[Rd] suggestion for R >= 3.0: computer-readable CHANGELOG

Dirk Eddelbuettel edd at debian.org
Fri Apr 17 17:20:18 CEST 2009


Salut Romain,

On 17 April 2009 at 16:36, Romain Francois wrote:
| I agree with the usefulness of having this available, but there is 
| absolutely no way people are going to log such information in a 
| systematic fashion. In the other hand, if you have version 1 and version 

There is: you gently prod and later force them.  

We had this discussion fourteen or so year ago in Debian when
debian/changelog was not yet mandatory.  We should recommendend it now (and
have 'R CMD check' issue a warning if it is missing.

The factof the matter is that for R, we do have a) gatekeppers and we also
have b) enforced minimum standards (of passing R CMD checl).  We now 'just'
need to agree that we plan to raise the standard in an ever so slight way in
order to get additional information into packages.

To be clear:  I think this should still be voluntary. Some packages authors
don't want to, some won't -- but there will hopefully be enough to play along
at fist, and I hope that this will over time sway 

| 2 of some package, then why not do some programmatic investigation of 
| the code to get (some of) these answers.
| 
| This looks like CRANberries, but working at the object level instead of 
| working at the file level, but then you can imagine to parse the 
| package, dump each function/object/class in its own file, and cranberry 
| that. Is CRANberry an R package ?

CRANberries is ultra-simple and just 200 lines of R code. It basically
compares what the DESCRIPTION parser and available.packages() et al provide
with a stateful snapshot of that same info via SQLite. It then uses standard
tools like diff and diffstat to generate comparisons between tarballs. It
does NOT really peek into tarballs to read / parse / analyse source content.
And I have no plan to add that either.

Because CRANberries lives off a local mirror and needs a local db, I haven't
really worked on abstracting things out to put it onto r-forge. I will,
eventually, and have sent the code to 'interested parties' like Duncan when
he was building the R NEWS RSS feed.

Hth, Dirk

-- 
Three out of two people have difficulties with fractions.



More information about the R-devel mailing list