tcrayford

class RubbishController
  protected
  def redirect_to(path)
    puts "redirecting_to #{path}"
  end

  def render(renderable)
    puts "rendering #{renderable}"
  end
  def current_user

tcrayford

class FuckYouPrivate
  def initialize(thing)
    @thing = thing
  end
  def method_missing(message, *args, &block)
    @thing.send(message, *args, &block)
  end
end

tcrayford

A Controlled Explosion

Well designed Rails controller methods follow very similar execution paths. Raptor is an experimental web framework designed to explode the Rails controller, reconstructing it inside the router.

This talk is a dive into Raptor's design, eventual goals and a small demonstration of what Raptor looks like today.

tcrayford

#logback.classic pattern: %d{HH:mm:ss.SSS} [%thread] %-5level %logger{36} - %msg%n
#logback.classic pattern: %-5p [%d{ISO8601,Europe/London}] %c: %m
%ex
#logback.classic pattern: %-5p [%d{ISO8601,UTC}] %c: %m
%ex
DEBUG [2012-05-25 23:57:42,351] org.eclipse.jetty.util.log: Logging to Logger[org.eclipse.jetty.util.log] via org.eclipse.jetty.util.log.Slf4jLog
DEBUG [2012-05-25 23:57:42,381] org.hibernate.validator.engine.resolver.DefaultTraversableResolver: Cannot find javax.persistence.Persistence on classpath. Assuming non JPA 2 environment. All properties will per default be traversable.
DEBUG [2012-05-25 23:57:42,382] org.hibernate.validator.xml.ValidationXmlParser: Trying to load META-INF/validation.xml for XML based Validator configuration.
DEBUG [2012-05-25 23:57:42,382] org.hibernate.validator.xml.ValidationXmlParser: No META-INF/validation.xml found. Using annotation based configuration only
WARN  [2012-05-25 23:57:42,446] com.yammer.dropwizard.config.ServerFactory: 

tcrayford

  specify "the first time rspec has made me sad" do
    class CustomError < RuntimeError; end
    expect do
      raise CustomError.new('what')
    end.to raise_error(CustomError.new('what'))
  end

tcrayford

  specify "the first time rspec has made me sad" do
    class CustomError < RuntimeError; end
    expect do
      raise CustomError.new('what')
    end.to raise(CustomError.new('what'))
  end

tcrayford

1.9.3p125 :001 > class Foo < RuntimeError; def to_str; "other_string"; end; end
 => nil 
1.9.3p125 :002 > begin raise Foo.new; rescue Foo => e; end
RuntimeError: other_string
	from (irb):2
	from /Users/tcrayford/.rvm/rubies/ruby-1.9.3-p125/bin/irb:16:in `<main>'

tcrayford

Note, the "don't stub external APIs" is a rule, but like all design/testing rules it *should* be broken with good judgement. Furthermore, like all design/testing rules, the best way to learn when to break it is to follow it rigorously until it leads you to a bad place.

Problems
--------

- Stubbing external APIs gives you no design feedback: you cannot change the third party API (unless you monkey patch, but seriously, Fuck That Shit. Since stubbing/mocking are primarily a design tool (for me), having to stub/mock something I cannot/will not change seems pointless.
- Stubbing a third party API relies on you having intimate understanding of the docs, including error conditions, timeouts etc. I'm not confident in my ability to successfully stub a third party API on each conditions.

The Solution
------------

tcrayford

# with dependency injection, and slightly different route syntax:
module WebCrawler
  Routes = Raptor::App.new(self) do
    queue = #... some sort of persistent queue initialized here
    crawler = WebCrawler.new(queue)

    path "to-be-crawled" do
      create crawler, :enqueue!, :require => :logged_in
    end
  end

tcrayford

module Kernel
  def an(value)
    Thingy.new(value)
  end

  def none
    Nothing.new
  end
end