Colin Paul Adams wrote:
> It seems that truncated_to_integer_64 is not enough in itself, and (in
> ISE 6.0) isn't sufficiently accurate anyway.
>
> My problem is converting 1e100 to an integer.
>
> The debugger displays this as something like 1.000000000000000006e100
> (I have not counted the the exact number of zeros).
>
> Calling truncated_to_integer_64 on this produces -9223372036854775808.
>
> The alternative (which I have adopted) is to convert to MA_DECIMAL
> first, using make_from_string ({DOUBLE}.out), works, but gives an
> inaccurate value (because of the 6). I guess this is acceptable, as
> doubles are inherently inaccurate.
>
It isn't a fault with ISE, but with computer arithmetic in general.
64-bit integers cover the range -9,223,372,036,854,775,808 to
+9,223,372,036,854,775,807, so 1e100 is not representable as an integer
(you would need 512-bit integers to represent 1e100).
So your method is probably the best general, unless you wanted to test
if your double is smaller than max_int, and treat accordingly if above
or below.
Mark
|
