Econometrics and Free Software by Bruno Rodrigues.
RSS feed for blog post updates.
Follow me on Mastodon, twitter, or check out my Github.
Check out my package that adds logging to R functions, {chronicler}.
Or read my free ebooks, to learn some R and build reproducible analytical pipelines..
You can also watch my youtube channel or find the slides to the talks I've given here.
Buy me a coffee, my kids don't let me sleep.

Unit testing with R

R

I've been introduced to unit testing while working with colleagues on quite a big project for which we use Python.

At first I was a bit skeptical about the need of writing unit tests, but now I must admit that I am seduced by the idea and by the huge time savings it allows. Naturally, I was wondering if the same could be achieved with R, and was quite happy to find out that it also possible to write unit tests in R using a package called testthat.

Unit tests (Not to be confused with unit root tests for time series) are small functions that test your code and help you make sure everything is alright. I'm going to show how the testthat packages works with a very trivial example, that might not do justice to the idea of unit testing. But you'll hopefully see why writing unit tests is not a waste of your time, especially if your project gets very complex (if you're writing a package for example).

First, you'll need to download and install testthat. Some dependencies will also be installed.

Now, you'll need a function to test. Let's suppose you've written a function that returns the nth Fibonacci number:

Fibonacci <- function(n){
    a <- 0
    b <- 1
    for (i in 1:n){
        temp <- b
        b <- a
        a <- a + temp
    }
    return(a)
}

You then save this function in a file, let's call it fibo.R. What you'll probably do once you've written this function, is to try it out:

Fibonacci(5)
## [1] 5

You'll see that the function returns the right result and continue programming. The idea behind unit testing is write a bunch of functions that you can run after you make changes to your code, just to check that everything is still running as it should.

Let's create a script called test_fibo.R and write the following code in it:

test_that("Test Fibo(15)",{
  phi <- (1 + sqrt(5))/2
  psi <- (1 - sqrt(5))/2
  expect_equal(Fibonacci(15), (phi**15 - psi**15)/sqrt(5))
})

The code above uses Binet's formula, a closed form formula that gives the nth Fibonacci number and compares it our implementation of the algorithm. If you didn't know about Binet's formula, you could simply compute some numbers by hand and compare them to what your function returns, for example. The function expect_equal is a function from the package testthat and does exactly what it tells. We expect the result of our implementation to be equal to the result of Binet's Formula. The file test_fibo.R can contain as many tests as you need. Also, the file that contains the tests must start with the string test, so that testthat knows with files it has to run.

Now, we're almost done, create yet another script, let's call it run_tests.R and write the following code in it:

library(testthat) 

source("path/to/fibo.R")

test_results <- test_dir("path/to/tests", reporter="summary")

After running these lines, and if everything goes well, you should see a message like this:

> library(testthat)
> source("path/to/fibo.R")
> test_results <- test_dir("path/to/tests", reporter="summary")

.
Your tests are dandy! 

Notice the small . over the message? This means that one test was run successfully. You'll get one dot per successful test. If you take a look at test_results you'll see this:

> test_results
         file context          test nb failed skipped error  user system  real
1 test_fibo.R         Test Fibo(15)  1      0   FALSE FALSE 0.004      0 0.006

You'll see each file and each function inside the files that were tested, and also whether the test was skipped, failed etc. This may seem overkill for such a simple function, but imagine that you write dozens of functions that get more and more complex over time. You might have to change a lot of lines because as time goes by you add new functionality, but don't want to break what was working. Running your unit tests each time you make changes can help you pinpoint regressions in your code. Unit tests can also help you start with your code. It can happen that sometimes you don't know exactly how to start; well you could start by writing a unit test that returns the result you want to have and then try to write the code to make that unit test pass. This is called test-driven development.

I hope that this post motivated you to write unit tests and make you a better R programmer!