[Matrix-SIG] Numeric Nits--a slightly different proposal

[David J. C. Beach]

On Thu, Jul 15, 1999 at 06:33:35PM -0400, Rick White wrote:
> >An expression involving two incompatible
> >forced objects would raise an exception.
> 
> I can't agree with that, though.  I'm afraid that will result
> in more ugly, hard-to-read code with lots of explicit casts.
> I believe that an operation between two forced objects should
> obey the normal promotion rules, with the lower forced type
> being promoted to the higher forced type.

Yes.  Then maybe there should be a ForcedForcedFloat32 which is
"stronger" (more forceful?) than a plain ol' ForcedFloat32.  Doesn't
this problem just start to repeat itself?  Do numeric types need a
"degree of force" (e.g., a Force5Float32 beats out a Force2Float64, so
the result is then typed as a Force5Float32 because 5 is "stronger"
than 2)?  Where is this going?

Will adding forced types break number-crunching code that counts on
working with a greater precision internally?  Are there any examples
of packages that use such a mechanism already in existence?

Forgive me for raising this point, I've mostly been a passive observer
of NumPy... This issue caught my attention.

Dave

-- 
David J. C. Beach
<beach@verinet.com>