2009/1/16 Prof Brian Ripley <rip...@stats.ox.ac.uk>:
> On Fri, 16 Jan 2009, Felix Andrews wrote:
>
>> list(...),
>>
>> I am looking for a way to interrupt R from a callback: specifically,
>> to interrupt plotting (typically on a cairoDevice, but would be good
>> if it worked with other devices too). Of course, one can interrupt R
>> nicely from the console with Ctrl-C (or Esc in Rgui), but I need to do
>> it from a GUI. Callbacks run in a new thread, so obviously stop() etc
>> will not work. I tried to look into how Rgui does it...
>> https://svn.r-project.org/R/trunk/src/gnuwin32/psignal.c
>
> Actually no, Rgui does not use that code (but Rterm does).
>
>> ?...but it is beyond me.
>
> In detail this is OS-specific, but what the front-ends do is to set a flag
> or send a signal to R: look e.g. at rterm.c:
>
> static void my_onintr(int nSig)
> {
> UserBreak = 1;
> PostThreadMessage(mainThreadId,0,0,0);
> }
>
> or onintr in errors.c (which is called when UserBreak = 1, from the ain
> thread). In due course R_interrupts_pending gets set and then at an
> appropriate spot R stops and unwinds the currrent evaluations.
>
> The relevant Windows code is even in 'Writing R Extensions'.
>
> So depending on your OS you can raise a signal or set a flag. You can raise
> a signal on Windows too, so 'raise' is the most portable solution -- but
> beware that threading models differ.
Dear Professor Ripley,
Many thanks for your information-rich response.
I have not looked at serious C code before, so I find the R source
code quite confusing.
This is what I have come up with (as my package's src/interrupt.c):
#include <signal.h>
#ifdef WIN32
extern int UserBreak;
#endif
void do_interrupt(void)
{
#ifdef WIN32
UserBreak = 1;
//raise(SIGBREAK);
#else
raise(SIGINT);
#endif
}
This seems to work under Rgui and Rterm on windows XP. I have not
tested it on linux or macOS. Does it look reasonable?
>
>> Any help would be appreciated.
>>
>> -Felix
>>
>> --
>> Felix Andrews / 安福立
>> http://www.neurofractal.org/felix/
>
> [Those seeking technical advice will increase their chances of help by
> giving their affiliation(s) or at least explaining the purpose of the
> request so that potential helpers know it is non-commercial. You may not
> get the benefit of the doubt next time.]
Fair enough; this is for the playwith package, which is GPL'd, and
(sadly) has no commercial links or prospects. I am affiliated in
various ways with the Australian National University.
--
Felix Andrews / 安福立
http://www.neurofractal.org/felix/
3358 543D AAC6 22C2 D336 80D9 360B 72DD 3E4C F5D8
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
