The FINAL function

The FINAL aggregate support function performs the post-iteration tasks for the user-defined aggregate.

The following table summarizes possible post-iteration tasks.

Table 1. Post-iteration tasks for the final aggregate support function
Post-iteration task	More information
Type conversion of the final state into the return type of the user-defined aggregate	Return an aggregate result different from the aggregate state
Post-iteration calculations that the user-defined aggregate might need	Aggregate support functions that the algorithm requires
Deallocation of memory that the INIT aggregate support function has allocated	Using a pointer-valued state for a UDA

The FINAL function is an optional aggregate support function. If your user-defined aggregate does not require one of the tasks in Table 1, you do not need to define a FINAL function. When you omit the FINAL function from your user-defined aggregate, the database server returns the final aggregate state as the return value of the user-defined aggregate. If this state does not match the expected data type of the aggregate return value, the database server generates a data type mismatch error.

Important: In general, the FINAL support function must not deallocate the aggregate state. Only for a pointer-valued state (in which the aggregate support functions must handle all state management) does the FINAL support function need to deallocate the state. For more information, see Using a pointer-valued state for a UDA.

To declare a FINAL function as a C function, use the following syntax:

agg_type final_func(final_state)
   agg_state final_state;

agg_type: The data type of the aggregate result, which is what the user-defined aggregate returns to the SQL statement in which it was invoked.
final_state: The final aggregate state, as previous calls to the INIT and ITER functions have generated.
final_func: The name of the FINAL aggregate support function.

In the execution of a UDA, the database server calls the FINAL function after all iterations of the ITER function are complete.

The following code fragment shows the aggregate support functions that handle an INTEGER argument for the SQSUM user-defined aggregate (which Table 1 describes).

Figure 1. The FINAL aggregate support function for SQSUM1 on INTEGER

/* SQSUM1 FINAL support function on INTEGER */
mi_integer final_sqsum1(state)
   mi_integer state;
{
   /* Calculate square of sum */
   state *= state;

   return (state);
}

For other aggregate support functions of SQSUM1, see Figure 1, Figure 1, and Figure 1.