On 14/09/2013 06:28, I wrote:
> 2) Checking row_count for a statement is ugly and cumbersome, so
> often it just isn't checked. I often use RETURNING TRUE INTO STRICT _OK
> for DML, but that a) requires an extra variable, and b) isn't possible
> if 0 rows affected is not an error in the application logic.
The b) part here wasn't exactly true; you could use RETURNING TRUE INTO
OK as INSERT/UPDATE/DELETE with RETURNING .. INTO raises an exception if
it returns more than one row, but it's even more convoluted than the
RETURNING TRUE INTO STRICT _OK, and it's repetitive.
Regards,
Marko Tiikkaja