Re: referential integrity without trigger

In article <302F3CB4-1087-4AAD-A23A-C9AE1C3FDFD9(at)weisshuhn(dot)de>,
Alexander Presber <aljoscha(at)weisshuhn(dot)de> writes:

> Hello everybody,
> Assuming I want to empty and refill table A (with roughly the same
> content, preferrably in one transaction) and don't want to completely
> empty a dependent table B but still keep referential integrity after
> the commit.

> Without disabling A's on-delete-trigger B will be be emptied on
> commit, even when I inserted exactly the same data into A that I
> deleted an instant before. That is because the trigger gets called on
> commit, no matter if the deleted rows have "reappeared".

> If I disable the trigger, My referential integrity is most likely
> corrupted.
> Is there a clever, general scheme to "recheck" and enforce foreign
> key contraints, after the responsible triggers have been disabled and
> reenabled?

> I hope this makes sense to you.

Not quite? Why do you use an explicit trigger for checking
referential integrity? Can't you just use a foreign key with "ON
DELETE NO ACTION DEFERRABLE INITIALLY DEFERRED"?