"Bruce McFarling" <agila61@[EMAIL PROTECTED]
> wrote in message
news:aa760fbd-ac90-47e9-80a3-5f2092e4fa7b@[EMAIL PROTECTED]
> On Mar 29, 4:15 pm, an...@[EMAIL PROTECTED]
(Anton Ertl)
> wrote:
> > Bruce McFarling <agil...@[EMAIL PROTECTED]
> writes:
> > >If FICL is a C-based forth, it should be straightforward to add
> > >REPRESENT. sprintf itself would be the core function
>
> > REPRESENT is pretty obviously modeled after ecvt(); they missed some
> > things in the transfer, as Ed has pointed out, and they also did not
> > provide an fcvt() variant.
>
> I just looked at a page describing ecvt() and fcvt() and I don't see
> what is missing from REPRESENT together with already existing CORE
> facilities if someone wants to reconstruct them. The signed exponent
> is returned, the sign, and the representation of the absolute value of
> the significant normalized between 0 and 1, which makes it
> straightforward to place the decimal point anywhere desired and adjust
> the exponent to suit.
There are several deficiencies in the standard REPRESENT but the main
one is that it can't round-up the entire significand (i.e. the fraction)
to a
whole number. This is required for functions like F.R to work properly.
For the story on REPRESENT and a solution see:
ftp://ftp.taygeta.com/pub/Forth/Applications/ANS/Represent_21.txt
|
