Writing tests for R packages without testthat and RUnit

As a beginner R package author who wanted to include tests in his package, I turned to Google. If you do the same, you will find multiple tutorials for the testthat package and maybe even a few for the RUnit package. If you look very closely, you will notice there is also a testit package. If you are reading this, that probably means you, like me, found testthat annoying and RUnit hard to decipher. So, why are these packages so popular? It's a positive feedback loop. testthat will be popular by default since it was written by Hadley Wickham. Quickly, testthat comes to dominate search results which makes new users think that this is the only method which further encourages future package authors to use testthat.

But guess what, you don't need to use those packages! The testit package looks nice and simple. If I were going to use a package, I'd go with testit, but for this blog, I'd like to show you how to do testing without any specific package. It's actually extremely simple. Let's do some examples.

For our first example, we just want to make sure some R code will run successfully. Perhaps you had an issue in a previous version where some R code crashed or returned an error which you didn't want. To perform this test, just create an R file in the tests/ folder of your package. You can name it whatever you want. The contents should just be to load your package and then run whatever R code you want:

library(MyPackage)
# Make sure x = 12 works for function MyFunction
MyFunction(12)

That's all it takes.

OK. Now maybe MyFunction(12) ought to return a specific value. Well, we can pretty simply just adapt the code to call an error if the right value doesn't come up:

library(MyPackage)
# Make sure x = 12 works for function MyFunction and gives correct output
a <- MyFunction(12)

if (a != 13)
{
    stop("MyFunction(12) doesn't equal 13")
}

Great! Now what if your function returns a big pile of output or a list and you aren't interested in capturing that output and writing the appropriate if statement to check? Well, if you are certain that the function is currently working, then just change your test file to something like this:

library(MyPackage)
# Make sure x = 12 works for function MyFunction and gives correct output
MyFunction(12)

Now create an Rout file: R CMD BATCH my_test_file.R. This creates the file my_test_file.Rout. It's probably a good idea to look over that Rout file just to make sure things look as they should. Now, rename that Rout file as my_test_file.Rout.save and make sure it is in the tests/ folder. To be clear, the base name should be the same as the .R file, and the extension should be .Rout.save. Now, when you run your tests, the output of my_test_file.R will be compared to the already existing my_test_file.Rout.save.

Pretty simple, eh?