Spring-hateoas: 0.23.0 problem with link generation with null values

[johann-sonntagbauer]

/* Controller */

@RequestMapping(value = "/{entityContext}/{searchSpecId}/{formLayoutId}", method = RequestMethod.GET, produces = {MediaType.APPLICATION_JSON_UTF8_VALUE})
    public HttpEntity<EntityTable> executeFilterByFormLayout(@PathVariable String entityContext,
            @PathVariable String searchSpecId,
            @PathVariable String formLayoutId,
            @RequestParam(required = false) String eql,
            @RequestParam(required = false) Integer maxResults,
            @RequestParam(required = false) Integer lastIndex,
            @RequestParam(required = false, defaultValue = "false") Boolean counting) {
        ...
    }

/* Link Generation */

  linkTo(
    methodOn(SearchFXApiController.class)
      .executeFilterByFormLayout(resource.getEntityContext(), resource.getSearchSpec(), formLayoutId, null, null, null, null)
  ).withRel(REL_SUBMIT_FORM);

results in following link:
http://localhost:8080/api/v1/fx/search/course/COURSEGUIDE_LEARNING/DEFAULT{?eql,maxResults,lastIndex,counting}

It seems that all null values are collected within the {?} block at the end

[johann-sonntagbauer]

Added an minimal breaking testcase in the testsuite.
Please take a look at the referenced commit.

[otrosien]

I think this is intended, as Link internally holds an UriTemplate. use expand() if you want to pre-fill the values.
The methodOn signature should be seen as "provide the necessary parameter-types to uniquely identify which method you are referring to", not a way to pass the real parameter values down to the Link.

[johann-sonntagbauer]

hm could be. The current situation with the api is kind of confusing.
I simply thought of this beeing a bug, because the api was working with 0.21 and broke during the update.

[johann-sonntagbauer]

@otrosien You are on to something :) I've added link = link.expand(); to the testcase and now it works. But the situation is kind a wired because the api doesn't reflect if I get an UriTemplate or an final Uri. Is the best practicse to always call link.expand() ?

[odrotbohm]

A Link never made a guarantee the href is a URI. That's why getHref() doesn't return URI in the first place. The effect you see is a side effect of our added support of partial URI template expansion on ControllerLinkBuilder. With that in place, dropping unexpanded request parameters is not an option as it'd create inconsistent URI templates with some parameters kept and some dropped.

Link.expand() still keeps the behavior it always had: rejecting non-optional template parameters (like e.g. path segments) and dropping request parameter template variables.

[johann-sonntagbauer]

@olivergierke Thank you for the advice. I have not found any hint about that behaviour in the documentation. Maybe that point should be made more obviouse in the docs with an example.

[johann-sonntagbauer]

so the pattern to use should be

@Test
public void linksToMethodWithPathVariableAndParameter() {
  Link link = linkTo(methodOn(ControllerWithMethods.class).methodForNextPage(null, null, null)).withSelfRel();
  assertPointsToMockServer(link);

  Map<String, Object> args = new HashMap<String, Object>();
  args.put("id", 12);
  args.put("limit", 150);
  link = link.expand(args);
  assertThat(link.getHref(), endsWith("/something/12/foo?limit=150"));
}

[otrosien]

I actually find it quite handy to give out templated URIs, esp. with regards to paging, but you need to make sure all your clients support RFC 6570.

[johann-sonntagbauer]

@otrosien Cool, never heard of the RFC but we definitly will use it. Thank you very much.