[Rd] NEWS.md support on CRAN

[Prof Brian Ripley]

On 23/05/2015 13:14, Duncan Murdoch wrote:
> On 23/05/2015 4:05 AM, Kurt Hornik wrote:
>>>>>>> Duncan Murdoch writes:
>>
>>> On 22/05/2015 8:49 PM, Imanuel Costigan wrote:
>>>> Are there any plans for CRAN to support NEWS files in markdown? Bit of a hassle to go the the package’s Github (or other like) site to read NEWS.
>>
>>> Not as far as I know.  There have been discussions about increasing the
>>> support of Markdown, but so far the conclusion has been that it's too
>>> hard to do -- the support is not stable enough on all the platforms
>>> where R runs.
>>
>> There are actually two issues here.
>>
>> For CRAN, we could in principle take inst/NEWS.md files, convert these
>> to HTML using pandoc, and use the HTML for the package web page.  (Would
>> need the CRAN incoming checks to be taught about inst/NEWS.md.)
>>
>> However, we cannot use such files for utils::news() because we do not
>> (yet?) know how to reliably parse such files and extract the news items
>> (and hence cannot really compute on the news information).
>
> It would be quite easy to modify the news() parser to parse a suitably
> described Markdown format.  The main change from the current text parser
> would be to expect a prefix on the line introducing each version or
> category.
>
> I think the harder problem is display.  CRAN can run pandoc, but can
> users who install the package from source?  I would expect some obscure
> platforms (like Windows ;-) would not have it available.  I could add it
> to Rtools on Windows, but I think platforms that normally install tools
> from source will have more trouble, because it has an unusual
> prerequisite (Haskell).  We could fall back to displaying the NEWS.md
> file without processing, but that makes lots of code more complicated.

pandoc is only viable on platforms with a pre-compiled binary, which 
does not even cover all the CRAN check platforms.  (It is not easy to 
compile ab initio from source even on a mainstream Linux, but they 
mostly have pandoc binaries.)

The next problem is that using pandoc as the de facto standard for .md 
depends on the version of pandoc, and .md seems to have no way to 
specify the version required.  (If it does, people are failing to use it 
....)


-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Emeritus Professor of Applied Statistics, University of Oxford
1 South Parks Road, Oxford OX1 3TG, UK