Suggestion for a sort function

[jmoggridge]

In javascript, there is a built-in function Array.prototype.sort() that works on an array (like an R list) and takes a user-defined binary function to define ordering.

my_array.sort((a, b) => some_function(a, b))

Base R functions sort and order accept atomic vectors only. To sort lists, the user needs to use a combination of indexing with [ and order on some expression. For example, imagine we wanted to sort a list of vectors by length...

x <- list(letters[1:2], letters[1:4], letters, LETTERS) 
sorted_x <- my_list[order(sapply(my_list, length))]

The tidy approach could use dplyr::arrange on a tibble to sort the list-column.

sorted_x <- tibble(x = x) |> 
  mutate(length = sapply(x, length)) |> 
  arrange(length) |> 
  pull(x)

The proposed purrr function would sort a list based on a user-provided function to determine precedence.

# accept a binary function for complex comparisons
sorted_x <- sort_list(.x = x, .f = \(a,b) length(a) < length(b))

# accept a unary function that evaluates to a `sort`-able atomic vector
sorted_x <- sort_list(x, .f = \(a) length(a))

This is a reprex to illustrate a simple implementation and use case.

# very minimal implementation of proposed function using quicksort
sort_list <- function(.x, .f) {
  if (length(.x) <= 1)  return(.x)
  if (length(formals(.f)) == 1)  return(.x[order(sapply(.x, .f))])
  pivot <- .x[[1]]
  items <- .x[-1]
  has_precedence <- vapply(items, \(item) .f(item, pivot), logical(1L))  
  before <- sort_list(items[which(has_precedence)], .f)
  after <- sort_list(items[which(!has_precedence)], .f)
  c(before, list(pivot), after)
}

# eg, say need to sort a list of vectors by length
x <- list(letters, letters[1:2], letters[1:4], LETTERS) 

# use proposed function to sort list
x |> sort_list(\(a, b) length(a) < length(b))
x |> sort_list(\(a) length(a))

Thanks for your consideration! I'm a huge fan of purrr and tidyverse.

[py-b]

You can use this if you wish : https://py_b.gitlab.io/funprog/reference/sort_by.html
(a package of mine)