R devel - Jan 2012 - pointers on including SVN revision number in package info?

I'm trying to keep debugging of a development package relatively sane.

  I see that some packages manage to incorporate what appears to be
Subversion (SVN) revision information in the package description; for
example,
> library(MASS)
> sessionInfo()$otherPkgs$MASS$Revision
[1] "$Rev: 3016 $"

  which looks like an auto-generated revision number.

  On the other hand, the rgl package (for example) appears to manually
encode the SVN revision in the package number:
> library(rgl)
> sessionInfo()$otherPkgs$rgl$Version
[1] "0.92.829"

  I'd love an automatic strategy, if possible, so I can be lazy about
updating the DESCRIPTION file every time I commit a change to SVN ...
I searched the R extensions manual (and the r-forge manual), but I'm
sure I could have missed something ...

  Or can people suggest other useful strategies for keeping track of
which development (micro-)version a random user might be working with?

 thanks,
  Ben Bolker

On 06/01/2012 20:24, Ben Bolker wrote:
>
>    I'm trying to keep debugging of a development package relatively
sane.
>
>    I see that some packages manage to incorporate what appears to be
> Subversion (SVN) revision information in the package description; for
> example,
>
>> library(MASS)
>> sessionInfo()$otherPkgs$MASS$Revision
> [1] "$Rev: 3016 $"
>
>    which looks like an auto-generated revision number.
>
>    On the other hand, the rgl package (for example) appears to manually
> encode the SVN revision in the package number:
>
>> library(rgl)
>> sessionInfo()$otherPkgs$rgl$Version
> [1] "0.92.829"
>
>    I'd love an automatic strategy, if possible, so I can be lazy about
> updating the DESCRIPTION file every time I commit a change to SVN ...
> I searched the R extensions manual (and the r-forge manual), but I'm
> sure I could have missed something ...
The Subversion book?  The MASS/DESCRIPTION file has

gannet% svn proplist DESCRIPTION
Properties on 'DESCRIPTION':
   svn:keywords

Note this only tracks the version when DESCRIPTION is changed, not what 
anything else is updated.  But then I change the DESCRIPTION file 
immediately before release.
>    Or can people suggest other useful strategies for keeping track of
> which development (micro-)version a random user might be working with?
>
>   thanks,
>    Ben Bolker
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

On 12-01-06 3:24 PM, Ben Bolker wrote:
>
>    I'm trying to keep debugging of a development package relatively
sane.
>
>    I see that some packages manage to incorporate what appears to be
> Subversion (SVN) revision information in the package description; for
> example,
>
>> library(MASS)
>> sessionInfo()$otherPkgs$MASS$Revision
> [1] "$Rev: 3016 $"
>
>    which looks like an auto-generated revision number.
>
>    On the other hand, the rgl package (for example) appears to manually
> encode the SVN revision in the package number:
>
>> library(rgl)
>> sessionInfo()$otherPkgs$rgl$Version
> [1] "0.92.829"
Yes, those are done manually.
>    I'd love an automatic strategy, if possible, so I can be lazy about
> updating the DESCRIPTION file every time I commit a change to SVN ...
> I searched the R extensions manual (and the r-forge manual), but I'm
> sure I could have missed something ...
I'd like one that produced an R version number directly, but I think 
Dirk's script is the only way to do it, and I forget how to use such 
scripts five minutes after I write them.
>
>    Or can people suggest other useful strategies for keeping track of
> which development (micro-)version a random user might be working with?
R-forge sometimes gets out of sync in what it displays as the revision, 
the Version, and what it actually offers as the "Package source" (let 
alone binaries), so I find putting the revision into the version number 
very helpful.  But I often forget to do it...

Duncan Murdoch
>
>   thanks,
>    Ben Bolker
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel

Ben Bolker <bbolker <at> gmail.com> writes:
> 
> 
>   I'm trying to keep debugging of a development package relatively
sane.
> 
...
> 
>   Or can people suggest other useful strategies for keeping track of
> which development (micro-)version a random user might be working with?
> 
Rather than update the DESCRIPTION file, rgeos includes the SVN revision 
in the startup messages, by copying in ./configure (here ./configure.in):

if test -e ".svn" ; then
  svnversion -n > inst/SVN_VERSION
fi

and testing for the file in R/AAA.R:

  fn <- system.file("SVN_VERSION", package="rgeos")
  if (file.exists(fn)) {
    svn_version <- scan(system.file("SVN_VERSION",
package="rgeos"),
      what=character(1), sep="\n", quiet=TRUE)
  } else {
    svn_version <- "(unknown)"
  }

Clunky, and inst/SVN_VERSION isn't in svn itself, but it does work if the 
build machine uses SVN and is up-to-date. I think I'll try to use the
DESCRIPTION Revision tag too. Naturally, this is feasible in packages 
using ./configure already. The package is on R-Forge, so the implicated 
files can be browsed online.

Roger
>  thanks,
>   Ben Bolker
> 
>