Is it possible to "map" over tar_combine()

[jdtrat]

Hi! Happy New Year, and thank you so much for such a wonderful ecosystem of packages. My question may be related to #35, but I was wondering if there was a good way to use tarchetypes::tar_combine() for subsets of static branching targets created with tarchetypes::tar_map().

In the example below, I am mapping over data (mtcars and iris), subsetting them, and would like to combine the results into one data frame at the end. Because they don't have the same columns, I can't call dplyr::bind_rows() on all of them. What I would like is to be able to combine the targets first_mtcars and last_mtcars as combined_data_mtcars; similarly, I want to combine first_iris and last_iris as combined_data_iris. Is this possible with static branching or would I need to explore dynamic branching here?

Thanks so much 😄

targets::tar_dir({
  targets::tar_script({
    mapped <- tarchetypes::tar_map(
      targets::tar_target(
        first,
        head(data)
      ),
      targets::tar_target(
        last,
        tail(data)
      ),
      values = tibble::tibble(
        data = rlang::syms(c("iris", "mtcars"))
        )
    )
    
    combined <- tarchetypes::tar_combine(
      combined_data,
      mapped,
      command = dplyr::bind_rows(!!!.x)
    )
    
    list(mapped, combined)
  })
  targets::tar_manifest()
})
#> # A tibble: 5 × 3
#>   name          command                                                  pattern
#>   <chr>         <chr>                                                    <chr>  
#> 1 last_mtcars   "tail(mtcars)"                                           <NA>   
#> 2 last_iris     "tail(iris)"                                             <NA>   
#> 3 first_iris    "head(iris)"                                             <NA>   
#> 4 first_mtcars  "head(mtcars)"                                           <NA>   
#> 5 combined_data "dplyr::bind_rows(first_iris = first_iris, first_mtcars… <NA>

^{Created on 2022-01-09 by the reprex package (v2.0.1)}

[wlandau]

Thanks, @jdtrat. In your case, I think it would be convenient to put the tar_combine() target inside tar_map(). Sketch:

library(targets)
  
tar_script({
  library(tarchetypes)
  library(tidyselect)
  
  # https://armchairecology.blog/iris-dataset/
  # https://allisonhorst.github.io/palmerpenguins/
  library(palmerpenguins)
  
  subsets <- list(
    tar_target(
      first,
      head(data)
    ),
    tar_target(
      last,
      tail(data)
    )
  )
  
  combined <- tar_combine(
    combined,
    subsets,
    command = dplyr::bind_rows(!!!.x)
  )
  
  tar_map(
    list(subsets, combined),
    values = tibble::tibble(
      data = rlang::syms(c("mtcars", "penguins"))
    ),
    unlist = TRUE # Flattened list of targets with informative names.
  )
})

tar_visnetwork()

^{Created on 2022-01-09 by the reprex package (v2.0.1)}

In case that doesn't fit the use case, here is an alternative. tar_map(unlist = TRUE) gives you a flattened list of mapped targets with informative names, and you can use the names to choose which targets to combine. tidyselect::eval_select() makes this convenient for large numbers of targets.

library(targets)
  
tar_script({
  library(tarchetypes)
  library(tidyselect)
  
  # https://armchairecology.blog/iris-dataset/
  # https://allisonhorst.github.io/palmerpenguins/
  library(palmerpenguins)
  
  mapped <- tar_map(
    tar_target(
      first,
      head(data)
    ),
    tar_target(
      last,
      tail(data)
    ),
    values = tibble::tibble(
      data = rlang::syms(c("mtcars", "penguins"))
    ),
    unlist = TRUE # Flattened list of targets with informative names.
  )
  
  combined_mtcars <- tar_combine(
    combined_mtcars,
    mapped[eval_select(ends_with("mtcars"), mapped)],
    command = dplyr::bind_rows(!!!.x)
  )
  
  combined_penguins <- tar_combine(
    combined_penguins,
    mapped[eval_select(ends_with("penguins"), mapped)],
    command = dplyr::bind_rows(!!!.x)
  )
  
  list(mapped, combined_mtcars, combined_penguins)
})

tar_visnetwork()
#> Warning message:
#> Targets and globals must have unique names. Ignoring global objects that conflict with target names: combined_mtcars, combined_penguins. Warnings like this one are important, but if you must suppress them, you can do so with Sys.setenv(TAR_WARN = "false").

^{Created on 2022-01-09 by the reprex package (v2.0.1)}

[jdtrat]

This is exactly what I was looking for -- thank you so much @wlandau! The eval_select() trick is super helpful!