Merge pull request #4893 from edx/web-analytics/pass-ga-cookie

Re-use GA cookie when sending server-side events to Segment.io

common/djangoapps/student/models.py

@@ -756,6 +756,7 @@ class CourseEnrollment(models.Model):
                 tracker.emit(event_name, data)
 
                 if settings.FEATURES.get('SEGMENT_IO_LMS') and settings.SEGMENT_IO_LMS_KEY:
+                    tracking_context = tracker.get_tracker().resolve_context()
                     analytics.track(self.user_id, event_name, {
                         'category': 'conversion',
                         'label': self.course_id.to_deprecated_string(),
@@ -763,7 +764,12 @@ class CourseEnrollment(models.Model):
                         'course': self.course_id.course,
                         'run': self.course_id.run,
                         'mode': self.mode,
+                    }, context={
+                        'Google Analytics': {
+                            'clientId': tracking_context.get('client_id')
+                        }
                     })
+
         except:  # pylint: disable=bare-except
             if event_name and self.course_id:
                 log.exception('Unable to emit event %s for user %s and course %s', event_name, self.user.username, self.course_id)

common/djangoapps/track/middleware.py

@@ -87,7 +87,7 @@ class TrackMiddleware(object):
         Extract information from the request and add it to the tracking
         context.
 
-        The following fields are injected in to the context:
+        The following fields are injected into the context:
 
         * session - The Django session key that identifies the user's session.
         * user_id - The numeric ID for the logged in user.
@@ -96,6 +96,7 @@ class TrackMiddleware(object):
         * host - The "SERVER_NAME" header, which should be the name of the server running this code.
         * agent - The client browser identification string.
         * path - The path part of the requested URL.
+        * client_id - The unique key used by Google Analytics to identify a user
         """
         context = {
             'session': self.get_session_key(request),
@@ -105,6 +106,14 @@ class TrackMiddleware(object):
         for header_name, context_key in META_KEY_TO_CONTEXT_KEY.iteritems():
             context[context_key] = request.META.get(header_name, '')
 
+        # Google Analytics uses the clientId to keep track of unique visitors. A GA cookie looks like 
+        # this: _ga=GA1.2.1033501218.1368477899. The clientId is this part: 1033501218.1368477899.
+        google_analytics_cookie = request.COOKIES.get('_ga')
+        if google_analytics_cookie is None:
+            context['client_id'] = None
+        else:
+            context['client_id'] = '.'.join(google_analytics_cookie.split('.')[2:])
+
         context.update(contexts.course_context_from_url(request.build_absolute_uri()))
 
         tracker.get_tracker().enter_context(

common/djangoapps/track/shim.py

@@ -45,6 +45,10 @@ class LegacyFieldMappingProcessor(object):
 def remove_shim_context(event):
     if 'context' in event:
         context = event['context']
-        for field in CONTEXT_FIELDS_TO_INCLUDE:
+        # These fields are present elsewhere in the event at this point
+        context_fields_to_remove = set(CONTEXT_FIELDS_TO_INCLUDE)
+        # This field is only used for Segment.io web analytics and does not concern researchers
+        context_fields_to_remove.add('client_id')
+        for field in context_fields_to_remove:
             if field in context:
                 del context[field]

common/djangoapps/track/tests/test_middleware.py

@@ -64,6 +64,7 @@ class TrackMiddlewareTestCase(TestCase):
             'path': '/courses/',
             'org_id': '',
             'course_id': '',
+            'client_id': None,
         })
 
     def get_context_for_path(self, path):

common/djangoapps/track/tests/test_shim.py

@@ -50,6 +50,7 @@ class LegacyFieldMappingProcessorTestCase(TestCase):
             'course_id': sentinel.course_id,
             'org_id': sentinel.org_id,
             'event_type': sentinel.event_type,
+            'client_id': sentinel.client_id,
         }
         with django_tracker.context('test', context):
             django_tracker.emit(sentinel.name, data)

common/djangoapps/track/tests/test_views.py

@@ -144,6 +144,39 @@ class TestTrackViews(TestCase):
 
         self.mock_tracker.send.assert_called_once_with(expected_event)
 
+    @freeze_time(expected_time)
+    def test_server_track_with_middleware_and_google_analytics_cookie(self):
+        middleware = TrackMiddleware()
+        request = self.request_factory.get(self.path_with_course)
+        request.COOKIES['_ga'] = 'GA1.2.1033501218.1368477899'
+        middleware.process_request(request)
+        # The middleware emits an event, reset the mock to ignore it since we aren't testing that feature.
+        self.mock_tracker.reset_mock()
+        try:
+            views.server_track(request, str(sentinel.event_type), '{}')
+
+            expected_event = {
+                'username': 'anonymous',
+                'ip': '127.0.0.1',
+                'event_source': 'server',
+                'event_type': str(sentinel.event_type),
+                'event': '{}',
+                'agent': '',
+                'page': None,
+                'time': expected_time,
+                'host': 'testserver',
+                'context': {
+                    'user_id': '',
+                    'course_id': u'foo/bar/baz',
+                    'org_id': 'foo',
+                    'path': u'/courses/foo/bar/baz/xmod/'
+                },
+            }
+        finally:
+            middleware.process_response(request, None)
+
+        self.mock_tracker.send.assert_called_once_with(expected_event)
+
     @freeze_time(expected_time)
     def test_server_track_with_no_request(self):
         request = None