From: Guillaume Melquiond (gmelquio_at_[hidden])
Date: 2002-09-02 14:11:09
On Mon, 2 Sep 2002, David Bergman wrote:
> In fact, paragraph 3 of the Comparisons section seems to give an
> (implicit) opening for partially ordered types, since the whole "for
> all" argument would not be needed, were the type totally ordered; one
> would only need to compare the end points.
> I agree that there would be usage of this interval abstraction without
> all the arithmetic operations.
> One question: should we have iterators over an interval? I know it sound
> a bit Pythonish, but it would be nice to traverse (at least discrete)
> And, if "numbers" imply types with the given arithmetic operations
> defined, then one should definitely split the headers into "interval"
> and "interval_arithemtics", so we could use (partially) ordered types
> for interval operations such as comparison, slicing and traversal.
If I clearly understand you, you don't need the arithmetic part of the
library. But please remember that the name Interval Library is an
abbreviation of Interval Arithmetic Library. Its purpose is *arithmetic*.
What you propose is a complete change in the interface and the meaning of
the library. It was meant to provide an usable and efficient way of
replacing numbers by intervals in arithmetical operations.
If you only need "interval operations such as comparison, slicing and
traversal", then you should take a look at the proposal of Richard Peters
who has been working on "a library for representing ranges".
Please excuse me if you find my tone of voice uncalled-for. I just wanted
to say that you may be misunderstanding the original purpose of the
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk