🚀 go-pugleaf

In article <1992Sep30.034813.2589@waggen.twuug.com> alpha@waggen.twuug.com (Joe Wright) writes:
>Please remember that floats (and doubles) are approximates.
>The value of time_t must be precise...

It is quite possible to manipulate precise values using doubles, if you
have well-behaved floating-point hardware (e.g. IEEE-standard).  If your
number is an integer that will fit within a double's mantissa field,
which is usually 50ish bits, the representation *is* precise.  Properly
designed floating-point hardware will give precise results when such
quantities are added, subtracted, compared, etc., given that neither final
nor intermediate results are big enough to exceed the mantissa field.
IEEE-standard FP hardware is *required* to do this right.

We do a lot of accounting software using awk, which does all its arithmetic
using doubles.  (The crucial trick is to keep everything in cents rather
than fooling around with fractional dollars.)  I assure you, we get precise
results:  our department's accountant notices if a penny is missing.

Changing time_t to a double does have problems, but this isn't one of them.
-- 
There is nothing wrong with making      | Henry Spencer @ U of Toronto Zoology
mistakes, but... make *new* ones. -D.Sim|  henry@zoo.toronto.edu  utzoo!henry