How to deal with DOIs that include ">"

[maelle]

From https://stat.ethz.ch/pipermail/r-package-devel/2020q4/006236.html

If one has a DOI like <doi:10.1175/1520-0469(1981)038<1179:TSLROA>2.0.CO;2> in the package DESCRIPTION one needs to URLencode it for R CMD check to pass. I.e. "doi:10.1175/1520-0469(1981)038%3C1179:TSLROA%3E2.0.CO;2"

The problem is that the URL encoded version of the URL then gets written in the package-level manual page. And when going from Rd to HTML, the DOI is truncated after the "%".

Could roxygen2 transform DOIs to URLs and unencode their text?

[maelle]

Like what pkgdown does in some places actually https://github.com/r-lib/pkgdown/blob/810084e8a9abd091378a6b75fa9d7de9ce637aa1/R/html-build.R#L16 🤔 (the linkifying, but not the unescaping of the DOI, which is ok I suppose).

What one gets as index page with pkgdown in the absence of an index source is nice with the clickable DOIs.

[dieghernan]

Hi! I am exploring a PR for solving #1265 and I just came across this issue. Just for information, I found the following on https://github.com/wch/r-source/:

https://github.com/wch/r-source/blob/51cfc84111f51048bf9415f17a8910650492c6a2/src/library/tools/R/doitools.R#L32-L40

    dois <- character()
    pattern <- "<(DOI|doi):([^>]*)>"
    if(!is.na(v <- meta["Description"])) {
        m <- gregexpr(pattern, v)
        dois <- c(dois, .gregexec_at_pos(pattern, v, m, 3L))
    }
    ## DOI names may contain ">", but we need this as a delimiter when
    ## writing the names in <doi:name> style.  So at least ">" and hence
    ## also "%" must be percent encoded ...
    doi_db(utils::URLdecode(dois), rep.int("DESCRIPTION", length(dois)))