R help - Dec 2004 - hang-up during nlme call

Hello:
I recently asked the list a question regarding
warning.expression.  This is a different statement of
the problem.  

I am doing a large simulation experiment using nlme
and most of the data realizations run fine.  The
simulations stall, however, on a few particularly
noisy data sets.  What happens is that the nlme
function initially calls the nlm function and some
code in there generates an endless number of warning
messages.  These are: "Warning: Singular precision
matrix in level ...".  Becuse of this, my simulation
is caught in an endless loop and after quite a while R
crashes.

I do use try(nlme(...)), so if I could find a way to
flag these warning and make them into errors, such as
by calling stop(), then my code could move on to the
next realizaton of the simulation.  Some warnings are
produced that are not endless or otherwise
problematic, so I do not want to cause an error for
every warning.  

Does anyone have any ideas as to how to identify
warning messages and call stop() or otherwise induce
an error if the warnings are of a particular type?  

John

JJ <josh8912 <at> yahoo.com> writes:

: 
: Hello:
: I recently asked the list a question regarding
: warning.expression.  This is a different statement of
: the problem.  
: 
: I am doing a large simulation experiment using nlme
: and most of the data realizations run fine.  The
: simulations stall, however, on a few particularly
: noisy data sets.  What happens is that the nlme
: function initially calls the nlm function and some
: code in there generates an endless number of warning
: messages.  These are: "Warning: Singular precision
: matrix in level ...".  Becuse of this, my simulation
: is caught in an endless loop and after quite a while R
: crashes.
: 
: I do use try(nlme(...)), so if I could find a way to
: flag these warning and make them into errors, such as
: by calling stop(), then my code could move on to the
: next realizaton of the simulation.  Some warnings are
: produced that are not endless or otherwise
: problematic, so I do not want to cause an error for
: every warning.  
: 
: Does anyone have any ideas as to how to identify
: warning messages and call stop() or otherwise induce
: an error if the warnings are of a particular type?  

See ?withCallingHandlers .  Here is an example which raises an
error for an "X" warning but not other sorts of warnings.  

f <- function(x) if (x$message == "X") stop("X")
withCallingHandlers({
	print(10)
	warning("Y")  # this warning does not result in stop
	print(20)
	warning("X")  # this warning results in stop
	print(30)
}, warning = f)

Excellent Gabor.  Thanks.  I had seen the
"withCallingHandlers" function, but I did not
understand how it was to be used.  Thanks for the
example.
John