This course is archived!

This tutorial uses a deprecated micro-framework called Silex. The fundamentals of REST are still ?valid, but the code we use can't be used in a real application.

Buy Access to Course
26.

Filtering and HATEOAS (The Buzzword)

Share this awesome video!

|

Keep on Learning!

Similar to pagination is filtering. Let's say that I want to be able to search in the programmers collection by nickname. So I'll add a &nickname=2. Why 2? Well, my programmer nicknames aren't very interesting, but if I use 2, I'll want it to return Programmer2 and Programmer12. Of course right now, this doesn't have any affect yet.

Filtering: Use Query Parameters

Filtering is actually really easy. But the reason I wanted to cover it is that sometimes people wonder if they should get clever with their URLs when they're filtering and maybe come up with URLs like /api/programmers/filter/nickname2. Don't do that. If you're filtering, use a query parameter, end of story.

Coding up a Simple Filter

So how do we get this to work? It couldn't be simpler. At the top of listAction, I'll look for the nickname query parameter. And if it's present, we'll query in a special way. And if it's not, we'll do the normal findAll. I have a shortcut setup to query using MySQL LIKE. I'll pass it the value surrounded by the percent signs:

// ... lines 1 - 89
public function listAction(Request $request)
{
$nicknameFilter = $request->query->get('nickname');
if ($nicknameFilter) {
$programmers = $this->getProgrammerRepository()
->findAllLike(array('nickname' => '%'.$nicknameFilter.'%'));
} else {
$programmers = $this->getProgrammerRepository()->findAll();
}
// ... lines 99 - 123
}
// ... lines 125 - 214

And that should be it!

If we go back to the Hal Browser and hit go, we get nothing back! But we're actually on page 3, so click to go back to page 1. Hmm, now we have too many results! That's because we lost the &nickname query parameter. That's because I was lazy - if I have extra filters I should pass those to the 3rd argument of PaginatedRepresentation. If I do that, it'll show up in our pagination links.

I'll re-add &nickname=2 manually and this time, we see it's only returning Programmer2 and Programmer12, and it knows that there's only going to be one page of them. So that's it for filtering: use query parameters, do whatever logic you need for filtering, and pass the filter to PaginatedRepresentation, even though I didn't do it here. It's that easy.

Hypermedia as the Engine of Application State (HATEOAS)

One quick thing to notice is that even though we now support this nickname filter, there isn't any way for the API client to know this just by looking at the links. We have links to help them go through the pages, but no link that says: "Hey, if you want, you could pass a nickname query parameter to filter this collection". There's not really a way with HAL to do that. There are other formats that give you more options and ways to say "Hey, you can add the following 4 query parameters to this URI, and here's what each will do".

And this is one of those spots where REST can get frustrating. The purpose of adding links is to help make your API client's life easier. They are not intended to replace human-readable documentation. So even if you did find a clever way to include filtering information in your response, this would just be as a "nice feature" for your API. You should still document which endpoints have which filters and what they mean.

The term HATEOAS means: hypermedia as the engine of application state. And as cool as it is, it's at the heart of this confusion. In its most pure form, HATEOAS is an idea that seems to suggest that if your API has really good links, it doesn't need human-readable technical documentation. Instead, a client can follow the links in your response and other details you return to figure everything out.

But this is a dream, not a reality. As cool as our API is, it lacks a ton of details. For example, it doesn't have any information about filtering. Second, it doesn't tell you which HTTP methods each URL supports, nor what fields you should POST to /api/programmers in order to create a new resource. The links we have are nice, but they're nowhere near giving the API client everything it needs.

So think of links as a nice addition to your API, but not something that'll replace really nice human documentation.