ncss-tech/SoilKnowledgeBase

Repository v.s. Package Size -- new workflows to generate extdata artifacts

brownag opened this issue · 1 comments

inst/extdata is ignored by .Rbuildignore (i.e. it is not part of the package) -- but is downloaded when one does remotes::install_github -- as this will clone the whole repo. One would be better off doing git clone so at least they could have a repo instance with all the data. This can be slow, and with the recent addition (by the bot) of a bunch of the JSON output, that payload is a bit bigger.

It is pretty handy to be able to get the extdata reasonably quickly -- and not have to run refresh() on data that are not changing frequently. Therefore, I think a fetch_artifacts or similar could be used to obtain gzipped extdata subfolders from GitHub. Then we could .gitignore inst/extdata, perhaps, creating tools for local inspection.

Consider {targets} for Make-like workflows

  • Refactor create_* processes as sets of tar_target()
  • Archive targets for faster / less-redundant processing when inputs are the same as last run