[r6rs-discuss] Hashes in the number syntax

From: Paul Schlie <schlie>
Date: Fri Dec 15 11:46:09 2006

Kent writes:
> (format "~,20f" 1.0) ;=> "1.0000000000000000####"
> (format "~,5e" 5e-324) ;=> "5.#####e-324" ; denormalized value

Neil writes he effectively expects:
> 1XX => [50 .. 150] vs. [100 .. 200]

(which seems sensible as being able to specify an ambiguity about
 a value in many ways is likely more useful than presuming the
 ambiguity is specified as being specified away value in its signed
 direction)

However regardless, in fact (presuming an ieee double implementation),
the actual values stored are correspondingly:

 1.0 => 1.0...

 5e-324 => 4.94065645841250...e-324

Being effectively indistinguishable from:

 1.0000000000000000000000000000000000001 => 1.0...

 4.94065645841250e-324 => 4.94065645841250...e-324

It seems fairly clear that as all potentially distinguishing
information as may be attempted to be represented in a value's
textual representation is only as well preserved as facilitated
by its implementation format, any attempt to pretend otherwise
will most likely only result in confusion;

thereby any attempt to pretend that a value specified as having
a range is futile, as at the end of the day a single value will
be stored which is indistinguishable from values otherwise specified
without corresponding representational support, for which scheme
does not require.
Received on Fri Dec 15 2006 - 11:45:29 UTC