On Wednesday, 18 December 2013 at 02:17:06 UTC, Andrei
Alexandrescu wrote:
On 12/17/13 5:58 PM, "Ola Fosheim Grøstad"
<[email protected]>" wrote:
Why is that? I would think that 10 == Interval(10,10), but
interval(5,100).contains(10) ?
Yah, defining == my way would make it non-transitive.
True interval arithmetic is difficult to
implement though, since !(a<b) does not imply (a>=b) if I got
it right…
I didn't peruse the wikipedia page in detail but it seems a lot
of interval arithmetic is well-defined.
But it is counter-intuitive since an interval represent
uncertainty?
[5,5] == [5,5] => true
[1,5] != [1,5] => false? // counter-intuitive
[0,5] < [6,10] => true
[0,5] < [2,10] => uncertain
So you you might have to introduce a tri-boolean type
That would be good too for distinct reasons.
I wrote libraries for both tribool and fuzzybool for D 7 years
ago to see how D worked for ADTs. I guess I could rewrite those
to be more modern and make it available as a starting point. I
believe I also started on (or planned) to do fuzzy numbers. Fuzzy
numbers are related to intervals, but have a triangular
membership function. You use it for representing hazy values like
"pretty" or "intelligent" (the same person can be both smart and
stupid at the same time). There are more than one definition,
each with trade-offs.
e.g.:
fast = FuzzyNumber(40,60,70)
slow = FuzzyNumber(1,10,40)
However, one deficiency in D is (or was) that you cannot make
comparison operators return non-bool values?
Anyway, I think it might be a good idea to implement Intervals,
Fuzzynumber, Tribool and Fuzzylogic in a manner that is
consistent.
