Re: Revised Motion 26 decoration scheme
Vincent Lefevre wrote:
Anyway bare decorations don't make much sense. I would say that
they are purely algorithmical objects (i.e. something that could
be used in an algorithm, but with no clear meaning without this
context). Compare to bare intervals, which are true mathematical
objects.
Yes. That's a nice summary, IMHO.
In the B&B algorithm, for example, one computes with bare intervals as true
mathematical objects.
When an exception occurs, however, a bare decoration is returned. However,
the *only* practically useful purpose of the bare decoration at that point
is to ensure it propagates to the end of the evaluation of the DET, so the
original cause of the exception can be reported to the main part of the
algorithm.
So in that sense, bare decorations are an algorithmic tool for efficient
computer implementations. Thier main purpose is to ensure information about
a particular exception is propagated to the end of some lengthy computation.
Nate