After some experimenting and FAQ/docs re-reading...
* Joan Picanyol <lists-pgsql@biaix.org> [20040316 22:27]:
> create table a (id serial not null primary key, data int);
> create table b (id serial not null primary key, data int\
> ref1 int references a not null, ref2 int references a not null,\
> ref3 int references a);
> create view vw (data_a, data_b1, data_b2, data_b3) as select \
> (b.data as data_b, a1.data as data_a1, a2.data as data_a2,\
> a3.data as data_a3) from b, a a1, a a2, a a3 where b.ref1=a1.id\
> and b.ref2=a2.id and b.ref3=a3.id;
>
> And now I want to create an insert rule into vw with all my foreign keys
> properly set. I can't think of anything better that doing first the
> three inserts into a and then use nextval()-1, nextval()-2 etc for the
> fk values, but it looks really ugly (even though I could wrap up within
> a BEGIN(...)COMMIT).
I found out that I can't
create rule my_rule as on insert to vw instead do (begin; ...; commit;);
Together with FAQ 4.15.3 this makes me think that simultaneous inserts
are not an issue, however FAQ 4.15.4 makes me doubt again. In Chapter 34
it is said that
"The rule system is located between the parser and the planner."
however, this doesn't mean much to me. Can I assume that an insert to a
view is atomic regardless how many inserts it get rewritten onto?
tks
--
pica
