FIx: Indifference to I18n errors.format translation

[HeitorMC]

Fixes #1492

I didn't find anything related to internationalization tests 🤔, also both spec/unit/shoulda/matchers/active_model/allow_value_matcher_spec.rb and spec/unit/shoulda/matchers/active_model/validate_presence_of_matcher_spec.rb passed without errors.

@vsppedro @mcmire do you have any suggestions on how to test this case, or is this enough?

[vsppedro]

Hello @HeitorMC, thanks for your PR. I'll take a look at how to do a test for this ASAP. Probably tomorrow or next week.

[HeitorMC]

Hey, @vsppedro any updates on this case?

[mcmire]

Hi @HeitorMC. Sorry for the delay in responding. We fake i18n in tests. You can see existing usage of this in the following context in the tests for allow_value:

shoulda-matchers/spec/unit/shoulda/matchers/active_model/allow_value_matcher_spec.rb

Lines 417 to 441 in fababc3

	context 'when the validation error message was provided via i18n' do
	it 'passes given a valid value' do
	builder = builder_for_record_with_different_error_attribute_using_i18n
	expect(builder.record).
	to allow_value(builder.valid_value).
	for(builder.attribute_to_validate).
	with_message(
	builder.validation_message_key,
	against: builder.attribute_that_receives_error,
	)
	end
	it 'fails given an invalid value' do
	builder = builder_for_record_with_different_error_attribute_using_i18n
	invalid_value = "#{builder.valid_value} (invalid)"
	expect(builder.record).
	not_to allow_value(invalid_value).
	for(builder.attribute_to_validate).
	with_message(
	builder.validation_message_key,
	against: builder.attribute_that_receives_error,
	)
	end
	end
	end

It's a bit hard to see because of the indirection, but these tests verify that if with_message is given a symbol, shoulda-matchers uses the same i18n key to resolve that symbol as Rails does. You can see I18n being stubbed here:

shoulda-matchers/spec/support/unit/record_builder_with_i18n_validation_message.rb

Line 38 in fababc3

I18nFaker.stub_translation(keys, default_message)

So, I think in order to test your PR, you'd want to add a new context and a new test inside of that test that verifies that when with_message is not specified, but strict is, and errors.format is stubbed to return a certain string that's different than what Rails uses, then the test still passes. You don't have to create a new class like we've done for RecordBuilderWithI18nValidationMessage; I think you could keep it simpler and do something like this:

shoulda-matchers/spec/unit/shoulda/matchers/active_model/allow_value_matcher_spec.rb

Lines 70 to 71 in fababc3

	expect(validating_format(with: /abc/)).
	to allow_value('abcde').for(:attr)

I can suggest something a little more specific later, but let me know if that helps in the meantime.