How to use purrr:map() and rlang to emulate a pipe chain

Question

There are several packages such as leaflet or magick that take special objects (maps or images, respectively), and allow them to be modified/added to using a pipe chain.

I would like to get the same behavior using a list of tibbles with the function arguments, but I'm struggling with how to do it since the output from purrr::map() is a list (or dfr, etc, but not a leaflet map or image).

(Note that I found a way to do it in magick after a bit, but it only works in magick because of using some special functions in a way they weren't meant to be used, so I"m still looking for a general answer to this question for other packages like leaflet)

Reprex:

suppressPackageStartupMessages(require(tidyverse))
suppressPackageStartupMessages(require(magick))
suppressPackageStartupMessages(require(rlang))


x <- image_blank(100, 100, "yellow")

blackbox <- image_blank(10, 10, "black")
redbox <- image_blank(10, 10, "red")

locations <- tribble(~box, ~offset,
                      "redbox", "+0+0",
                      "redbox", "+90+0",
                      "redbox", "+0+90",
                      "redbox", "+90+90",
                      "blackbox", "+40+40",
                      "blackbox", "+50+50",
                      "blackbox", "+30+60",
                      "blackbox", "+60+30")

#preferred method:
locations %>% 
  mutate(row = row_number()) %>% 
  split(.$row) %>% 
  map(function(location) {
    x <- x %>% image_composite(eval_tidy(sym(location$box)), offset = location$offset, operator = "over")
  })
#> $`1`
#> # A tibble: 1 x 7
#>   format width height colorspace matte filesize density
#>   <chr>  <int>  <int> <chr>      <lgl>    <int> <chr>  
#> 1 png      100    100 sRGB       FALSE        0 72x72  
#> 
#> $`2`
#> # A tibble: 1 x 7
#>   format width height colorspace matte filesize density
#>   <chr>  <int>  <int> <chr>      <lgl>    <int> <chr>  
#> 1 png      100    100 sRGB       FALSE        0 72x72  
#> 
#> $`3`
#> # A tibble: 1 x 7
#>   format width height colorspace matte filesize density
#>   <chr>  <int>  <int> <chr>      <lgl>    <int> <chr>  
#> 1 png      100    100 sRGB       FALSE        0 72x72  
#> 
#> $`4`
#> # A tibble: 1 x 7
#>   format width height colorspace matte filesize density
#>   <chr>  <int>  <int> <chr>      <lgl>    <int> <chr>  
#> 1 png      100    100 sRGB       FALSE        0 72x72  
#> 
#> $`5`
#> # A tibble: 1 x 7
#>   format width height colorspace matte filesize density
#>   <chr>  <int>  <int> <chr>      <lgl>    <int> <chr>  
#> 1 png      100    100 sRGB       FALSE        0 72x72  
#> 
#> $`6`
#> # A tibble: 1 x 7
#>   format width height colorspace matte filesize density
#>   <chr>  <int>  <int> <chr>      <lgl>    <int> <chr>  
#> 1 png      100    100 sRGB       FALSE        0 72x72  
#> 
#> $`7`
#> # A tibble: 1 x 7
#>   format width height colorspace matte filesize density
#>   <chr>  <int>  <int> <chr>      <lgl>    <int> <chr>  
#> 1 png      100    100 sRGB       FALSE        0 72x72  
#> 
#> $`8`
#> # A tibble: 1 x 7
#>   format width height colorspace matte filesize density
#>   <chr>  <int>  <int> <chr>      <lgl>    <int> <chr>  
#> 1 png      100    100 sRGB       FALSE        0 72x72




#desired result I'm trying to emulate:
x %>% 
  image_composite(redbox, offset = "+0+0", operator = "over") %>% 
  image_composite(redbox, offset = "+90+0", operator = "over") %>% 
  image_composite(redbox, offset = "+0+90", operator = "over") %>% 
  image_composite(redbox, offset = "+90+90", operator = "over") %>% 
  image_composite(blackbox, offset = "+40+40", operator = "over") %>% 
  image_composite(blackbox, offset = "+50+50", operator = "over") %>% 
  image_composite(blackbox, offset = "+30+60", operator = "over") %>% 
  image_composite(blackbox, offset = "+60+30", operator = "over")

#as an aside, i figured out to do it part of the way with magick, but does not apply to the general question and doesn't fully emulate the above.
y <- image_blank(100, 100, "transparent")

locations %>% 
  mutate(row = row_number()) %>% 
  split(.$row) %>% 
  map(function(location) {
    y %>% image_composite(eval_tidy(sym(location$box)), offset = location$offset, operator = "over")
  }) %>% 
  reduce(c) %>% 
  image_flatten() %>% 
  image_background(color = "yellow") #also having to flatten an animation loses the transparency, so you can't add it on top of the background.

^{Created on 2019-01-27 by the reprex package (v0.2.1)}

score 2 · Accepted Answer · answered Apr 01 '19 at 22:10

2

You could use purrr::reduce2 :

library(purrr)
reduce2(locations$box, locations$offset,
    ~image_composite(., get(..2), offset = ..3, operator = "over"), .init = x)

answered Apr 01 '19 at 22:10

moodymudskipper

46,417
11
121
167

1

Ah yes, my solution basically reinvented the `reduce` operation using `partial`, `compose` and duct tape. >. – Artem Sokolov May 22 '19 at 17:26

Artem Sokolov · Answer 2 · 2019-04-01T15:21:00.643

You can use purrr::partial to generate a list of functions with pre-specified values for offset, operator, etc. These functions can then be aggregated into a single operation via purrr::compose:

funs <- map2( locations$box, locations$offset,
         ~partial(image_composite, composite_image=get(.x), 
                  offset=.y, operator="over") )

## Equivalent to x %>% funs[[1]]() %>% funs[[2]]() %>% ...
compose(!!!funs)(x)

How to use purrr:map() and rlang to emulate a pipe chain

2 Answers2