drake static branching: How to use .id within map() to increase visibility of dependency graph

Question

I am using a drake workflow to process ~100 files which are stored in a location with very long file names. These long file names make the dependency graph hard to read. Here is a minimal example:

# example setup
library(drake)
very_long_path <- "this_is_a_very_long_file_path_which_makes_the_dependency_graph_hard_to_read"
dir.create(very_long_path)
filenames <- paste0("file_", seq(4), ".csv")
for (file in filenames) {
    file.create(file.path(very_long_path, file))
}
files <- list.files(very_long_path, full.names = TRUE)
ids <- rlang::syms(filenames)

# my drake plan
plan <- drake_plan(
    raw = target(
        read.csv(file_in(!!file)),
        transform = map(file = !!files)
    )
)
plan

## A tibble: 4 x 2
#  target                                           command                                              
#  <chr>                                            <expr>                                               
#1 raw_this_is_a_very_long_file_path_which_makes_t~ readLines(file_in("this_is_a_very_long_file_path_whic~
#2 raw_this_is_a_very_long_file_path_which_makes_t~ readLines(file_in("this_is_a_very_long_file_path_whic~
#3 raw_this_is_a_very_long_file_path_which_makes_t~ readLines(file_in("this_is_a_very_long_file_path_whic~
#4 raw_this_is_a_very_long_file_path_which_makes_t~ readLines(file_in("this_is_a_very_long_file_path_whic~

vis_drake_graph(drake_config(plan)) ## very hard to read

I've read the following about .id in ?transformations:

Symbol or vector of symbols naming grouping variables to incorporate into target names. Useful for creating short target names. Set .id = FALSE to use integer indices as target name suffixes.

That's why I created ids in the code above in order to provide short names for the targets. But changing the plan as follows did not help:

plan <- drake_plan(
    raw = target(
        readLines(file_in(!!file)),
        transform = map(file = !!files,
                        .id = !!ids)
    )
)
plan

## A tibble: 4 x 2
#  target                                           command                                              
#  <chr>                                            <expr>                                               
#1 raw_this_is_a_very_long_file_path_which_makes_t~ readLines(file_in("this_is_a_very_long_file_path_whic~
#2 raw_this_is_a_very_long_file_path_which_makes_t~ readLines(file_in("this_is_a_very_long_file_path_whic~
#3 raw_this_is_a_very_long_file_path_which_makes_t~ readLines(file_in("this_is_a_very_long_file_path_whic~
#4 raw_this_is_a_very_long_file_path_which_makes_t~ readLines(file_in("this_is_a_very_long_file_path_whic~

From my understanding, ids is a vector of symbols, so I do not understand why this is not working. What am I missing? Is that even possible?

I also tried to insert ids as a character vector, without success. I know that I can set .id = FALSE to simply enumerate the elements of raw, but I really want to keep the file names.

score 2 · Accepted Answer · answered Jan 21 '20 at 13:43

You are very close. All you need to do is register ids as a grouping variable and then pass the grouping variable symbol to .id.

library(drake)
very_long_path <- "this_is_a_very_long_file_path_which_makes_the_dependency_graph_hard_to_read"
dir.create(very_long_path)

filenames <- paste0("file_", seq(4), ".csv")

for (file in filenames) {
  file.create(file.path(very_long_path, file))
}

files <- list.files(very_long_path, full.names = TRUE)
ids <- rlang::syms(filenames)

plan <- drake_plan(
  raw = target(
    read.csv(file_in(!!file)),
    transform = map(
      file = !!files,
      id_var = !!ids, # Register the grouping variable.
      .id = id_var    # Use the existing grouping variable.
    )
  )
)

plan
#> # A tibble: 4 x 2
#>   target        command                                                         
#>   <chr>         <expr>                                                          
#> 1 raw_file_1.c… read.csv(file_in("this_is_a_very_long_file_path_which_makes_the…
#> 2 raw_file_2.c… read.csv(file_in("this_is_a_very_long_file_path_which_makes_the…
#> 3 raw_file_3.c… read.csv(file_in("this_is_a_very_long_file_path_which_makes_the…
#> 4 raw_file_4.c… read.csv(file_in("this_is_a_very_long_file_path_which_makes_the…

plan$target
#> [1] "raw_file_1.csv" "raw_file_2.csv" "raw_file_3.csv" "raw_file_4.csv"

^{Created on 2020-01-21 by the reprex package (v0.3.0)}

After applying this, I observed the weird behavior that those targets always remain outdated (after successful `make()`). I tracked it down to symbols starting with digits (in the case above `ids` were like 20190131). Is this a bug or something I do not (yet) understand about symbols? — der_grund, Jan 23 '20 at 08:42
Odd. Are the IDs different in the plan each time you create the `ids` vector and call `drake_plan()`? Regardless, if it continues to be a problem, you can create another [`reprex`](https://github.com/tidyverse/reprex) and post an [issue](https://github.com/ropensci/drake/issues). — landau, Jan 23 '20 at 18:13

drake static branching: How to use .id within map() to increase visibility of dependency graph

1 Answers1