How can I add control flow into a mysql function? - mysql

Question

I want to optimise a mysql 5.7 function that reads settings from a table. the function returns a 1 or 2 if the date parsed in is in 'semester 1' or 'semester 2'. the dates for semester 1 and 2 change each year.
we have confirmed that dateIn is a valid date.
the function is:
DELIMITER //
CREATE function getSemester (dateIN date)
RETURNS INT DETERMINISTIC
BEGIN
DECLARE sem int;
select if( dateIN < a.mindate,1,2) into sem
from (SELECT min(date(value)) mindate FROM `settings` WHERE name = CONCAT(‘sem2_‘,year(dateIN),‘_start’) ) a;
return sem;
END//
DELIMITER ;
settings is defined as:
CREATE TABLE `settings` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`name` varchar(100) CHARACTER SET utf8 COLLATE utf8_unicode_ci NOT NULL,
`value` varchar(255) CHARACTER SET utf8 COLLATE utf8_unicode_ci NOT NULL,
PRIMARY KEY (`id`),
KEY `name` (`name`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
the settings data is:
INSERT INTO `mdl_sap_settings` (`id`, `name`, `value`)
VALUES
(4, 'sem2_2012_start', '2012/7/16'),
(15, 'sem2_2013_start', '2013/7/1'),
(25, 'sem2_2014_start', '2014/6/30'),
(29, 'sem2_2015_start', '2015/6/29'),
(37, 'sem2_2016_start', '2016/6/27'),
(42, 'sem2_2011_start', '2011/7/16'),
(50, 'sem2_2017_start', '2017/6/26'),
(56, 'sem2_2018_start', '2018/6/25'),
(63, 'sem2_2019_start', '2019/6/24');
the issue is the function is slow quite slow when called on 20,000 rows. I thought to optimise it by having some sort of flow control in the function something like:
if (year(dateIN) = 2012)
{
return dateIN < '2012-07-16' ? 1 : 2;
}
if (year(dateIN) = 2013)
{
return dateIN < '2013-07-01' ? 1 : 2;
}
... etc.
We need to keep the select as is query because if the code is not maintained we want it to return the correct values.
i was just wondering if this sort of control flow is possible in a mysql function, or is there an alternative way to optimise the function?

Answer 1

I don't have 20k records to test with but if I were you, I'd rewrite your function like this and see if that helps.
DELIMITER //
CREATE function getSemester (dateIN date)
RETURNS INT DETERMINISTIC
BEGIN
DECLARE sem int;
DECLARE nameByDate varchar(255);
SET nameByDate = (SELECT CONCAT('sem2_',year(dateIN),'_start') );
select if( dateIN < a.mindate,1,2) into sem
from (SELECT min(date(value)) mindate FROM `test`.`settings` WHERE name = nameByDate ) a;
return sem;
END//
DELIMITER ;
I'd avoid adding a concat in the where clause as there is a performance hit on large dataset. So I moved it out of where clause and assigned it to a variable once. If that doesn't help you can also try the query with explain and see if that provides any hints.
EXPLAIN select getSemester('2019/6/24');
Good luck.

Answer 2

With 20,000 rows in the results table, the my query run without the getSemester function in 157 msec.
with the version of getSemster in the question it was taking 1.1 seconds.
with this optimisation the query was running in 0.9 seconds.
DELIMITER //
CREATE function getSemester (dateIN date)
RETURNS tinyint DETERMINISTIC
BEGIN
DECLARE sem tinyint;
DECLARE sem2_start_label char(16);
set sem2_start_label = CONCAT('sem2_',year(dateIN),'_start');
select case year(dateIN)
when 2012 then
case when dateIN < date('2012-07-16') then 1 else 2 end
when 2013 then
case when dateIN < date('2013-07-01') then 1 else 2 end
when 2014 then
case when dateIN < date('2014-06-30') then 1 else 2 end
when 2015 then
case when dateIN < date('2015-06-29') then 1 else 2 end
when 2016 then
case when dateIN < date('2016-06-27') then 1 else 2 end
when 2017 then
case when dateIN < date('2017-06-26') then 1 else 2 end
when 2018 then
case when dateIN < date('2018-06-25') then 1 else 2 end
when 2019 then
case when dateIN < date('2019-06-24') then 1 else 2 end
else
case when dateIN < (SELECT min(date(value)) mindate FROM `settings` WHERE name = sem2_start_label ) then 1 else 2 end
end into sem;
return sem;
END//
DELIMITER ;
what i did was only select the date from the settings table when it was unknown -- seems to give a small optimisation.