R package: tests

Notes on tests, with the use of testthat package

Useful references:

• GSWEP4R: Ensuring Quality by Matt Secrest
• R packages (2e) Chapter 13, 14, 15 by Hadley Wickham and Jenny Bryan

Unit test: tests whether your function returns values as expected.

Benefits:

• Fewer bugs
• Better code structure, thhe functions need to be better designed

Example situations:

• missing values in the data;
• zero standard deviation (same values; or one value for the group)
• infinity

Use testthat in a package

1. Initialize

Create tests/testthat/ directory

usethis::use_testthat(3)

This creates the directory with

• an empty folder testthat where you write your tests
• an R script testthat.R where tests are run when R CMD check is run. Do not modify this file.

2. Create a test

Test files must have names that start with test. For example, a function is R/fn_name.R, then test is tests/testthat/test-fn_name.R.

usethis::use_test('testname')

3. Run a test

• testthat::test_file('tests/testthat/test-foofy.R')
• Run Tests button
• devtools::test() for entire test suite. Cmd + Shift + T
• devtools::check()

A workflow worked for me

1. Create a simple function
2. Create a test file immediately, with clear naming. Inside this test file, can write various tests for the same function.

• have at least a test that expects the correct result (expect_identical() or else)
• have at least a test that expects error, (expect_error()). Inside the original function, error should be thrown by rlang::abort().

3. Run test
4. Run test coverage, covr::package_coverage() or covr::code_coverage()
5. Check