Re: Stopping a program
- To: mathgroup at smc.vnet.net
- Subject: [mg93277] Re: Stopping a program
- From: carlos at colorado.edu
- Date: Sun, 2 Nov 2008 01:56:41 -0500 (EST)
- References: <200810300700.CAA00553@smc.vnet.net> <email@example.com>
On Oct 31, 2:04 am, Daniel Lichtblau <d... at wolfram.com> wrote: > car... at colorado.edu wrote: > > This is a revision of a previous posting "Looking for Stop". > > Unfortunately the answers were not helpful. My fault: I didnt > > explain what I am looking for in specific terms. So here it is. > > > Several application programs in Mathematica that I use in > > graduate courses are translations from old (1960-80s) Fortran and C > > programs. A medium-size Fortran program with, say, 10^6 statements > > contracts to about 40-100K Mathematica commands, and similarly in > > other high order languages. The reduction comes mostly from > > use of built-in functions. However, the overall hierarchical > > structure > > does not change much, e.g. the "main program" becomes a > > "Driver Script Cell" (DSC) and functions/routines become Modules. > > The typical structure is > > > DSC: > > Preprocessing Modules (written by students) > > Processing Modules (supplied) > > Postprocessing Modules (supplied) > > > There are typically 50-200 supplied modules, forming trees 3-6 > > levels deep, depending on application. Programs are run > > interactively by students doing homework or take-home exams. > > They write the problem-dependent preprocessing part of the DSC, > > save the NB, initialize the supplied modules (which are in > > untouchable initialization cells) and execute the DSC. > > > Now suppose an irrecoverable error is detected in a Module four levels > > deep. In Fortran or C, I give diagnostics and call a central error > > termination routine that also displays the "call trace" kept in a > > stack. > > In Mathematica I can only say Exit, which is a bit drastic. > > > What I was looking for is a way to transfer to the end of the DSC > > as if the program had terminated normally. The student looks at > > the error message, fixes the preprocessing part as needed, and > > tries again. As mentioned in the OP, I thought to put a Label: > > at the end of the DSC but AFAIK Mathematica does not allow > > forward control transfers. Specific suggestions are welcome. > > Mathematica certainly supports allowing Label to syntactically follow > corresponding Goto(s) (I have used it this way). But I doubt this is > what you want because they really need to be in the same Module, With, > or Block. (Maybe they don't, but I'd not bet on them working correctly > if not). > > You almost certainly can achieve what you have in mind with Throw at the > point of error, and Catch in the DSC. That's generally how we transfer > control back to caller in error situations. Other possibilities include > using Check in your DSC, and Return at points of error. But this > involves really knowing where Return will take you and thus is not > reliable if you are neted several layers deep. > > Daniel Lichtblau > Wolfram Research My experience with Catch and Throw in C++ (esp. write-only code written by PhD students) has been discouraging. I think the old fashioned way described by Don Knuth in his "Structured Programming with goto Statements" is still the best to cut through multiple layers. But it requires a compilable language. How about a simple addition to Mathematica 7: Stop [ options ]. One of the options, perhaps default, could be "Stop execution of the topmost executing cell and return control to front end". For 4.1 through 6.0 I can keep Exit.