build_lesson.R

library(tic)

BuildLesson <- R6::R6Class(
  "BuildLesson", inherit = TicStep,
  public = list(
    run = function() {
      build_status <- system("make clean-rmd; make lesson-md")
      if (build_status > 0)
        stop("Error during building process")

      file.remove(".gitignore")
      system("mkdir  _rendered")
      system("cp -r \`ls -A | grep -v '.git' | grep -v '_rendered' | grep -v '_site'\` _rendered")
    })
)


build_lesson <- function() {
    BuildLesson$new()
}


format_url_errors <- function(res_links, idx) {
    paste0("  - broken url:", res_links$urlname[idx], "\n",
           "          from:", res_links$parentname[idx], "\n")
}

CheckLinks <- R6::R6Class(
     "CheckLinks", inherit = TicStep,
     public = list(
       run = function() {
         ## While build_lesson() merely copies the content of the repo into a
         ## new folder that GitHub picks up to render (so the dynamically
         ## generated links such as "Edit on GitHub" are functional), here we
         ## actually need to generate the website so we can test the links.
         on.exit(system("rm -rf _rendered/_site"))

         system("cd _rendered && rvm 2.5.3 do ruby -S jekyll build")

         ## ignore JS file not included as part of rmarkdown
         ## ignore email addresses
         ## ignore embedded images
         link_status <- system("linkchecker --ignore-url=external.+js --ignore-url=^mailto: --ignore-url=^data: --no-warnings  --file-output=csv/link_res.csv _rendered/_site")
         message("linkchecker exit code: ", link_status)

         ## Because URLs can contain #, we first need to remove the commented
         ## lines in the generated CSV file. We can't rely on the comment
         ## argument of the read_delim function, see
         ## https://github.com/tidyverse/readr/issues/828
         csv_links_files <- readLines("link_res.csv")
         writeLines(csv_links_files[grepl("^[^#]", csv_links_files)],
                    con = "link_res_clean.csv",
                    sep = "\n")

         ## write output to CSV file and check error codes
         ## stop only for 404s
         res_links <- readr::read_delim("link_res_clean.csv", delim = ";")
         unique(res_links$result)
         res_404 <- grepl("^404|^gaierror", res_links$result)
         other_errors <- grepl("error", res_links$result, ignore.case = TRUE)
         if (any(other_errors))
           warning("These links might be problematic: \n",
                   format_url_errors(res_links, other_errors))
         if (any(res_404))
           stop("Some links are broken (check log to inspect): \n",
                format_url_errors(res_links, res_404))
         })
)

check_links <- function() {
  CheckLinks$new()
}