> > But executor can knows that somethig was already executed, we can mark
> > some already executed expressions in rewriter and not execute it again in
> > final executor... like:
> ...
> >
> > IMHO this is a good point for 7.2 ...
>
> But if instead of nextval() you had random(), would you still want to execute
> it
> only once ? And how should postgres know ?
Talking you still about RULEs?
...I don't undestand you. What is a 'NEW' in RULE? I (and probably more
users) expect that new data from tuple which go into original table. Right?
Not ... if you use sequence. IMHO it's not "feature" but nice bug that
crash your data integrity...
Karel