callback spaghetti (was scope of malloc)

[Richard Minner]

In article <207 at thor.UUCP> scjones at thor.UUCP (Larry Jones) writes:
>I don't know about Doug, but *I* sure reguard callbacks as both tangled
>control structures and unstructured braches.  In fact, they seem to be
>very closely related to the "altered goto" which seem to be held in more
>or less universal contempt.  Mind you, I'm not saying they aren't USEFUL,
>just that they're very hard to follow!  

Yep, qsort() is awfully confusing.  But seriously, I think I understand
what you mean, and indeed I have seen some horridly unstructured use
of callbacks, and of function pointers in general.  But I can say the
same about data pointers used to make incomprehensible data structures,
or almost any other language feature.  Very few things are *inherently*
unstructured (not even goto's!).  The spaghetti is put there by the
designer of the code, not the designer of the language.  (Granted,
some languages make it a lot easier than others.)  Also, a function
call of any sort is very unlike a goto of any sort, because a function
call *returns* (barring longjmp()).  Which brings up...

I'm hoping that it was the longjmp() out of the callback that
prompted the charge of `spaghetti' from Doug.  I'm not sure I'd
immediately call such a longjmp() spaghetti (though I might), but
I do strongly recommend that callbacks be specified to return
normally.  (It never occurred to me before to specify it explictly,
but I probably will in future.)

Function pointers, including their use as callbacks, are
not just useful; they are *extremely* useful.  This mainly
applies when your goals include reusability, generality
and extendability.  If the functions pointed to are reasonably
constrained, callbacks and the like really aren't that hard
to follow, once you get used to the idea.

-- 
Richard Minner  || {uunet,sun,well}!island!rtm     (916) 736-1323 ||
                || Island Graphics Corporation     Sacramento, CA ||