refactor+docs: set_api_key

[dshemetov]

closes #181

Changes:

• get_api_key() only reads from environment variable now, no R options
• save_api_key() now instructs the user to write their key and dynamically opens ~/.Renviron or proj_root/.Renviron
• update documentation to reflect this

[capnrefsmmat]

Just a couple ideas

[dshemetov]

Might be good to go.

[brookslogan]

Looks good. Made a few suggestions. There are still some references to "set" in the text; some definitely should still be "set", but not sure if others should be "save" or something else; I didn't see any place where I was convinced just "save" would be better though.

[brookslogan]

Suggested change

	#' We recommend setting your key with `save_api_key()`, which will modify your
	#' `.Renviron` file and will persist your key across R sessions (see `?Startup`
	#' for a description of `.Renviron` files). Alternatively, you can modify the
	#' environment variable in the shell or with `Sys.setenv()`, but these will not
	#' persist across sessions.
	#' We recommend setting your key with `save_api_key()`, which will modify an
	#' applicable `.Renviron` file, which will be read in automatically when you
	#' start future R sessions (see [`?Startup`][base::Startup] for details on
	#' `.Renviron` files). Alternatively, you can modify the environment variable at
	#' the command line before/while launching R, or inside an R session with
	#' [`Sys.setenv()`], but these will not persist across sessions.

[brookslogan]

GitHub's just randomly not highlighting changes for me in some places. This reworded "your", "persist your key across", "for a description of", and "in the shell", plus added links to ?Startup and Sys.setenv().

[dshemetov]

Btw, ?Startup and Sys.setenv() is auto-linked by roxygen (see here for example).

[brookslogan]

One thing that would be nice, but I imagine might actually be stupidly tricky, is to first tell the user what file is being modified, before they press enter to confirm.

[EDIT: just saw the conversation about this above. I did not notice the usethis messages about the path before. Maybe okay but I still hope we can move to something that doesn't require this manual editing step eventually.]

[dshemetov]

Yea, that'll be hard. I think we found the minimal effort, maximal user gain solution, so we should probably just go with it.

I thought about manually appending to .Renviron with write(paste("DELPHI_EPIDATA_KEY=", key), environ_path, append=TRUE), but got stuck at trying to find the OS-independent user .Renviron path again. rappdirs would only point me to config or cache dirs, but not user environs. There's probably another package that we sifted through in the other thread that has something, but I don't think digging in on this is worth it compared to other things we could work on. Let's wait til we get feedback actual users before optimizing.

[brookslogan]

btw, this document command looks super handy with GitHub suggestions. Very cool. (And doesn't seem like there should be conflicts if you need to rebase local changes made in parallel to the doc commits.)

[dshemetov]

Right? It works well enough. (Wonder if we can have CI trigger from its commits too, because sometimes people push without documenting NAMESPACE changes, the CI breaks, but then the CI document pushes a change, but the status of the CI doesn't change to passing. But tuning GH CI often feels like another "pencil straightening" exercise to me.)