How to look for high complexity parts
I want to visualize the messiness of code
When I usually develop with Ruby, I sometimes encounter code that I find difficult to read. I wish I could refactor it freely, but software development at work is often aimed at maximizing profit for a specific period of time, so the investment ratio between product development and debt return is often a tradeoff between business impact and difficulty of development.
If the product will be finished in a few months, there may be a decision to keep going even if it is hard to read. On the other hand, if the company plans to put more effort into the product in the future, refactoring may be a more successful investment because it maximizes development efficiency * development time.
The quality of code and the code to be refactored is often talked about in terms of ambiguity. It would be better to have a certain measure of success to make the discussion with confidence.
I introduce Flog this time because it seems to make such triage easier.
Install Flog
You can use Sinatra or Hanami as long as the code is in Ruby, but since I think this is the most common use case, I will proceed on the assumption that you are using Ruby on Rails in your web development. Redmine code(commit: e8db9db8948d928ab8e9dae75c261967676291a8) is used.
Install Flog.
# Gemfile group :developmentdogem"flog"end
bundle install
Run flog. I think that app or lib will be used in most cases when dir is specified.
$ bundle exec flog app/
38137.0: flog total
15.5: flog/method average
456.0: ApplicationHelper#parse_redmine_links app/helpers/application_helper.rb:1088-1284
348.4: Query#sql_for_field app/models/query.rb:1218-1446
264.4: IssuesHelper#show_detail app/helpers/issues_helper.rb:524-658
251.6: Issue#none
189.4: User#none
185.1: Project#none
172.7: IssueImport#build_object app/models/issue_import.rb:113-241
167.9: Project#copy_issues app/models/project.rb:1122-1225
162.9: IssueQuery#initialize_available_filters app/models/issue_query.rb:147-280
160.8: Principal#none
158.8: IssuesController#bulk_edit app/controllers/issues_controller.rb:257-341
144.4: WorkflowTransition::replace_transitions app/models/workflow_transition.rb:23-91
141.3: Repository::Cvs#fetch_changesets app/models/repository/cvs.rb:128-200
136.9: ActivitiesController#index app/controllers/activities_controller.rb:25-85
136.0: Issue#validate_issue app/models/issue.rb:751-815
131.1: Issue#safe_attributes= app/models/issue.rb:555-633
128.0: IssuesHelper#render_descendants_tree app/helpers/issues_helper.rb:92-144
(the rest continues)In this example, flog score under app is output.
flog score, class name#method name, file name, and line number are output. flog total is the total score and average is the average.
Internally, flog score seems to measure Abc size. By default, only top 60% is displayed.
Flog threshold change, grouping
Thresholds can be optionally changed.
$ bundle exec flog -t 10 app/ # top 10%
38137.0: flog total
15.5: flog/method average
456.0: ApplicationHelper#parse_redmine_links app/helpers/application_helper.rb:1088-1284
348.4: Query#sql_for_field app/models/query.rb:1218-1446
264.4: IssuesHelper#show_detail app/helpers/issues_helper.rb:524-658
251.6: Issue#none
189.4: User#none
185.1: Project#none
172.7: IssueImport#build_object app/models/issue_import.rb:113-241
167.9: Project#copy_issues app/models/project.rb:1122-1225
162.9: IssueQuery#initialize_available_filters app/models/issue_query.rb:147-280
160.8: Principal#none
158.8: IssuesController#bulk_edit app/controllers/issues_controller.rb:257-341
144.4: WorkflowTransition::replace_transitions app/models/workflow_transition.rb:23-91
141.3: Repository::Cvs#fetch_changesets app/models/repository/cvs.rb:128-200
136.9: ActivitiesController#index app/controllers/activities_controller.rb:25-85
136.0: Issue#validate_issue app/models/issue.rb:751-815
131.1: Issue#safe_attributes= app/models/issue.rb:555-633
128.0: IssuesHelper#render_descendants_tree app/helpers/issues_helper.rb:92-144
124.4: Query#statement app/models/query.rb:969-1026
123.9: AccountController#lost_password app/controllers/account_controller.rb:56-132
121.0: QueriesHelper#retrieve_query app/helpers/queries_helper.rb:345-395
118.9: IssuesController#bulk_update app/controllers/issues_controller.rb:343-424
Grouping (and sorting) by class is also possible.
$ bundle exec flog -g app/
38137.0: flog total
15.5: flog/method average
1735.8: Issue total
251.6: Issue#none
136.0: Issue#validate_issue app/models/issue.rb:751-815
131.1: Issue#safe_attributes= app/models/issue.rb:555-633
112.3: Issue#after_create_from_copy app/models/issue.rb:1728-1797
102.4: Issue#recalculate_attributes_for app/models/issue.rb:1826-1877
61.9: Issue#project= app/models/issue.rb:413-452
61.6: Issue#workflow_rule_by_attribute app/models/issue.rb:685-732
60.5: Issue#validate_required_fields app/models/issue.rb:818-837
59.8: Issue#new_statuses_allowed_to app/models/issue.rb:1052-1093
52.7: Issue::visible_condition app/models/issue.rb:131-163
45.4: Issue#reschedule_on! app/models/issue.rb:1389-1423
43.1: Issue::update_versions app/models/issue.rb:1883-1899
42.6: Issue::load_visible_relations app/models/issue.rb:1227-1245
41.3: Issue#would_reschedule? app/models/issue.rb:1321-1338
41.2: Issue#css_classes app/models/issue.rb:1452-1466
37.5: Issue#after_project_change app/models/issue.rb:1697-1723
36.3: Issue::load_visible_last_notes app/models/issue.rb:1277-1294
35.9: Issue#notified_users app/models/issue.rb:1115-1129
34.7: Issue#copy_from app/models/issue.rb:295-323
34.6: Issue::cross_project_scope app/models/issue.rb:1535-1560
34.2: Issue#create_parent_issue_journal app/models/issue.rb:2007-2031
33.8: Issue#visible? app/models/issue.rb:166-188
33.0: Issue::load_visible_total_spent_hours app/models/issue.rb:1214-1224
32.6: Issue::load_visible_last_updated_by app/models/issue.rb:1258-1274
28.0: Issue#valid_parent_project? app/models/issue.rb:1517-1532
27.2: Issue::allowed_target_trackers app/models/issue.rb:1659-1676
26.1: Issue#visible_journals_with_index app/models/issue.rb:909-925
25.8: Issue#behind_schedule? app/models/issue.rb:974-979
25.6: Issue#assignable_versions app/models/issue.rb:999-1016
23.8: Issue::count_and_group_by app/models/issue.rb:1598-1617
23.3: Issue#tracker= app/models/issue.rb:379-393
1578.2: ApplicationHelper total
456.0: ApplicationHelper#parse_redmine_links app/helpers/application_helper.rb:1088-1284
113.6: ApplicationHelper#parse_wiki_links app/helpers/application_helper.rb:974-1038
85.3: ApplicationHelper#progress_bar app/helpers/application_helper.rb:1570-1602
76.8: ApplicationHelper#none
68.4: ApplicationHelper#format_object app/helpers/application_helper.rb:254-313
59.6: ApplicationHelper#render_page_hierarchy app/helpers/application_helper.rb:453-481
57.9: ApplicationHelper#page_header_title app/helpers/application_helper.rb:744-770
57.6: ApplicationHelper#render_project_nested_lists app/helpers/application_helper.rb:422-451
53.9: ApplicationHelper#textilizable app/helpers/application_helper.rb:848-893
53.0: ApplicationHelper#replace_toc app/helpers/application_helper.rb:1434-1465
43.7: ApplicationHelper#render_projects_for_jump_box app/helpers/application_helper.rb:533-568
42.9: ApplicationHelper#parse_non_pre_blocks app/helpers/application_helper.rb:895-925
42.8: ApplicationHelper#link_to_principal app/helpers/application_helper.rb:58-79
42.6: ApplicationHelper#project_tree_options_for_select app/helpers/application_helper.rb:595-616
41.1: ApplicationHelper#parse_inline_attachments app/helpers/application_helper.rb:936-962
40.8: ApplicationHelper#body_css_classes app/helpers/application_helper.rb:818-833
36.4: ApplicationHelper#render_project_jump_box app/helpers/application_helper.rb:571-593
35.8: ApplicationHelper#principals_options_for_select app/helpers/application_helper.rb:644-661
33.9: ApplicationHelper#link_to_issue app/helpers/application_helper.rb:104-122
32.0: ApplicationHelper#include_calendar_headers_tags app/helpers/application_helper.rb:1633-1659
28.3: ApplicationHelper#principals_check_box_tags app/helpers/application_helper.rb:625-641
27.2: ApplicationHelper#inject_macros app/helpers/application_helper.rb:1412-1429
25.4: ApplicationHelper#parse_headings app/helpers/application_helper.rb:1352-1373
23.1: ApplicationHelper#title app/helpers/application_helper.rb:773-783
1433.4: Query total
348.4: Query#sql_for_field app/models/query.rb:1218-1446
124.4: Query#statement app/models/query.rb:969-1026
104.3: Query#project_statement app/models/query.rb:935-967
(the rest continues)There does not seem to be an option to summarize by class name only, so grep is used. (Actually I would like to know the average rather than the total if possible.)
$ bundle exec flog -g app/ |grep total |grep ':'
38137.0: flog total
1735.8: Issue total
33.0: Issue::load_visible_total_spent_hours app/models/issue.rb:1214-1224
1578.2: ApplicationHelper total
1433.4: Query total
27.8: Query#total_with_scope app/models/query.rb:1099-1112
1036.1: IssuesController total
917.1: IssueQuery total
909.9: IssuesHelper total
757.1: MailHandler total
732.5: Project total
657.5: User total
564.8: QueriesHelper total
24.9: QueriesHelper#available_totalable_columns_tags app/helpers/queries_helper.rb:113-128
442.4: RepositoriesController total
424.5: ApplicationController total
408.6: Mailer total
354.4: UsersController total
344.9: WikiController total
312.9: TimelogController total
311.4: RepositoriesHelper total
305.8: TimeEntry total
303.0: Attachment total
295.5: ProjectsController total
276.9: AccountController total
275.6: IssueImport total
254.2: Journal total
244.6: Repository::Cvs total
239.5: TimeEntryQuery total
232.0: Repository total
225.6: VersionsController total
205.2: Principal total
189.1: Version total
180.2: Setting total
174.1: WatchersController total
167.3: WorkflowsController total
166.8: Changeset total
156.7: WikiPage total
152.0: SearchHelper total
151.5: CustomField total
145.6: AuthSourceLdap total
144.4: WorkflowTransition total
143.0: Import total
(the rest continues)Redmine is project management software. It is natural that the code around issues, projects and emails will be the majority. Data analysis often produces results that fit the obvious intuition.
It would be easier to read if some sort of cutout was made.
Flog score and future change-based triage
Now that we have the score, we can triage it. There are other code metrics such as Cyclomatic Complexity and Perceived Complexity. Although whether or not they are completely consistent with actual complexity may have the room of improvement, I believe flog score.
It is an option to fix what is wrong locally on a method-by-method basis, but I personally think it is better to look at it on a class-by-class basis or on a concept-by-concept basis such as a combination of controller and model. It helps us understand what kind of domain handling is complicated.
I make a DA table with future changes and flog scores to determine.
| Item A | Item B | Item C | |
|---|---|---|---|
| flog score | 1500 | 800 | 50 |
| development plan | ○ | ◎ | △ |
I think it would be a good idea to refactor Item B first, which has a reasonably bad score and is planned to be developed.
By the way, in team development, it would be difficult to make major modifications to a part that is currently being actively changed. It would cause conflicts. The successful strategy is to go for areas that are a little more settled, have plans to resume development, and where the code is hard to read. That is another point.
Can CI prevent this in advance?
In other point of view, although many companies are doing it, it would be interesting if CI can show the difference from the default branch.
According to a Microsoft Research study I read earlier, Code Complexity was a higher indicator of Recall than Code Coverage as a factor in predicting bugs.
In my opinion, software bugs are similar to medical decisions. Recall is more important than Precision. When I say software bug, a bug refers to a rare phenomenon, and if a large percentage of the total code is okay, it may be similar to medical data.
While there are many companies that outputs whether or not the tests are passing in CI, there are not many companies that outputs Abc Size. "The pull request you made raised the code complexity score 10 points compared to the original value of existing files. 20 points higher than the repository average for new files added. " The warnings like this would raise the level of awareness of clean code.
Future Work
I would like to do something equivalent to flog since rubocop seems to be able to measure Cyclomatic Complexity and Perceived Complexity.
