`build_rmd()` fails with some output format

[cderv]

devtools::build_rmd() is supposed to be a generic function to build any rmd

Currently it will error on some format like rmardown::pdf_document() (which could be used in vignettes too) because of

devtools/R/build-readme.R

Lines 33 to 34 in daca0cf

	# Ensure rendering github_document() doesn't generate HTML file
	output_options$html_preview <- FALSE

> devtools::build_rmd("vignettes/demo.Rmd")
ℹ Installing dummyPkg in temporary library
ℹ Building C:/Users/chris/Documents/dummyPkg/vignettes/demo.Rmd
Error in rmarkdown::pdf_document(..., template = template) : 
  argument inutilisé (html_preview = FALSE)
Error:
! ! in callr subprocess.
Caused by error in `rmarkdown::pdf_document(..., template = template)`:
! argument inutilisé (html_preview = FALSE)

This is because html_preview is not part of pdf_document() arguments, and pdf_document() does not have an ignore ... where those unused argument can go.

We could add this ... argument, like rmarkdown::html_document_base() has, but really this option html_preview is only to be used with github_document() IMO. So maybe setting this option was supposed to be part of build_readme() and not build_rmd() ?

[jennybc]

I'm not so sure that build_rmd() was intended to be a generic function to build any .Rmd. I think the origin story is more that we like having build_readme() and it made sense to implement that as a special case of something slightly more generic. I suspect that the way things are is what's best, for the devtools use cases, than trying to make build_rmd() more general.

That being said, if you want to make a PR with your proposed change, we can discuss further. How is this coming up for you?

[hadley]

Since we're moving to make devtools as minimal possible, I'd recommend we move towards making build_readme() internal — I created it in #2201 for reasons that are no longer obvious. I don't think we see folks building random Rmds routinely as part of standard development practice so it's reasonable to un-export (while leaving it available internally to power build_readme())

[jennybc]

@hadley I think you've got a typo 👆. Is the first build_readme() meant to be build_rmd()? The one you're proposing to mark as internal.

[jennybc]

I created it in #2201 for reasons that are no longer obvious

Maybe to preview vignettes or articles?

[hadley]

Hmmm, it doesn't predate pkgdown, but maybe I was running into some problem debugging a vignette failure? But I'm not sure this simulation would be close enough to what R CMD check does to be useful.