This project provides some APIs to ease creating REST representations that follow the HATEOAS principle when working with Spring and especially Spring MVC. The core problem it tries to address is link creation and representation assembly.
As representations for REST web services are usually rendered in either XML or JSON the natural choice of technology to achieve this is either JAXB, JSON or both in combination. To follow HATEOAS principles you need to incorporate links into those representation. Spring HATEOAS provides a set of useful types to ease working with those.
The Link
value object follows the Atom link definition and consists of a rel
and an href
attribute. It contains a few constants for well known rels such as self
, next
etc. The XML representation will render in the Atom namespace.
Link link = new Link("http://localhost:8080/something");
assertThat(link.getHref(), is("http://localhost:8080/something"));
assertThat(link.getRel(), is(Link.SELF));
Link link = new Link("http://localhost:8080/something", "my-rel");
assertThat(link.getHref(), is("http://localhost:8080/something"));
assertThat(link.getRel(), is("my-rel"));
As pretty much every representation of a resource will contain some links (at least the self
one) we provide a base class to actually inherit from when designing representation classes.
class PersonResource extends ResourceSupport {
String firstname;
String lastname;
}
Inheriting from ResourceSupport
will allow adding links easily:
PersonResource resource = new PersonResource();
resource.firstname = "Dave";
resource.lastname = "Matthews";
resource.add(new Link("http://myhost/people"));
This would render as follows in JSON:
{ firstname : "Dave",
lastname : "Matthews",
links : [ { rel : "self", href : "http://myhost/people" } ] }
… or slightly more verbose in XML …
<person xmlns:atom="http://www.w3.org/2005/Atom">
<firstname>Dave</firstname>
<lastname>Matthews</lastname>
<links>
<atom:link rel="self" href="http://myhost/people" />
</links>
</person>
You can also easily access links contained in that resource:
Link selfLink = new Link("http://myhost/people");
assertThat(resource.getId(), is(selfLink);
assertThat(resource.getLink(Link.SELF), is(selfLink));
Now we've got the domain vocabulary in place, but the main challenge remains: how to create the actual URIs to be wrapped into Link
s in a less fragile way. Right now we'd have to duplicate URI strings all over the place which is brittle and unmaintainable.
Assume you have your Spring MVC controllers implemented as follows:
@Controller
@RequestMapping("/people")
class PersonController {
@RequestMapping(method = RequestMethod.GET)
public HttpEntity<PersonResource> showAll() { … }
@RequestMapping(value = "/{person}", method = RequestMethod.GET)
public HttpEntity<PersonResource> show(@PathVariable Long person) { … }
}
We see two conventions here. There's a collection resource exposed through the controller class' @RequestMapping
annotation with individual elements of that collections exposed as direct sub resource. The collection resource might be exposed at a simple URI (as just shown) or more complex ones like /people/{id}/addresses
.
Let's say you would like to actually link to the collection resource of all people. Following the approach from up above would cause two problems:
- To create an absolute URI you'd need to lookup the protocol, hostname, port, servlet base etc. This is cumbersome and requires ugly manual string concatenation code.
- You probably don't want to concatenate the
/people
on top of your base URI because you'd have to maintain the information in multiple places then. Change the mapping, change all the clients pointing to it.
Spring Hateoas now provides a ControllerLinkBuilder
that allows to create links by pointing to controller classes:
import static org.sfw.hateoas.mvc.ControllerLinkBuilder.*;
Link link = linkTo(PersonController.class).withRel("people");
assertThat(link.getRel(), is("people"));
assertThat(link.getHref(), endsWith("/people"));
The ControllerLinkBuilder
uses Spring's ServletUriComponentsBuilder
under the hood to obtain the basic URI information from the current request. Assuming your application runs at http://localhost:8080/your-app
This will be exactly the URI you're constructing additional parts on top. The builder now inspects the given controller class for its root mapping and thus end up with http://localhost:8080/your-app/people
. You can also easily build more nested links as well:
Person person = new Person(1L, "Dave", "Matthews");
// /person / 1
Link link = linkTo(PersonController.class).slash(person.getId()).withSelfRel();
assertThat(link.getRel(), is(Link.SELF));
assertThat(link.getHref(), endsWith("/people/1"));
If your domain class implements the Identifiable
interface the slash(…)
method will rather invoke getId()
on the given object instead of toString()
. Thus the just shown link creation can be abbreviated to:
class Person implements Identifiable<Long> {
public Long getId() { … }
}
Link link = linkTo(PersonController.class).slash(person).withSelfRel();
The builder also allows creating URI instances to build up e.g. response header values:
HttpHeaders headers = new HttpHeaders();
headers.setLocation(linkTo(PersonController.class).slash(person).toUri());
return new ResponseEntity<PersonResource>(headers, HttpStatus.CREATED);
As of version 0.4 you can even easily build links pointing to methods or creating dummy controller method invocations. The first approach is to hand a Method
instance to the ControllerLinkBuilder
:
Method method = PersonController.class.getMethod("show", Long.class);
Link link = linkTo(method, 2L).withSelfRel();
assertThat(link.getHref(), is("/people/2")));
This is still a bit dissatisfying as we have to get a Method
instance first, which throws an exception and is generally quite cumbersome. At least we don't repeat the mapping. An even better approach is to have a dummy method invocation of the target method on a controller proxy we can create easily using the methodOn(…)
helper.
Link link = linkTo(methodOn(PersonController.class).show(2L)).withSelfRel();
assertThat(link.getHref(), is("/people/2")));
methodOn(…)
creates a proxy of the controller class that is recording the method invocation and exposed it in a proxy created for the return type of the method. This allows the fluent expression of the method we want to obtain the mapping for. However there are a few constraints on the methods that can be obtained using this technique:
- The return type has to be capable of proxying as we need to expose the method invocation on it.
- The parameters handed into the methods are generally neglected, except the ones referred to through
@PathVariable
as they make up the URI.
So far we have created links by pointing to the web-framework implementations (i.e. Spring MVC controllers or JAX-RS resource classes) and inspected the mapping. In many cases these classes essentially read and write representations backed by a model class.
The EntityLinks
interfaces now exposes API to lookup Link
s or LinkBuilder
s based on the model types. The methods essentially return links to either point to the collection resource (e.g. /people
) or a single resource (e.g. /people/1
).
EntityLinks links = …;
LinkBuilder builder = links.linkFor(CustomerResource.class);
Link link = links.linkToSingleResource(CustomerResource.class, 1L);
EntityLinks
is available for dependency injection by activating @EnableEntityLinks
in your Spring MVC configuration. Activating this functionality will cause all your Spring MVC controllers and JAX-RS resource implementations available in the current ApplicationContext
being inspected for the @ExposesResourceFor(…)
annotation. The annotation exposes which model type the controller manages. Beyond that we assume you follow the URI mapping convention of a class level base mapping and assuming you have controller methods handling an appended /{id}
. Here's an example implementation of an EntityLinks
capable controller:
@Controller
@ExposesResourceFor(Order.class)
@RequestMapping("/orders")
class OrderController {
@RequestMapping
ResponseEntity orders(…) { … }
@RequestMapping("/{id}")
ResponseEntity order(@PathVariable("id") … ) { … }
}
The controller exposes that it manages Order
instances and exposes handler methods that are mapped to our convention. Enabling EntityLinks
through @EnableEntityLinks
in your Spring MVC configuration you can now go ahead and create links to the just shown controller as follows.
@Controller
class PaymentController {
@Autowired EntityLinks entityLinks;
@RequestMapping(…, method = HttpMethod.PUT)
ResponseEntity payment(@PathVariable Long orderId) {
Link link = entityLinks.linkToSingleResource(Order.class, orderId);
…
}
}
As you can see you can refer to the link Order
instances are handled at without even referring to the OrderController
.
As the mapping from an entity to a resource type will have to be used in multiple places it makes sense to create a dedicated class responsible for doing so. The conversion will of course contain very custom steps but also a few boilerplate ones:
- Instantiation of the resource class
- Adding a link with rel
self
pointing to the resource that gets rendered.
Spring Hateoas now provides a ResourceAssemblerSupport
base class that helps reducing the amount of code needed to be written:
class PersonResourceAssembler extends ResourceAssemblerSupport<Person, PersonResource> {
public PersonResourceAssembler() {
super(PersonController.class, PersonResource.class);
}
@Override
public PersonResource toResource(Person person) {
PersonResource resource = createResource(person);
// … do further mapping
return resource;
}
}
Setting the class up like this gives you the following benefits: there are a hand full of createResource(…)
methods that will allow you to create an instance of the resource and have it a Link
with a rel of self
added to it. The href of that link is determined by the configured controllers request mapping plus the id of the Identifiable
(e.g. /people/1
in our case). The resource type gets instantiated by reflection and expects a no-arg constructor. Simply override instantiateResource(…)
in case you'd like to use a dedicated constructor or avoid the reflection performance overhead.
The assembler can then be used to either assemble a single resource or an Iterable
of them:
Person person = new Person(…);
Iterable<Person> people = Collections.singletonList(person);
PersonResourceAssembler assembler = new PersonResourceAssembler();
PersonResource resource = assembler.toResource(person);
List<PersonResource> resources = assembler.toResources(people);
To enable the ResourceSupport
subtypes be rendered according to the specification of various hypermedia representations types, the support for a particular hypermedia representation format can be activated through @EnableHypermediaSupport
. The annotation takes a HypermediaType
enumeration as argument. Currently we support HAL as well as a default rendering. Using the annotation triggers the following:
- registers necessary Jackson modules to render
Resource
/Resources
in the hypermedia specific format. - if JSONPath is on the classpath, it automatically registers a
LinkDiscoverer
instance to lookup links by theirrel
s in plain JSON representations (see below). - enables
@EnableEntityLinks
by default (see above), will automatically pick upEntityLinks
implementations and bundle them into aDelegatingEntityLinks
instance available for autowiring. - automatically picks up all
RelProvider
implementations in theApplicationContext
and bundles them into aDelegatingRelProvider
available for autowiring. Registers providers to consider@Relation
on domain types as well as Spring MVC controllers. If EVO inflector is on the classpath collection rels are derived using the pluralizing algorithm implemented in the library (see below).
When working with hypermedia enabled representations, a common task is to find a link with a particular relation type in them. Spring HATEOAS provides JSONPath based implementations of the LinkDiscoverer
interface for either the default representation rendering or HAL out of the box. When using @EnableHypermediaSupport
we automatically expose an instance supporting the configured hypermedia type as Spring bean.
Alternatively you can simply setup and use an instance like this:
String content = "{'_links' : { 'foo' : { 'href' : '/foo/bar' }}}";
LinkDiscoverer discoverer = new HalLinkDiscoverer();
Link link = discoverer.findLinkWithRel("foo", content);
assertThat(link.getRel(), is("foo"));
assertThat(link.getHref(), is("/foo/bar"));
When building links you usually need to determine the relation type to be used for the link. In most cases the relation type is directly associated with a (domain) type. We encapsulate the detailed algorithm to lookup the relation types behind a RelProvider
API that allows to determine the relation types for single and collection resources. Here's the algorithm the relation type is looked up:
- If the type is annotated with
@Relation
we use the values configured in the annotation. - if not, we default to the uncapitalized simple class name plus an appended
List
for the collection rel. - in case the EVO inflector JAR is in the classpath, we rather use the plural of the single resource rel provided by the pluralizing algorithm.
@Controller
classes annotated with@ExposesResourceFor
(see section on EntityLinks for details) will transparently lookup the relation types for the type configured in the annotation, so that you can userelProvider.getSingleResourceRelFor(MyController.class)
and get the relation type of the domain type exposed.
A RelProvider
is exposed as Spring bean when using @EnableHypermediaSupport
automatically. You can plug in custom providers by simply implementing the interface and exposing them as Spring bean in turn.
The Web Linking RFC describes registered and extension link relation types. Registered rels are well-known strings registered with the IANA registry of link relation types. Extension rels can be used by applications that do not wish to register a relation type. They are a URI that uniquely identifies the relation type. The rel URI can be serialized as a compact URI or Curie. E.g. a curie ex:persons
stands for the link relation type http://example.com#persons
if ex
is defined as http://example.com#
. If curies are used, the base URI must be present in the response scope.
The rels created by the default RelProvider are extension relation types and as such must be URIs, which can cause a lot of overhead. The CurieProvider
API takes care of that: it allows to define a base URI as URI template and a prefix which stands for that base URI. If a CurieProvider
is present, the RelProvider prepends all rels with the curie prefix. Furthermore a curies
link is automatically added to the HAL resource.
The configuration below defines a default curie provider.
@Configuration
@EnableWebMvc
@EnableHypermediaSupport(type= {HypermediaType.HAL})
public class Config {
@Bean
public CurieProvider curieProvider() {
return new DefaultCurieProvider("ex",
new UriTemplate("http://www.example.com{#rel}"));
}
}
Note that now the prefix ex:
automatically appears before all rels which are not registered with IANA, as in ex:orders
. Clients can use the curies
link to resolve a curie to its full form:
{
_links : {
self: { href: "http://myhost/person/1" },
curies: {
name: "ex",
href: "http://example.com#{rel}",
templated: true
},
"ex:orders" : { href: "http://myhost/person/1/orders" }
},
firstname : "Dave",
lastname : "Matthews"
}
Since the purpose of the CurieProvider
API is to allow for automatic curie creation, you can define only one CurieProvider
bean per application scope.
As of version 0.11 Spring HATEOAS provides an API for client side service traversal inspired by the Traverson JavaScript library.
Map<String, Object> parameters = new HashMap<>();
parameters.put("user", 27);
Traverson traverson = new Traverson(new URI("http://localhost:8080/api/"), MediaTypes.HAL_JSON);
String name = traverson.follow("movies", "movie", "actor").
withTemplateParameters(parameters).
toObject("$.name");
You set up a Traverson
instance by pointing it to a REST server and configure the media types you want to set as Accept
header. You then go ahead and define the relation names you want to discover and follow. relation names can either be simple names or JSONPath expressions (starting with an $
).
The sample then hands a parameter map into the execution. The parameters will be used to expand URIs found during the traversal that are templated. The traversal is concluded by accessing the representation of the final traversal. In the case of the sample we evaluate a JSONPath expression to access the actor's name.