rlang: using function argument in default assignment function

Using rlang, I'd like to have a function that works both when directly called and when passed arguments as part of constructing another function argument by default, e.g.:

refdf = data.frame(x=1:100, y=runif(100,-1,1))

test.helper <- function(z, df) {
  qz <- enquo(z)
  range(eval_tidy(qz, df))
}
test.helper(y, refdf) # works

test.main <- function(z, df, def = test.helper(z, df)) {
  print(def)
}
test.main(y, refdf)
# doesn't work:  Error in eval_tidy(qz, df) : object 'y' not found

If instead, I do

refdf = data.frame(x=1:100, y=runif(100,-1,1))

test.helper <- function(z, df) {
  qz <- as_quosure(z)
  range(eval_tidy(qz, df))
}
test.helper(y, refdf)
# doesn't work: Error in is_quosure(x) : object 'y' not found

test.main <- function(z, df, def = test.helper(enquo(z), df)) {
  print(def)
}
test.main(y, refdf)
# now works

I feel like I'm missing something about what gets quoted when; is there an alternative syntax I can use to make both work? I know I could define a separate test.helper_quo or some such, but I'd really like to use the test.helper in the signature (as an extra hint to users about what functions are available).

Solution

This should work

library(rlang)

test.helper <- function(z, df) {
  qz <- enquo(z)
  range(eval_tidy(qz, df))
}
test.helper(y, refdf) # works

test.main <- function(z, df, def = test.helper(!!enquo(z), df)) {
  print(def)
}
test.main(y, refdf) # works

# or with rlang >= 0.4.0
test.main <- function(z, df, def = test.helper({{z}}, df)) {
  print(def)
}
test.main(y, refdf) # works

Note that in def, we need to capture the quosure passed as z and then expand that into the the call the test.helper so the it's own enquo will be able to see the original symbol.