Suggestion: Use string literals instead of named routes or URL helpers in tests

[ohbarye]

Hi. This is a proposal about named routes usage in tests. I'd like to hear what you think.

Proposal

Use string literals instead of named routes or URL helpers in tests.

# bad
get photos_path
get photo_path(id)
get edit_photo_path(id)

# good
get "/photos"
get "/photos/#{id}"
get "/photos/#{id}/edit"

and so on.

Reason

Integration test (e.g. request specs, system specs) is to mimic user eyes and actions. A user could see "/photos", but not "photos_path". I agree with the following idea.

The more your tests resemble the way your software is used, the more confidence they can give you.
https://twitter.com/kentcdodds/status/977018512689455106

From a programmer's view,

• the bad pattern slightly increases cognitive load that forces me to build actual URLs in my mind.
• the bad pattern sometimes discourages Test Driven Development.
  • In TDD, a test comes first. At this point, I know actual URL but I don't know the named routes because the routes.rb has not been edited yet. (=I don't care about named routes at that moment).

Possible Objection

named routes possibly have merits in some points.

• You don't have to edit test cases when you edit routes.rb. It automatically follows up.
• It could be consistent to use named routes in tests if you use named routes in production code.

References

• https://guides.rubyonrails.org/routing.html

[pirj]

There's some controversy there in the Rails Testing Rails Applications guide.

Integration:

class BlogFlowTest < ActionDispatch::IntegrationTest
  test "can see the welcome page" do
    get "/"

System:

class ArticlesTest < ApplicationSystemTestCase
  test "viewing the index" do
    visit articles_path

Wondering why is it so.

If you don't write route tests and use URL helpers in system/integration, the mapping is not covered. It is risky to have a green CI after changing routes.