Richard A. O'Keefe wrote:
<...>


> The problem I pointed out is purely and simply this:
>     equality between rationals is compatible with rational hashes
> AND equality between floats is compatible with float hashes
> AND equality between rationals and floats is defined (required by the
>     ANSI standard) and behaves just the way a Fortran/Pascal/C hacker
>     would expect
> BUT rational/float equality is not consistent with rational/float hashes.
> 
> A collection with rational keys is not a problem.
> A collection with floating point keys is NOT a problem.
> A collection with mixed keys *IS* a problem.

Viewn in this light (together with snipped in depth explanations) I second
your proposal now.


Greetings,

Stephan

> 
> 
> 
> 


-- 
Stephan Rudlof (sr at evolgo.de)
   "Genius doesn't work on an assembly line basis.
    You can't simply say, 'Today I will be brilliant.'"
    -- Kirk, "The Ultimate Computer", stardate 4731.3