Re: UUID's as primary keys

Martijn van Oosterhout wrote:
> On Wed, Jun 28, 2006 at 10:22:45AM +0200, Thomas Hallgren wrote:
>> Thanks. That would of course work but at the same time it increases the
>> complexity of my app. Yet another component to install and keep track
>> of. It's also a bit of an overkill since the only thing I need is an
>> opaque bit storage. Why is it that PostgreSQL lack a fixed length binary
>> type similar to the RAW type in Oracle? ISTM that could be very useful
>> and not very hard to implement.
>
> AIUI, it can't be done because of a basic rule of the type system: the
> typmod can't be necessary to interpret the binary representation of a
> value. For something like RAW(16) the type would be the oid for "raw"
> and the typmod would be 16. However, when reading the value from a disk
> page, you're not given the typmod, so you have no way of determining
> the length.
>
OK. I thought you always had a type descriptor handy when reading the binary representation.
I've noticed that the typmod is expected in some receive functions (bpcharrecv and
numeric_recv for instance). Are you saying that there are times when you don't use that?

Regards,
Thomas Hallgren