David says:
I've been trying to say that generic algorithms *can't* worry about the range lifetime; they have to assume that the iterators they're passed will not be invalidated before they're destroyed. ... Bleah (sorry). Iterators in general shouldn't be burdened with the requirement to know the limits of the sequence, and if you go back and ... My advice: make sure you keep in mind the whole problem domain. You can't understand the reason that C++ iterators are unlike those in other languages without looking at the algorithms.
What you are saying about iterators is all true, and they work wonderfully. The problem that you acknowledged, which is exactly the problem that I have when designing difference_range, is that a _pair_ of iterators often has redundant information, which in the case of stacking difference_ranges leads either to storage bloat or to storage lifetime issues because we need a place to store this redundant information. The universal end iterator fixes this problem for forward ranges, but not for bidirectional or random access ranges. Giovanni says:
Anyways, the solution is simple: if you want your adaptor to survive the original wrapped range, you have >to store a copy of the wrapped range in the adaptor. As an added benefit, the adaptor size will be the >smallest possible for deep sequences of stacked adaptors.
This is a good solution, but RangeEx is not designed that way, and it is a big departure from regular containers. I am happy with it if we decide that this is the way to push for the next C++ standard:-) Arno -- Dr. Arno Schoedl · aschoedl@think-cell.com Technical Director think-cell Software GmbH · Invalidenstr. 34 · 10115 Berlin, Germany http://www.think-cell.com · phone +49-30-666473-10 · toll-free (US) +1-800-891-8091 Directors: Dr. Markus Hannebauer, Dr. Arno Schoedl · Amtsgericht Charlottenburg, HRB 85229