exspes@bath.ac.uk (P E Smee) (04/09/90)
In article <16406@haddock.ima.isc.com> karl@haddock.ima.isc.com (Karl Heuer) writes: >In article <1990Apr6.102819.9379@bath.ac.uk> exspes@bath.ac.uk (P E Smee) writes: >>In fact, I tend to the opinion that any N-possibility operation (function, >>whatever) should offer an N+1st value which means 'I couldn't tell'. ... >>Consider the ERROR value to be an 'out-of-bandwidth' value which is >>coincidentally returned down the same channel -- much as EOF in get*. > >That can work when N is small, but you start to have problems with a function >whose valid return values span the entire universe of its natural type. The >`getchar()' function was kludged in by declaring it `int' instead of the more >natural `char', but there's no such simple remedy for `atoi()'. Too true. I think my *real* opinion is that I don't really mind whether you add out-of-band error values or not, as long as you're consistent. (Just in case I ever have to take over the code. :-) I assume that anyone who uses TRUE/FALSE/ERROR 'two-states' will write their code to work with them. It also demonstrates a weakness which is the price you pay for C's (usually handy) orientation of 'everything's a function, and you don't have to worry about protecting arguments because it's pass-by-value'. That approach has a lot of good points, but the down side is that there is no 'easy and aesthetically pleasing' way of getting both a return value, and a success/failure indication, out of a function. (Have to design one when I invent MY language. :-) -- Paul Smee, Computing Service, University of Bristol, Bristol BS8 1UD, UK P.Smee@bristol.ac.uk - ..!uunet!ukc!bsmail!p.smee - Tel +44 272 303132
brnstnd@stealth.acf.nyu.edu (04/11/90)
In article <1990Apr9.090340.7216@bath.ac.uk> exspes@bath.ac.uk (P E Smee) writes: > That approach has a lot of good points, but the down side is that there > is no 'easy and aesthetically pleasing' way of getting both a return > value, and a success/failure indication, out of a function. Actually, it isn't too hard. Just think how you would implement Ada's exception handling, which, like mostly everything else in Ada, is much too high level for me to appreciate. Along with the other information on the stack you pass a pointer to a statement---one per exception handler, or just one for all exception handlers. The called function raises an exception by doing the callee's half of function returns, then jumping to the exception-handling statement. (If there's just one, it uses a register to indicate which exception it wants.) The exception-handling statement does the caller's half of function returns and that's it. The most natural way to do this in C is to pass an appropriately initialized jmp_buf as a parameter to the function; everything else is straightforward. Q will have a nicer solution---trust me. ---Dan