In dirkschumacher/listcomp: List Comprehensions

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

List comprehensions

The package implements list comprehensions as purely syntactic sugar with a minor runtime overhead. It constructs nested for-loops and executes the byte-compiled loops to collect the results.

Installation

remotes::install_github("dirkschumacher/listcomp")

install.packages("listcomp")

Example

This is a basic example which shows you how to solve a common problem:

library(listcomp)
head(gen_list(c(x, y), x = 1:100, y = 1:100, z = 1:100, x < 5, y < 5, z == x + y))

gen_list(c(x, y), x = 1:10, y = x:5, x < 2)

This is how the code looks like:

lst_verbose <- function(expr, ...) {
  deparse(listcomp:::translate(rlang::enquo(expr), rlang::enquos(...)))
}
lst_verbose(c(x, y), x = 1:10, y = x:5, x < 2)

You can also burn in external variables

z <- 10
gen_list(c(x, y), x = 1:!!z, y = x:5, x < 2)

It also supports parallel iteration by passing a list of named sequences

gen_list(c(i, j, k), list(i = 1:10, j = 1:10), k = 1:5, i < 3, k < 3)

The code then looks like this:

lst_verbose(c(i, j, k), list(i = 1:10, j = 1:10), k = 1:5, i < 3, k < 3)

It is quite fast, but the order of filter conditions also greatly determines the execution time. Sometimes, ahead of time compiling is slower than running it right away.

bench::mark(
  a = gen_list(c(x, y), x = 1:100, y = 1:100, z = 1:100, x < 5, y < 5, z == x + y),
  b = gen_list(c(x, y), x = 1:100, x < 5, y = 1:100, y < 5, z = 1:100, z == x + y),
  c = gen_list(c(x, y), x = 1:100, y = 1:100, z = 1:100, x < 5, y < 5, z == x + y, .compile = FALSE),
  d = gen_list(c(x, y), x = 1:100, x < 5, y = 1:100, y < 5, z = 1:100, z == x + y, .compile = FALSE)
)

How slow is it compared to a for loop and lapply for a very simple example?

bench::mark(
  a = gen_list(x * 2, x = 1:1000, x**2 < 100),
  b = gen_list(x * 2, x = 1:1000, x**2 < 100, .compile = FALSE),
  c = lapply(Filter(function(x) x**2 < 100, 1:1000), function(x) x * 2),
  d = {
    res <- list()
    for (x in 1:1000) {
      if (x**2 >= 100) next
      res[[length(res) + 1]] <- x * 2
    }
    res
  }, 
  time_unit = "ms"
)

Related packages

lc Uses a similar syntax as listcomp
comprehenr Uses a similar code generation approach as listcomp but with a different syntax.
listcompr Uses a similar syntax as listcomp and offers special generator functions for lists, vectors, data.frames and matrices.