On Fri, Sep 9, 2022 at 5:21 AM Amit Kapila <amit.kapila16@gmail.com> wrote:
> So, why shouldn't a "FOR ALL TABLES IN SCHEMA" publication follow a
> similar behavior?
Surely that is not the same case at all. If you're publishing
everything, there's no point in also having a specific list of things
that you want published, but when you're publishing only some things,
there is. If my wife tells me to wash everything in the laundry basket
and also my nice pants, and I discover that my nice pants already
happen to be in the laundry basket, I do not tell her:
ERROR: my nice pants are already in the laundry basket
It feels like a mistake to me that there's any catalog representation
at all for a table that is published because it is part of a schema.
The way a feature like this should work is that the schema should be
labelled as published, and we should discover which tables are part of
it at any given time as we go. We shouldn't need separate catalog
entries for each table in the schema just because the schema is
published. But if we do have such catalog entries, surely there should
be a difference between the catalog entry that gets created when the
table is individually published and the one that gets created when the
containing schema is published. We have such tracking in other cases
(coninhcount, conislocal; attinhcount, attislocal).
In my opinion, this shouldn't have been committed working the way it does.
--
Robert Haas
EDB: http://www.enterprisedb.com
