Suggest a replacement in messages of `package_hooks_linter`?

[etiennebacher]

Some lintr messages show the potential replacement for a linter, e.g. in any_duplicated_linter:

library(lintr)
lint(
  text = "any(duplicated(x), na.rm = TRUE)",
  linters = any_duplicated_linter()
)
#> <text>:1:1: warning: [any_duplicated_linter] anyDuplicated(x, ...) > 0 is better than any(duplicated(x), ...).
#> any(duplicated(x), na.rm = TRUE)
#> ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

However, some messages are not very helpful because they don’t recommend any replacement. For example, package_hooks_linter says Don't use cat() in .onAttach().:

lint(
  text = ".onLoad <- function(lib, pkg) cat('hi')",
  linters = package_hooks_linter()
)
#> <text>:1:31: warning: [package_hooks_linter] Don't use cat() in .onLoad().
#> .onLoad <- function(lib, pkg) cat('hi')
#>                               ^~~

Is it possible to suggest the replacement for this linter? My reflex was to use message() instead but this is also flagged. Same with packageStartupMessage() (only flagged in .onLoad(), not in .onAttach()) so I don’t know what should be used here.

[MichaelChirico]

messages should be produced by .onAttach, not .onLoad

[etiennebacher]

Then I think messages from .onLoad should say something like:

Don't use cat() in .onLoad(). Use packageStartupMessage() in .onAttach() instead.

and messages from .onAttach() could say:

Don't use cat() in .onAttach(). Use packageStartupMessage() instead.

[AshesITR]

I suggest using a more active lint message for .onAttach:

Use packageStartupMessage() instead of cat() in .onAttach().