The cost of the operation is likely related to the # of qualifying rows * the number of operations. You may be able to reduce the time slightly if you can simplify the number of steps in the expression. Here I have a compound date field I can express as a between clause rather than the compound logic show below.
| Code: |
mysql> select sum(case when lo_orderdate between 19920101 and 19920630
-> then 1 else 0 end * lo_extendedprice) 1992_Q1_Q2_sum, count(*)
-> from lineorder where lo_orderdate <= 19921231;
+--------------------+----------+
| 1992_Q1_Q2_sum | count(*) |
+--------------------+----------+
| 138804076164817.00 | 91248844 |
+--------------------+----------+
1 row in set (5.90 sec)
mysql>
mysql>
mysql> select sum(case when (year = 1992 and month <= 6 ) and (year = 1992 and month >= 1)
-> then 1 else 0 end * lo_extendedprice) 1992_Q1_Q2_sum, count(*)
-> from lineorder where lo_orderdate <= 19921231;
+--------------------+----------+
| 1992_Q1_Q2_sum | count(*) |
+--------------------+----------+
| 138804076164817.00 | 91248844 |
+--------------------+----------+
1 row in set (7.81 sec)
mysql> select sum(lo_extendedprice) 1992_Q1_Q2_sum, count(*)
-> from lineorder where lo_orderdate <= 19921231;
+--------------------+----------+
| 1992_Q1_Q2_sum | count(*) |
+--------------------+----------+
| 279113147753549.00 | 91248844 |
+--------------------+----------+
1 row in set (2.29 sec)
mysql> select sum(1 * lo_extendedprice) 1992_Q1_Q2_sum, count(*) from lineorder where lo_orderdate <= 19921231;
+--------------------+----------+
| 1992_Q1_Q2_sum | count(*) |
+--------------------+----------+
| 279113147753549.00 | 91248844 |
+--------------------+----------+
1 row in set (4.54 sec)
|
There is definitely some overhead for inclusion of even a trivial function expression.
I'm not sure what else to recommend for you without access to the data model/queries.
Thanks - Jim Tommaney
Chief Product Architect
.... 
...........
