Monday, January 7, 2013

Dependency Injection vs. Dependency Injection Framework

Yesterday, David Heinemeier Hansson wrote a blog: ‘Dependency injection is not a virtue’. There is a lot of things mixed together in this blog, all finally put together with the statement “I'm a Ruby programmer”. I have found this moves people into more of a sports team cheering mindset and does little to help keep conversation rationale and productive.

So let's take a step back, and start to unpack the meaning behind 'dependency injection'

In the article, DHH uses the ruby example of time, saying that

Time.stub(:now) { Time.new(2012, 12, 24) }
article.publish!
assert_equal 24, article.published_at.day

is preferable to

article.publish! Time.new(2012, 12, 24)
assert_equal 24, article.published_at.day

Let’s skip the opinion part of which is preferable, and start by pointing out that both of these actually are examples of dependency injections, albeit via different implementations [Monkey patch vs. parameter passing].

So I wanted to break down the first misconception my experience has shown to be common in the programming community; The confusion between Dependency Injection and Dependency Injection Frameworks.

disclaimer: as each framework is unique, this is a fairly blanket statement I am about to make. It may not apply to your framework.

Dependency Injection (Concept)

Dependency Injection (DI) is a general concept which addresses polymorphism. The idea is that you will have different implementations for things, and you need some way of switching which implementation you use. The programming world needed a label to refer to that concept and chose Dependency Injection. There are many many forms of Dependency Injection, and they offer different pro’s & con’s for different scenarios. Off the top of my head, I can think of quite a few methods for Dependency Injection:

  • parameters
  • inheritance
  • mocks
  • byte code manipulation
  • Global Variables
  • Factory Pattern
  • Callbacks
  • Inversion of control (IoC)
  • Monkey Patching
  • Dependency Injection Frameworks
  • Reflection

There are many more and as programming evolves many more will be created.

Dependency Injection Framework (Post Compiler DI)

Dependency Injection Frameworks, on the other hand, have seemed to evolved to support a specific type of scenario. Namely

“How do I inject a polymorphic instances after compilation time?”

You might ask, why would I want to do this? There are many answers, the best being Plugins (You do not want to have to recompile your web browser to allow for a plugin to work). There are, of course, many more reasons.

With non-compiled languages, it get’s even weirder to think of “compile time” but the concept is still valid, especially if you are not at the top of the runtime stack.

Dependency Injection Frameworks usually achieve their DI via a combination of factories, reflection and some sort of runtime configuration setup (files and metadata are common).

and they can be a bit of a heavy handed solution if you are using them for DI when you aren’t really concerned about what happens after compilation time.

Was DHH talking about Dependency Injection Frameworks?

It is worth noting the DHH never actually mentions dependency injection frameworks. Although, there is definitely a lot of mention of them is the conversations afterwards. There is a lot more in this article that is worth talking about, but let’s save that for a different blog…

No comments: