vec_depth fails if list has values of class "name"

[yogat3ch]

Hi Team purrr,
Grateful for the work ya'll have done on this package. I've used purrr - 974 times since I first became aware of it. Needless to say, it's an essential part of my R toolkit - so thank you!

I'm actually trying to come up with a solution for Advanced R 18.5.3 Q4 and have just used rapply to convert an expression to a series of lists with:
.calls <- rapply(as.list(x), as.list, classes = "call", how = "list")
I would like to map over the resulting list using purrr with something like the following:
purrr::map_depth(.calls, purrr::vec_depth(.calls), call_match_lgl, pattern = "fn")

But I'm finding that an element of class name apparently causes vec_depth to error.

Here's a simple reprex to illustrate:

purrr::vec_depth(as.list(rlang::call2(mean, 1:10, na.rm = TRUE)))

It seems like modifying it like this might make it return results match with what one would expect when visualizing with lobstr::ast:

vec_depth <- function(x) {
  if (is_null(x)) {
    0L
  } else if (is_atomic(x)) {
    1L
  } else if (rlang::is_symbol(x) || rlang::is_function(x)) {
    1L
  } else if (is_list(x) || rlang::is_call(x)) {
    depths <- map_int(as.list(x), vec_depth)
    1L + max(depths, 0L)
  } else {
    abort("`x` must be a vector")
  }
}

There might be some exceptions where this would have unexpected results though.
Thoughts?

[hadley]

Somewhat more minimal reprex:

purrr::vec_depth(list(x = quote(x)))
#> Error in `.f()` at purrr/R/map.R:204:2:
#> ! `x` must be a vector

^{Created on 2022-08-24 by the reprex package (v2.0.1)}

It feels like the final else should just return 0. I don't think this function needs to handle all the other types of object that you might index into with [ or [[. (So the documentation should also be updated)

[hadley]

Maybe this should be pluck_depth()?

[yogat3ch]

Maybe this should be pluck_depth()?

If making a function to do the task as described in the example, that would certainly make sense as a name!

[yogat3ch]

Checked out the changes and they look solid! Thanks @hadley & @lionel- 🙌