Peter Eisentraut <peter_e@gmx.net> writes:
> (The workaround here could be to convert numerics that are wider
> than floats to `infinity' :-)
Hmm, that's actually not a bad idea. Infinity is a pretty portable
notion these days, what with nearly everyone toeing the IEEE float
line ... so we could have numeric->float generate a NaN if possible
and only resort to an elog() on machines without NaN.
OTOH, no mathematician will accept the notion that 1e1000 is the
same as infinity ;-). Mathematical purity would probably favor
the elog.
Comments anyone?
regards, tom lane