Oracle SQL Tuning with function-based indexes Oracle Tips by Burleson Consulting

Oracle function-based indexes are a super silver bullet for optimizing your SQL to do a minimum amount of I/O to get the rows that you need.  Here are details on using function based indexes. 

Issues with function-based indexes

One of the most common Oracle SQL tuning problems are full-table scans that are imposed by invalidating a WHERE clause predicate with an Oracle built-in Function (BIF).  This problem is especially prevalent when constraining SQL queries for date ranges because of the intrinsic encryption of the relational database DATE datatype.

For example, these WHERE clause predicates might invoke an unexpected full-table scan:

WHERE trunc(ship_date) > trunc(sysdate-7);

WHERE to_char(ship_date,’YYYY-MM-DD’) = ‘2004-01-04’;

Even though the ship_date column may have an index, the trunc and to_char built-in functions will invalidate the index, causing sub-optimal execution with unnecessary I/O:

1. A less selective index would be used.

2. A full-table scan might be performed.

To avoid the index invalidation issues you have several options:

1. Fix Predicate - You can rewrite the query to leave the ship_date predicate alone:

WHERE ship_date >= trunc(ship_date-7) + 1;

WHERE ship_date = to_date(‘2004-01-04’,’YYYY-MM-DD’);

2. Call FBI - Create a function-based index on the predicate:

Create index trunc_ship_date on product(trunc(ship_date));

Create index char_ship_date on product(to_char(ship_date,’YYYY-MM-DD’);

3. Add Column - You could add a redundant column to the table. This is an old Oracle7 trick that was used before the invention of function-based indexes.

alter table product add trunc_ship_date varchar 40;

update product set trunc_ship_date = trunc(ship_date);

In sum, function-based indexes (FBIs) are a Godsend for the Oracle DBA who has to remove unnecessary full-table scans or fix sub-optimal index choices. Using an FBI can positively effect all SQL statements that have the matching predicate and speed-up an entire database.

Once an FBI is created, you need to create CBO statistics, but beware that there are numerous bugs and issues when analyzing a function-based index.  See these important notes on statistics and function-based indexes.

After creating a function-based index, you want to re-gather table statistics and get extended statistics on the function.

Create an FBI on emp_nbr column with NULL values:

create index
    emp_null_emp_nbr_idx
on
    emp
    (nvl(ename,o));

To make the function-based index work, we must now analyze the index.  Depending you your release, these methods are appropriate:

 EXEC DBMS_STATS.gather_index_stats('EMP', ‘emp_null_emp_nbr_idx’);  

In this case, we analyze the function-based index using the method_opt argument set to FOR ALL HIDDEN COLUMNS:

 exec dbms_stats.gather_table_stats(
    ownname=>null,
    tabname=> 'EMP',
    estimate_percent=>null,
    cascade=>true,
    method_opt=> ‘FOR ALL HIDDEN COLUMNS SIZE 1′
);     

In Oracle 11g and beyond, Oracle recommends analyzing “extended” statistics on function-based indexes:

 begin
   dbms_stats.gather_table_stats (
      ownname    => 'SCOTT',
      tabname    => 'EMP',
      method_opt => 'for all columns size skewonly for columns (nvl(ename,o))'  
);
end;

Also, see these related notes on FBI:

 

If you like Oracle tuning, you may enjoy my new book "Oracle Tuning: The Definitive Reference", over 900 pages of BC's favorite tuning tips & scripts.  You can buy it direct from the publisher for 30%-off and get instant access to the code depot of Oracle tuning scripts.