You don't inject for tests

http://blog.arkency.com/2013/01/you-dont-inject-for-tests/

by persei8 • 14 years ago

19 13 14 years ago

blog.arkency.com

fideloper• 14 years ago

I "discovered" DI when learning about Unit Testing, but the benefits to maintainability were obvious from the get-go.

I'm not sure many people argued that it wasn't about maintainability. I'd say they just ignored that part of the argument.

The best explanations of Unit Tests I've seen also explain how it helps improve maintainability, so perhaps that's why Unit Testing and Maintainability has always gone hand-in-hand in my thinking.

zby• 14 years ago

It's like removing magic constants from business logic into config files :)

andrzejkrzywda2• 14 years ago

Yeah, I was thinking about this recently. All of the "good patterns" turn your code into data. At the end it's more of declaring than coding the logic.

martinced• 14 years ago

No need to read such a post...

Title says: "You don't inject for tests"

And, before opening the link, you know what you're going to find.

Someone confusing "testing" and "unit testing".

First thing that came to my mind (while reading the header on HN) was "There are several kind of testing..." but... Curiosity killed the cat. Let's open the link.

Ouch, first sentence:

"What is unit testing for?"

OK, closed.

Don't waste your time reading this crap ; )

jordan0day• 14 years ago

I'm not sure I understand what you're so upset about?

"Don't waste your time reading this crap" because of some imprecise verbiage?

According to the typical "testing pyramid", unit testing should be the most common type of testing done -- is it really so unforgivable that someone would use a general term (test) for what they do most often (unit test)?

yashke• 14 years ago

I scoped my blog just to unit tests, because it's short note rather than article. DI can be useful on integration and acceptance level too - for simple apps you should be able to configure whole app or its parts as easy as you run it on production. Without DI you have to do some magic with class / module constants, which can populate on other parts.

wyuenho• 14 years ago

I think we are starting to argue about semantics here. This example is simply a very common way of coding to an interface, which BTW, duck-typing makes this so much easier.

I think through out the whole discussion about DI in the past week or so is that people don't talk about the injectors themselves enough. Too many pseudo-experts talking about inconsequential and totally unrelated programming principles and paradigms.

I don't think people naturally object to coding to the interface, which this example illustrates very well its utility. I think detractors simply just hate the injectors, the people who use the bad injectors indiscriminately, and the stubborn people who insist on doing DI on an entire app.

My guess is, most people automatically jump to a DI framework not really because of DI, you can do DI with just a good old factory and may be a good configuration library. Most people jump to DI because of a framework, and the niceties that said framework offers. After a while, people just start confusing the effect with the cause.

Can we please start talking about what is DI, when would one find it useful, and what DI approach is suitable under what circumstances now?

RyanZAG• 14 years ago

Can someone not work out a way to 'unit test' dependency injection vs other methods so we have some facts to compare? These discussions feel like people arguing with each other over which color is the best, with everyone chiming in with random anecdotes and feelings.

Red must be the best - it goes the fastest!

franzwong• 14 years ago

It would be quite hard to draw a straight line to separate goals of software creation and maintenance.

mimog2• 14 years ago

I test to verify that my code works.. Why would unit test show me where object orientation breaks down?

andrzejkrzywda2• 14 years ago

When things are hard to unit test, it might be a smell that there is something wrong with your OOP code.

mimog2• 14 years ago

But that's why you start by writing the tests and then evolve the test and the production code in parallel, so that you don't end up with code that is hard to test. I don't think that just because something is testable means that the design incorporates good OO practices.

btilly• 14 years ago

Let me turn this around.

In a dynamic language, unit testing is a bad reason to use dependency injection. There are lots of valid reasons to use it. But you don't need it for unit testing.

Returning to DHH's example, if you just want to get the current time in Ruby code, just get it. You can use the dynamic nature of the language to stub that out and unit test it. But you don't need to use dependency injection to enable that. And you really don't need to use a heavy framework.

If you have a different reason to use dependency injection, though, by all means go ahead.

You don't inject for tests

Related Stories