Re: How do indexes work on non-unique values ?

>> I don't understand how this works for indexes on foreign keys,
which I understand to be beneficial or how it would work for my initial example <<

This is a product specific question and you posted it in the Theory newsgroup. In fact, in Standard SQL, there are no indexes -- we felt that this was too implementation specific to put into a language description.

Sybase and other products build a PK-FK index that "pre-joins" the referenced and referencing tables -- big boost in performance. SQL Server has only one index on the PK table and no way of building a multi-table index.

Nucleus uses compressed bit vectors for the entire database, so all columns are automatically indexed. Teradata uses hashing, so the longest possible fetch takes two seeks. Etc. Received on Fri Jan 24 2003 - 15:06:00 CST