Why Rx?

When adding an API to your web service, you have to choose how to encode the data you send across the line. XML is one common choice for this, but it can grow arcane and cumbersome pretty quickly. Lots of webservice authors want to avoid thinking about XML, and instead choose formats that provide a few simple data types that correspond to common data structures in modern programming languages. In other words, JSON and YAML.

Unfortunately, while these formats make it easy to pass around complex data structures, they lack a system for validation. XML has XML Schemas and RELAX NG, but these are complicated and sometimes confusing standards. They're not very portable to the kind of data structure provided by JSON, and if you wanted to avoid XML as a data encoding, writing more XML to validate the first XML is probably even less appealing.

Rx is meant to provide a system for data validation that matches up with JSON-style data structures and is as easy to work with as JSON itself.

Simple Example

type: //rec
  from: //str
  to  : 
    type  : //arr
    length: { min: 1 }
    contents: { type: //str }
  subj: //str
  body: //str
  user: /example.com/account
  "type": "//rec",
  "required": {
    "from": "//str",
    "to"  : {
      "type"    : "//arr",
      "length"  : { "min": 1 },
      "contents": { "type": "//str" }
    "subj": "//str",
    "body": "//str"
  "optional": {
    "user": "/example.com/account"

This sample shows a schema to validate a simple data structure composed of a map (a hash, dict, js Object, or the like) with at least four and maybe five entries. Each entry in the map has its own schema for validation, even though some of the schemata are as simple as "must be a string."

You can see the example schema expressed in either YAML or JSON by clicking the button.

Using Rx

Right now, implementations of Rx have been written in Perl, JavaScript, Ruby, Python, and PHP. You can find them and play with them in their git repository.

Hacking on Rx

One of Rx's strengths is its simplicity. It should be easy to implement in any language into which JSON can be easily loaded, and one of its design goals was to keep its core types simple to implement on any such language. The core tests for proving the acceptability of an Rx implementation are written in JSON, so new implementations need only write a loader and runner for the test suite.

To have a look at the existing Rx code and test suite, check out its git repository.

Join the mailing list.