What is the fastest way to read the header of a delim?

colnames_delim.R

readr_lazy_colnames <- function(path, delim = "|") {
  readr::read_delim(
    path,
    delim = delim,
    lazy = TRUE,
    progress = FALSE,
    show_col_types = FALSE
  ) |>
    colnames()
}

arrow_colnames <- function(path, delim = "|") {
  arrow::open_delim_dataset(path, delim = delim) |>
    colnames()
}

arrow_schema_colnames <- function(path, delim = "|") {
  arrow::open_delim_dataset(path, delim = delim)$schema |>
    names()
}

arrow_nodf_colnames <- function(path, delim = "|") {
  arrow::read_delim_arrow(path, delim = delim, as_data_frame = FALSE) |>
    colnames()
}

readr_nmax_colnames <- function(path, delim = "|") {
  readr::read_delim(
    path,
    delim = delim,
    n_max = 0,
    progress = FALSE,
    show_col_types = FALSE
  ) |>
    colnames()
}

readlines_colnames <- function(path, delim = "|") {
  strsplit(readLines(path, n = 1), split = delim, fixed = TRUE)[[1]]
}

vroom_colnames <- function(path, delim = "|") {
  vroom::vroom(
    path,
    delim = delim,
    n_max = 0 ,
    show_col_types = FALSE,
    progress = FALSE
  ) |>
    colnames()
}

dt_colnames <- function(path, delim = "|") {
  data.table::fread(
    nrow = 0,
    sep = delim,
    header = TRUE,
    cmd = sprintf("head -n 1 %s", path)
  ) |>
    colnames()
}

# Big csv file (pipe separated in my case, not provided! 
big_csv <- "~/Downloads/wcvp_taxon.csv"

bench::mark(
  readr_lazy_colnames(big_csv),
  readr_nmax_colnames(big_csv),
  arrow_colnames(big_csv),
  arrow_schema_colnames(big_csv),
  arrow_nodf_colnames(big_csv),
  readlines_colnames(big_csv),
  vroom_colnames(big_csv),
  dt_colnames(big_csv)
)

expression             min   median `itr/sec` mem_alloc `gc/sec` n_itr  n_gc total_time result
  <bch:expr>        <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl> <int> <dbl>   <bch:tm> <list>
1 readr_lazy_colna…  208.3ms 214.54ms      4.58    6.01MB     0        3     0      655ms <chr> 
2 readr_nmax_colna… 531.01µs 600.75µs   1505.      6.31KB     9.13   659     4      438ms <chr> 
3 arrow_colnames(b…   5.93ms   7.56ms    123.     22.92MB     2.05    60     1      488ms <chr> 
4 arrow_schema_col…   6.23ms   7.36ms    136.     11.66KB     4.23    64     2      472ms <chr> 
5 arrow_nodf_colna… 245.64ms 248.67ms      4.01    3.79MB     0        3     0      749ms <chr> 
6 readlines_colnam…  11.73µs  12.54µs  64088.        520B     6.41  9999     1      156ms <chr> 
7 vroom_colnames(b… 506.75µs 564.36µs   1581.    466.03KB     9.11   694     4      439ms <chr> 
8 dt_colnames(big_…  868.4µs 954.75µs    987.      1.65MB     2.04   483     1      489ms <chr> 

Using readlines is by far the fastest way to do this