[Rd] hook for when R quits

[Duncan Murdoch]

On 11/03/2011 3:11 PM, Henrik Bengtsson wrote:
> On Fri, Mar 11, 2011 at 11:07 AM, Prof Brian Ripley
> <ripley at stats.ox.ac.uk>  wrote:
> >  On Fri, 11 Mar 2011, Duncan Murdoch wrote:
> >
> >>  On 11/03/2011 1:37 PM, Michael Lawrence wrote:
> >>>
> >>>  Thanks for the suggestion, but I don't think that R finalizes all of its
> >>>  objects when it quits. At least a simple test suggests that on Linux.
> >>
> >>  Did you use onexit=TRUE?  On Windows that appears to work...
>
> Agree - here an object finalizer is more appropriate (different from
> an end-of-session hook).
>
> >
> >  It does work: RODBC makes extensive use of it, for exactly the purpose you
> >  describe (or rather, the C_level equivalent of 'it').
>
> In help(reg.finalizer) it says:
>
>    'onexit': logical: should the finalizer be run if the object is
> still uncollected at the end of the R session?
>
> What environments, objects, search path etc are available when the
> finalizer is called this way when R exits?  Is safe to always add
> 'onexit=TRUE' (which now defaults to FALSE), or should I expect an
> "exceptional" R system that the finalizer needs to account for?

To be very conservative, I would assume that nothing in R is available 
other than things that are stored in that environment.  That's probably 
not true, but it would be safe.  This mechanism is for finalizing things 
that R doesn't know about.

If you want to be less conservative, then you could look at the source 
code:  currently R_RunExitFinalizers is the very first
part of the cleanup.  I don't know if we guarantee this though.

Duncan Murdoch

> Is there any further documentation on what happens when an R session
> shuts down and in which order?
>
> /Henrik
>
> >
> >>
> >>  Duncan Murdoch
> >>
> >>>  Michael
> >>>
> >>>  On Fri, Mar 11, 2011 at 10:19 AM, Jeffrey
> >>>  Ryan<jeffrey.ryan at lemnica.com>wrote:
> >>>
> >>>  >    Take a look at reg.finalizer.  You'd have to create an object
> >>>  >    internally that would persist until R exits - and a related function
> >>>  >    to handle cleanup of course.
> >>>  >
> >>>  >    HTH
> >>>  >    Jeff
> >>>  >
> >>>  >    On Fri, Mar 11, 2011 at 12:08 PM, Michael Lawrence
> >>>  >    <lawrence.michael at gene.com>    wrote:
> >>>  >    >    Hi,
> >>>  >    >
> >>>  >    >    Is there any way that a package can listen for when R quits? The Qt
> >>>  >  stuff
> >>>  >    is
> >>>  >    >    hooking into platform-specific event loops and when those die
> >>>  >    unexpectedly
> >>>  >    >    (from the perspective of Qt), it aborts, causing an annoying error
> >>>  >    dialog.
> >>>  >    >    If we could catch when R is killed, we could cleanup, like we do
> >>>  >  with
> >>>  >    >    .onUnload.
> >>>  >    >
> >>>  >    >    Thanks,
> >>>  >    >    Michael
> >>>  >    >
> >>>  >    >           [[alternative HTML version deleted]]
> >>>  >    >
> >>>  >    >    ______________________________________________
> >>>  >    >    R-devel at r-project.org mailing list
> >>>  >    >    https://stat.ethz.ch/mailman/listinfo/r-devel
> >>>  >    >
> >>>  >
> >>>  >
> >>>  >
> >>>  >    --
> >>>  >    Jeffrey Ryan
> >>>  >    jeffrey.ryan at lemnica.com
> >>>  >
> >>>  >    www.lemnica.com
> >>>  >
> >>>
> >>>          [[alternative HTML version deleted]]
> >>>
> >>>  ______________________________________________
> >>>  R-devel at r-project.org mailing list
> >>>  https://stat.ethz.ch/mailman/listinfo/r-devel
> >>
> >>  ______________________________________________
> >>  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
> >
> >  ______________________________________________
> >  R-devel at r-project.org mailing list
> >  https://stat.ethz.ch/mailman/listinfo/r-devel
> >