7 Dec
2008
7 Dec
'08
6:13 p.m.
From: Neal Becker
But, the limitations on floating point comparisons are/were always there, and those who care should already be aware of them.
Even discussions during this review show how people are *not* aware of the limitations and are very surprised to learn about them. I think the lack of the awarness is common because the behaviour of FP is far from what one could call "normal" or intuitive.
All that happens, I think, is that when you assign a result to the constrained float type, it may or may not violate the constraint if it is very close to the limit. Sounds to me that it's still a useful construct.
OK, I'll explain this in the docs. Best regards, Robert