Skip to content

Conversation

nicolas-grekas
Copy link
Member

@nicolas-grekas nicolas-grekas commented Aug 26, 2025

Q A
Branch? 7.4
Bug fix? no
New feature? yes
Deprecations? no
Issues -
License MIT

Prerequisite for #61288

At the moment, validation attributes are read at runtime when framework.validation.enable_attributes is true.

This means they don't fit for bundles nor can't they be warmed up.

This PR fixes both issues by using a new validator.attribute_metadata resource tag, that's turned into a list of classes to parse for attributes at compile-time.

For apps, the tag is added by autoconfiguration: any Constraint-derived attributes found on a class in the src/ folder will trigger the rule to add the tag.

For bundles (and for apps if they want to), the tag is added by explicit service configuration. In an "eat your own dog-food" spirit, this capability is used to declare the constraints of the Form class: instead of loading the validation.xml file, we now declare this service resource:

        ->set('validator.form.attribute_metadata', Form::class)
            ->resourceTag('validator.attribute_metadata')

This reads the attributes added to the Form class:

#[AssertForm()]
#[Traverse(false)]
class Form implements \IteratorAggregate, FormInterface, ClearableErrorsInterface

Bundles can do the same and replace their XML files by attributes.

As a next step, we could also deprecate runtime-discovery of attributes. This could be worth it if this discovery has a measurable performance impact. To be measured if one wants to dig this idea.

Side note: I'm hoping this could allow removing the yaml and xml config formats one day. For serialization metadata also (PR coming). BUT, this doesn't (yet) cover the use case of overriding metadata defined by bundles. For that, apps still have to use xml or yaml in config/validation/. I have an idea to cover this, coming to a next PR if it works.

(failures unrelated)

@carsonbot carsonbot added this to the 7.4 milestone Aug 26, 2025
@nicolas-grekas nicolas-grekas force-pushed the validator.attribute_metadata branch 3 times, most recently from 289783f to 8be45c5 Compare August 26, 2025 08:37
@nicolas-grekas nicolas-grekas changed the title [Validator] Allow using attributes on classes as constraint metadata configuration [Validator] Allow using attributes to declare compile-time constraint metadata Aug 26, 2025
@nicolas-grekas nicolas-grekas force-pushed the validator.attribute_metadata branch 3 times, most recently from defd051 to ea317e8 Compare August 26, 2025 16:01
@nicolas-grekas nicolas-grekas force-pushed the validator.attribute_metadata branch 2 times, most recently from d3490c8 to 9d08608 Compare August 27, 2025 10:05
@nicolas-grekas nicolas-grekas force-pushed the validator.attribute_metadata branch 2 times, most recently from 3dc6ebf to 35c205e Compare August 27, 2025 12:41
@@ -27,6 +27,7 @@
*
* @author Bernhard Schussek <bschussek@gmail.com>
*/
#[\Attribute(\Attribute::TARGET_CLASS | \Attribute::TARGET_PROPERTY | \Attribute::TARGET_METHOD | \Attribute::IS_REPEATABLE)]
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why adding this here ? The abstract class cannot be used as an attribute. And child classes defining concrete constraints will have to specify the proper target anyway.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's required by the engine in relation to registerAttributeForAutoconfiguration

@nicolas-grekas nicolas-grekas force-pushed the validator.attribute_metadata branch 3 times, most recently from ca89d23 to dbe819b Compare September 3, 2025 15:37
@nicolas-grekas nicolas-grekas force-pushed the validator.attribute_metadata branch from dbe819b to 0197b50 Compare September 3, 2025 15:41
Copy link
Member

@GromNaN GromNaN left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@nicolas-grekas nicolas-grekas merged commit c539c77 into symfony:7.4 Sep 3, 2025
10 of 12 checks passed
@nicolas-grekas nicolas-grekas deleted the validator.attribute_metadata branch September 3, 2025 16:45
nicolas-grekas added a commit that referenced this pull request Sep 3, 2025
…-time serialization metadata (nicolas-grekas)

This PR was merged into the 7.4 branch.

Discussion
----------

[Serializer] Allow using attributes to declare compile-time serialization metadata

| Q             | A
| ------------- | ---
| Branch?       | 7.4
| Bug fix?      | no
| New feature?  | yes
| Deprecations? | yes
| Issues        | -
| License       | MIT

Very close to #61528
Prerequisite for #61287

At the moment, serialization attributes are read at runtime when `framework.serialization.enable_attributes` is true.

This means they don't fit for bundles nor can't they be warmed up.

This PR fixes both issues by using a new `serializer.attribute_metadata` resource tag, that's turned into a list of classes to parse for attributes at compile-time.

For bundles and for apps, the tag is added by explicit service configuration:
```php
        ->set('my_bundle.api_resource.product', Product::class)
            ->resourceTag('serializer.attribute_metadata')
```

Unlike validation where we have constraint attributes to auto-discover service resources, serialization doesn't have any corresponding hooks. We do have a few like `#[DiscriminatorMap]` of `#[Groups]`, but relying on those would miss many more classes that are meant for serialization. Maybe we could introduce an attribute that'd hint that some class is serializable by the component (and require the attribute at some point in the future?)

Commits
-------

bc6e054 [Serializer] Allow using attributes to declare compile-time serialization metadata
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

6 participants