I am starting to implement testing for an R package, and have been using the testthat package. Note that I am new to testing, so perhaps my approach is off.
I have a function that currently fails the 16th time it is executed, and before fixing this, I want to write a regression test that will catch it if it re-appears.
e.g., the following always throws the same error message:
for i in (1:17) myfun()
myfun does not return anything, it only has a side-effect of opening a database connection. It is clear to me that I can write a test that expects an error and passes if it is returned:
expect_error(for (i in 1:17) myfun())
But I don't quite get how to write a test to ensure that the error does not occur. As it is not obvious, perhaps my approach is wrong. I can figure out how to write more specific tests, but I would like to start with this one.
What type of test would I write to make sure that such an error does not appear?
edit due to changes in testthat
Since version 3.2.0 (2023-10) there are functions to test lack of conditions (https://testthat.r-lib.org/reference/expect_no_error.html):
expect_no_error(my_fun())
expect_no_warning(my_fun())
expect_no_message(my_fun())
expect_no_condition(my_fun())
Minus: it does not support info argument.
Since version 0.11 (via RStudio blog) there is direct support to testing lack of errors:
expect_error(myfun(), NA)
Same for catching warning and message:
expect_warning(myfun(), NA)
expect_message(myfun(), NA)
Side note: There is an info parameter in expect_xxx functions to pass additional info. So you can do:
for (i in 1:17) expect_error(myfun(), NA, info = paste("i =", i))