Re: Assignment of valid collation for SET operations on queries with UNKNOWN types.

Michael Paquier <michael(dot)paquier(at)gmail(dot)com> writes:
> As unknown is a pseudo type, I don't think you need
> TYPCATEGORY_UNKNOWN in pg_type.h or even the mention to the unknown
> type in catalogs.sgml as that becomes a pseudo-type.

I wondered whether to remove TYPCATEGORY_UNKNOWN but thought it was an
unnecessary change. "unknown" is different from the other pseudotypes
in that type resolution treats it very specially, so it doesn't seem
unreasonable for it to continue to have its own typcategory. Also,
since type resolution sometimes takes into account whether types are
of the same category or not, I'm a bit worried about whether moving
"unknown" into the pseudotype category might have unexpected side effects.

> The table of Pseudo-Types needs to be updated as well with unknown in
> datatype.sgml.

Check.

> For domains, it is still necessary to add an extra check in pg_upgrade
> and fail the upgrade if one of the domains declared uses the type
> unknown. Domains are not listed in pg_class, and are only present in
> pg_type. If you don't do that, the binary restore would just fail.

Meh. I think this would largely be a useless check --- who would
create such a domain? Also, it's not like the system will crash and
burn if we don't check for it, it will just fail a bit further into
the pg_upgrade process. That's not like the matview situation where
it would appear to go through and then you'd have a broken matview.

regards, tom lane