JAX-RS 2.0 pagination with Link header

Your REST API is looking good, everything is ready, but here comes a paginated resource. There could be various reasons to do it, be it search, listing, etc.

In this matter several choices to implement that appear :

  • Do nothing that’s the client’s job to control everything to iterate over pages

  • Wrap the items within an envelope in the HTTP body, this envelope will also contain field having information on the pagination. You can find variation on the exact implementation some choose json-api, some their home brewed solution. This is probably the most common approach as this does require less HTTP knowledge.

    And it is likely preferred by pure Javascript clients as there’s no need to peek outside the body.

  • Use the HTTP response with pagination meta-data being placed in the headers. That’s the path chosen by Github in their v3 API. That’s also this option that this blog post will explore.

More specifically the Github API is using the Link header form the HTTP standard. It allows express relations and how to reach them for a resource.

API implementors that want separate concerns between the content and the pagination could be interested by this article that show a basic setup in footsteps of Github and with JAX-RS 2.

The RFC 5988 says :

In this specification, a link is a typed connection between two resources that are identified by Internationalised Resource Identifiers (IRIs) [RFC3987], and is comprised of:

  • A context IRI,
  • a link relation type (Section 4),
  • a target IRI, and
  • optionally, target attributes.

A link can be viewed as a statement of the form “{context IRI} has a {relation type} resource at {target IRI}, which has {target attributes}”.

The RFC gives this example :

Link: <http://example.com/TheBook/chapter2>; rel="previous"; title="previous chapter"

The Link header is a way to advertise to the client what are the URIs of this resource’s relations. This header can also contains several links and so several relations. These relations allows a client to navigate from a resource to another ; GitHub chose this approach with their V3 API. In other words the HTTP response is the envelope of the payload.

JAX-RS, Do it yourself

How to implement this pagination approach with JAX-RS ?

Since JAX-RS 2.0 the Link header is specifically supported via the Link class and via the .links(Link...) in the Response builder.

Response.ok()
        .links(Link.valueOf("<https://host:port/path?q=foo&page=72>; rel=\"previous results\""),
               Link.valueOf("<https://host:port/path?q=foo&page=74>; rel=\"next results\""))
        .entity(page_73_results)
        .build());

In the above example, a link is created by parsing an actual Link string with the target IRI and other meta-data like the relation parameter.

In order to generate the target IRI, be it relative or absolute, the code would need the actual URI data in order to generate all constituent parts of the relation. We’ll need JAX-RS’s UriInfo class. As a reminder:

Given the following URI http://localhost:59520/path/to/search?foo=bar&qix=zzz, then

  • uriInfo.getAbsolutePath() will return http://localhost:59520/path/to/search
  • uriInfo.getBaseUri() will return http://localhost:59520/
  • uriInfo.getPath() will return path/to/search
  • uriInfo.getQueryParameters() will return a MultivaluedMap with
    "foo" : ["bar"]
    "qix" : ["zzz"]
    
  • uriInfo.getRequestUri() will return http://localhost:59520/path/to/search?foo=bar&qix=zzz

A JAX-RS resource will typically be written like that.

public static String X_TOTAL_COUNT = "X-Total-Count";

@Context
UriInfo uriInfo;

@GET @Path("search")
public void search(@BeanParam QueryParams queryParams,
                   @Suspended AsyncResponse asyncResponse) {

    SamePaginated results = // get paginated results
    LinkPagination linkPagination = new LinkPagination(results.currentPageIndex(),
                                                       results.pageCount())

    asyncResponse.resume(Response.ok()
                                 .links(linkPagination.toLinks(uriInfo)
                                                      .toArray(Link[]::new))
                                 .header(X_TOTAL_COUNT, results.totalCount())
                                 .entity(results.currentPage())
                                 .build());
}

Notice the class LinkPagination that has the responsibility to create the JAX-RS Link instances that have to be used in the response builder.

import javax.ws.rs.core.Link;
import javax.ws.rs.core.UriInfo;
import java.util.stream.Stream;

public class LinkPagination {
    public static final String PREV_REL = "prev";
    public static final String NEXT_REL = "next";
    public static final String FIRST_REL = "first";
    public static final String LAST_REL = "last";
    public static final String PAGE_QUERY_PARAM = "page";
    public static final int FIRST_PAGE = 1

    public final int pageCount;
    public final int currentPageIndex;

    public LinkPagination(int currentPageIndex, int pageCount) { /* ... */ }

    public Stream<Link> toLinks(UriInfo uriInfo) {
        if (currentPageIndex == 1 && pageCount == 1) {
            return Stream.empty();
        }

        Stream.Builder<Link> linkStreamBuilder = Stream.builder();

        if (currentPageIndex > 1) {
            linkStreamBuilder.accept(
                Link.fromUriBuilder(uriInfo.getRequestUriBuilder()
                                           .replaceQueryParam(PAGE_QUERY_PARAM,
                                                              currentPageIndex - 1))
                    .rel(PREV_REL)
                    .build());
        }

        if (currentPageIndex < pageCount) {
            linkStreamBuilder.accept(
                Link.fromUriBuilder(uriInfo.getRequestUriBuilder()
                                           .replaceQueryParam(PAGE_QUERY_PARAM,
                                                              currentPageIndex + 1))
                    .rel(NEXT_REL)
                    .build());
        }

        linkStreamBuilder.accept(
            Link.fromUriBuilder(uriInfo.getRequestUriBuilder()
                                       .replaceQueryParam(PAGE_QUERY_PARAM,
                                                          FIRST_PAGE))
                .rel(FIRST_REL)
                .build());

        linkStreamBuilder.accept(
            Link.fromUriBuilder(uriInfo.getRequestUriBuilder()
                                       .replaceQueryParam(PAGE_QUERY_PARAM,
                                                          pageCount))
                .rel(LAST_REL)
                .build());

        return linkStreamBuilder.build();
    }
}

The toLinks() method’s job is to determine which page relations exists and emit them if necessary. As a side: when there a Link header there’s usually self Link, the implementation chose not to emit this self relation.

uriInfo is used to build the link URI, the Uri builder has useful method that allows replace query params replaceQueryParam(paramName,value). This approach require however to make sure the parameter have the same name ; however as developers we understand that pagination is cross cutting concern and as such we understand that pagination parameters have to be the same across the application. This resource uses the bean param class QueryParams that use a constant for page param.

@Data
public class QueryParams {
    @QueryParam("q") String question;

    @Min(value = 1, message = "page start at 1")
    @QueryParam(PAGE_QUERY_PARAM) @DefaultValue("1") Integer page;
}

Obviously this naming constraint could be improved in various ways outside the scope of this blog post.

However if there’s multiple paginated resources, then we want to avoid repeating this code in each JAX-RS resource. Having multiple places with such code make more likely to be borken, or harder to benefit from bugfix or updates in general. It’s time for a better design.

JAX-RS, a more generic approach

The version 2 of JAX-RS alos brought us filters, they are applied on the request pipeline. There’s two type of filters, one integrates in the request processing, the second integrates on the response processing. This design need to filter the response. The following code implements ContainerResponseFilter to identify a paginated payload and to emit the links header in the response.

@Provider
public class LinkPaginationContainerResponseFilter implements ContainerResponseFilter {

    public static final String X_TOTAL_COUNT = "X-Total-Count";
    public static final String X_PAGE_COUNT = "X-Page-Count";

    @Override
    public void filter(ContainerRequestContext requestContext,
                       ContainerResponseContext responseContext) {

        if (!(responseContext.getEntity() instanceof Paginated)) {
            return;
        }

        UriInfo uriInfo = requestContext.getUriInfo();
        Paginated entity = (Paginated) responseContext.getEntity();

        responseContext.setEntity(entity.currentPage);
        responseContext.getHeaders()
                       .addAll(LINK,
                               new LinkPagination(
                                   entity.currentPageIndex(),
                                   entity.pageCount()
                               ).toLinks(uriInfo).toArray(Link[]::new)
                       );
        responseContext.getHeaders().add(X_TOTAL_COUNT, entity.totalCount());
        responseContext.getHeaders().add(X_PAGE_COUNT, entity.pageCount());
    }
}

This code uses the request’s UriInfo in order to build relation links.

As one can notice the API of ContainerResponseContext is a bit less comfortable to use than the Response builder API in the regard of Link headers (there’s no .links() equivalent).

To activate this filter the entity has to implement Paginated, let’s declare it on the business DTO SomePaginated.

public interface Paginated<T> {
    T currentPage();
    int currentPageIndex();
    int pageCount();
    int totalCount();
}
public class SomePaginated implements Paginated<Page> { /* ... */ }

Then the resource code can be written in much more simplier way :

@GET @Path("search")
public void search(@BeanParam QueryParams queryParams,
                   @Suspended AsyncResponse asyncResponse) {

    SomePaginated results = // get paginated results
    asyncResponse.resume(results);
}

Any future resource can then return any object representing a pagination by making the entity class implements Paginated. The filter will be the only place where bugfix or improvements will take place.

Things to consider

  • Wether the link IRI should be absolute with authority or relative.

    e.g. It may be that there’s a front that serves rendered pages on a different domain than the domain of the API.

  • Weither the enclosing type of the payload is a collection type or an object.

    An array has the merit to actually represent a collection of items. However the server may want to expose business meta-data, related to the items returned, like A/B testing flags, or search flags if is a search, etc. This information may be exposed via headers, but is it the right place for business data ?

  • That last point lead to another thought, if meta-data are in the payload does the pagination data must be there as well ?

    My opinion is it shouldn’t. The pagination is elated to the protocol and how to access pages of a resource, while other business related data isn’t. Besides having this pagination data in the response headers don’t force someone to parse the payload to access and use them.


Blog post on the same topic but focusing specifically on how to handle the pagination with a DB and only with Jersey.