"codeWarrior" <gpatnude@xxxxxxxxxxx> writes:
> BTW: I was thinking more along the lines of:
> SELECT PT.*, PA.* FROM pg_attribute PA
> JOIN pg_type PT ON PA.attrelid = PT.typrelid
> JOIN pg_class PC ON PC.oid = PA.attrelid
> WHERE PT.typnamespace = 2200 AND PA.attnum > 0 AND PC.relkind IN ('r', 'c',
> 'v');
> Perhaps this should be made into a view in the information_schema.....
It seems like the COLUMN_UDT_USAGE view ought to handle this ...
except perhaps that COLUMN_UDT_USAGE for some reason substitutes the
underlying type if the column type is a domain. I can't see the
justification for that --- Peter, what part of the spec led you to do
it that way? There is probably more than one way to define the spec's
notion of "user-defined types" in terms of Postgres notions, but I
don't see where you got this particular definition.
regards, tom lane
