CSC/ECE 517 Spring 2017 E1711

From Expertiza_Wiki
Jump to navigation Jump to search

E1711. Refactor delayed_mailer.rb and scheduled_task.rb


Overview

Introduction to Expertiza

Expertiza is a web based tool that allows instructors to create collaborative assignments where students can provide peer feedback on each others work. It provides instructors a system to manage assignments and different courses.

Scope of project E1711

The goal of E1711 was to refactor code found in the files delayed_mailer.rb and scheduled_task.rb.

The following issues were identified as the scope of this project:


  • scheduled_task.rb duplicated most of its code from delayed_mailer.rb. Our objective is to reduce/remove the duplication in keeping with the DRY principle.
  • The perform method uses a giant case statement, instead we were to incorporate the use of polymorphism
  • The mail_signed_up_users is long and should be broken into smaller and better named methods
  • Add/modify test cases as we added/removed/modified areas of the code


Information about the assignment can be found on this document

Test Plan

As this was a purely refactoring effort, our testing consisted of confirming existing tests did not break, modifying existing tests to match our changes, or adding tests as needed. The below sections will cover each objective in detail and will include information on how testing was done for each of those changes. See each of the Our Testing sections.


Remove duplicated code between delayed_mailer.rb and scheduled_task.rb

Background of Objective

The main objective of this task was to reduce duplicated code as recommended by DRY principle.

After a deep investigation into understanding why the same code would appear in two separate files, we concluded that the authors of scheduled_task.rb found that delayed_mailer.rb already integrated with a Ruby gem called delayed_job. delayed_job is used to run tasks asynchronously in the background. It was used in delayed_mailer to add an outgoing email to a queue to send out in the future. Instead of recreating similar infrastructure, the authors copied over the file to a different location, added on the new desired functionality, and renamed it. The authors also pointed all existing code to use scheduled_task.rb instead of delayed_mailer.rb

Our Changes

We found that the vast majority of the code in both the files applied to sending emails so we decided to merge the scheduled_task.rb into delayed_mailer.rb by porting the enhancements made for the scheduled tasks into delayed_mailer.rb. Doing so allowed us to meet the objective of DRYing out the code. However, we still have a situation where scheduled tasks feature has code in a section devoted to mailer code. We recommend that a future project take up the needed changes to pull out the scheduled task code out of the mailer file and make it a standalone feature as it should have been originally. See details in the Future Refactoring Opportunities below.

Existing Tests

We did not find any existing automated testing targeting this area of the code. All existing tests in the spec file associated with this feature were copies of the tests in the delayed mailer feature and did not exercise any of the new code that was introduced with adding the new types of scheduled tasks.

Our Testing

We went back to the demonstration of this feature and manually tested all the same scenarios. We verified that no functionality presented in the scheduled tasks demonstration was broken. Tests were added for the delayed mailer feature (covered in more details in the following sections). We recommend tests focused towards scheduled tasks are added once the scheduled tasks code is removed from the delayed mailer file back into an independent file.


Refactor the perform mehtod

Background of Objective

The objective of this task was to clean up this method. It was identified as being badly named, long, and not taking advantage of good design. It is essentially a giant case statement where one of the variables passed into the DelayedMailer constructor determines the people who receives the email.. Some suggested ways to refactor this function was first rename the perform to better describe what it does and to use polymorphism in place of the case statement. The original thought was that polymorphism could be used to determine what was sent out in each email based on the type of the email.

Our Changes

The first item we looked at was to rename the method. A generic name such as perform usually means that the method is doing too much or does not have a clear objective. Unfortunately, we found we could not rename the method because the delayed_job gem requires a method named perform in order to work on a custom job. See documentation here

The second part of the planned refactoring was to use polymorphism to determine the content of the email. After simplifying much of the code in the file, we discovered that the content of the email was independent of the variables used to initialize the object. At that point, we weighed the complexity of new code to use polymorphism versus a significantly more simplified case statement, and because of the lack of existing tests, decided that the simplified case statement was not only short, it is also very simple to follow the flow through the code. We instead focused on DRYing out this code as a better return on immediate investment as opposed to adding polymorphism.

Existing Tests

There were six existing Rspec tests for delayed_mailer.rb, but the existing tests only verify that a new delayed job was added to the queue based on the type of deadline specified when creating a new DelayedMalier. We deemed the existing testing as adequate for this section of the file as the perform method calls other methods to find the email addresses and send the email.

Our Testing

We did not change existing testing for the perform method as the existing testing verified that a job to send out emails was added to the queue. We did, however, verify that our code changes did not break this portion of the code by continuously running the existing test cases.


Refactor mail_signed_up_users method

Background of Objective

The primary objective of this task was that the method is long and could be broken into smaller appropriately named methods.

Our Changes

We replaced the existing single method mail_signed_up_users with a shorter version and a new method find_team_members_email_for_all_topics. We also heavily modified (including renaming) the existing method getTeamMembersMail, and it is now called find_team_members_email. As the naming of the methods suggest, there are still opportunities for DRYing the two functions to get team member emails, but exiting code makes it extremely difficult to combine the two functions exactly though they are extremely similar at first glance. Merging the two would require refactoring areas outside the scope of this assignment and a whole new set of test cases to verify a lot of basic functionality is not broken in the process. We recommend this is taken up as a part of a future assignment. See Future Refactoring Opportunities below.

Existing Tests

We did not find any existing automated testing targeting this area of the code. None of the six existing tests verified if the emails are being retrieved from the database so changes in the database schema would break the feature.

Our Testing

We added test cases for each of the methods we changed. There are now 10 test cases that are covering this file. The four new test cases focus on verifying that the functions can access the database to protect against schema changes and making sure the right methods are called. Tests were not added for code that was not within the scope of this assignment but we recommend the remainder of the functions also get similar test cases. See details in Future Refactoring Opportunities below.

Future Refactoring Opportunities

Once we started to refactor the code, we found many additional opportunities to refactor the files that was beyond the planned scope. We note down some of the opportunities here and leave it to future projects to address.

1. Scheduled Task feature does not belong in the mailers. It needs to be a standalone file with new tests added.

2. There are still methods that look like they should be collapsed into one function and opportunities for DRYing out code once other tightly coupled methods are also refactored.

3. There were practically no tests for this area of the code. We added tests for what we modified but much code still remains that is not being tested.

4. There are many calls to the database for the same information in different methods. These repeated calls could probably be reduced to one call and save the result from the database.