Unexpected behavior of lmap_if when .f alters structure of .x and a function is supplied to .else

[TimTeaFan]

I was playing around with the lmap_if function and encountered an unexpected behaviour when a function is provided to .else while the function in .f is altering the structure of the object that lmap_if is applied to. This behavior might be intended, in this case please close this issue.

# setup
library(purrr)

Here is a simple function, similar to the one in the documentation. This function just duplicates
a list element.

list_rep <- function(x) {
  out <- rep_len(x, 2)
  names(out) <- paste0(names(x), 1:2)
  out
}

This is the example list from the documentation:

x <- list(a = 1:4, b = letters[5:7], c = 8:9, d = letters[10])

Lets apply lmap_if and say we want to duplicate each numeric list element and we want to
replace all other elements with list(z = 1).

lmap_if(x, 
        .p = is.numeric,
        .f = list_rep,
        .else = ~ list(z = 1))
#> $a1
#> [1] 1 2 3 4
#> 
#> $z
#> [1] 1
#> 
#> $b
#> [1] "e" "f" "g"
#> 
#> $z
#> [1] 1
#> 
#> $c2
#> [1] 8 9
#> 
#> $d
#> [1] "j"

The output is somewhat unexpected. Rewritting lmap_if as a loop - at least the way I thought it would work - would yield the following result.

out <- vector("list", length = length(x))

for (i in seq_along(x)) {
  out[[i]] <- if (is.numeric(x[[i]])) {
    list_rep(x[i]) 
  } else {
    list(z = 1)
  }
}
out <- flatten(out)
out
#> $a1
#> [1] 1 2 3 4
#> 
#> $a2
#> [1] 1 2 3 4
#> 
#> $z
#> [1] 1
#> 
#> $c1
#> [1] 8 9
#> 
#> $c2
#> [1] 8 9
#> 
#> $z
#> [1] 1

The reason for this unexpected behavior is that lmap_if creates an selector based on the function in .p and then applies lmap_at first on all selected elements (which might alter the structure of .x) and then applies lmap_at again at the result of the first operation using all elements which were not selected. However, since the selector was created before the first call to lmap_at it doesn’t necessarily match the structure of .x anymore. Is this behavior intended and was just my mental model of what lmap_if is doing wrong?

^{Created on 2021-12-12 by the reprex package (v0.3.0)}

[hadley]

Somewhat more minimal reprex:

library(purrr)

x <- list("a", 99)
str(lmap_if(x, is.character, ~ list(1, 2), .else = ~ list(3, 4)))
#> List of 4
#>  $ : num 1
#>  $ : num 3
#>  $ : num 4
#>  $ : num 99

^{Created on 2022-08-27 by the reprex package (v2.0.1)}