Flows Documentation

Initalization

When a user loads the home timetable page, FeatureFlowView inside of timetable.utils is used to handle the request. On initial page load, the frontend requires some data to initialize the redux state, like information about the current user, the list of possible semesters for the school, and the list of student integrations. This initial data is created inside of the view, and passed in as a single json string in the response context:

class FeatureFlowView(ValidateSubdomainMixin, APIView):

    def get(self, request, *args, **kwargs):
        # ...gather values for init_data

        init_data = {
            'school': self.school,
            'currentUser': get_user_dict(self.school, self.student, sem),
            'currentSemester': curr_sem_index,
            'allSemesters': all_semesters,
            'uses12HrTime': self.school in AM_PM_SCHOOLS,
            'studentIntegrations': integrations,
            'examSupportedSemesters': map(all_semesters.index,
                                          final_exams_available.get(self.school, [])),

            'featureFlow': dict(feature_flow, name=self.feature_name)
        }

        return render(request, 'timetable.html', {'init_data': json.dumps(init_data)})

which makes the init_data variable accessible in timetable.html. This dumped json string is then passed to the frontend as a global variable:

<script type="text/javascript">
  var initData = "{{init_data|escapejs}}";
</script>

And then parsed inside of the setup() function in init.jsx

const setup = () => (dispatch) => {
    initData = JSON.parse(initData);

    // pass init data into the redux state
    dispatch({ type: ActionTypes.INIT_STATE, data: initData });

    // do other logic with initData...
};

In other words, the data that the frontend requires is retrieved/calculated inside of FeatureFlowView, and then passed to the frontend as global variable initData. The frontend then does any logic it needs based on that data inside of setup() in init.jsx. Any data that needs to be reused later on from initData should be passed in to the redux state so that the only global variable uses appear in setup().

Feature Flows

One such piece of data that is passed to the frontend is a featureFlow object. This object is obtained as the return value of .get_feature_flow(), in addition to a name: self.feature_name key value pair. In the default implementation, this is just the dictionary {name: None}:

class FeatureFlowView(ValidateSubdomainMixin, APIView):
    feature_name = None

    def get_feature_flow(self, request, *args, **kwargs):
        return {}

    def get(self, request, *args, **kwargs):
        ...
        feature_flow = self.get_feature_flow(request, *args, **kwargs)
        init_data = {
            ...
            'featureFlow': dict(feature_flow, name=self.feature_name)
        }

        return render(request, 'timetable.html', {'init_data': json.dumps(init_data)})

This feature flow value can be used to store any extra information that the frontend needs for any endpoints that would require initial data to be loaded. For example, when loading a timetable share link, the frontend also needs to get data about the timetable that is being shared - instead of making a request to the backend after page load, this information can be provided by the backend directly by passing this information in the feature flow. It is easy to write new views that pass different data and have custom logic by subclassing FeatureFlowView and overwriting the get_feature_flow() method and the .feature_name class attribute.

Having this data all stored under the key featureFlow in init_data ensures two things. Firstly, it makes explicit that there can only be one feature flow in play at a time (we can’t load a timetable share link and a course share link at the same time), and secondly, it allows the frontend to know where to look for any feature data and act accordingly. In practice, this is done by switching on the name of the feature flow:

const setup = () => (dispatch) => {
    initData = JSON.parse(initData);

    dispatch({ type: ActionTypes.INIT_STATE, data: initData });

    // do other logic with initData...

    dispatch(handleFlows(initData.featureFlow));
};

const handleFlows = featureFlow => (dispatch) => {
    switch (featureFlow.name) {
        case 'SIGNUP':
            dispatch({ type: ActionTypes.TRIGGER_SIGNUP_MODAL });
            break;
        case 'USER_ACQ':
            dispatch({ type: ActionTypes.TRIGGER_ACQUISITION_MODAL });
            break;
        case 'SHARE_TIMETABLE':
            dispatch({ type: ActionTypes.CACHED_TT_LOADED });
            dispatch(lockTimetable(featureFlow.sharedTimetable, true, initData.currentUser.isLoggedIn));
            break;
        // ... etc.
        default:
            // unexpected feature name
            break;
  }
};

Example

To help understand how feature flows work, let’s go through the code for an example feature flow: course sharing. In order to implement course sharing, we want to create a new view/endpoint that retrieves course data based on the url and passes it to the frontend, which would then update the redux state and dispatch an action to open the course modal.

We start be defining a new endpoint for this feature flow:

url(r'course/(?P<code>.+?)/(?P<sem_name>.+?)/(?P<year>.+?)/*$',
                   courses.views.CourseModal.as_view())

Then we create a new FeatureFlowView for this endpoint which needs to do two things: define a name for the feature flow, which the frontend look at to determine what action to do, and return the course data that the frontend needs inside of get_feature_flow():

class CourseModal(FeatureFlowView):
    feature_name = "SHARE_COURSE"

    def get_feature_flow(self, request, code, sem_name, year):
        semester, _ = Semester.objects.get_or_create(name=sem_name, year=year)
        code = code.upper()
        course = get_object_or_404(Course, school=self.school, code=code)
        course_json = get_detailed_course_json(self.school, course, semester, self.student)

        # analytics
        SharedCourseView.objects.create(
            student=self.student,
            shared_course=course,
        ).save()

        return {'sharedCourse': course_json, 'semester': semester}

The frontend can now add a new case in handleFlows to perform logic for this feature flow:

const handleFlows = featureFlow => (dispatch) => {
    switch (featureFlow.name) {
        ...
        case 'SHARE_COURSE':
            dispatch(setCourseInfo(featureFlow.sharedCourse));
            dispatch(fetchCourseClassmates(featureFlow.sharedCourse.id));
            break;
        // ... etc.
        default:
            // unexpected feature name
            break;
  }
};

Shortcuts

Some feature flows don’t require any extra data - they simply require the frontend to know that a feature flow is being run. For example, for the signup feature flow, loading the page at /signup should simply open the signup modal, which requires no extra logic or data other than knowing that it should occur. We could do this by writing a new view:

class SignupModal(FeatureFlowView):
    feature_name = "SIGNUP"

We do not need to implement .get_feature_flow() since the frontend doesn’t require any extra data and the default implementation already returns an empty dictionary. We can simplify this by simply declaring this view directly inside of the urls file:

url(r'^signup/*$/', FeatureFlowView.as_view(feature_name='SIGNUP')

see https://github.com/noahpresler/semesterly/pull/838 for the original pull request implementing feature flows