Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

A Level 2 query



P1788

I am redrafting Level 2 text to take into account our decisions to support Explicit and Implicit interval types. I have a quandary about functions that take an interval and return a point value such as
  midpoint, radius, magnitude etc.       (*)

What floating point (FP) format should the result have?

For an inf-sup interval type there is no problem. It is defined in terms of some number format F (which will be a FP format in practice), and its nonempty intervals xx are precisely those whose endpoints are F-numbers. So functions (*) should return an F-number too.

What about other interval types, e.g. a mid-rad one, which is implicit? There is no FP format inherently associated to it. I feel that the less is specified, the better, but there is a minimum that must be specified, to avoid silly differences between implementations. 

I suggest there shall be a FP format that is part of the definition of any interval type, and all functions (*) shall return a value of that format.

Anyone have a better idea?

John Pryce