Skip to main content

The best dependency is no dependency

Before we get into all of the techniques we can use to invert a dependency, let's look at the most powerful technique we have for dealing with unruly dependencies: get rid of them.

Take our previous example:

// api.js

import { findUserByEmail } from './database'

export const getUserName = async (email) => {
  const user = await findUserByEmail(email)

  if (!user) return "User Not Found"
  if (!user.name) return "No Name Provided"
  if (user.title) return `${user.title}. ${user.name}`

  return user.name
}

If we break out the logic we want to test into a separate function, we can then test that function without any "mocks" at all.

The result looks like this:

// api.js

import { findUserByEmail } from './database'

export const formatUserName = (user) => {
  if (!user) return "User Not Found"
  if (!user.name) return "No Name Provided"
  if (user.title) return `${user.title}. ${user.name}`

  return user.name
}

export const getUserName = async (email) => {
  const user = await findUserByEmail(email)

  // Test this function directly
  return formatUserName(user)
}

Now, we can test formatUserName directly by passing in different values like null or { name: 'Arthur'} and asserting about the result.

This is one of the most powerful tools in our software design toolkit, and should usually be the first one we reach for.

But, now we aren't testing everything!​

How do we test that getUserName correctly calls findUserByEmail!?​

You have a few options:

  • Live with it. In this case, the invocation of findUserByEmail is not the important logic, and is trivial enough that we can assume it is correct.
  • Combine this technique with one of the following techniques, and write tests that assert that the expected arguments are passed and that any exceptional cases are properly handled.