ellipsis

Superseded by rlang. All the functionality in ellipsis is now provided by rlang.

Adding ... to a function is a powerful technique because it allows you to accept any number of additional arguments. Unfortunately it comes with a big downside: any misspelled or extraneous arguments will be silently ignored. This package provides tools for making ... safer:

• check_dots_evaluated() errors if any components of ... are not evaluated. This allows an S3 generic to state that it expects every input to be evaluated.

• check_dots_unnamed() errors if any components of ... are named. This allows you to collect arbitrary unnamed arguments, warning if the user misspells a named argument.

• check_dots_empty() errors if ... is used. This allows you to use ... to force the user to supply full argument names, while still warning if an argument name is misspelled.

Thanks to Jenny Bryan for the idea, and Lionel Henry for the heart of the implementation.

Installation

Install the released version from CRAN:

install.packages("ellipsis")

Or the development version from GitHub:

devtools::install_github("r-lib/ellipsis")

Example

mean() is a little dangerous because you might expect it to work like sum():

sum(1, 2, 3, 4)
#> [1] 10
mean(1, 2, 3, 4)
#> [1] 1

This silently returns the incorrect result because mean() has arguments x and .... The ... silently swallows up the additional arguments. We can use ellipsis::check_dots_used() to check that every input to ... is actually used:

safe_mean <- function(x, ..., trim = 0, na.rm = FALSE) {
  ellipsis::check_dots_used()
  mean(x, ..., trim = trim, na.rm = na.rm)
}

safe_mean(1, 2, 3, 4)
#> Error in `safe_mean()`:
#> ! Arguments in `...` must be used.
#> ✖ Problematic arguments:
#> • ..1 = 2
#> • ..2 = 3
#> • ..3 = 4
#> ℹ Did you misspell an argument name?