[r6rs-discuss] Hashes in the number syntax

[dyb at cs.indiana.edu]

The floating-point printing algorithm described in "Printing
floating-point numbers quickly and accurately" (Burger/Dybvig PLDI '96)
uses hash digits to indicate less than the requested amount of precision. 
This algorithm is used by Chez Scheme's implementation of format, so that,
for example:

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

So while I agree that the hash digits are not particularly useful for
data-entry input, they are useful for output.  Furthermore, if the output
routines can produce hash digits, it's also useful for the reader to
accept them.

Kent