Use of target names and target object names

[ghost]

Hello,

I have just recently switched from drake to targets and I'm finding the use of target objects in the global environment alongside the names argument inside the target() function confusing.

I'm trying to use the tar_plan() to try to overcome this, If I don't assign my tar_map() to an object called 'ecafs' and instead try to combine 'ecaf' as defined in the tar_target() nested inside the tar_map() it doesn't know it exists. Any advice would be much appreciated.

# correct code ---------------------------------------------
tar_plan(
    ecafs <- tar_map(
        values = data.table(
            as_at = as.character(quarters$as_at),
            start = as.character(quarters$start),
            end = as.character(quarters$end),
            year = as.character(quarters$year),
            quarter = as.character(quarters$quarter),
            names_ = paste(quarters$year, quarters$quarter, sep = '_')
        ),
        names = 'names_',
        tar_target(
            ecaf,
            ecaf_fn(
                as_at_ = as_at,
                start_ = start,
                end_ = end,
                year_ = year,
                quarter_ = quarter
            )
        ),
        unlist = F
    ),
    
    tar_combine(ecaf_combined,
                                 ecafs[[1]],
                                 command = bind_rows(!!!.x))
)

# incorrect code ---------------------------------------------
tar_plan(
    tar_map(
        values = data.table(
            as_at = as.character(quarters$as_at),
            start = as.character(quarters$start),
            end = as.character(quarters$end),
            year = as.character(quarters$year),
            quarter = as.character(quarters$quarter),
            names_ = paste(quarters$year, quarters$quarter, sep = '_')
        ),
        names = 'names_',
        tar_target(
            ecaf,
            ecaf_fn(
                as_at_ = as_at,
                start_ = start,
                end_ = end,
                year_ = year,
                quarter_ = quarter
            )
        ),
        unlist = F
    ),

    tar_combine(ecaf_combined,
                                 ecaf,
                                 command = bind_rows(!!!.x))
)

[wlandau]

Yes, static branching with target objects is an adjustment. But unfortunately, the functionality in tar_plan() is too limited to help. All tar_plan() does is make certain kinds of simpler pipelines shorter, e.g. tar_plan(target1 = command1(), target2 = command2()). I recommend something like this for _targets.R:

# _targets.R
library(targets)
library(tarchetypes)
ecafs <- tar_map(
  values = data.table(
    as_at = as.character(quarters$as_at),
    start = as.character(quarters$start),
    end = as.character(quarters$end),
    year = as.character(quarters$year),
    quarter = as.character(quarters$quarter),
    names_ = paste(quarters$year, quarters$quarter, sep = '_')
  ),
  names = 'names_',
  tar_target(
    ecaf,
    ecaf_fn(
      as_at_ = as_at,
      start_ = start,
      end_ = end,
      year_ = year,
      quarter_ = quarter
    )
  )
)
combined <- tar_combine(ecaf_combined, ecafs)
list(ecafs, combined)

Correct use with tar_plan() will look almost the same.

# _targets.R
library(targets)
library(tarchetypes)
ecafs <- tar_map(
  values = data.table(
    as_at = as.character(quarters$as_at),
    start = as.character(quarters$start),
    end = as.character(quarters$end),
    year = as.character(quarters$year),
    quarter = as.character(quarters$quarter),
    names_ = paste(quarters$year, quarters$quarter, sep = '_')
  ),
  names = 'names_',
  tar_target(
    ecaf,
    ecaf_fn(
      as_at_ = as_at,
      start_ = start,
      end_ = end,
      year_ = year,
      quarter_ = quarter
    )
  )
)
combined <- tar_combine(ecaf_combined, ecafs)
tar_plan(ecafs, combined)