Use stable version of {purrr}

[IndrajeetPatil]

Benchmarks:

• map_dfr() vs map() + list_rbind()

internal_map_dfr <- function(.x, .f, ...) {
  .f <- purrr::as_mapper(.f, ...)
  res <- purrr::map(.x, .f, ...)
  vctrs::vec_rbind(!!!res)
}

x <- list(
  a = data.frame(x = rnorm(1e6)),
  b = data.frame(y = rnorm(1e6))
)

bench::mark(
  internal_map_dfr(x, ~ .x),
  purrr::map(x, ~ .x) |> purrr::list_rbind()
)
#> # A tibble: 2 × 6
#>   expression                             min median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>                          <bch:> <bch:>     <dbl> <bch:byt>    <dbl>
#> 1 internal_map_dfr(x, ~.x)            4.93ms 5.08ms      168.    46.3MB     364.
#> 2 purrr::list_rbind(purrr::map(x, ~.… 4.81ms    5ms      199.    45.8MB     388.

^{Created on 2024-05-23 with reprex v2.1.0}

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 9e9d8ec is merged into main:

• ✔️cache_applying: 155ms -> 156ms [-1.65%, +1.95%]
• ✔️cache_recording: 523ms -> 525ms [-0.28%, +1.17%]
• ✔️without_cache: 993ms -> 994ms [-0.22%, +0.45%]

Further explanation regarding interpretation and methodology can be found in the documentation.

[IndrajeetPatil]

Not used anywhere.

[lorenzwalthert]

Good simplification, thanks. Not sure 1.0.2 is the first release that has this functionality, but since it's already a year old, I guess we can require it without causing people troubles...