CSC/ECE 517 Spring 2022 - E2243. Refactor student teams functionality

This page contains information about Expertiza Issue E2243. Refactor student teams functionality which was a project in CSC517 Spring 2022.

Please see below for a description of the design of the project.

Background

The Expertiza project is software to create reusable learning objects through peer review. It also supports team projects, and the submission of almost any document type, including URLs and wiki pages.

Users & Participants: Everyone who took the CSC517 OODD class at NC State is a user in Expertiza, and instructors can add users as participants in an assignment so that they can access it to contribute, make changes and submit their work to be graded.

Almost all of the works within Assignments is done by participants, not users. All the tables use participant_id to track contributions to assignments, except teams_users. In the documentation, it can be noted that the team_users table references users instead of participants. This anomaly in the teams_users relation causes problems with how student teams are rendered in the UI, and it doesn’t mix well with the new functionality that was recently introduced.

teams_users should be changed to teams_participants. the team_id, user_id mapping in this table should be changed to team_id, participant_id. All the references of user_id should be updated to use participant_id.

Given Requirements

Changing the schema

• We need to rename the table table teams_users to teams_participants.
• Create a column that would be called “participant_id”, which will be a foreign key that references the participants table.
• Create a migration that fetches the assignment_id for each tuple from the teams table. We would find the participant with the help of the assignment_id we just got from the teams table and the user_id that is already present in the current table, and store its “id” in the participant_id column we created in the last step.
• Get rid of the user_id column.

Refactoring the teams rendering (/views/student_teams/view.html.erb)

The logic in the team's view (where participants can see their team members) that iterates and fetches information of all the team members and displays them would be refactored so that it matches our new design. The view is quite bloated and has lots of functionalities written on it. We’ll be making partials and placing those functionality in their specific partials.

Other Changes

Update logic for all the crud functionalities in teams_users controller (wherever applicable). Rename the teams_users_controller to teams_participants_controller (and all its references)

Schema Changes

In teams_users table, we can add a new column participant_id which is a foreign key from participants table and remove the column user_id. We plan to remove the user_id column only after data migration since the user_id field is used predominantly in several other controllers in a variety of ways, and as such its removal must be done carefully.

The participant_id field in the the teams_users relation is computed as follows:

• Finding Assignment ID By Team ID

The teams_users table contains team_id, user_id and duty_id but not the participant_id. Since the same user_id can be associated with multiple participant_ids (since the same user can participate in multiple assignments and gets a unique participant_id in each assignment they take part in), we need to get the user_ids in the assignment for which we need the particular user's participant_id. For this reason, we need to first find the assignment_id.

• Finding Participant ID By Assignment ID and User ID

With the assignment_id retrieved, we can find the list of participant_ids that participated in the assignment, and using the user_id, get the participant_id which is saved in the participants table.

ER diagram for updated schema

Data Migration

Analysis

The Participants relation contains the user_id attribute which is a foreign key from users table. teams_users contains user_id which is also a foreign key from users table. Although the user_id field corresponds to same attribute, we cannot use this attribute alone to find the respective participant id because user_id is not a primary key in the participants table. i.e. same user_id can be associated with multiple participants. This is because in expertiza, all the users are stored in users table but, for each assignment, for each user, an entry is added in participants table. So, each user will have an entry in participants tables for all the assignments.

An Assignment contains a list of participants (An Assignment has many Participants). Since a user can have only one entry in the participants table for each assignment, (assignment_id, user_id) pair will have only one participant id associated with it.

Proposed Implementation

Finding Assignment ID By Team ID

The teams_users table contains team_id, user_id and duty_id. Each team is associated with an assignment. The parent_id attribute in the teams table corresponds to assignment id. So, team_id in the teams_users table can be used to get the Team object using which we can fetch the assignment ID (parent_id).

# Code snippet to fetch assignment_id based on team_id
assignment_id = Team.find(team_id).select(:parent_id)

Finding Participant ID By Assignment ID and User ID

The Assignment object can be fetched using the assignment ID. Each assignment contains a list of participants. The Participant object contains a user_id attribute. Since, a user can have only one entry in the Participants table for an assignment, we can use user_id to find the user's participant_id from this list.

# Code snippet to fetch participant_id based on assignment_id and user_id
participant_id = Assignment.find(assignment_id).participants.find_by(:user_id user_id)

Brute Force Data Migration

The above mentioned code snippets can be used for each tuple in the teams_users table.

# Brute Force Code Snippet
TeamsUser.all.each do |teamsUser|
    assignment_id = Team.find(teamsUser.team_id).select(:parent_id)
    participant_id = Assignment.find(assignment_id).participants.find_by(:user_id teamsUser.user_id)
    teamsUser.update_attribute(:participant_id, participant_id)
end

Efficient Data Migration

In the brute force approach, database calls are made multiple times inside the loop. Moving the database calls out of the loop will give us a more efficient solution.

Steps:

1. Load all the TeamUser objects.
2. Load all the Team objects for all the team_ids in TeamUser object
3. Create a map with team_id as key and assignment_id as value. The resultant map usage: map[team_id] = assignment_id
4. Load all the Assignment objects for all the assignment_ids from the map in the previous step. Also eager fetch participants along with the assignments.
5. Create a nested map with assignment_id and user_id as key and participant_id as value. The resultant map usage: map[assignment_id][user_id] = participant_id
6. For all TeamUser objects, find the participant_id from the maps generated from the previous steps and update the participant_id attribute.

In this approach, we make a total of 3 database calls which is considerably better than the previous brute-force approach.

Code Changes

The user_id attribute is used in multiple places within the application. Since, the new schema changes do not include user_id attribute in teams_users table, all the references of user_id must be updated to use participant_id, which is its new equivalent instead.

Impacted Files

1. teams_users_controller
2. suggestion_controller
3. assessment360_controller
4. grades_controller
5. join_team_requests_controller
6. review_mapping_controller
7. student_teams_controller

The respective test cases associated with these files must be fixed before fixing the code so that we can make sure that the proposed refactoring does not raise any new bugs.