testthat can be used to add automated tests and checks to your work (in software development these are known as “unit tests”). The book R Packages has a chapter explaining the testing process and how to use testthat.
You can think of tests as a set of rules that your code must follow. Testing tools such as testthat run your code and notify you if the results break these rules.
We will set up some simple tests to check the pre-processing carried out in Exercise 2 and the discount() helper function created in Exercise 3.
Setting up a test
Test code is saved in the tests directory. Delete 1.R and add test.R with the following code
# Give tests in this file a name
context("Basic checks")
# Test data
test_that("Correct business lines are included", {
expect_equal(ncol(cashflows_to_use), 6)
})
# Test discount() helper function
test_that("discount() works correctly", {
# Discount 3 payments of 1,000 at 5% p.a.
expect_equal(discount(c(1000, 1000, 1000), 0.05, 1),
1000*(1.05^-0.5 + 1.05^-1.5 + 1.05^-2.5))
# Discount 3 payments of 1,000 (paid monthly) at 3% p.a.
expect_equal(discount(c(1000, 1000, 1000), 0.03, 12),
1000*(1.03^-(1/24) + 1.03^-(1/12 + 1/24) + 1.03^-(2/12 + 1/24)))
# Discount 10 payments of 1,000 at 0%
expect_equal(discount(rep(1000, 10), 0, 1),
10000)
})
The format for the tests above is expect_equal(A, B) i.e. the code A is expected to be equal to the value B when the test is run.
The first block of code checks that cashflows from all six business lines are included in the analysis (it checks the number of columns in the cashflows_to_use tibble/data frame equals 6).
The second tests that the output from the discount() helper function is as expected for some simple test inputs.
Running the tests
Run test.project() to carry out the tests and display a summary of the results in a table. testhat provides additional information if a test (or tests) fail.
> test.project()
Project name: cashflow_analysis
Loading project configuration
Autoloading packages
Loading package: reshape2
Loading package: plyr
Loading package: tidyverse
Loading package: stringr
Loading package: lubridate
Loading package: forecast
Autoloading helper functions
Running helper script: globals.R
Running helper script: helpers.R
Autoloading data
Munging data
Running preprocessing script: 01-A.R
√ | OK F W S | Context
√ | 4 | Basic checks
== Results =====================================================================
OK: 4
Failed: 0
Warnings: 0
Skipped: 0
Keep up the good work.
A message will be displayed if a test (or tests) fail.
You can display a full summary of the tests by running
> (test_dir("tests", reporter = "list"))
file context test nb failed skipped error warning user system real
1 test.R Basic checks Correct business lines are included 1 0 FALSE FALSE 0 0.02 0 0.02
2 test.R Basic checks discount() works correctly 3 0 FALSE FALSE 0 0.00 0 0.00
Continuous integration
The process of running tests can be automated using continuous integration tools such as Travis CI. These services build your project and run all tests when you “commit” your changes using a version control tool such as GitHub.
You can also use code coverage tools such as Codecov to monitor how much of your code is subject to automatic testing. These tools can highlight when parts of your code are not being checked, prompting you to write more unit tests.