2012/4/5 Stas Malyshev <smalys...@sugarcrm.com> > Hi! > > >> it's a wonderful mechanism with more uses than simply reporting errors > >> - the aspect of transferring control is what I find really interesting > >> about exceptions. > > Exceptions should not be used for flow control. [...] > > While exceptions themselves may not be suitable for a general-purpose control structure, they do embody one. It's an early-return mechanism that you also see in some of the other control structures mentioned in this thread (Python's generators, Common LISP's conditions). Even the "return" statement itself could be considered a simpler form of it, with the difference that "return" can only go up a single invocation while "throw" can exit an arbitrary number of them (also, "throw" is limited in the type of value it can return).
The argument against any early-return mechanism is that it creates a too-complex control flow. This is the same argument from the structured programming wars about having multiple exit points<http://c2.com/cgi/wiki?SingleFunctionExitPoint> from a function, in support of having a single return at the end of functions. It certainly can produce messy, hard to understand code, but it can also be quite powerful, simplifying certain problems. Perhaps there's some middle ground that balances clarity, simplicity and power.
