Re: Motion 46: finalise interval literals, amendments NO RATIONALS
Jürgen proposes to make rational literals optional, and also wrote:
> p1788 is about intervals not about symbolic computing
about the comment mentioning the omission of a notation for "pi".
Having a name for Pi has nothing to do with symbolic computation. It
shares with simple rationals such as 1/3 the property that it allows
mathematical constants to be specified in a precision-independent
manner, so that the maximum available precision can be exploited.
There is a work-around of course: use expressions such as:
interval tmp1 = [1]
interval tmp2 = [3]
interval Thrd = tmp2/tmp3
or
interval tmp = atan([1])
interval Pi = [4]*tmp
Mathematical constants with greatest-available precision are important when
transforming values through many operations, even when those values are
initially imprecise -- ESPECIALLY so if the transformations are non-linear.
For this purpose it is sufficient to allow rationals with bounded numerator
and denominator, e.g. the environment's common integral type. That should
not cause too much grief on any platform that supports division!
Michel.
---Sent: 2013-07-11 15:45:59 UTC