Joe Conway <mail@joeconway.com> writes:
> Tom Lane wrote:
>> If we can't use an OID to identify a data type
>> completely, we're going to have lots of problems.
> You only really need two pieces of information to uniquely identify an
> array data type -- the OID of the (leaf-node) scalar elements, and the
> fact that what you have is an array.
Yes, but replacing one piece of data (type OID) with two (type OID plus
array flag bit) is going to be notationally horrible. What's more it
will force knowledge about the existence of array types into a lot of
places that don't have to care right now. For something that is
ultimately just a flange on the side of SQL, that doesn't seem like a
good tradeoff.
regards, tom lane