Date: 2000-03-07 15:08:20
Intervals certainly would be a good addition for boost. My group at
Notre Dame implemented such an interval arithmetic class last year at
the request of some chemical engineers we work with... intervals are
very important for optimization problems.
Looking at the interval proposal there's some missing functionality.
There are a number of miscellaneous functions that are commonly used
with intervals such as hull(), bisect(), diameter(), distance(),
intersection(), predecessor(), successor(), empty(). Also it would
probably be good to include a specialization for numeric_limits. I've
attached our implementation... probably the thing to do is merge this
functionality into the proposal. Also it includes the assembly code
for Solaris rounding modes.
There are some issues that need carefull attention, the behaviour of
divide by zero, what semmantics to choose for operator< if any,
etc. It would probably be good to pick the brains of the interval
computing community (there's a mailing list reliable_computing that
you can get to from the web page previously mentioned). They have some
C++ implementations as well, though they have not paid much attention
to creating a class that is C++ standard compatible. This of course
is important so that we can write generic code, for instance the
interval class should be usable in MTL with no changes.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk