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_used()
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.
Install the released version from CRAN:
install.packages("ellipsis")
Or the development version from GitHub:
devtools::install_github("r-lib/ellipsis")
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: 3 components of `...` were not used.
#>
#> We detected these problematic arguments:
#> * `..1`
#> * `..2`
#> * `..3`
#>
#> Did you misspecify an argument?