On 4/27/06, Strobhen <strobhen@xxxxxxxxx> wrote:
Hey, I am trying to figure out some unexpected behavior in Postgresql. When I create a rule that fires on a table after an update, and if that rule has a SELECT statement in it, it seems to be attempting to fire (on an empty set) regardless of how the conditional evaluates after an update. The result being that if I run an update on a table with such a rule, instead of getting a message along the lines of "UPDATE (# of rows)" I get the column names of the select statement with no rows and the message "row number -1 is out of range 0..-1". So first off, is having a select statement (I'm actually trying to run a function) inside a rule that fires on an update considered bad practice? I could do this through a trigger, but a rule just seems more natural.
<snip>
When that rule should never fire (the id hasn't changed). If I change the conditional of the rule to something that must always be false (like false, or 1 = 0), it will still behave in this manner.
Does anyone know what's going on here? I'm experiencing an identical situation, and it doesn't seem logical. If it evaluates to false, why on earth is the function result set attempting to be returned? Maybe not a bug, but definitely unexpected behavior Thanks, Steve
