|
Boost : |
From: Andrew Koenig (ark_at_[hidden])
Date: 2002-08-23 13:22:16
Peter> Agreed.
>> The same argument applies to [].
Peter> A weaker version of the argument. Only random access iterators
Peter> have [], so users cannot expect all iterators to supply it.
Right. But they can expect random-access iterators to supply it.
Peter> Theory aside, do users really expect [] from iterators? I have
Peter> never used iterator::[], although I do use iterator::->
Peter> frequently.
I would expect that most users of iterators do not actually require
random-access iterators. There are only three operations that
random-access iterators, and not other iterators, are required to
support:
iter[integer]
iter + integer (and integer + iter)
iter - iter
Either of the first two operations can substitute for the other,
so the real question is whether the second is much more common than
the first. I suspect it isn't.
-- Andrew Koenig, ark_at_[hidden], http://www.research.att.com/info/ark
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk