Andre Poenitz wrote: > It is good enough if we are happy with input iterator semantics. No '--'
[forward iterator semantics would be more applicable here]. > there. For backwards iteration we could just use another iterator class. Yes, but why? We give up bidi iterator semantics for no reason. > In fact, the only place where we seem to need iteration backwards is in > 'search backwards' and there we've always a suitable 'valid' start. We may (circumstantially) have no need now... but it would be so easy to comply with bidi iterators requirements (at virtually no cost)... [if you insist with the 'empty stack' past-the-end, we could just add an inset * member and use that on operator-- (and possibly even operator++) for an empty stack. Then the past-the-end (and possibly also before-start) position would be DocumentIterator(inset)] PS: OT, should I rename iterators.[Ch] to pariterator.[Ch] to make things more uniform? Alfredo
