Methods of speeding up a MySQL Query which has 12 subqueries - mysql

Question

The following query returns the correct data but I'd like to see if there's a better way of doing this. The query should return the number of cases for each month within a 12 month period where a record exists within the past 2 months. The idea is to get number of accounts that ordered during the month in question and at least one of the previous 2 months. Also, please note that every value in the table for data_date will always be the 1st of the month.
SELECT
sum(
case
WHEN a.data_date = '2013-03-01'
and exists(
select 1 from sales mth1
where mth1.client_id = a.client_id
and
data_date BETWEEN '2013-01-01'
and
'2013-02-01'
)
then case_qty
ELSE 0 END
) AS M1 ,
sum(
case
WHEN a.data_date = '2013-04-01'
and exists(
select 1 from sales mth2
where mth2.client_id = a.client_id
and mth2.data_date BETWEEN '2013-02-01'
and
'2013-03-01'
)
then case_qty
ELSE 0 END
) AS M2 ,
sum( case WHEN a.data_date = '2013-05-01' and exists( select 1 from sales mth3 where mth3.client_id = a.client_id and mth3.data_date BETWEEN '2013-03-01' and '2013-04-01' ) then case_qty ELSE 0 END) AS M3 ,
sum( case WHEN a.data_date = '2013-06-01' and exists( select 1 from sales mth4 where mth4.client_id = a.client_id and mth4.data_date BETWEEN '2013-04-01' and '2013-05-01' ) then case_qty ELSE 0 END) AS M4 ,
sum( case WHEN a.data_date = '2013-07-01' and exists( select 1 from sales mth5 where mth5.client_id = a.client_id and mth5.data_date BETWEEN '2013-05-01' and '2013-06-01' ) then case_qty ELSE 0 END) AS M5 ,
sum( case WHEN a.data_date = '2013-08-01' and exists( select 1 from sales mth6 where mth6.client_id = a.client_id and mth6.data_date BETWEEN '2013-06-01' and '2013-07-01' ) then case_qty ELSE 0 END) AS M6 ,
sum( case WHEN a.data_date = '2013-09-01' and exists( select 1 from sales mth7 where mth7.client_id = a.client_id and mth7.data_date BETWEEN '2013-07-01' and '2013-08-01' ) then case_qty ELSE 0 END) AS M7 ,
sum( case WHEN a.data_date = '2013-10-01' and exists( select 1 from sales mth8 where mth8.client_id = a.client_id and mth8.data_date BETWEEN '2013-08-01' and '2013-09-01' ) then case_qty ELSE 0 END) AS M8 ,
sum( case WHEN a.data_date = '2013-11-01' and exists( select 1 from sales mth9 where mth9.client_id = a.client_id and mth9.data_date BETWEEN '2013-09-01' and '2013-10-01' ) then case_qty ELSE 0 END) AS M9 ,
sum( case WHEN a.data_date = '2013-12-01' and exists( select 1 from sales mth10 where mth10.client_id = a.client_id and mth10.data_date BETWEEN '2013-10-01' and '2013-12-01' ) then case_qty ELSE 0 END) AS M10 ,
sum( case WHEN a.data_date = '2014-01-01' and exists( select 1 from sales mth11 where mth11.client_id = a.client_id and mth11.data_date BETWEEN '2013-11-01' and '2013-12-01' ) then case_qty ELSE 0 END) AS M11
FROM sales as a
INNER JOIN Products AS P ON P.product_id = a.product_id
WHERE a.client_id IN ('123')
AND a.data_date BETWEEN '2013-03-01' AND '2013-12-01' AND a.case_qty > 0;
Here's a screen shot of the explain
Here's a screen shot of the indexes

Answer 1

change
data_date BETWEEN '2013-01-01' and '2013-02-01'
to
data_date in ('2013-01-01', '2013-01-02',.....,...'2013-02-01' )
but i would turn this whole query into sum cases then use a wrapper to pull out the accounts i need.

Answer 2

From your query it seems like you want the years sales for a product in an entire year. Or something to that affect. So from your data you are almost trying to create a pivot. I had a simlar problem and solved it like this
select
p.product_id,
s.client_id,
sum(case when DATE_FORMAT(a.data_date,'%Y%m') in (201301) then ifnull(case_qty,0) else 0 end) period1,
sum(case when DATE_FORMAT(a.data_date,'%Y%m') in (201301, 201302) then ifnull(case_qty,0) else 0 end) period2,
sum(case when DATE_FORMAT(a.data_date,'%Y%m') in (201301, 201302, 201303) then ifnull(case_qty,0) else 0 end) period3,
sum(case when DATE_FORMAT(a.data_date,'%Y%m') in (201302, 201303, 201304) then ifnull(case_qty,0) else 0 end) period4,
sum(case when DATE_FORMAT(a.data_date,'%Y%m') in (201303, 201304, 201305) then ifnull(case_qty,0) else 0 end) period5,
.
.
.
up to period12
from
sales s
where
a.data_date between '2013-03-01' and '2013-12-31' and
s.client_id = some_value
group by
p.product_id,
s.client_id
Note that the performance of the query has increased drastically since you are only doing a single scan of your sales table (depending on what's in your where clause and indexes). To speed it up you would need indexes on say client_id and data_date for example.
Ideally this query would be run for a report or something where the start and end data is fixed and all the user can change is the year of that date. I wasn't sure if you wanted to group by product_id since in your query you aren't, but you can always remove it to get total per client.
I have adjusted the query based on my understanding in your comments. I am not 100% sure if you can use the keyword 'in' in a case statement. You could alternatively have more case statements.