Should {downlit} be smarter about detecting the pkgdown website URL from DESCRIPTION?

[IndrajeetPatil]

Currently, it simply picks the first URL from all URLs specified in the DESCRIPTION file.

downlit/R/link.R

Line 343 in c0a8f64

urls[[1]]

But I guess not everyone is aware that, if there are multiple URLs specified, the first URL should be the pkgdown website.

For example:

downlit:::package_urls("styler")
#> [1] "https://github.com/r-lib/styler" "https://styler.r-lib.org"
downlit::href_package("styler")
#> [1] "https://github.com/r-lib/styler"

^{Created on 2024-04-22 with reprex v2.1.0}

So I am wondering if there could be a way to detect the pkgdown website URL, instead of just picking the first URL.

[olivroy]

I use a custom logic (based on experience mostly)

The best way is to check for the existence of {url}/pkgdown.yml

However, based on experience, I created these criteria to determine which URL seems to be the pkgdown site in my browse_pkg() function.

 # remove cran, github.com urls from potential pkgdown candidates
  pkgdown_candidates <-
    stringr::str_subset(
      urls,
      pattern = "github.com/.*/|cran\\.|contact.html",
      negate = TRUE
    )

  # If there are more than one link.
  if (length(pkgdown_candidates) > 1) {
    # using common pkgdown URLs pattern to identify it
    pkgdown <- grep("github.io|docs.ropensci|r.igraph", urls, value = TRUE)

    if (length(pkgdown) > 1) {
     # avoid linking papers, or other packages
     # some packages link more than one website, use the one that contains the package name
      pkgdown <- stringr::str_subset(pkgdown_candidate, package_name)
    } else {
      pkgdown <- pkgdown_candidate
 
   }
}

It is a bit ad-hoc and based on experience, but seems to be working well!

[hadley]

We could add something using remote_metadata() since that already figures out where an actual pkgdown site is.