CSC/ECE 517 Spring 2020 E2009 Refactor assignment

Introduction

Expertiza is an open source web application project based on Ruby on rails framework. Expertiza allows instructors to add assignments and students to upload their submissions. The assignment.rb model file consists of some basic CRUD operations along with some methods which help calculate scores and export details etc. The goal of the project is to refactor assignment.rb file to follow good coding practices.

About Assignments

Assignments is the most important base class in expertiza. It enables students to submit their assignments, also aiding TA's and profs to access and grade the assignments and also gives support for peer reviews. This model is used for all the backend operations and DB querying related to Assignments. Assignments can be submitted, reviewed by other peers and scores assigned and accessed, keeping in mind the deadline constraints too.

File:Assignments.jpg

Below is screenshot of questionnaire on expertiza. It shows all the kinds of questionnaires we can create on expertiza.

Refactoring assignment.rb

Some of the coding issues with the assignment.rb file are

1) Methods performing more than one tasks, resulting in long methods

2) Methods with multiple conditions and loops resulting in increased cyclomatic and cognitive complexity

3) Large number of methods in one file.

4) No proper naming conventions in some places.

Approach

The approach we took to refactor this file, is to go through the issues generated by code climate and fix the smaller issues first. This gave us an idea about what the code is doing and gave us a head start to fix bigger issues. 69 issues were found on code climate and through this project, 30-35 issues have been resolved. Few of the issues that were resolved was detected by rubocop.

Code climate gives different metrics that indicates the code quality. For methods, some of the metrics the code climate gives are

1) Assignment Branch Condition (ABC) size - It is computed by counting the number of assignments, branches and conditions in a section of code. Specifically ABC size = sqrt(A*A + B*B + C*C), where A - number of assignments, B - number of branches, C - number of conditions.

2) Cyclomatic complexity - It is a quantitative measure of the number of linearly independent paths through a program's source code (or a method). It gives a measure of how difficult a code is to test. Higher the number of branches in a method, higher the number of independent paths and hence higher the cyclomatic complexity.

An if statement (or unless or ?:) increases the complexity by one. An else branch does not, since it doesn't add a decision point. The && operator (or keyword and) can be converted to a nested if statement, and ||/or is shorthand for a sequence of ifs, so they also add one. Loops can be said to have an exit condition, so they add one.

3) Cognitive complexity - Is a measure of how difficult a unit of code is to intuitively understand. Generally methods with higher cyclomatic complexity will have higher cognitive complexity also.

4) Perceived complexity - It is complexity score that's a measure of the complexity the reader experiences when looking at a method. In contrast to the Cyclomatic Complexity, this metric considers `else` nodes as adding complexity.

All the metrics are quite interlinked, so aiming to reduce one complexity will also reduce other metrics.

Refactoring longer methods

The longer methods are refactored mostly using extract method. Longer methods generally has higher Assignment Branch Size and higher complexity metrics. The methods refactored using Extract method are scores method and review_questionnaire_id method.

Refactor scores method

The scores method is one of the biggest methods in assignment.rb file. The code climate gives the Assignment Branch Condition size as 131.2/15, number of lines of code as 48 and cyclomatic complexity of 8/6.

The scores method computes and returns the scores of participants and teams as a hash. The participant scores are directly got by calling a method in Participant class. The logic to compute the scores of teams has an outer each loop with one if-else block inside and 2 each loops inside the if block. This resulted in 3 levels of nesting and many lines of code inside the each loop.

Below is the each block before refactoring

index = 0
    self.teams.each do |team|
      scores[:teams][index.to_s.to_sym] = {}
      scores[:teams][index.to_s.to_sym][:team] = team
      if self.varying_rubrics_by_round?
        grades_by_rounds = {}
        total_score = 0
        total_num_of_assessments = 0 # calculate grades for each rounds
        (1..self.num_review_rounds).each do |i|
          assessments = ReviewResponseMap.get_responses_for_team_round(team, i)
          round_sym = ("review" + i.to_s).to_sym
          grades_by_rounds[round_sym] = Answer.compute_scores(assessments, questions[round_sym])
          total_num_of_assessments += assessments.size
          total_score += grades_by_rounds[round_sym][:avg] * assessments.size.to_f unless grades_by_rounds[round_sym][:avg].nil?
        end
        # merge the grades from multiple rounds
        scores[:teams][index.to_s.to_sym][:scores] = {}
        scores[:teams][index.to_s.to_sym][:scores][:max] = -999_999_999
        scores[:teams][index.to_s.to_sym][:scores][:min] = 999_999_999
        scores[:teams][index.to_s.to_sym][:scores][:avg] = 0
        (1..self.num_review_rounds).each do |i|
          round_sym = ("review" + i.to_s).to_sym
          if !grades_by_rounds[round_sym][:max].nil? && scores[:teams][index.to_s.to_sym][:scores][:max] < grades_by_rounds[round_sym][:max]
            scores[:teams][index.to_s.to_sym][:scores][:max] = grades_by_rounds[round_sym][:max]
          end
          if !grades_by_rounds[round_sym][:min].nil? && scores[:teams][index.to_s.to_sym][:scores][:min] > grades_by_rounds[round_sym][:min]
            scores[:teams][index.to_s.to_sym][:scores][:min] = grades_by_rounds[round_sym][:min]
          end
        end
        if total_num_of_assessments != 0
          scores[:teams][index.to_s.to_sym][:scores][:avg] = total_score / total_num_of_assessments
        else
          scores[:teams][index.to_s.to_sym][:scores][:avg] = nil
          scores[:teams][index.to_s.to_sym][:scores][:max] = 0
          scores[:teams][index.to_s.to_sym][:scores][:min] = 0
        end
      else
        assessments = ReviewResponseMap.get_assessments_for(team)
        scores[:teams][index.to_s.to_sym][:scores] = Answer.compute_scores(assessments, questions[:review])
      end
      index += 1
    end

When examined carefully the logic inside the if block performs 2 distinct tasks. The if condition is for the assignments with varying rubrics. Hence logic inside the if blocks is first it computes the grades from different rounds and then as second step it merges the grades from different rounds to get max, min and avg scores. So two methods are extracted out of this block : 1) compute_grades_by_rounds - computes and returns the grades from different rounds 2) merge_grades_by_rounds - merges the grades from different rounds as computed in step 1 to compute max, min and avg scores.

After refactoring

  index = 0
    self.teams.each do |team|
      scores[:teams][index.to_s.to_sym] = {:team => team, :scores => {}}
      if self.varying_rubrics_by_round?
        grades_by_rounds, total_num_of_assessments, total_score = compute_grades_by_rounds(questions, team)
        scores[:teams][index.to_s.to_sym][:scores] = merge_grades_by_rounds(grades_by_rounds, total_num_of_assessments, total_score)
      else
        assessments = ReviewResponseMap.get_assessments_for(team)
        scores[:teams][index.to_s.to_sym][:scores] = Answer.compute_scores(assessments, questions[:review])
      end
      index += 1
    end

Hence after refactoring the number of lines of code reduced to 17.

Refactor review_questionnaire_id method

This method as the name implies returns the review questionnaire ID. Before refactoring

  def review_questionnaire_id(round = nil)
    # Get the round it's in from the next duedates
    if round.nil?
      next_due_date = DueDate.get_next_due_date(self.id)
      round = next_due_date.try(:round)
    end
    # for program 1 like assignment, if same rubric is used in both rounds,
    # the 'used_in_round' field in 'assignment_questionnaires' will be null,
    # since one field can only store one integer
    # if rev_q_ids is empty, Expertiza will try to find questionnaire whose type is 'ReviewQuestionnaire'.
    rev_q_ids = if round.nil?
                  AssignmentQuestionnaire.where(assignment_id: self.id)
                else
                  AssignmentQuestionnaire.where(assignment_id: self.id, used_in_round: round)
                end
    if rev_q_ids.empty?
      AssignmentQuestionnaire.where(assignment_id: self.id).find_each do |aq|
        rev_q_ids << aq if aq.questionnaire.type == "ReviewQuestionnaire"
      end
    end
    review_questionnaire_id = nil
    rev_q_ids.each do |rqid|
      next if rqid.questionnaire_id.nil?
      rtype = Questionnaire.find(rqid.questionnaire_id).type
      if rtype == 'ReviewQuestionnaire'
        review_questionnaire_id = rqid.questionnaire_id
        break
      end
    end
    review_questionnaire_id
  end

If a block of code inside a method is commented it is a good indication to extract that part of code as another method. In this method, such a part is computing rev_q_ids. (rev_q_ids is not following good naming convention, but probably review_questionnaire_id variable is already taken for the actual review_questionnaire_id, so developer would have given rev_q_ids). Hence after extracting get_rev_q_ids as separate method, below is the method after refactoring

  def review_questionnaire_id(round = nil)
    # Get the round it's in from the next duedates
    if round.nil?
      next_due_date = DueDate.get_next_due_date(self.id)
      round = next_due_date.try(:round)
    end

    rev_q_ids = get_questionnaire_ids(round)
    review_questionnaire_id = nil
    rev_q_ids.each do |rqid|
      next if rqid.questionnaire_id.nil?
      rtype = Questionnaire.find(rqid.questionnaire_id).type
      if rtype == 'ReviewQuestionnaire'
        review_questionnaire_id = rqid.questionnaire_id
        break
      end
    end
    review_questionnaire_id
  end

Refactoring to reduce code complexity

1) Refactoring export_details_fields The export_details_fields has many if statements as can be seen below. Each if statement increases the cyclomatic complexity by 1.

Before Refactoring

 def self.export_details_fields(detail_options)	  def self.export_details_fields(detail_options)
   fields = []	    fields = []
   fields << 'Team ID / Author ID' if detail_options['team_id'] == 'true'	    EXPORT_FIELDS.each do |key, value|
   fields << 'Reviewee (Team / Student Name)' if detail_options['team_name'] == 'true'	      fields << value if detail_options[key]=='true'
   fields << 'Reviewer' if detail_options['reviewer'] == 'true'	    end
   fields << 'Question / Criterion' if detail_options['question'] == 'true'	
   fields << 'Question ID' if detail_options['question_id'] == 'true'	
   fields << 'Answer / Comment ID' if detail_options['comment_id'] == 'true'	
   fields << 'Answer / Comment' if detail_options['comments'] == 'true'	
   fields << 'Score' if detail_options['score'] == 'true'	
   fields	    fields
 end

This can be refactored as below

After Refactoring

  EXPORT_FIELDS={'team_id'=>'Team ID / Author ID', 'team_name'=>'Reviewee (Team / Student Name)','reviewer'=>'Reviewer','question'=>'Question / Criterion','question_id'=>'Question ID','comment_id'=>'Answer / Comment ID','comments'=>'Answer / Comment','score'=>'Score' }.freeze
 def self.export_details_fields(detail_options)
   fields = []
   EXPORT_FIELDS.each do |key, value|
     fields << value if detail_options[key]=='true'
   end

2) Refactoring delete method

The delete method has the following lines of code. Each line has call to each method.

Before Refactoring

   self.invitations.each(&:destroy)
   self.teams.each(&:delete)
   self.participants.each(&:delete)
   self.due_dates.each(&:destroy)
   self.assignment_questionnaires.each(&:destroy)

This can be refactored as below

After Refactoring

   DELETE_INSTANCES=['invitations','teams','participants','due_dates','assignment_questionnaires']

   DELETE_INSTANCES.each do |instance|
     self.instance_eval(instance).each(&:destroy)
   end

Reducing cyclomatic complexity and improving code reuse

1) Three methods in assignment.rb file has the following check

if self.staggered_deadline and topic_id.nil?

Each predicate in if condition count towards one decision point and hence increase in cyclomatic complexity. Also the same condition check is done in 3 to 4 places. Hence we can write a separate method like below and call in all places the conditions self.staggered_deadline and topic_id.nil? are checked together.

def staggered_and_no_topic?(topic_id)
   self.staggered_deadline? and topic_id.nil?
end

2) Similarly the method valid_num_review has if condition with 3 predicates as can be seen below

Before Refactoring

if self.num_reviews_allowed && self.num_reviews_allowed != -1 && self.num_reviews_allowed < self.num_reviews_required
    self.errors.add(:message, "Num of reviews required cannot be greater than number of reviews allowed")
elsif self.num_metareviews_allowed && self.num_metareviews_allowed != -1 && self.num_metareviews_allowed < self.num_metareviews_required
    self.errors.add(:message, "Number of Meta-Reviews required cannot be greater than number of meta-reviews allowed")

Here the 2 if conditions are quite similar. Also both are checking 3 conditions. Hence the condition check part can written as a separate function as below

After Refactoring

#returns true if reviews required is greater than reviews allowed
  def num_reviews_greater?(reviews_required, reviews_allowed)
    reviews_allowed && reviews_allowed != -1 && reviews_required > reviews_allowed
  end

if num_reviews_greater?(self.num_reviews_required, self.num_reviews_allowed)
   self.errors.add(:message, "Num of reviews required cannot be greater than number of reviews allowed")
elsif num_reviews_greater?(self.num_metareviews_required, self.num_metareviews_allowed)
   self.errors.add(:message, "Number of Meta-Reviews required cannot be greater than number of meta-reviews allowed")
end

By this way, it is both reducing cyclomatic complexity and improving code reuse.

Refactoring for other good coding practices

1. Removing unused variables

2. Changing variable/function names

Before Refactoring

   if @map.assignment.has_badge?
   @courses = Assignment.set_courses_to_assignment(current_user)

After Refactoring

   if @map.assignment.badge?
   @courses = Assignment.assign_courses_to_assignment(current_user)

3. Avoiding multi-line ternary operators

4. Using Guard Clause instead of wrapping the code inside a conditional expression.
A guard clause is simply a check that immediately exits the function, either with a return statement or an exception.

Before Refactoring

 team[:scores] ?
       tcsv.push(team[:scores][:max], team[:scores][:min], team[:scores][:avg]) :
       tcsv.push('---', '---', '---')

After Refactoring

if team[:scores]
        tcsv.push(team[:scores][:max], team[:scores][:min], team[:scores][:avg])
      else
        tcsv.push('---', '---', '---')
      end

unless self.staggered_deadline?

return due_date.deadline_name unless self.staggered_deadline?

Test Plan

Our test plan includes rspec tests and testing the GUI in the browser. The rspec tests for the assignment.rb file is pretty much complete with tests written for all the methods called from outside the file. The methods added as part of refactoring are given a private scope and added in assignment.rb since it is not called from outside. Testing the public methods implicitly tests these private methods. Below is a screenshot our rspec test results.