Re: proposal to toss out text2interval.
Richard Fateman wrote:
> Thus instead of double_text2interval("[a,b]") we would use
> A := <<double floating literal in the default locale>>
> B := << ditto>>
> result := double__nums2interval(A,B);
What aspect of "containment" don't you (Richard) understand????
As I mentioned in an earlier reply, it *would* be possible for a
language with first-class interval support to tag numeric values
with an exactness indication -- ideally a signed one, but if we
stay within standard numeric formats such as BFP64, ORing in the
inexactness in the low-order bit could work. (With DFP128, as
used in Cobol, one would have to use IBM's round-for-reround (aka
round-to-prepare-for-shorter-precision), to handle a low-order 5
correctly.)
But we would also like to accomodate existing languages where
interval support is provided through libraries and macros.
The only generic way to do that is via text2interval().
There is no problem supporting your bounded-precision bigfloat format.
For exact-rational however you would have to provide a documented
precision cutoff, in order for the hull() operation to be well-defined.
Michel.
---Sent: 2013-03-01 19:30:41 UTC