"Ross J. Reedstrom" <reedstrm@rice.edu> writes:
>> Well, you clearly don't want to be fooled by temp relations. I was
>> sorta visualizing a check based on relation OIDs instead of names...
> Well, when I did a test implementation of OID filenames, lo these many
> moons ago, I hacked around this problem by adding the (fixed) shared
> system table oids to the static array that is searched for matches by
> IsSharedSystemRelationName.
> Admittedly, a hack, but it got past all the regresion tests.
Not a hack at all, IMHO, since all the shared system rels have
nailed-down OIDs. It's pure historical artifact that
IsSharedSystemRelationName wasn't IsSharedSystemRelationOID in the
first place.
We probably want to be thinking about merging the "shared system
relation" concept together with the "tablespace" concept once we start
to implement tablespaces. But for now, I think testing the OIDs is
fine.
regards, tom lane