The CRAN Task View system is a fine project which Achim Zeileis initiated almost a decade ago. It is described in a short R Journal article in Volume 5, Number 1. I have been editor / maintainer of the Finance task view essentially since the very beginning of these CRAN Task Views, and added the High-Performance Computing one in the fall of 2008. Many, many people have helped by sending suggestions or even patches; email continues to be the main venue for the changes.
The maintainers of the Web Technologies task view were, at least as far as I know, the first to make the jump to maintaining the task view on GitHub. Karthik and I briefly talked about this when he was in town a few weeks ago for our joint Software Carpentry workshop at Northwestern.
So the topic had been on my mind, but it was only today that I realized that the near-limitless amount of awesome that is pandoc can probably help with maintenance. The task view code by Achim neatly converts the very regular, very XML, very boring original format into somewhat-CRAN-website-specific html. Pandoc, being as versatile as it is, can then make (GitHub-flavoured) markdown out of this, and with a minimal amount of sed magic, we get what we need.
And hence we now have these two new repos:
- CRAN Task View for Finance, on GitHub
- CRAN Task View for High-Performance Computing with R, on GitHub
Contributions are now most welcome by pull request. You can run the included converter scripts, it differs between both repos only by one constant for the task view / file name. As an illustration, the one for Finance is below.
#!/usr/bin/r ## if you do not have /usr/bin/r from littler, just use Rscript ctv <- "Finance" ctvfile <- paste0(ctv, ".ctv") htmlfile <- paste0(ctv, ".html") mdfile <- "README.md" ## load packages suppressMessages(library(XML)) # called by ctv suppressMessages(library(ctv)) r <- getOption("repos") # set CRAN mirror r["CRAN"] <- "http://cran.rstudio.com" options(repos=r) check_ctv_packages(ctvfile) # run the check ## create html file from ctv file ctv2html(read.ctv(ctvfile), htmlfile) ### these look atrocious, but are pretty straight forward. read them one by one ### - start from the htmlfile cmd <- paste0("cat ", htmlfile, ### - in lines of the form ^<a href="Word">Word.html</a> ### - capture the 'Word' and insert it into a larger URL containing an absolute reference to task view 'Word' " | sed -e 's|^<a href="\([a-zA-Z]*\)\.html|<a href="http://cran.rstudio.com/web/views/\1.html"|' | ", ### - call pandoc, specifying html as input and github-flavoured markdown as output "pandoc -s -r html -w markdown_github | ", ### - deal with the header by removing extra ||, replacing |** with ** and **| with **: "sed -e's/||//g' -e's/|\*\*/\*\*/g' -e's/\*\*|/\*\* /g' -e's/|$/ /g' ", ### - make the implicit URL to packages explicit "-e's|../packages/|http://cran.rstudio.com/web/packages/|g' ", ### - write out mdfile "> ", mdfile) system(cmd) # run the conversion unlink(htmlfile) # remove temporary html file cat("Done.n")
I am quite pleased with this setup—so a quick thanks towards the maintainers of the Web Technologies task view; of course to Achim for creating CRAN Task Views in the first place, and maintaining them all those years; as always to John MacFarlance for the magic that is pandoc; and last but not least of course to anybody who has contributed to the CRAN Task Views.