Tree-based router library. Useful for (specifically) for Rails and Rack, but probably generally useful for anyone interested in doing routing. Based on Ilya Grigorik suggestion, turns out looking up in a hash and following a tree is faster than Krauter's massive regex approach.


  • Understands single and path-globbing variables

  • Understands arbitrary regex variables

  • Arbitrary HTTP header requirements

  • No optimization phase, so routes are always alterable after the fact

  • Understands Proc and Regex transformations, validations

  • Really, really fast

  • Relatively light and happy code-base, should be easy and fun to alter (it hovers around 1,000 LOC, 800 for the core)

  • Interface and implementation are separate, encouraging cross-pollination

  • Works in 1.9!

Route format

From the rdoc:

Creates a route from path and options


A path consists a mix of dynamic and static parts delimited by /


Dynamic parts are prefixed with either :, *. :variable matches only one part of the path, whereas *variable can match one or more parts.

Example: /path/:variable/path would match

  • /path/test/path

  • /path/something_else/path

  • /path/one_more/path

In the above examples, 'test', 'something_else' and 'one_more' respectively would be bound to the key :variable. However, /path/test/one_more/path would not be matched.

Example: /path/*variable/path would match

  • /path/one/two/three/path

  • /path/four/five/path

In the above examples, ['one', 'two', 'three'] and ['four', 'five'] respectively would be bound to the key :variable.

As well, variables can have a regex matcher.

Example: /product/{:id,\d+} would match

  • /product/123

  • /product/4521

But not

  • /product/AE-35

As well, the same logic applies for * variables as well, where only parts matchable by the supplied regex will actually be bound to the variable

Variables can also have a greedy regex matcher. These matchers ignore all delimiters, and continue matching for as long as much as their regex allows.

Example: /product/{!id,hello/world|hello} would match

  • /product/hello/world

  • /product/hello


Static parts of literal character sequences. For instance, /path/something.html would match only the same path. As well, static parts can have a regex pattern in them as well, such as /path/something.{html|xml} which would match only /path/something.html and /path/something.xml

Optional sections

Sections of a route can be marked as optional by surrounding it with brackets. For instance, in the above static example, /path/something(.html) would match both /path/something and /path/something.html.

One and only one sections

Sections of a route can be marked as “one and only one” by surrounding it with brackets and separating parts of the route with pipes. For instance, the path, /path/something(.xml|.html) would only match /path/something.xml and /path/something.html. Generally its more efficent to use one and only sections over using regex.


  • requirements - After transformation, tests the condition using ===. If it returns false, it raises an Usher::ValidationException

  • conditions - Accepts any of the request_methods specificied in the construction of Usher. This can be either a string or a regular expression.

  • Any other key is interpreted as a requirement for the variable of its name.


script/plugin install git://


require 'usher'
app = proc do |env|
  body = "Hi there #{env['usher.params'][:name]}"
    200,          # Status code
    {             # Response headers
      'Content-Type' => 'text/plain',
      'Content-Length' => body.size.to_s,
    [body]        # Response body

routes = Usher::Interface.for(:rack) do

run routes

>> curl
<< Hi there samueltanders


  • add support for () optional parts

  • Add support for arbitrary HTTP header checks

  • Emit exceptions inline with relevant interfaces

  • More RDoc! (optionally cowbell)


  • Make it integrate with merb

  • Make it integrate with rails3

  • Create decent DSL for use with rack

(Let me show you to your request)