Suppress variant from non-submission problem-check events if not rerandomize.

Change-Id: I44b132e333e20fb32d032722406b61748a5a71fd

edx/analytics/tasks/answer_dist.py

@@ -175,7 +175,7 @@ class LastProblemCheckEventMixin(object):
                     if answer_id not in correct_map:
                         log.error("Unexpected answer_id %s not in correct_map: %s", answer_id, event)
                         continue
-                    correctness = correct_map[answer_id].get('correctness') in ['correct']
+                    correctness = correct_map[answer_id].get('correctness') == 'correct'
 
                     variant = event.get('state', {}).get('seed')
 
@@ -275,7 +275,8 @@ class AnswerDistributionPerCourseMixin(object):
         problem_id = most_recent_answer.get('problem_id')
         problem_display_name = most_recent_answer.get('problem_display_name')
         most_recent_question = most_recent_answer.get('question', '')
-        answer_uses_code = ('answer_value_id' in most_recent_answer)
+        answer_uses_value_id = ('answer_value_id' in most_recent_answer)
+        answer_uses_variant = (most_recent_answer.get('variant', '') != '')
         answer_dist = {}
         for _timestamp, value_string in reversed(values):
             answer = json.loads(value_string)
@@ -291,7 +292,7 @@ class AnswerDistributionPerCourseMixin(object):
             # We only want this from the most recent answer that has
             # this value.
             if answer_grouping_key not in answer_dist:
-                if answer_uses_code:
+                if answer_uses_value_id:
                     # The most recent overall answer indicates that
                     # the code should be returned as such.  If this
                     # particular answer did not have 'submission'
@@ -317,19 +318,21 @@ class AnswerDistributionPerCourseMixin(object):
                 # If there is no variant, then the question should be
                 # the same, and we want to go with the most recently
                 # defined value.
-                if answer.get('variant'):
+                if answer_uses_variant:
                     question = answer.get('question', '')
+                    variant = answer.get('variant') or ''
                 else:
                     question = most_recent_question
+                    variant = ''
 
                 # Key values here should match those used in get_column_order().
                 answer_dist[answer_grouping_key] = {
                     'ModuleID': problem_id,
                     'PartID': answer_id,
-                    'ValueID': value_id,
-                    'AnswerValue': answer_value,
-                    'Variant': answer.get('variant'),
-                    'Problem Display Name': problem_display_name,
+                    'ValueID': value_id or '',
+                    'AnswerValue': answer_value or '',
+                    'Variant': variant,
+                    'Problem Display Name': problem_display_name or '',
                     'Question': question,
                     'Correct Answer': '1' if answer.get('correct') else '0',
                     'Count': 0,

edx/analytics/tasks/launchers/remote.py

@@ -28,6 +28,7 @@ def main():
 
     sys.exit(return_code)
 
+
 def run_task_playbook(arguments, uid):
     """
     Execute the ansible playbook that triggers and monitors the remote task execution.
@@ -39,6 +40,7 @@ def run_task_playbook(arguments, uid):
     extra_vars = convert_args_to_extra_vars(arguments, uid)
     return run_ansible(('task.yml', '-e', extra_vars), arguments.verbose, executable='ansible-playbook')
 
+
 def convert_args_to_extra_vars(arguments, uid):
     """
     Generate the set of variables that need to be passed in to ansible since they are expected to be set by the
@@ -60,6 +62,7 @@ def convert_args_to_extra_vars(arguments, uid):
         extra_vars['wait_for_task'] = True
     return ' '.join(["{}='{}'".format(k, extra_vars[k]) for k in extra_vars])
 
+
 def run_ansible(args, verbose, executable='ansible'):
     """
     Execute ansible passing in the provided arguments.
@@ -95,6 +98,7 @@ def run_ansible(args, verbose, executable='ansible'):
 
     return proc.returncode
 
+
 def download_logs(arguments, uid):
     """
     Connect to the remote machine and download the logs produced by luigi.

edx/analytics/tasks/tests/test_answer_dist.py

@@ -443,13 +443,13 @@ class AnswerDistributionPerCourseReduceTest(unittest.TestCase):
     def _get_expected_output(self, answer_data, **kwargs):
         """Get an expected reducer output based on the input."""
         expected_output = {
-            "Problem Display Name": answer_data.get('problem_display_name'),
+            "Problem Display Name": answer_data.get('problem_display_name') or "",
             "Count": 1,
             "PartID": self.answer_id,
-            "Question": answer_data.get('question'),
-            "AnswerValue": answer_data.get('answer'),
+            "Question": answer_data.get('question') or "",
+            "AnswerValue": answer_data.get('answer') or answer_data.get('answer_value_id') or "",
             "ValueID": "",
-            "Variant": answer_data.get('variant'),
+            "Variant": answer_data.get('variant') or "",
             "Correct Answer": "1" if answer_data['correct'] else '0',
             "ModuleID": self.problem_id,
         }
@@ -551,6 +551,23 @@ class AnswerDistributionPerCourseReduceTest(unittest.TestCase):
         expected_output_2 = self._get_expected_output(answer_data_2)
         self._check_output([input_data_1, input_data_2], (expected_output_1, expected_output_2))
 
+    def test_two_answer_event_different_old_and_new(self):
+        answer_data_1 = self._get_non_submission_answer_data(answer_value_id="first")
+        answer_data_2 = self._get_answer_data(problem_display_name=self.problem_display_name)
+        input_data_1 = (self.earlier_timestamp, json.dumps(answer_data_1))
+        input_data_2 = (self.timestamp, json.dumps(answer_data_2))
+        expected_output_2 = self._get_expected_output(answer_data_2)
+        # An older non-submission-based event should inherit some
+        # information from a newer submission-based event.
+        # In particular, the Variant, the Question, and Problem Display Name.
+        expected_output_1 = self._get_expected_output(
+            answer_data_1,
+            Variant="",
+            Question=expected_output_2['Question'],
+        )
+        expected_output_1['Problem Display Name'] = expected_output_2['Problem Display Name']
+        self._check_output([input_data_1, input_data_2], (expected_output_1, expected_output_2))
+
     def test_two_answer_event_different_variant(self):
         answer_data_1 = self._get_answer_data(variant=123)
         answer_data_2 = self._get_answer_data(variant=456)