[Rd] error handling

Don Wingate Don.Wingate@intellichem.com
Tue, 7 Nov 2000 11:10:36 -0800

Hello R developers,

Platform: Windows 2000.
Compiler: GNU GCC - 2.95.2 (CRTDLL)

I have a C function for evaluating commands using eval(SEXP,SEXP) which is
included in a personal make of R.dll. (Similiar to R_Proxy_evaluate() in
rproxy_impl.c)  Using SETJMP to catch errors works but there is a problem.
For example, the following code:

if (SETJMP(R_ToplevelContext->cjmpbuf)) //  R_ToplevelContext->cjmpbuf
	return FALSE;
R_CurrentExpr = eval(R_CurrentExpr , rho);

will appear to work the first time there is an error, but the second error
causes an infinite loop. I have identified the location where the loop
occurs. It is in the function jump_to_toplevel() in errors.c inside the for

	for (c = R_GlobalContext; c; c = c->nextcontext) { ... }

I have found a work around for this but is seems cludgy and I don't know the
POTENTIAL PROBLEMS that might be introduced. It involves setting up a
context and forcing R_GlobalContext's nextcontext to 0, as in the following
(by the way, this same approach doesn't work in the example above, which is
why I set up a context):

begincontext(&thiscontext, 0, CTXT_GENERIC, R_NilValue, R_NilValue,
R_GlobalContext->nextcontext = 0;
if (!SETJMP(thiscontext.cjmpbuf))
	R_CurrentExpr = eval(R_CurrentExpr , rho);

This code solves my problem, however, in certain cases my calling C++ client
application cannot catch the error (exception?) that is generated, and my
program hard crashes. (A bug in Visual Studio, I think.) For this reason,
using SETJMP(R_ToplevelContext->cjmpbuf) as in the first example would
probably be preferable in my case because it doesn't generate an exception
which must be caught by the C++ client.

Note: I am forcing an error to test this by simply calling matrix(x) on a
variable x which doesn't exist.

I realize that R is meant to run in its own event loop and that this is not
a straight-forward problem. However, if anyone able to provide any insight
into what might be going or how to handle it on I would be greatful.

Don Wingate

r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch