Archives of the TeradataForum

Message Posted: Tue, 09 Sep 2003 @ 13:17:49 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

First suggestion: buy a round of stiff drinks for everybody. Unfortunately, the techniques for rounding depend on the data types you're working with. If your inputs and outputs are floating-point, the floating-point unit may produce results that surprise you; in such a case, you will probably have to use a value for "one-half" that is slightly larger than 0.5 in order to make things come out predictably. (This technique is used in the Perl module Math::Round, available from CPAN.)

If your inputs and outputs are packed decimal or some other fixed-point form, the best approach is probably this:

intermediate = input * 10000; /* Multiply by the appropriate power of 10 */

rounded = floor(intermediate + 0.5); /* The usual "round up" formula */

final = rounded / 10000; /* Divide by the appropriate power of 10 */

These examples are puzzling. Is the present rounding mistakenly dividing things by 10?

	<Prev	Next>		<<First	<Prev	Next>	Last>>