ess and debug package

[Stephen Eglen]

Dear Mark,

 > Thanks for the report. I'd suggest trying 'options(
 > debug.command.recall=FALSE)' before running the debugger (perhaps
 > in '.First')-- please let me know if this works.

Thanks, that helps.  The following sequence now works under ESS for
me:

library(debug)
options( debug.command.recall=FALSE)
mtrace(glm)
glm(stupid.args)

now the Tk window appears, with D(1)> in the ESS window.


 > I've never had any joy with emacs, and I haven't been able to test
 > debug under ESS, so I'd particularly appreciate feedback if (when)
 > there are problems. The doco for 'savehistory' says "doesn't work
 > in batch use of Rterm" which I guess may be ESS's modus
 > operandi.

Yes, Martin Maechler pointed this out to me in response to my email,
that since ESS uses "R --no-readline", there will be no readline
history.

 > Is it the case that 'interactive()' returns FALSE in ESS?
 > If so, I can patch the debug code easily.
Unfortunately not:

> interactive()
[1] TRUE

One way to test whether R is running under ESS is to check the value
of the STERM option, since ESS sets this:

> options("STERM")
$STERM
[1] "iESS"

[Although we had some debate a while ago about changing the exact
value of STERM.]  Let me know if you want any new versions tested
under ESS.

As an aside, when I saw your package announcement, my immediate
thought was whether we could make use of an Emacs window (rather than
a Tk window) to show the function being debugged.  Emacs already has
similar support for other programming environments (e.g running gdb
under C, lisp debugger).  But that's a low priority for now!

best wishes,

Stephen