I googled this a lot many times but I didn't get the exact explanation for the same.
I am working on a complex database structures (in Oracle 10g) where I hardly have a primary key on one single column except for the static tables.
Now my question is consider a composite primary key ID (LXI, VCODE, IVID, GHID). Since it's a primary key, Oracle will provide a default index.
Will I get ONE (system generated) single index for the primary key itself or for its sub-columns also?
Asking this because I am retrieving data (around millions of records) based on individual columns as well. Now if system generates the indices for the individual columns as well. Why my query runs pretty faster than how it actually runs when I explicitly define indices for each individual column.
Please give a satisfactory answer
Thanks in advance
A primary key is a non-NULL unique key. In your case, the unique index has four columns, LXI, VCODE, IVID GHID in the order of declaration.
If you have a condition on VCODE but not on LXI, then most databases would not use the index. Oracle has a special type of index scan called the "skip scan", which allows for this very situation. It is described in the documentation.
I would expect an index skip scan to be a bit slower than an index range scan on individual columns. However, which is better might also depend on the complexity of the where clause. For instance, three equality conditions on VCODE, IVID and GHID connected by AND might be a great example for the skip scan. And, such an index would cover the WHERE clause -- a great efficiency -- and better than one-column indexes.
As a note: index skip scans were introduced in Oracle 9i, so they are available in Oracle 10.