Make example runner less clever #191
Conversation
There was a problem hiding this comment.
I would use fs::dir_delete. I hate that unlink just fails silently.
It's a bit of a shame the r-lib packages don't handle this well. Might be worth raising the issue with pkgload/withr/devtools.
pkgload does a new.environment, and withr uses do.call to run on.exit in that environment. I guess no one exits that environment, and I don't even know if there's an API for doing that (didn't find anything in the rlang package). Maybe if pkgload had a do.call(return, list(), env=env) line it would work?
https://github.com/r-lib/pkgload/blob/96d1a746f12ebf0a54157ed2a256bdc32358f9f6/R/dev-example.R#L38
https://github.com/r-lib/withr/blob/82daa902b654edabdd523e09d9a3b1b7216f7cc3/R/defer.R#L108-L111
R/cleanup.R
Outdated
| @@ -42,6 +42,7 @@ | |||
| ##' | |||
| ##' @export | |||
| ##' @examples | |||
| ##' message(sprintf("Running in %s", format(environment()))) | |||
There was a problem hiding this comment.
indeed, trying to find out what withr/devtools were thinking they had as their environment - turns out it was exactly what I expected them to have, but it just did not work
Co-authored-by: Paul Liétar <[email protected]>
After a short fight, I can't get the withr deferred events to behave nicely under
devtools::run_examples, which is I presume a similar environment topkgdown- which leads to the silently failing examples on the docs. I've updated this to be very simple now and not try and clean up automatically at all, which should mean that the docs now run with no problems.