FR: add env argument to is_missing()

[DanChaltiel]

Hi,

In some cases, it can be interesting to know if an argument is missing from the caller frame.

For instance, I'd like to propose a standard way to handle arguments through options when they are missing. Here, being able to check the missingness from inside a helper would be very handy.

Here is a reprex with a clumsy workaround:

library(rlang)

is_miss_bad = function(x, env=parent.frame()){
  is_missing(x)
}
is_miss_good = function(x, env=parent.frame()){
  xname = caller_arg(x)
  eval(parse(text=paste0("missing(",xname,")")), envir=env)
}

f = function(a=2){
  cat("bad:", is_miss_bad(a), "\n")
  cat("good:", is_miss_good(a))
  invisible()
}

f()
#> bad: FALSE 
#> good: TRUE

^{Created on 2024-05-30 with reprex v2.1.0}

Note that is_miss_bad() would actually work fine if a didn't have a default value.

Would you consider adding this feature to rlang::is_missing()?

[lionel-]

I think I'd rather fix either is_missing() or missing() in base. This env argument doesn't seem ergonomic.

[DanChaltiel]

I edited my post to show the proper FR demand.
However, I am not sure to understand, what do you mean by "fix in base"?
Also, what would you suggest instead of env?
If you don't think this would be a nice improvement to rlang::is_missing(), would you have a suggestion on a better way to do that than eval(parse(...))?

[lionel-]

The C API of R is about to gain an env_binding_type() function that will return an enum, with the missing argument being one of the possible kinds (see https://gist.github.com/lionel-/1ebcbd5ec69c0775d514c329522408a3 for more information).

So I think we could expose this on the R level, and you could use that to detect missing arguments in a caller frame. We could add individual variants too, such as:

env_has_missing <- function(env, nm) {
  env_binding_type(env, nm) == "missing"
}

But I'm not sure we want to add so many predicates.