编写自定义聚合函数

Question

Lately I've been practicing and studying PL/pgSQL. I'm stuck on creating a custom aggregate function.

The code below works just fine, but I'm not able to write an aggregate function of this type:

SELECT my_aggregate_function(column) from table;

Here is the code of the "partially working custom aggregate":

CREATE OR REPLACE FUNCTION searchMinValue (numeric[]) RETURNS numeric   AS $$
  DECLARE 
     i numeric;
     minVal numeric;
  BEGIN
    minVal := $1[1];
    IF ARRAY_LENGTH($1,1) > 0 THEN --Checking whether the array is empty or not
  <<confrontoMinimo>>
   FOREACH i IN ARRAY $1 LOOP --Looping through the entire array, passed as parameter
       IF minVal >= i THEN
           minVal := i;
       END IF;
   END LOOP confrontoMinimo;
   ELSE
    RAISE NOTICE 'Invalid parameter % passed to the aggregate function',$1;
   --Raising exception if the parameter passed as argument points to null.
   RAISE EXCEPTION 'Cannot find Max value. Parameter % is null', $1
   USING HINT = 'You cannot pass a null array! Check the passed parameter';
END IF;
RETURN minVal;
END;
$$ LANGUAGE plpgsql;

CREATE AGGREGATE searchMinValueArray (numeric)
(
sfunc = array_append,
stype = numeric[],
finalfunc = searchMinValue,
initCond = '{}'
); 

 with w(v) as (select 5 union all select 2 union all select 3)
 select min(v) "Normal Aggregate", searchMinValueArray(v) "My Customed Aggregate" from w;

As I said before, I'd like to call my custom aggregate function in this way:

SELECT my_aggregate_function(column) from table;

where the table is Customers and the column is salary of type numeric .

Answer 1

You are fundamentally doing the right thing, but there are still some problems with your implementation:

1. Your aggregate will error out if the table is empty. Rather, it should return NULL .

2. If the first value you are aggregating is NULL , your aggregate will do the wrong thing. if

    minVal := $1[1]; 

   sets minVal to NULL , then all future comparisons minVal >= i will not be TRUE , and the end result will be NULL , which is not what you want.

3. You collect all values in an array, which will become quite big if you aggregate many rows. First this puts you in danger of running out of memory with the array, and then the performance of the aggregate will not be as good as it could be.

   It is much better to perform the comparison in SFUNC : start with INITCOND = NULL , use STYPE = numeric and perform the aggregation whenever a you process a new value. The relevant part of SFUNC would look like this:

    IF $1 IS NULL OR $1 > $2 RETURN $2; ELSE RETURN $1; END IF; 

   That way you don't need a FINALFUNC at all.