Trying: REST api designΒΆ

Tags: django, nelenschuurmans

Two months ago I made my opinion about SOAP and REST clear. Happily, I now have a small project where I have to actually make a REST interface. So here’s some preliminary thinkwork. Suggestions welcome!

My thought on what a proper REST api ought to be about:

  • Linked data. No magic URLs which you have to know exactly. From a root URL, you should be able to find everything you need by just following the URLs. So not a list of IDs of something that you have to hand-craft a URL with, but simply a list of URLs with the ID already included.
  • Json. In principle, json is returned. No html, no xml. Json is the de-facto simple data exchange format nowadays. Images, html, wms: those are alternative representations that can be requested.
  • Self-describing. Adding a couple of ‘name’ and ‘explanation’ attributes to the json helps a lot in explaining the data. It aids discovery and debugging.
  • Resource-based. The ‘normal’ part of the URL is for identifying resources. Extra parameters (?width=100&height=80) are for small tweaks and adjustments.
  • Identify resources, not actions. Give every resource a URL, restrict the use of verbs. So not /get_document?id=42, but /documents/42.

In our case, in the end we have to display data about a certain filter/parameter/location combination (from “FEWS”, a water management database). We do that via a “JDBC” coupling.

To cut to the chase, this is my initial brainstormy url structure. Everything postfixed with _id is an id (or slug) that depends on the actual item being requested:


Every call returns json, except for the csv/table/png representations being requested of the actual filter/parameter/location “end points”.

The api/jdbc/filter/parameter resources return a json with a list of items within it, so /api/ returns a list of available jdbcs, for instance. Those list items provide the name and url of the jdbcs it points at, see this partial json example:

[{"name": "Production FEWS at N&S",
  "url": ""},

In this way, it is easy to construct a web interface based on these json files. You do not need to know the available IDs beforehand, you can simply ask the system for the current list. And requesting the next set of information is simply a matter of “following the url”.

Every json return will have a fixed structure:

{"info": {"name": "One-line name of current resource",
          "explanation": "Explanation of contents, for programmers"},
 "alternative_representations": [{"type": "csv",
                                  "url": "..."},
                                 {"type": "png",
                                  "url": "..."}],
 "data": [...the actual data, like list of subitems or actual data ...]}

Everything with a url parameter can also have an arguments parameter:

"url": "....",
"arguments": {"period": "period in days counted backwards from now",
              "width": "graph width in pixels"
              "height": "graph height in pixels"}

All in all, a call to get a png from the production jdbc, filter 32, parameter 2, location 134, for the last month, would be:
My son looking at a sluice
blog comments powered by Disqus logo

About me

My name is Reinout van Rees and I work a lot with Python (programming language) and Django (website framework). I live in The Netherlands and I'm happily married to Annie van Rees-Kooiman.

Weblog feeds

Most of my website content is in my weblog. You can keep up to date by subscribing to the automatic feeds (for instance with Google reader):