[Rd] trySilent (was try.Silent)
Sat Jan 11 15:16:02 2003
On Wed, 8 Jan 2003, Kurt Hornik wrote:
> >>>>> writes:
> > We could, and it had crossed my mind. Thanky ou for raising it.
> > However, we could not do that before 1.7.0 (new feature, and
> > anyway 1.6.2 is in final testing for release in a couple of days).
> > We have been considering having package methods in the default package
> > list for 1.7.0, which would provide this. However, my preference would be
> > to have this as an argument to try, as in try(..., silent=TRUE)
> > Note that trySilent actually does two things:
> > i) it sets options(show.error.messages = FALSE)
> > ii) it sets options(error = empty.dump) (or equivalently) to suppress
> > any error handler
> > and one might just want just one of those two. Do perhaps
> > try(..., silent = FALSE, nodump= TRUE)
> > (or defaulting nodump to silent) might be a comprehensive interface.
> > Any preferences?
> [Anticipating Luke's comment, or at least trying to ...]
> Wouldn't it be better to more generally have a way of executing with
> options temporarily reset in a 'safe' way, a la
> withOptions(EXPR, list(show.error.messages = FALSE, ...))
In the (probably not so) longer run we need that; we will also need to
think through how try and try(...,silent=TRUE) should work within a
more sophisticated error handling system. But I think the simpler
interface that allows just try(..., silent=TRUE) is fine and can be
implemented in a way that works reasonably now and also later when we
have better options management and better error handling. (In writing
the documentation we should make sure not to commit to implementation
details, just the funtionality.) I would prefer not to add separate
control over frame dumping; fine control like that is I think better
left to the error handling system.
University of Iowa Phone: 319-335-3386
Department of Statistics and Fax: 319-335-3017
241 Schaeffer Hall email: email@example.com
Iowa City, IA 52242 WWW: http://www.stat.uiowa.edu