Monk consists of several components:

Monk Binary

At the core of Monk is a command line tool for starting new projects.  Think of the rails command, and what it does and you have a good idea of what the monk command does.  It lists, manages and instantiates new skeletons for your project. The Monk binary can instatiate any skeleton you want, which means it is a great way to start projects in any language.

Monk Skeletons

A Monk skeleton is a set of files defining the structure of a new project.  There are almost no rules here, the new project can be Sinatra, Django, an ircbot setup, or anything else you want, heck, it could even be Rails. 

Monk Default Skeleton

The default Monk skeleton is where lots of work has gone in, and where you’ll find the meat of Monk.  It’s a set of amazing tools forming a full-stack Sinatra based web development platform.

• Sinatra - super lightweight web DSL. You know what this is (and if you don’t, go read the rest of this blog, and check it out)
• Ohm - data access layer using Redis, a stupid fast key-value database.  Super simple models for your app
• Monk Glue - a Reloader, Logger, and Settings handler for your new application. Don’t worry about it, Sinatra Logging finally made easy…  Monk Glue is destined to keep growing too, including a handful of mix & match helpers for common tasks.
• Rack-Test, Contest, Stories, Webrat, Faker, Spawn - An amazing unit and integration test stack, allowing both low level unit tests of your code, and high level integration tests, similar in style to Cucumber, but without that nasty repeating-yourself part.  You owe it to yourself to try the stories & webrat approach to testing your Sinatra application.
• Dependencies - trivial management of all the gems you rely on. Just define a dependencies file, and the dependencies gem will handle vendoring and unpacking your requirements.

Monk Glue

Logging - a perennial problem and hassle in setting up new projects is now simple with Monk, and Monk Glue. 

get '/' do
  logger.info("Entered homepage route")
  "homepage text"
end

Reloader - an alternative to the classic Sinatra reloading (with weird require related issues), and Shotgun, which causes a fork() each request, which is fairly heavyweight.  Glue’s Reloader is the happy midpoint, between the other two solutions. It reloads everything, but in the same Ruby process, making it much faster.

Settings - a minimalist implementation of the settings.yml file, with environment support. Easily switch up your database settings, your email address, url and more.  Bonus Feature: the default skeleton creates an automatic settings.yml file when you first run monk:start. 

get '/' do
  @foobar = settings(:foobar)
  "the value of the foobar setting is:" + @foobar
end

Ohm Persistence Layer

Ohm is a lightweight data management layer that uses Redis as it’s backend. Redis is a fast key-value database, and Ohm makes mappings that are simple to define, and automates the nasty parts of key-value databases like reverse lookups (indexes). Ohm isn’t technically an ORM, since there’s no relational database backing it, but it fills the same niche as ActiveRecord or Datamapper, storing your data for future use.

Awesome Reddit Clone Example

Check out http://news.monkrb.com/, the code is located at http://github.com/monkrb/reddit-clone

Pay attention for a future blog post digging further into this example code, and dissecting it.

Get Monk & Find out More

Install Monk

gem install monk

Links

• Official Monk Website - http://monkrb.com
• Monk Github: http://www.github.com/monkrb/monk
• Monk Default Skeleton Github:  http://www.github.com/monkrb/skeleton
• Monk Glue Github: http://www.github.com/monkrb/glue
• Ohm Documentation: http://ohm.keyvalue.org/
• Ohm Github: http://github.com/soveran/ohm
• Monk Mailing List Link: http://groups.google.com/group/monkrb

In plain Rails, it’s simple:

Rails::Initializer.run do |config|
  config.time_zone = 'Pacific Time (US & Canada)'
end

But I wasn’t running Rails proper, and didn’t want to pull in the whole Rails boot sequence. So the Rails::Initializer call didn’t work. To do it manually, time zone config turns out it’s a 3 lines of setup calls:

require 'activerecord'
Time.zone = "Pacific Time (US & Canada)"
ActiveRecord::Base.time_zone_aware_attributes = true
ActiveRecord::Base.default_timezone = "Pacific Time (US & Canada)"

When you do that, time zone attributes work both going into the database, and coming back out. If you miss the last line, they’ll just not work coming out, which is a damn confusing thing to deal with.

Ohh, and one other thing I ran into when getting this setup is that AR pulls in Active Support, which in turn pulls in Builder (to hack xml support or something). Just watch out for that, you’ll get weird crashes if you don’t have builder installed. Honestly, I just went and commented out the areas of Active Support that did it, I didn’t want builder anyway.

Tokyo Cabinet - A Key Value database that’s stupidly fast 
Tokyo Tyrant - A network layer on top of Tokyo Cabinet
Tokyo Dystopia - A full text search (inverted index) on top of Tokyo Cabinet 

Official Docs:

These include mostly the official C API docs.  They are useful if you skim through them, not for the C definitions, but to get an idea of what is possible.  You will get lots of sparse explanations, but it’s a start.

• Tokyo Cabinet: http://tokyocabinet.sourceforge.net/
• Tokyo Tyrant: http://tokyocabinet.sourceforge.net/tyrantdoc/
• Tokyo Dystopia: http://tokyocabinet.sourceforge.net/dystopiadoc/

Ruby Libraries:

• Official Ruby Lib (forked to github & gemified) http://github.com/careo/tokyocabinet-ruby
• Official Ruby Lib (forked to github & gemified) http://github.com/careo/tokyotyrant-ruby
• Rufus Library (FFI Based) http://github.com/jmettraux/rufus-tokyo/tree/master
• Datamapper Adapter  http://github.com/makoto/dm-tokyo-cabinet-adapter
• Speed Comparison (At the bottom): http://www.igvita.com/2009/02/13/tokyo-cabinet-beyond-key-value-store/

More About Rufus Library:

• Rufus 0.1.5 Release Notes - http://jmettraux.wordpress.com/2009/02/13/rufus-tokyo-015-hail-to-the-tyrant/
• Rufus 0.1.9 Release Notes - http://jmettraux.wordpress.com/2009/02/27/rufus-tokyo-018-rufusedo/

OSX Installation:

Macports version of Tokyo Cabinet and Tokyo Tyrant are broken.  The current state is that there are incompatible versions of cabinet & tyrant, and that there are Portfile problems as well, where the checksum doesn’t match the downloaded file.  I eventually gave up, and just used the instructions at http://openwferu.rubyforge.org/tokyo.html to install it manually (the version numbers in that post are old, but otherwise accurate).  Those instructions install it to /usr/local/bin, so you may need to add it to your path.

Key Value Store Theory:

• This post talks about Redis, but is a good introduction to the thinking needed for Key Value storage, as opposed to a relational approach. http://code.google.com/p/redis/wiki/TwitterAlikeExample
• http://antirez.com/post/Sorting-in-key-value-data-model.html

More Cabinet Docs:

• A quick intro into the storage and theory of Cabinet.  Not much actionable content here, but useful background. http://www.scribd.com/doc/12016121/Tokyo-Cabinet-and-Tokyo-Tyrant-Presentation

Cabinet Storage Engines:

I’m still vague on the real differences in this section. I’ll expand out as I learn and understand more. What I do know is that B-Tree and Table have additional methods on them, and that Tyrant doesn’t support at least the B-Tree ones (at a C level, it’s not a problem with the Ruby bindings).

• Hash - Standard hash structure for storing a key. 
• B-Tree - Tree structure to store keys. Allows for ordering, and multiple values.
• Table - ?? Allows storing of arbitrary keys, and indexes (like CouchDB?).
• ?? - I think array is another one, for fixed length values, but I haven’t gone into that one yet

Other Items:

Lightcloud is some sort of distributed layer on top of tyrant, which gives some horizontal scaling ability. I haven’t looked much further than that.

• Lightcloud - http://opensource.plurk.com/LightCloud/
• Lightcloud Ruby - http://github.com/mitchellh/lightcloud
• Django Tokyo Sessions - http://pypi.python.org/pypi/django-tokyo-sessions/0.1.0

That’s enough for now…. I’ll try to keep this updated as I learn.

Sinatra makes amazing web service endpoints.  The lightweight “nothing in your way” approach is great. Rendering alternate media types isn’t immediately obvious though. Here’s a little snippet that demonstrates rendering JSON. XML would work in a very similar way.

Here’s a quick summary of changes since the 0.2.2 release:

• Add sinatra.gemspec w/ support for github gem builds. Forks can now
  enable the build gem option in github to get free username-sinatra.gem
  builds: gem install username-sinatra.gem — source=http://gems.github.com/
• Require Rack 0.4; removes frozen rack dir.
• Basic RSpec support; require ‘sinatra/test/rspec’ instead of
  ‘sinatra/test/spec’ to use. [avdi]
• before filters can modify request environment vars used for
  routing (e.g., PATH_INFO, REQUEST_METHOD, etc.) for URL rewriting
  type functionality.
• In-file templates now uses @@ instead of ## as template separator.
• Top-level environment test predicates: development?, test?, production?
• Top-level “set”, “enable”, and “disable” methods for tweaking
  app options. [rtomayko] — This allows the syntax in my post about options
• Top-level “use” method for building Rack middleware pipelines
  leading to app. see readme for usage. [rtomayko] — this is the use syntax in my post about sessions
• New “reload” option - set false to disable reloading in development.
• New “host” option - host/ip to bind to [cschneid]
• New “app_file” option - override the file to reload in development
  mode [cschneid] — This makes embedding Sinatra inside other apps easier to do during development. More info in my app_file post.
• Development error/not_found page cleanup [sr, adamwiggins]
• Remove a bunch of core extensions (String#to_param, String#from_param,
  Hash#from_params, Hash#to_params, Hash#symbolize_keys, Hash#pass)
• Various grammar and formatting fixes to README; additions on
  community and contributing [cypher]
• Build RDoc using Hanna template: http://sinatra.rubyforge.org/api/
• Specs, documentation and fixes for splat’n routes [vic] — See how to use splat routes in my post on it.
• Fix whitespace errors across all source files. [rtomayko]
• Fix streaming issues with Mongrel (body not closed). [bmizerany]
• Fix various issues with environment not being set properly (configure
  blocks not running, error pages not registering, etc.) [cypher]
• Fix to allow locals to be passed to ERB templates [cschneid] — More Details
• Fix locking issues causing random errors during reload in development.
• Fix for escaped paths not resolving static files [Matthew Walker]

The module:

module Sinatra
  module Authorization
 
  def auth
    @auth ||= Rack::Auth::Basic::Request.new(request.env)
  end
 
  def unauthorized!(realm="myApp.com")
    header 'WWW-Authenticate' => %(Basic realm="#{realm}")
    throw :halt, [ 401, 'Authorization Required' ]
  end
 
  def bad_request!
    throw :halt, [ 400, 'Bad Request' ]
  end
 
  def authorized?
    request.env['REMOTE_USER']
  end
 
  def authorize(username, password)
    # Insert your logic here to determine if username/password is good
    false
  end
 
  def require_administrative_privileges
    return if authorized?
    unauthorized! unless auth.provided?
    bad_request! unless auth.basic?
    unauthorized! unless authorize(*auth.credentials)
    request.env['REMOTE_USER'] = auth.username
  end
 
  def admin?
    authorized?
  end
 
  end
end

To use the module:

require 'authorization'
 
helpers do
  include Sinatra::Authorization
end
 
get '/admin' do
  require_administrative_privileges
  # Do private stuff
end

EDIT: Thanks to foca on #sinatra for the nicer setup of authorize

License: Because this was extracted out of Wink, follow the Wink license (found here).

Sinatra sends a 302 response code as a redirect. According to the spec, 302 shouldn’t change the request method, but you can see a note saying that most clients do change it. Apparently the mobile browser that person was using did things correctly (instead of the mainstream misinterpretation).

The fix for this in the spec is 2 different response codes: 303 and 307. 303 resets to GET, 307 keeps the same method.

To force sinatra to send a different response code, it’s very simple:

redirect '/', 303 # forces the 303 return code
 
## OR:
 
redirect '/', 307 # forces the 307 return code

Many thanks to Sydd for working through this, he did most of the hunting to find the specification details and to figure out what was going on.

1) Install Lighttpd and Thin

# Figure out lighttpd yourself, it should be handled by your 
# linux distro's package manager
 
# For thin:
gem install thin

2) Create your rackup file - the “require ‘app’” line should require the actual sinatra app you have written.

require 'sinatra'
 
Sinatra::Application.default_options.merge!(
  :run => false,
  :env => :production
)
 
require 'app'
run Sinatra.application

3) Setup a config.yml - change the /path/to/my/app path to reflect reality.

---
    environment: production
    chdir: /path/to/my/app
    address: 127.0.0.1
    user: root
    group: root
    port: 4567
    pid: /path/to/my/app/thin.pid
    rackup: /path/to/my/app/config.ru
    log: /path/to/my/app/thin.log
    max_conns: 1024
    timeout: 30
    max_persistent_conns: 512
    daemonize: true

4) Setup lighttpd.conf - change mydomain to reflect reality. Also make sure the first port here matches up with the port setting in config.yml.

$HTTP["host"] =~ "(www\.)?mydomain\.com"  {
        proxy.balance = "fair"
        proxy.server =  ("/" =>
                                (
                                        ( "host" => "127.0.0.1", "port" => 4567 ),
                                        ( "host" => "127.0.0.1", "port" => 4568 )
                                )
                        )
}

5) Start thin and your application . I have a rake script so I can just call “rake start” rather than typing this in. It’s just a few lines to do that.

thin -s 2 -C config.yml -R config.ru start

Go to mydomain.com/ and see the result! Everything should be setup now, check it out at the domain you setup in your lighttpd.conf file.

Variations - nginx - I haven’t looked into the config file syntax, but there isn’t any reason that this exact same approach wouldn’t work with nginx. Thin and the rackup file stay the same, while the nginx layer has to be configured to reverse proxy to the thin install.

Variations - More Thin instances - To add more thin instances, change the -s 2 parameter on the thin start command to be how ever many servers you want. Then be sure lighttpd proxies to all of them by adding more lines to the proxy statements. Then restart lighttpd and you’re done.

1) URL Options

get '/post/:id' do
  post_id = params[:id]
end
 
# Several options, period gets interpreted as a separator between the two.
get '/author/:id.:format' do
  author_id = params[:id]
  format = params[:format]
end
 
# Several options, separated by '/'
get '/author/:author_id/comments/:comment_id' do
  author_id = params[:author_id]
  comment_id = params[:comment_id]
end

2) URL Parameters - the classic ?foo=bar&baz=quux

# this will get the post id out of /author?id=2
get '/author' do
  author_id = params[:id]
end
 
# Works the same way for the POST method (and DELETE and PUT of course)
post '/author' do
  author_id = params[:id]
end

3) Raw Posted Data - if the POSTed content is in the “application/x-www-form-urlencoded” format, it will get parsed as above, with the params[:id] method of getting the data. But occasionally, you want the whole chunk of raw data. This is useful for images, or big chunks of xml.

post '/data' do
  raw = request.env["rack.input"].read
end

Lets first look at the simple use case:

require 'sinatra'
 
get '/' do
    # Get the string representation
    cookie = request.cookies["thing"]
 
    # Set a default
    cookie ||= 0
 
    # Convert to an integer 
    cookie = cookie.to_i
 
    # Do something with the value
    cookie += 1
 
    # Reset the cookie
    set_cookie("thing", cookie)
 
    # Render something
    "Thing is now: #{cookie}"
end

Setting a path, expiration date, or domain gets a little more complicated - see the source code for set_cookie if you want to dig deeper.

set_cookie("thing", { :domain => myDomain,
                      :path => myPath,
                      :expires => Date.new } )

That’s the easy stuff with cookies - It can also serialize Array objects, separating them with ampersands (&), but when they come back, it doesn’t deserialize or split them in any way, it hands you the raw, encoded string for your parsing pleasure.

I’ll look into how serialization and data handling needs to occur in a later post. I’m guessing that it’s the user’s job to split and un-encode, but I haven’t verified that yet.