[Rd] ChangeLog for R packages {was "Wish list"}

[Prof Brian Ripley]

On Wed, 21 Jan 2004, Martin Maechler wrote:

> >>>>> "BDR" == Prof Brian Ripley <ripley at stats.ox.ac.uk>
> >>>>>     on Tue, 20 Jan 2004 17:26:58 +0000 (GMT) writes:
> 
>     BDR> Folks, Have you looked at R's own ChangeLog recently?
> 
>     BDR> You can insist on a ChangeLog, but some of us may be
>     BDR> tempted to follow R's own example ....
> 
> what do you mean?  I think I haven't understood....
> 
> We've followed the rule to use
>  NEWS +  "informative (but sometimes terse)" CVS messages.
> 
> Using a *manual* ChangeLog is a different approach and doing both
> (or even all three) doesn't make sense at all.  That's why Greg
> mentioned the "cvs2cl" tools where ChangeLog is autogenerated
> from the CVS logs.
> 
> However, for (non-core) R *packages* things are very different:  
>     The majority of package authors will never use CVS (and if
>     they were urged to, they'd use almost empty log messages),
> but encouraging them to use a ChangeLog file in a simple format
> seems feasible IMO.

Encourage or force?

Quite a few packages do already have a ChangeLog, and I took the thrust of 
this thread to be to require one or at least nag for one.  It started as

Dirk> R CMD check could warn if there is not ChangeLog or Changes file, and
Dirk> thereby suggest that authors provide them.

My point is that we cannot know if there is useful information in the 
ChangeLog, and all too frequently (from wider experience than R) it is 
either

manual and incomplete (and missing the change I am searching for)
automatic and far too detailed for useful searching (and with records of 
   all the less-than accurate CVS messages).

-- 
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