Using a user-defined type as a primary key

Question

Suppose I have a system where I have metadata such as:

table: 
====== 
   key 
   name 
   address 
   ... 

Then suppose I have a user-defined type described as so:

datasource 
datasource-key 

A) are there systems where it's possible to have keys based on user-defined types?
B) if so, how do you decompose the keys into a form suitable for querying?
C) is this a case where I'm just better off with a composite primary key?

Answer 1

Use a composite primary key if this is what the model says

• if you have to decompose into components to query then you've already killed performance
• any opaque user defined type may give false duplicates because 2 different inputs may give same output

I've not tried it, but SQL Server will probably allow it. However, a primary key is an index so it may not, definitely not if the user type is neither deterministic nor schemabound

Although, I think I'm mixing up types and udfs in my thinking...