Archives of the TeradataForum

Message Posted: Wed, 17 Mar 2004 @ 10:11:56 GMT

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

Thats a fair point. You would need to manage the signs separately, and the product calculation would become:

zeroifnull(exp(sum(ln(abs(nullifzero(col)))))) as n_product

Note this also accomodates zeros in this form.

The sign of the product is positive if the number of negative terms in the product is even, else the product is negative, so a further column of:

sum(
  case
     when col<0 then 1
     else 0
  end
) as n_neg,

to count the number of negative elements in col and then the sign itself should be:

-1*(n_neg mod 2) as signum,

so, finally, the product would be

signum * n_product as product

I hope that generalises the result to a more useful state.

Regards

John

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