Expose a generic fixture accessor for fixture names that may conflict with Minitest

[casperisfine]

assert_equal "Ruby on Rails", web_sites(:rubyonrails).name
assert_equal "Ruby on Rails", fixture(:web_sites, :rubyonrails).name

This was brought to me by someone with a Metadata model. The fixtures accessor being metadata which conflicts with the metadata method recently added in Minitest.

cc @zk-1

[ghiculescu]

This is clashing with Minitest::Spec. Here is an example:

class MyTest < ActiveSupport::TestCase
  let(:user) { users(:alex) }
  let(:fixture) { SomeFixtureObject.new } 

  it "passes" { assert_not_nil user }
end

fixture will be redefined here as a method that takes no arguments. Here it will get called when you access users(:alex). And the test will fail with:

ArgumentError: wrong number of arguments (given 2, expected 0)
    minitest (5.17.0) lib/minitest/spec.rb:256:in `block in let'
    rails (5411787a155e) activerecord/lib/active_record/test_fixtures.rb:258:in `method_missing'

A workaround would be to not ever have a let(:fixture), but I wonder if there's a way to avoid having to do that. (My lazy suggestion would be to rename the method added in this PR to _fixture 😆 )

[byroot]

My lazy suggestion would be to rename the method added in this PR to _fixture 😆

I'd rather not because it's meant to be public API, so a leading underscore would be wrong.

However what we can do:

• Alias it as _active_record_fixture.
• Use that alias internally.

You want to PR that?

[byroot]

NB: that alias should be :nodoc: and private.

[ghiculescu]

👍 can do

[ghiculescu]

I'd rather not because it's meant to be public API, so a leading underscore would be wrong.

Hmm, should it be showing in https://edgeapi.rubyonrails.org/classes/ActiveRecord/TestFixtures.html ?

[byroot]

Yeah, it should but apparently it doesn't. Not sure how we can make a private method documented.

[ghiculescu]

I assume you'd have to make all the methods public and then :nodoc: most of them. Not much of that file is public atm.

[justinko]

Yeah, it should but apparently it doesn't. Not sure how we can make a private method documented.

#52054