Preferred approach for making the `{targets}` R package detect changes to functions and exported datasets from my own developed R data package

Question

I am a beginner with the {targets} package, and I was wondering what is the right setup to register dependencies to functions (and datasets) developed by myself in an R data package.

My idea is to use {targets} to develop the somewhat involved workflow of generating several exported datasets, and files on disk, for this hypothetical R data package of mine: {MyRDataPackage}. And I would like those functions that generate these datasets/files to data-raw/ to be exported functions from the package itself, i.e. I would rather not have them sourced (as in source("R/functions.R")) in _targets.R.

By reading from Chapter 6.3 Dependencies , I got the feeling I could take this approach:

# _targets.R
tar_option_set(envir = getNamespace("MyRDataPackage"))

but reading a bit further, namely in Chapter 6.5 Packages-based invalidation, it seems I could also pass my {MyRDataPackage} to the imports argument:

# _targets.R
tar_option_set(
  packages = c("MyRDataPackage"),
  imports = c("MyRDataPackage")
)

So my question is: is either approach is fine? Or, are there reasons to prefer one over the other?

score 1 · Accepted Answer · answered Aug 06 '21 at 17:16

The guidance in Section 6.5 is the current recommendation. 6.3 is outdated but was just updated in https://github.com/ropensci-books/targets/commit/a9661e642beb174383222af16c1a599ae10a4735. Also answered at https://github.com/ropensci/targets/discussions/586\#discussioncomment-1140345.

Preferred approach for making the `{targets}` R package detect changes to functions and exported datasets from my own developed R data package

1 Answers1