Bean Validation

Update to the Bean Validation 2.0 Public Review Draft

Posted by Gunnar Morling    |    24 May 2017    release

While the Bean Validation 2.0 spec (JSR 380) has been put up for Public Review last month, we’ve continued to address some more outstanding issues, added some clarifications etc.

Of course we wanted to get out these improvements as quickly as possible, so we’ve published an update to the Public Review draft (the JCP rules explicitly foresee the possibility of updates during the Public Review phase). The updated draft can be found here on beanvalidation.org or you can download it from jcp.org.

As always the updated API is deployed to Maven Central, using the GAV coordinates are javax.validation:validation-api:2.0.0.Beta2. Alternatively, it’s part of the ZIP that can be downloaded from from jcp.org.

What has changed since the original Public Review Draft?

Value extractors got more flexible and can extract values from non-generic wrapper types now.

This allows to put all the constraints for numeric types (e.g. @Min, @Max, @DecimalMin, @DecimalMax etc.) to the optional numeric wrapper types in Java 8: OptionalInt, OptionalLong and OptionalDouble (BVAL-579). Also other JVM languages with their own numeric types benefit from that, as one can define value extractors for such types and apply all existing numeric constraints to them. The key to this is the new attribute @ExtractedValue#type() which allows to specify the type of the wrapped element for extractors of non-generic types.

Another improvement related to constraints on the elements of (generic) containers is the possibility to build property paths leading to a container element using the node builder API exposed via ConstraintValidatorContext (BVAL-592). For instance this will let you declare the container type and type argument index of a container element from within a custom class-level constraint validator.

The following validator for the ValidUser constraint will ensure that an entry for the WORK key is in the addresses map if the user’s type is EMPLOYEE. If not, it will create a property path pointing to the map’s value, with the container type (Map.class) and type argument index (1, representing the V type parameter, as it’s about the map value):

@ValidUser
public class User {

    private UserType type;
    private Map<AddressType, Address> addressesByType;

    // getters, setters etc. ...
}
@Constraint(validatedBy=UserValidator.class)
public @interface ValidUser {
    // constraint attributes ...
}
public class UserValidator implements ConstraintValidator<ValidUser, User> {

    @Override
    public boolean isValid(User user, ConstraintValidatorContext context) {
        context.disableDefaultConstraintViolation();

        if ( user.getType() == UserType.EMPLOYEE &&
                user.getAddressesByType().get( AddressType.WORK ) == null ) {
            context.buildConstraintViolationWithTemplate( "Work address needed for employee" )
                .addContainerElementNode( "addressesByType", Map.class, 1 )
                    .inIterable()
                        .atKey( AddressType.WORK )
                .addConstraintViolation();

            return false;
        }

        return true;
    }
}

One further improvement to mention is that the specification now recommends the module name "java.validation" (BVAL-517), should Bean Validation providers wish to distribute the API as a module for the Java Platform Module System (JPMS) as currently developed under JSR 376. This could be used within a module-info.java descriptor or via the manifest entry "Automatic-Module-Name" when preparing the module to be used as an automatic module.

Note that this is a non-binding recommendation as of Bean Validation 2.0. A mandatory module name (which could be "java.validation" or something else) — and potentially other requirements relating to the module system — will be mandated in a future revision of the spec, once the module module system has been finalized and best practices around modules have emerged.

The complete list of all issues can be found in the release notes. There are also HTML diffs which highlight what has changed since the original Public Review Draft (2.0.0.Beta1) and since Bean Validation 1.1.

New releases of reference implementation and TCK

Together with the specification we’ve also released a corresponding version of the reference implementation Hibernate Validator 6. Refer to the announcement blog post for more details. This release lets you try out all the new features added in Bean Validation 2.0.

Also the TCK (test compatibility kit) has been updated and released. You can learn more in the TCK documentation and the release notes.

What can you do to help?

The Public Review phase still runs for a few more days, followed by the Public Review Ballot (the voting by the JCP executive committee) until June 12th. So it’s the perfect time for reviewing the spec changes, we are very curious about your feedback. Also it’d be of great help if you tried out the reference implementation in your applications and let us know how it works. Any feedback is appreciated!

To post your feedback, just add a comment below, send a message to our mailing list or post in the Bean Validation forum. If you find a bug or have a specific feature request, please raise it in the issue tracker.

Everything in Bean Validation is open source, so you also can send in actual patches: to the API, the spec document or the TCK. If you are interested, you can find out all the details to get started in the contribution guide.