From: Avdi Grimm <avdi@...>
Date: 2008-04-10T06:51:41+09:00
Subject: Re: dependency injection with Needle gem

On Wed, Apr 9, 2008 at 4:44 PM, Daniel Berger <djberg96@gmail.com> wrote:
>  DI has no place in Ruby. So says Jamis Buck himself:
>
>  http://weblog.jamisbuck.org/2007/7/29/net-ssh-revisited
>
>  "I've since learned my lesson and have come to understand that
>  although DI is a nifty technique that works well in some environments,
>  it is wholly unnecessary in Ruby. Ruby sports an agility that Java
>  will never know, and can get by without such crutches."

There's a difference between dependency injection the pattern, and
dependency-injection tools/frameworks.

It's still a good idea to write classes that have their collaborators
passed in, rather than finding or creating them internally.

One of my favorite examples:

# 1. Not so great
class Whatsit
  def initialize
    @timestamp = Time.now
  end
end

# 2. better
class Whatsit
  def initialize(clock=Time)
    @timestamp = clock.now
  end
end

#2 is better because it's decoupled from the nondeterministic system
clock, and thus easier to test.

Sure, with ruby Mock Object libraries you can do things like:

  Time.stub!(:now).and_return("TIMESTAMP")

...but this runs the risk of breaking unrelated code that also happens
to be called in the test.   And no, that's not a theoretical risk;
I've had real tests fail because I needed to deterministically test
time-based code and some other code (like, say, RSpec itself) depended
on Time.now.

DI frameworks may be overkill in Ruby, but DI is still a good way of
breaking up your dependencies and keeping things decoupled.

-- 
Avdi

Home: http://avdi.org
Developer Blog: http://avdi.org/devblog/
Twitter: http://twitter.com/avdi
Journal: http://avdi.livejournal.com