Well, this is simply by definition. There can not be two "primary" conditions, just like there can not be two "latest" versions.
Every table can contain more than one unique
keys, but if you decide to have a primary key, this is just one of these unique
keys, the "one" you deem the "most important", which identifies every record uniquely.
If you have a table and come to the conclusion that your primary key does not uniquely identify each record (also meaning that there can't be two records with the same values for the primary key), you have chosen the wrong primary key, as by definition, the fields of the primary key must uniquely define each record.
That, however, does not mean there can be no other combination of fields uniquely identifying the record! This is where a second feature kicks in: referential integrity.
You can "link" tables using their primary key. For example: If you have a Customer
table and an Orders
table, where the Customers
table has a primary key on the customer number and the Orders
table has a primary key on the order number and the customer number, that means:
- Every customer can be identified uniquely by his customer number
- Every order is uniquely identified by the order number and the customer number
You can then link the two tables on the customer number. The DB system then ensures several things, among which is the fact that you can not remove a customer who has orders in your database without first removing the orders. Otherwise, you would have orders without being able to find out the customer data, which would violate your database's referential integrity.
If you had two primary keys, the system would not know on which to ensure referential integrity, so you'd have to tell the system which key to use - which would make one of the primary keys more important, which would make it the "primary key" (!) of the primary keys.