check

{check} is yet another package to deal with assertions. It’s only a proof of concept.

{assertive}, {assertthat}, {ensurer}, {assertr}, {checkmate} … are more mature packages that will do the the same job.

Installation

remotes::install_gitub("moodyudskipper/check")

How it works

Your functions will look like this:

hello <- function(x) {
  check(
    "`x` must be a character vector of length 1."
    # we can add more
    )
  paste("hello", x)
}

hello("world")
#> [1] "hello world"

hello(c("it's", "me"))
#> Error: `x` must be a character vector of length 1.

• The error messages are the code, so it’s very readable.
• Assertion patterns are defined for the package and are reused in different functions, so it is both enforcing consistency and leaving complete flexibility to the users.

To do so :

• Paste check::setup() in an a R file of your package. It will create local copies of check::check, check::set_check_fun, and a check_funs environment

• Right below define assertions using set_check_fun(...)

For instance have a “check.R” file containing :

check::setup() # unnecessary if you want to test it outside of a package

# don't namespace this one! (no `::`)
set_check_fun(
  "`{var}` must be a {type} vector of length {length}." = {
      val <- get(var, env)
      is.atomic(val) && is(val, type) && length(val) == length
  }
  # we can add more
  )

If you forget what patterns to use, then type check_funs$ and auto-complete will show you what you’ve defined. If you misspell a pattern, check will suggest the closest fit to you.

Caveats

• A bit more setup than packages mentioned above. We could provide a check_funs environment with some patterns usable out of the box, for most common checks.
• Doesn’t currently tell you more about the problematic arguments (e.g. failing example above doesn’t tell you that it fails because x is of length 2)