Hibernate.orgCommunity Documentation

Chapter 6. Creating custom constraints

6.1. Creating a simple constraint
6.1.1. The constraint annotation
6.1.2. The constraint validator
6.1.3. The error message
6.1.4. Using the constraint
6.2. Class-level constraints
6.2.1. Custom property paths
6.3. Cross-parameter constraints
6.4. Constraint composition

The Bean Validation API defines a whole set of standard constraint annotations such as @NotNull, @Size etc. In cases where these buit-in constraints are not sufficient, you cean easily create custom constraints tailored to your specific validation requirements.

To create a custom constraint, the following three steps are required:

This section shows how to write a constraint annotation which can be used to ensure that a given string is either completely upper case or lower case. Later on this constraint will be applied to the licensePlate field of the Car class from Chapter 1, Getting started to ensure, that the field is always an upper-case string.

The first thing needed is a way to express the two case modes. While you could use String constants, a better approach is using a Java 5 enum for that purpose:

The next step is to define the actual constraint annotation. If you've never designed an annotation before, this may look a bit scary, but actually it's not that hard:

An annotation type is defined using the @interface keyword. All attributes of an annotation type are declared in a method-like manner. The specification of the Bean Validation API demands, that any constraint annotation defines

  • an attribute message that returns the default key for creating error messages in case the constraint is violated

  • an attribute groups that allows the specification of validation groups, to which this constraint belongs (see Chapter 5, Grouping constraints). This must default to an empty array of type Class<?>.

  • an attribute payload that can be used by clients of the Bean Validation API to assign custom payload objects to a constraint. This attribute is not used by the API itself. An example for a custom payload could be the definition of a severity:

    public class Severity {
        public interface Info extends Payload {
        public interface Error extends Payload {
    public class ContactDetails {
        @NotNull(message = "Name is mandatory", payload = Severity.Error.class)
        private String name;
        @NotNull(message = "Phone number not specified, but not mandatory",
                payload = Severity.Info.class)
        private String phoneNumber;
        // ...

    Now a client can after the validation of a ContactDetails instance access the severity of a constraint using ConstraintViolation.getConstraintDescriptor().getPayload() and adjust its behaviour depending on the severity.

Besides these three mandatory attributes there is another one, value, allowing for the required case mode to be specified. The name value is a special one, which can be omitted when using the annotation, if it is the only attribute specified, as e.g. in @CheckCase(CaseMode.UPPER).

In addition, the constraint annotation is decorated with a couple of meta annotations:

  • @Target({ FIELD, METHOD, PARAMETER, ANNOTATION_TYPE }): Defines the supported target element types for the constraint. @CheckCase may be used on fields (element type FIELD), JavaBeans properties as well as method return values (METHOD) and method/constructor parameters (PARAMETER). The element type ANNOTATION_TYPE allows for the creation of composed constraints (see Section 6.4, “Constraint composition”) based on @CheckCase.

    When creating a class-level constraint (see Section 2.1.3, “Class-level constraints”), the element type TYPE would have to be used. Constraints targetting the return value of a constructor need to support the element type CONSTRUCTOR. Cross-parameter constraints (see Section 6.3, “Cross-parameter constraints”) which are used to validate all the parameters of a method or constructor together, must support METHOD or CONSTRUCTOR, respectively.

  • @Retention(RUNTIME): Specifies, that annotations of this type will be available at runtime by the means of reflection

  • @Constraint(validatedBy = CheckCaseValidator.class): Marks the annotation type as constraint annotation and specifies the validator to be used to validate elements annotated with @CheckCase. If a constraint may be used on several data types, several validators may be specified, one for each data type.

  • @Documented: Says, that the use of @CheckCase will be contained in the JavaDoc of elements annotated with it

Finally, there is an inner annotation type named List. This annotation allows to specify several @CheckCase annotations on the same element, e.g. with different validation groups and messages. While also another name could be used, the Bean Validation specification recommends to use the name List and make the annotation an inner annotation of the corresponding constraint type.

Having defined the annotation, you need to create a constraint validator, which is able to validate elements with a @CheckCase annotation. To do so, implement the interface ConstraintValidator as shown below:

The ConstraintValidator interface defines two type parameters which are set in the implementation. The first one specifies the annotation type to be validated (CheckCase), the second one the type of elements, which the validator can handle (String). In case a constraint supports several data types, a ConstraintValidator for each allowed type has to be implemented and registered at the constraint annotation as shown above.

The implementation of the validator is straightforward. The initialize() method gives you access to the attribute values of the validated constraint and allows you to store them in a field of the validator as shown in the example.

The isValid() method contains the actual validation logic. For @CheckCase this is the check whether a given string is either completely lower case or upper case, depending on the case mode retrieved in initialize(). Note that the Bean Validation specification recommends to consider null values as being valid. If null is not a valid value for an element, it should be annotated with @NotNull explicitly.

Example 6.3, “Implementing a constraint validator for the constraint @CheckCase” relies on the default error message generation by just returning true or false from the isValid() method. Using the passed ConstraintValidatorContext object it is possible to either add additional error messages or completely disable the default error message generation and solely define custom error messages. The ConstraintValidatorContext API is modeled as fluent interface and is best demonstrated with an example:

Example 6.4, “Using ConstraintValidatorContext to define custom error messages” shows how you can disable the default error message generation and add a custom error message using a specified message template. In this example the use of the ConstraintValidatorContext results in the same error message as the default error message generation.


It is important to add each configured constraint violation by calling addConstraintViolation(). Only after that the new constraint violation will be created.

Refer to Section 6.2.1, “Custom property paths” to learn how to use the ConstraintValidatorContext API to control the property path of constraint violations for class-level constraints.

You can now use the constraint in the Car class from the Chapter 1, Getting started chapter to specify that the licensePlate field should only contain upper-case strings:

Finally, Example 6.7, “Validating objects with the @CheckCase constraint” demonstrates how validating a Car object with an invalid license plate causes the @CheckCase constraint to be violated.

As discussed earlier, constraints can also be applied on the class level to validate the state of an entire object. Class-level constraints are defined in the same was as are property constraints. Example 6.8, “Implementing a class-level constraint” shows constraint annotation and validator of the @ValidPassengerCount constraint you already saw in use in Example 2.3, “Class-level constraint”.

As the example demonstrates, you need to use the element type TYPE in the @Target annotation. This allows the constraint to be put on type definitions. The validator of the constraint in the example receives a Car in the isValid() method and can access the complete object state to decide whether the given instance is valid or not.

By default the constraint violation for a class-level constraint is reported on the level of the annotated type, e.g. Car.

In some cases it is preferable though that the violation's property path refers to one of the involved properties. For instance you might want to report the @ValidPassengerCount constraint against the passengers property instead of the Car bean.

Example 6.9, “Adding a new ConstraintViolation with custom property path” shows how this can be done by using the constraint validator context passed to isValid() to build a custom constraint violation with a property node for the property passengers. Note that you also could add several property nodes, pointing to a sub-entity of the validated bean.

Bean Validation distinguishes between two different kinds of constraints.

Generic constraints (which have been discussed so far) apply to the annotated element, e.g. a type, field, method parameter or return value etc. Cross-parameter constraints, in contrast, apply to the array of parameters of a method or constructor and can be used to express validation logic which depends on several parameter values.

In order to define a cross-parameter constraint, its validator class must be annotated with @SupportedValidationTarget(ValidationTarget.PARAMETERS). The type parameter T from the ConstraintValidator interface must resolve to either Object or Object[] in order to receive the array of method/constructor arguments in the isValid() method.

The following example shows the definition of a cross-parameter constraint which can be used to check that two Date parameters of a method are in the correct order:

The definition of a cross-parameter constraint isn't any different from defining a generic constraint, i.e. it must specify the members message(), groups() and payload() and be annotated with @Constraint. This meta annotation also specifies the corresponding validator, which is shown in Example 6.11, “Generic and cross-parameter constraint”. Note that besides the element types METHOD and CONSTRUCTOR also ANNOTATION_TYPE is specified as target of the annotation, in order to enable the creation of composed constraints based on @ConsistentDateParameters (see Section 6.4, “Constraint composition”).


Cross-parameter constraints are specified directly on the declaration of a method or constructor, which is also the case for return value constraints. In order to improve code readability, it is therefore recommended to chose constraint names - such as @ConsistentDateParameters - which make the constraint target apparent.

As discussed above, the validation target PARAMETERS must be configured for a cross-parameter validator by using the @SupportedValidationTarget annotation. Since a cross-parameter constraint could be applied to any method or constructor, it is considered a best practice to check for the expected number and types of parameters in the validator implementation.

As with generic constraints, null parameters should be considered valid and @NotNull on the individual parameters should be used to make sure that parameters are not null.


Similar to class-level constraints, you can create custom constraint violations on single parameters instead of all parameters when validating a cross-parameter constraint. Just obtain a node builder from the ConstraintValidatorContext passed to isValid() and add a parameter node by calling addParameterNode(). In the example you could use this to create a constraint violation on the end date parameter of the validated method.

In rare situations a constraint is both, generic and cross-parameter. This is the case if a constraint has a validator class which is annotated with @SupportedValidationTarget({ValidationTarget.PARAMETERS, ValidationTarget.ANNOTATED_ELEMENT}) or if it has a generic and a cross-parameter validator class.

When declaring such a constraint on a method which has parameters and also a return value, the intended constraint target can't be determined. Constraints which are generic and cross-parameter at the same time, must therefore define a member validationAppliesTo() which allows the constraint user to specify the constraint's target as shown in Example 6.12, “Generic and cross-parameter constraint”.

The @ScriptAssert constraint has two validators (not shown), a generic and a cross-parameter one and thus defines the member validationAppliesTo(). The default value IMPLICIT allows to derive the target automatically in situations where this is possible (e.g. if the constraint is declared on a field or on a method which has parameters but no return value).

If the target can not be determined implicitly, it must be set by the user to either PARAMETERS or RETURN_VALUE as shown in Example 6.13, “Specifying the target for a generic and cross-parameter constraint”.

Looking at the licensePlate field of the Car class in Example 6.6, “Applying the @CheckCase constraint”, you see three constraint annotations already. In complexer scenarios, where even more constraints could be applied to one element, this might become a bit confusing easily. Furthermore, if there was a licensePlate field in another class, you would have to copy all constraint declarations to the other class as well, violating the DRY principle.

You can address this kind of problem by creating higher level constraints, composed from several basic constraints. Example 6.14, “Creating a composing constraint @ValidLicensePlate” shows a composed constraint annotation which comprises the constraints @NotNull, @Size and @CheckCase:

To create a composed constraint, simply annotate the constraint declaration with its comprising constraints. If the composed constraint itself requires a validator, this validator is to be specified within the @Constraint annotation. For composed constraints which don't need an additional validator such as @ValidLicensePlate, just set validatedBy() to an empty array.

Using the new composed constraint at the licensePlate field is fully equivalent to the previous version, where the three constraints were declared directly at the field itself:

The set of ConstraintViolations retrieved when validating a Car instance will contain an entry for each violated composing constraint of the @ValidLicensePlate constraint. If you rather prefer a single ConstraintViolation in case any of the composing constraints is violated, the @ReportAsSingleViolation meta constraint can be used as follows: