webmachine for Ruby travis

webmachine-ruby is a port of Webmachine, which is written in Erlang. The goal of both projects is to expose interesting parts of the HTTP protocol to your application in a declarative way. This means that you are less concerned with handling requests directly and more with describing the behavior of the resources that make up your application. Webmachine is not a web framework per se, but more of a toolkit for building HTTP-friendly applications. For example, it does not provide a templating engine or a persistence layer; those choices are up to you.

A Note about Rack

Webmachine has a Rack adapter -- thanks to Jamis Buck -- but when using it, we recommend you ensure that NO middleware is used. The behaviors that are encapsulated in Webmachine could be broken by middlewares that sit above it, and there is no way to detect them at runtime. Caveat implementor. That said, Webmachine should behave properly when given a clear stack.

Getting Started

Webmachine is very young, but it's still easy to construct an application for it!

require 'webmachine'
# Require any of the files that contain your resources here
require 'my_resource'

# Create an application which encompasses routes and configruation
MyApp = Webmachine::Application.new do |app|
  app.routes do
    # Point all URIs at the MyResource class
    add ['*'], MyResource
  end
end

# Start the server, binds to port 8080 using WEBrick
MyApp.run

Your resource will look something like this:

class MyResource < Webmachine::Resource
  def to_html
    "<html><body>Hello, world!</body></html>"
  end
end

Run the first file and your application is up. That's all there is to it! If you want to customize your resource more, look at the available callbacks in lib/webmachine/resource/callbacks.rb. For example, you might want to enable "gzip" compression on your resource, for which you can simply add an encodings_provided callback method:

class MyResource < Webmachine::Resource
  def encodings_provided
    {"gzip" => :encode_gzip, "identity" => :encode_identity}
  end

  def to_html
    "<html><body>Hello, world!</body></html>"
  end
end

There are many other HTTP features exposed to your resource through Webmachine::Resource::Callbacks. Give them a try!

Application/Configurator

There's a configurator that allows you to set the ip address and port bindings as well as a different webserver adapter. You can also add your routes in a block (as shown above). Both of these call return the Webmachine::Application instance, so you could chain them if you like. If you don't want to create your own separate application object, Webmachine.application will return a global one.

require 'webmachine'
require 'my_resource'

Webmachine.application.routes do
  add ['*'], MyResource
end

Webmachine.application.configure do |config|
  config.ip = '127.0.0.1'
  config.port = 3000
  config.adapter = :Mongrel
end

# Start the server.
Webmachine.application.run

Webmachine includes adapters for Webrick, Mongrel, Reel, and Hatetepe. Additionally, the Rack adapter lets it run on any webserver that provides a Rack interface. It also lets it run on Shotgun (example).

Visual debugger

It can be hard to understand all of the decisions that Webmachine makes when servicing a request to your resource, which is why we have the "visual debugger". In development, you can turn on tracing of the decision graph for a resource by implementing the #trace? callback so that it returns true:

class MyTracedResource < Webmachine::Resource
  def trace?
    true
  end

  # The rest of your callbacks...
end

Then enable the visual debugger resource by adding a route to your configuration:

Webmachine.application.routes do
  # This can be any path as long as it ends with '*'
  add ['trace', '*'], Webmachine::Trace::TraceResource
  # The rest of your routes...
end

Now when you visit your traced resource, a trace of the request process will be recorded in memory. Open your browser to /trace to list the recorded traces and inspect the result. The response from your traced resource will also include the X-Webmachine-Trace-Id that you can use to lookup the trace. It might look something like this:

preview calls at decision

Refer to examples/debugger.rb for an example of how to enable the debugger.

Features

  • Handles the hard parts of content negotiation, conditional requests, and response codes for you.
  • Most callbacks can interrupt the decision flow by returning an integer response code. You generally only want to do this when new information comes to light, requiring a modification of the response.
  • Supports WEBrick and Mongrel (1.2pre+), and a Rack shim. Other host servers are being investigated.
  • Streaming/chunked response bodies are permitted as Enumerables, Procs, or Fibers!
  • Unlike the Erlang original, it does real Language negotiation.
  • Includes the visual debugger so you can look through the decision graph to determine how your resources are behaving.

Caveats

  • The Reel adapter might fail with a SystemStackError on MRI (< 2.0) due to its limited fiber stack size. The only known solution is to switch to JRuby, Rubinius or MRI 2.0.

Documentation & Finding Help

LICENSE

webmachine-ruby is licensed under the Apache v2.0 license. See LICENSE for details.

Changelog

1.1.0 January 12, 2013

1.1.0 is a major feature release that adds the Reel and Hatetepe adapters, support for "weak" entity tags, streaming IO response bodies, better error handling, a shortcut for spinning up specific resources, and a bunch of bugfixes. Added Tony Arcieri, Sebastian Edwards, Russell Garner, Justin McPherson, Paweł Pacana, and Nicholas Young as contributors. Thank you for your contributions!

  • Added Reel adapter.
  • The trace resource now opens static files in binary mode to ensure compatibility on Windows.
  • The trace resource uses absolute URIs for its traces.
  • Added Hatetepe adapter.
  • Added direct weak entity tag support.
  • Related libraries are linked from the README.
  • Removed some circular requires.
  • Fixed documentation for the valid_content_headers? callback.
  • Fixed Headers initialization by downcasing incoming header names.
  • Added a Headers#fetch method.
  • Conventionally "truthy" and "falsey" values (non-nil, non-false) can now be returned from callbacks that expect a boolean return value.
  • Updated to the latest RSpec.
  • Added support for IO response bodies (minimal).
  • Moved streaming encoders to their own module for clarity.
  • Added Resource#run that starts up a web server with default configuration options and the catch-all route to the resource.
  • The exception handling flow was improved, clarifying the handle_exception and finish_request callbacks.
  • Fix incompatibilities with Rack.
  • The request URI will not be initialized with parts that are not present in the HTTP request.
  • The tracing will now commit to storage after the response has been traced.

1.0.0 July 7, 2012

1.0.0 is a major feature release that finally includes the visual debugger, some nice cookie support, and some new extension points. Added Peter Johanson and Armin Joellenbeck as contributors. Thank you for your contributions!

  • A cookie parsing and manipulation API was added.
  • Conneg headers now accept any amount of whitespace around commas, including none.
  • Callbacks#handle_exception was added so that resources can handle exceptions that they generate and produce more friendly responses.
  • Chunked and non-chunked response bodies in the Rack adapter were fixed.
  • The WEBrick example was updated to use the new API.
  • Dispatcher was refactored so that you can modify how resources are initialized before dispatching occurs.
  • Route now includes the Translation module so that exception messages are properly rendered.
  • The visual debugger was added (more details in the README).
  • The Content-Length header will always be set inside Webmachine and is no longer reliant on the adapter to set it.

0.4.2 March 22, 2012

0.4.2 is a bugfix release that corrects a few minor issues. Added Lars Gierth and Rob Gleeson as contributors. Thank you for your contributions!

  • I always intended for Webmachine-Ruby to be Apache licensed, but now that is explicit.
  • When the #process_post callback returns an invalid value, that will now be inspected in the raised exception's message.
  • Route bindings are now applied to the Request object before the Resource class is instantiated. This means you can inspect them inside the #initialize method of your resource.
  • Some NameError exceptions and scope problems in the Mongrel adapter were resolved.
  • URL-encoded = characters in the query string decoded in the proper order.

0.4.1 February 8, 2012

0.4.1 is a bugfix release that corrects a few minor issues. Added Sam Goldman as a contributor. Thank you for your contributions!

  • Updated README with Webmachine::Application examples.
  • The CGI env vars CONTENT_LENGTH and CONTENT_TYPE are now being correctly converted into their Webmachine equivalents.
  • The request body given via the Rack and Mongrel adapters now responds to #to_s and #each so it can be treated like a String or Enumerable that yields chunks.

0.4.0 February 5, 2012

0.4.0 includes some important refactorings, isolating the idea of global state into an Application object with its own Dispatcher and configuration, and making Adapters into real classes with a consistent interface. It also adds some query methods on the Request object for the HTTP method and scheme and Route guards (matching predicates). Added Michael Maltese, Emmanuel Gomez, and Bernerd Schaefer as committers. Thank you for your contributions!

  • Fixed Request#query to handle nil values for the URI query accessor.
  • Webmachine::Dispatcher is a real class rather than a module with state.
  • Webmachine::Application is a class that includes its own dispatcher and configuration. The default instance is accessible via Webmachine.application.
  • Webmachine::Adapter is now the superclass of all implemented adapters so that they have a uniform interface.
  • The Mongrel spec is skipped on JRuby since version 1.2 (pre-release) doesn't work. Direct Mongrel support may be removed in a later release.
  • Webmachine::Dispatcher::Route now accepts guards, which may be expressed as lambdas/procs or any object responding to call preceding the Resource class in the route definition, or as a trailing block. All guards will be passed the Request object when matching the route and should return a truthy or falsey value (without side-effects).

0.3.0 November 9, 2011

0.3.0 introduces some new features, refactorings, and now has 100% documentation coverage! Among the new features are minimal Rack compatibility, streaming responses via Fibers and a friendlier route definition syntax. Added Jamis Buck as a committer. Thank you for your contributions!

  • Chunked bodies are now wrapped in a way that works on webservers that don't automatically produce them.
  • HTTP Basic Authentication is easy to add to resources, just include Webmachine::Resource::Authentication.
  • Routes are a little less painful to add, you can now specify them with Webmachine.routes which will be evaled into the Dispatcher.
  • The new default port is 8080.
  • Rack is minimally supported as a host server. Don't put middleware above Webmachine!
  • Fibers can be used as streamed response bodies.
  • Dispatcher#add_route will now return the added Route instance.
  • The header-conversion code for CGI-style servers has been extracted into Webmachine::Headers.
  • Route#path_spec is now public so that applications can inspect existing routes, perhaps for URL generation.
  • Request#query now uses CGI.unescape so '+' characters are correctly parsed.
  • YARD documentation has 100% coverage.

0.2.0 September 11, 2011

0.2.0 includes an adapter for Mongrel and a central place for configuration as well as numerous bugfixes. Added Ian Plosker and Bernd Ahlers as committers. Thank you for your contributions!

  • Acceptable media types are matched less strictly, which has implications on both responses and PUT requests. See the discussion on the commit.
  • Resources now receive a callback after the language has been negotiated, so they can decide what to do with it.
  • Added Webmachine::Configuration so we can more easily support more than one host server/adapter.
  • Added Mongrel adapter, supporting 1.2pre+.
  • Media type headers are more lax about whitespace following semicolons.
  • Fix some problems with callable response bodies.
  • Make sure String response bodies get a Content-Length header added and streaming responses get chunked encoding.
  • Numerous refactorings, including extracting MediaType into its own top-level class.

0.1.0 August 25, 2011

This is the initial release. Most things work, but only WEBrick is supported.