suggests: Declare when Suggested Packages are Needed

R-CMD-check Codecov test coverage

By adding dependencies to the “Suggests” field of a package’s DESCRIPTION file, and then declaring that they are needed within any dependent functionality, you can often significantly reduce the number of “hard” dependencies (i.e. Depends/Imports) required by your package.

{suggests} aims to make that workflow as minimal and painless as possible, primarily via the need() function, which provides a lightweight, simple, speedy way to prompt your users to install missing suggested packages.


You can install the development version of {suggests} from GitHub with:

# install.packages("remotes")

Or directly from R-universe:

install.packages("suggests", repos = "")


You can declare that one or more packages are needed for subsequent functionality with need():

read_data <- function(path, clean_names = FALSE) {

  # Call suggests::need() as early as possible, to avoid wasted work
  if (isTRUE(clean_names))

  output <- utils::read.csv(path)

  if (isTRUE(clean_names))
    output <- janitor::clean_names(output)


You can also make sure a minimum version is available by appending >=[version]:


Additionally, find_pkgs() is a quick-and-dirty diagnostic tool to find dependency usage within top-level expressions (e.g. declared functions) in R scripts within a development package. This can be useful when looking for good candidates for dependencies which could be moved from Imports to Suggests in the DESCRIPTION file.

Given a path, it returns a data frame with one row per distinct top-level expression where a package is used. Packages used in the fewest places are listed first.

find_deps(system.file("demopkg", package = "suggests"))
  package_used              in_file          in_expr
1        stats R/median_first_ten.R median_first_ten
2        tools R/median_first_ten.R median_first_ten
3        utils R/median_first_ten.R median_first_ten


The *_installed() functions from {rlang} provide similar functionality - more flexible, but less lightweight than this package.

You could avoid the need for any package-checking dependencies, if you’re willing to write slightly more code yourself! See the Dependencies: In Practice chapter of R Packages (Wickham & Bryan).