data_api.v0.models import SequentialOpenDistribution
 from analytics_data_api.v0.serializers import SequentialOpenDistributionSerializer
+from analytics_data_api.utils import consolidate_answers
 
 
 class ProblemResponseAnswerDistributionView(generics.ListAPIView):
@@ -36,15 +43,30 @@ class ProblemResponseAnswerDistributionView(generics.ListAPIView):
             * variant: For randomized problems, the random seed used. If problem
               is not randomized, value is null.
             * created: The date the count was computed.
+
+    **Parameters**
+
+        You can request consolidation of response counts for erroneously randomized problems.
+
+        consolidate_variants -- If True, attempt to consolidate responses, otherwise, do not.
+
     """
 
-    serializer_class = ProblemResponseAnswerDistributionSerializer
+    serializer_class = ConsolidatedAnswerDistributionSerializer
     allow_empty = False
 
     def get_queryset(self):
         """Select all the answer distribution response having to do with this usage of the problem."""
         problem_id = self.kwargs.get('problem_id')
-        return ProblemResponseAnswerDistribution.objects.filter(module_id=problem_id)
+
+        queryset = ProblemResponseAnswerDistribution.objects.filter(module_id=problem_id).order_by('part_id')
+
+        consolidated_rows = []
+
+        for _, part in groupby(queryset, lambda x: x.part_id):
+            consolidated_rows += consolidate_answers(list(part))
+
+        return consolidated_rows
 
 
 class GradeDistributionView(generics.ListAPIView):