Dropwizard Validation

Dropwizard comes with a host of validation tools out of the box to allow endpoints to return meaningful error messages when constraints are violated. Hibernate Validator is packaged with Dropwizard, so what can be done in Hibernate Validator, can be done with Dropwizard.

Validations

Almost anything can be validated on resource endpoints. To give a quick example, the following endpoint doesn’t allow a null or empty name query parameter.

@GET
public String find(@QueryParam("name") @NotEmpty String arg) {
    // ...
}

If a client sends an empty or nonexistent name query param, Dropwizard will respond with a 400 Bad Request code with the error: query param name may not be empty.

Additionally, annotations such as HeaderParam, CookieParam, FormParam, etc, can be constrained with violations giving descriptive errors and 400 status codes.

Constraining Entities

If we’re accepting client-provided Person, we probably want to ensure that the name field of the object isn’t null or blank in the request. We can do this as follows:

public class Person {

    @NotEmpty // ensure that name isn't null or blank
    private final String name;

    @JsonCreator
    public Person(@JsonProperty("name") String name) {
        this.name = name;
    }

    @JsonProperty("name")
    public String getName() {
        return name;
    }
}

Then, in our resource class, we can add the @Valid annotation to the Person annotation:

@PUT
public Person replace(@NotNull @Valid Person person) {
    // ...
}

If the name field is missing, Dropwizard will return a 422 Unprocessable Entity response detailing the validation errors: name may not be empty

Note

You don’t need @Valid when the type you are validating can be validated directly (int, String, Integer). If a class has fields that need validating, then instances of the class must be marked @Valid. For more information, see the Hibernate Validator documentation on Object graphs and Cascaded validation.

Since our entity is also annotated with @NotNull, Dropwizard will also guard against null input with a response stating that the body must not be null.

Constraints on optional types

If an entity, field, or parameter is not strictly required, it can be wrapped in an Optional<T>, but the wrapped value can still be constrained by setting the annotation parameter payload = Unwrapping.Unwrap.class. If the optional container is empty, then the constraints are not applied.

Numeric optional types, such as OptionalDouble, OptionalInt, and OptionalLong do not require explicit unwrapping.

Note

Be careful when using constraints with validation annotations on Optional<String> parameters as there is a subtle, but important distinction between null and empty. If a client requests bar?q=, the query parameter q will evaluate to Optional.of(""). If you want q to evaluate to Optional.empty() in this situation, change the type to NonEmptyStringParam.

Note

Parameter types such as IntParam and NonEmptyStringParam can also be constrained.

There is a caveat regarding payload = Unwrapping.Unwrap.class and *Param types, as there still are some cumbersome situations when constraints need to be applied to the container and the value.

@POST
// The @NotNull is supposed to mean that the parameter is required but the Max(3) is supposed to
// apply to the contained integer. Currently, this code will fail saying that Max can't
// be applied on an IntParam
public List<Person> createNum(@QueryParam("num")
                              @NotNull(payload = Unwrapping.Unwrap.class)
                              @Max(value = 3, payload = Unwrapping.Unwrap.class) IntParam num) {
    // ...
}

@GET
// Similarly, the underlying validation framework can't unwrap nested types (an integer wrapped
// in an IntParam wrapped in an Optional), regardless if `Unwrapping.Unwrap.class` is used
public Person retrieve(@QueryParam("num") @Max(3) Optional<IntParam> num) {
    // ...
}

To work around these limitations, if the parameter is required check for it in the endpoint and throw an exception, else use @DefaultValue or move the Optional<T> into the endpoint.

@POST
// Workaround to handle required int params and validations
public List<Person> createNum(@QueryParam("num") @Max(3) IntParam num) {
    if (num == null) {
        throw new BadRequestException("query param num must not be null");
    }
    // ...
}

@GET
// Workaround to handle optional int params and validations with DefaultValue
public Person retrieve(@QueryParam("num") @DefaultValue("0") @Max(3) IntParam num) {
    // ...
}

@GET
// Workaround to handle optional int params and validations with Optional
public Person retrieve2(@QueryParam("num") @Max(3) IntParam num) {
    Optional.fromNullable(num);
    // ...
}

Enum Constraints

Given the following enum:

public enum Choice {
    OptionA,
    OptionB,
    OptionC
}

And the endpoint:

@GET
public String getEnum(@NotNull @QueryParam("choice") Choice choice) {
    return choice.toString();
}

One can expect Dropwizard not only to ensure that the query parameter exists, but to also provide the client a list of valid options query param choice must be one of [OptionA, OptionB, OptionC] when an invalid parameter is provided. The enum that the query parameter is deserialized into is first attempted on the enum’s name() field and then toString(). During the case insensitive comparisons, the query parameter has whitespace removed with dashes and dots normalized to underscores. This logic is also used when deserializing request body’s that contain enums.

Return Value Validations

It’s reasonable to want to make guarantees to clients regarding the server response. For example, you may want to assert that no response will ever be null, and if an endpoint creates a Person that the person is valid.

@POST
@NotNull
@Valid
public Person create() {
    return new Person(null);
}

In this instance, instead of returning someone with a null name, Dropwizard will return an HTTP 500 Internal Server Error with the error server response name may not be empty, so the client knows the server failed through no fault of their own.

Analogous to an empty request body, an empty entity annotated with @NotNull will return server response may not be null

Warning

Be careful when using return value constraints when endpoints satisfy all of the following:

Function name starts with get
No arguments
The return value has validation constraints

If an endpoint satisfies these conditions, whenever a request is processed by the resource that endpoint will be additionally invoked. To give a concrete example:

@Path("/")
public class ValidatedResource {
    private AtomicLong counter = new AtomicLong();

    @GET
    @Path("/foo")
    @NotEmpty
    public String getFoo() {
        counter.getAndIncrement();
        return "";
    }

    @GET
    @Path("/bar")
    public String getBar() {
        return "";
    }
}

If a /foo is requested then counter will have increment by 2, and if /bar is requested then counter will increment by 1. It is our hope that such endpoints are few, far between, and documented thoroughly.

Limitations

Jersey allows for BeanParam to have setters with *Param annotations. While nice for simple transformations it does obstruct validation, so clients won’t receive as instructive of error messages. The following example shows the behavior:

@Path("/root")
@Produces(MediaType.APPLICATION_JSON)
public class Resource {

    @GET
    @Path("params")
    public String getBean(@Valid @BeanParam MyBeanParams params) {
        return params.getField();
    }

    public static class MyBeanParams {
        @NotEmpty
        private String field;

        public String getField() {
            return field;
        }

        @QueryParam("foo")
        public void setField(String field) {
            this.field = Strings.nullToEmpty(field).trim();
        }
    }
}

A client submitting the query parameter foo as blank will receive the following error message:

{"errors":["getBean.arg0.field may not be empty"]}

Workarounds include:

Name BeanParam fields the same as the *Param annotation values
Supply validation message on annotation: @NotEmpty(message = "query param foo must not be empty")
Perform transformations and validations on *Param inside endpoint

The same kind of limitation applies for Configuration objects:

public class MyConfiguration extends Configuration {
    @NotNull
    @JsonProperty("foo")
    private String baz;
}

Even though the property’s name is foo, the error when property is null will be:

* baz may not be null

Annotations

In addition to the annotations defined in Hibernate Validator, Dropwizard contains another set of annotations, which are briefly shown below.

public class Person {
    @NotEmpty
    private final String name;

    @NotEmpty
    @OneOf(value = {"m", "f"}, ignoreCase = true, ignoreWhitespace = true)
    // @OneOf forces a value to value within certain values.
    private final String gender;

    @Min(value = 0, payload = Unwrapping.Unwrap.class)
    @Max(value = 10, payload = Unwrapping.Unwrap.class)
    // The integer contained, if present, can attain a min value of 0 and a max of 10.
    private final Optional<Integer> animals;

    @JsonCreator
    public Person(@JsonProperty("name") String name) {
        this.name = name;
    }

    @JsonProperty("name")
    public String getName() {
        return name;
    }

    // Method that must return true for the object to be valid
    @ValidationMethod(message="name may not be Coda")
    @JsonIgnore
    public boolean isNotCoda() {
        return !"Coda".equals(name);
    }
}

The reason why Dropwizard defines @ValidationMethod is that more complex validations (for example, cross-field comparisons) are often hard to do using declarative annotations. Adding @ValidationMethod to any boolean-returning method which begins with is is a short and simple workaround:

Note

Due to the rather daft JavaBeans conventions, when using @ValidationMethod, the method must begin with is (e.g., #isValidPortRange(). This is a limitation of Hibernate Validator, not Dropwizard.

Validating Grouped Constraints with `@Validated`

The @Validated annotation allows for validation groups to be specifically set, instead of the default group. This is useful when different endpoints share the same entity but may have different validation requirements.

Going back to our favorite Person class. Let’s say in the initial version of our API, name has to be non-empty, but realized that business requirements changed and a name can’t be longer than 5 letters. Instead of switching out the API from unsuspecting clients, we can accept both versions of the API but at different endpoints.

// We're going to create a group of validations for each version of our API
public interface Version1Checks { }

// Our second version will extend Hibernate Validator Default class so that any validation
// annotation without an explicit group will also be validated with this version
public interface Version2Checks extends Default { }

public class Person {
    @NotEmpty(groups = Version1Checks.class)
    @Length(max = 5, groups = Version2Checks.class)
    private String name;

    @JsonCreator
    public Person(@JsonProperty("name") String name) {
        this.name = name;
    }

    @JsonProperty
    public String getName() {
        return name;
    }
}

@Path("/person")
@Produces(MediaType.APPLICATION_JSON)
public class PersonResource {

    // For the v1 endpoint, we'll validate with the version1 class, so we'll need to change the
    // group of the NotNull annotation from the default of Default.class to Version1Checks.class
    @POST
    @Path("/v1")
    public void createPersonV1(
        @NotNull(groups = Version1Checks.class)
        @Valid
        @Validated(Version1Checks.class)
        Person person
    ) {
        // implementation ...
    }

    // For the v2 endpoint, we'll validate with version1 and version2, which implicitly
    // adds in the Default.class.
    @POST
    @Path("/v2")
    public void createPersonV2(
        @NotNull
        @Valid
        @Validated({Version1Checks.class, Version2Checks.class})
        Person person
    ) {
        // implementation ...
    }
}

Now when clients hit /person/v1 the Person entity will be checked by all the constraints that are a part of the Version1Checks group. If /person/v2 is hit, then all validations are performed.

Warning

If the Version1Checks group wasn’t set for the @NotNull annotation for the v1 endpoint, the annotation would not have had any effect and a null pointer exception would have occurred when a property of a person is accessed. Dropwizard tries to protect against this class of bug by disallowing multiple @Validated annotations on an endpoint that contain different groups.

Testing

It is critical to test the constraints so that you can ensure the assumptions about the data hold and see what kinds of error messages clients will receive for bad input. The recommended way for testing annotations is through Testing Resources, as Dropwizard does a bit of magic behind the scenes when a constraint violation occurs to set the response’s status code and ensure that the error messages are user friendly.

@Test
public void personNeedsAName() {
    // Tests what happens when a person with a null name is sent to
    // the endpoint.
    final Response post = resources.target("/person/v1").request()
            .post(Entity.json(new Person(null)));

    // Clients will receive a 422 on bad request entity
    assertThat(post.getStatus()).isEqualTo(422);

    // Check to make sure that errors are correct and human readable
    ValidationErrorMessage msg = post.readEntity(ValidationErrorMessage.class);
    assertThat(msg.getErrors())
            .containsOnly("name may not be empty");
}

Extending

While Dropwizard provides good defaults for validation error messages, one can customize the response through an ExceptionMapper<JerseyViolationException>:

/** Return a generic response depending on if it is a client or server error */
public class MyJerseyViolationExceptionMapper implements ExceptionMapper<JerseyViolationException> {
    @Override
    public Response toResponse(final JerseyViolationException exception) {
        final Set<ConstraintViolation<?>> violations = exception.getConstraintViolations();
        final Invocable invocable = exception.getInvocable();
        final int status = ConstraintMessage.determineStatus(violations, invocable);
        return Response.status(status)
                .type(MediaType.TEXT_PLAIN_TYPE)
                .entity(status >= 500 ? "Server error" : "Client error")
                .build();
    }
}

To register MyJerseyViolationExceptionMapper and have it override the default:

@Override
public void run(final MyConfiguration conf, final Environment env) {
    env.jersey().register(new MyJerseyViolationExceptionMapper());
    env.jersey().register(new Resource());
}

Dropwizard calculates the validation error message through ConstraintMessage.getMessage.

If you need to validate entities outside of resource endpoints, the validator can be accessed in the Environment when the application is first ran.

Validator validator = environment.getValidator();
Set<ConstraintViolation> errors = validator.validate(/* instance of class */)

Dropwizard Validation

Validations

Constraining Entities

Constraints on optional types

Enum Constraints

Return Value Validations

Limitations

Annotations

Validating Grouped Constraints with @Validated

Testing

Extending

Validating Grouped Constraints with `@Validated`