Re: Motion to finalise interval literals

[Richard Fateman <fateman@xxxxxxxxxxxxxxxxx>]

On 6/5/2013 8:15 AM, Michel Hack wrote:
> John Pryce is proposing:
> > - with the addition of the singleton interval form [x]
> >    which is equivalent to [x,x].
> What if there is no such singleton, e.g. [0.1] in Binary64?
>
> I seem to recall that we explicitly avoided this form two or three years
> ago precisely because of this issue.
>
> The syntax would be ok if it was not required to deliver a singleton, but
> could produce an interval containing only two representable numbers which
> enclose a non-representable mathematical value.
>
> Note that such an interval could be semi-bounded if the mathematical value
> exceeds MAXREAL in absolute value, e.g. [1e400] for Binary64.
>
> Michel.
> ---Sent: 2013-06-05 15:22:09 UTC

I'm not sure where this is heading, but if text2interval is made 
optional, then
Sun's  version of converting text to interval could, under a different name,
be non-controversial.  (As I recall, but did not check,  a distinction 
is made
there between say  0.1 and 0.100  etc).

I have no need whatsoever for text2interval
in lisp,  but then I have exact rationals, and the possibility of 
writing programs
that round values up and down  separate from intervals, e.g. 
nextup_double(1/10).


RJF