[Validator] Allow to use translation_domain false for validators and to use custom translation domain by constraints

[VincentLanglet]

Q	A
Branch?	6.3
Bug fix?	no
New feature?	yes
Deprecations?	no
Tickets	Fix #...
License	MIT
Doc PR	symfony/symfony-docs#...

Currently it's possible to write

$this->context->addViolation($constraint->message);
$this->context->addViolation($constraint->message, ['customParameter' => $value]);

but it's not allowed to pass a custom translation_domain (or to disable the translation by passing false).

Adding a third parameter would allow this. (And changing the ?string type to string|false|null).

Wdyt about this feature ? How can we make it fully BC ?

[carsonbot]

Hey!

To help keep things organized, we don't allow "Draft" pull requests. Could you please click the "ready for review" button or close this PR and open a new one when you are done?

Note that a pull request does not have to be "perfect" or "ready for merge" when you first open it. We just want it to be ready for a first review.

Cheers!

Carsonbot

[carsonbot]

Hey!

I think @norkunas has recently worked with this code. Maybe they can help review this?

Cheers!

Carsonbot

[VincentLanglet]

Friendly ping @nicolas-grekas, how can I provide BC for this situation ?

[nicolas-grekas]

not sure about your questions for BC 😬

[VincentLanglet]

not sure about your questions for BC 😬

If I change

public function setTranslationDomain(?string $translationDomain): static

to

public function setTranslationDomain(string|false|null $translationDomain): static

in the ValidatorBuilder, this is not really BC because

• ValidatorBuilder is not final
• ValidatorBuilder is not internal

So someone could extends the ValidatorBuilder and override the setTranslationDomain with the "old" signature.

If I change

public function setTranslationDomain(string $translationDomain): static;

to

public function setTranslationDomain(string|false $translationDomain): static;

in the ConstraintViolationBuilderInterface, this is not a BC because old implementation won't support false.
But for this, the solution I thought about is to move the method to an @method annotation.

[VincentLanglet]

not sure about your questions for BC 😬

I change my PR to be BC and resolved your request changes @nicolas-grekas :)

[VincentLanglet]

Tests failure doesn't seem related

[VincentLanglet]

Would you have time to take a look/review this @nicolas-grekas @fabpot ?
Thanks a lot !

[nicolas-grekas]

Parameter validator.translation_domain is also used in FormTypeCsrfExtension and in UploadValidatorExtension so I suppose we should make them accept false also?
How do you expect to disable translation otherwise for builders?
Or maybe this change is not needed for builders?

[VincentLanglet]

Parameter validator.translation_domain is also used in FormTypeCsrfExtension and in UploadValidatorExtension so I suppose we should make them accept false also? How do you expect to disable translation otherwise for builders? Or maybe this change is not needed for builders?

If this (#48852 (comment)) is a BC break, it cannot be done for Builder then. So validator.translation_domain cannot be set to false, because of the setTranslationDomain call in

->set('validator.builder', ValidatorBuilder::class)
            ->factory([Validation::class, 'createValidatorBuilder'])
            ->call('setConstraintValidatorFactory', [
                service('validator.validator_factory'),
            ])
            ->call('setTranslator', [
                service('translator')->ignoreOnInvalid(),
            ])
            ->call('setTranslationDomain', [
                param('validator.translation_domain'),
            ])

It's maybe not needed for Builder then ; I removed it.

The second BC break (#48852 (comment)) is more annoying ; any way to avoid it but still allowing calling setTranslationDomain(false) ?

[stof]

Passing a custom translation domain is already possible through ->buildViolation('my_message')->setTranslationDomain('custom')->addViolation()

[nicolas-grekas]

May I ask why you need this feature BTW?
Why isn't configuring validation.translation_domain enough?

[VincentLanglet]

May I ask why you need this feature BTW? Why isn't configuring validation.translation_domain enough?

Sure.
The validation.translation_domain allows to set a custom Translation domain.
As mentioned by stof, it's also possible with

->buildViolation('my_message')->setTranslationDomain('custom')->addViolation()

But currently, there is no way to use the Validator features without the translator.
In the same way I can pass translation_domain => false to not try to translate a form label because it's hardcoded ; I'd like to not try to translate an validator error message. For this:

->buildViolation('Message')->disableTranslation()->addViolation()

would be perfect

[nicolas-grekas]

So, we don't need to change addViolation, right?
What about allowing validation.translation_domain to be false and injecting an IdentityTranslator when that's the case?
All this could be resolved at the bundle level then.

[VincentLanglet]

So, we don't need to change addViolation, right? What about allowing validation.translation_domain to be false and injecting an IdentityTranslator when that's the case? All this could be resolved at the bundle level then.

Changing validation.translation_domain to false would be globally no ?
I don't fully know the impact of this config but wouldn't it:

• Have impact on all the existing constraints (like the Symfony one or FormTypeCsrfExtension/UploadValidatorExtension like you mentioned) => Which I don't want to.
• Be globally for all my custom validation constraints => Which I like to avoid, because I only have some of my constraint which doesn't require translations/translators.

To me validation.translation_domain is a global config, while addViolation would be a local config.

[nicolas-grekas]

only have some of my constraint which doesn't require translations/translators.

what's the issue with still processing them through the translator?
aren't you looking for a micro-optimization?

[VincentLanglet]

only have some of my constraint which doesn't require translations/translators.

what's the issue with still processing them through the translator? aren't you looking for a micro-optimization?

Can't the same argument be used for the option translation_domain => false of form labels ?
It feel inconsistent to be able to disable the translation for forms but not for validations of forms.

The issue with having them processed by the translator are:

• It can conflict with existing translations keys
• It's considered as a missing translation by the profiler

The second point is annoying when you look for all the missing translations, either

• with https://symfony.com/doc/6.3/translation.html#how-to-find-missing-or-unused-translation-messages-2
• with a custom implementation which makes functional tests failing because of missing translations (following suggestion like Add an option to fail tests when a translation is missing. #36792 (comment))

[nicolas-grekas]

Parameter validator.translation_domain is also used in FormTypeCsrfExtension and in UploadValidatorExtension so I suppose we should make them accept false also?

please have a look at this concern

please also add needed changelog+upgrade entries

{%- if translation_domain is same as(false) -%}

should we update the template so that they use strtr?

[stof]

I'm not sure we should support false for the global configuration (in the context of the framework, that would break all bundles providing validation messages for instance). And if you have a project were you really don't want to use translations anywhere, you can inject an IdentityTranslator (which is what FrameworkBundle does if you disable the translator component for instance).

I also think we could avoid changing the signature of addViolation, keeping the case of disabling translations or changing the domain as a more advanced case that requires using the violation builder (and also without changing the signature of buildViolation itself).

If we do these 2 things, it simplifies the code a lot as the ExecutionContext does not need to change anymore and we can add ->disableTranslation in the ViolationBuilder.

[VincentLanglet]

{%- if translation_domain is same as(false) -%}

should we update the template so that they use strtr?

This might be a good idea since it would be consistent with the review of this PR and with the behavior of the IdentityTranslator. But I would say it's better to scope this to another PR (that I can do after if wanted).

I'm not sure we should support false for the global configuration (in the context of the framework, that would break all bundles providing validation messages for instance). And if you have a project were you really don't want to use translations anywhere, you can inject an IdentityTranslator (which is what FrameworkBundle does if you disable the translator component for instance).

I agree

I also think we could avoid changing the signature of addViolation, keeping the case of disabling translations or changing the domain as a more advanced case that requires using the violation builder (and also without changing the signature of buildViolation itself).

If we do these 2 things, it simplifies the code a lot as the ExecutionContext does not need to change anymore and we can add ->disableTranslation in the ViolationBuilder.

Since @nicolas-grekas was thinking the same, I reverted this change.

[VincentLanglet]

I squashed the commit and rebase the branch.

Is everything ok on this PR @nicolas-grekas ? :)

[nicolas-grekas]

LGTM yes, thanks. Can you please add some tests?

[VincentLanglet]

LGTM yes, thanks. Can you please add some tests?

Sure, I added one. Do you think this is enough or is there more to test ?

[nicolas-grekas]

Thank you @VincentLanglet.