Question

The following query is working as expected.
But the summary table from where the data is selected is too large. The query takes a lot of time.

SELECT s.date_time AS date_time,SUM(s.requests) AS sum_requests,SUM(s.impressions) AS sum_views,SUM(s.clicks) AS sum_clicks,SUM(s.conversions) AS sum_conversions,
SUM(s.total_revenue) AS sum_revenue,SUM(s.total_revenue)  AS sum_spend,
SUM(s.total_basket_value) AS sum_bv,SUM(s.total_num_items) AS sum_num_items,SUM(s.total_pubrevenue) AS sum_pubrevenue
FROM ox_data_summary_ad_hourly AS s 
INNER JOIN ox_banners AS d ON (d.bannerid=s.ad_id) 
INNER JOIN ox_campaigns AS m ON (m.campaignid=d.campaignid) 
INNER JOIN ox_clients AS a ON (a.clientid=m.clientid) 
LEFT JOIN ox_zones AS z ON (z.zoneid=s.zone_id) 
LEFT JOIN ox_affiliates AS p ON (p.affiliateid=z.affiliateid)
WHERE a.agencyid = 3 AND z.zone_type = 'application' AND z.delivery = 8 
AND s.date_time>='2012-07-01 00:00:00' AND s.date_time<='2012-07-01 23:59:59' 
GROUP BY date_time

I need to simplify the joins. Is there any way to create a temporary table for all joins and then join that table to the base table.
This is what I have tried so far.

drop table if exists test.helper;
create table test.helper 
select d.network_type , d.is_network, d.bannerid, p.agencyid, z.zone_type, z.delivery , z.zoneid
FROM ox_banners as d
INNER JOIN ox_campaigns AS m ON (m.campaignid=d.campaignid) 
INNER JOIN ox_clients AS a ON (a.clientid=m.clientid) 
left JOIN ox_agency as oa ON (oa.account_id = a.clientid)
left JOIN ox_affiliates AS p ON (p.agencyid = oa.agencyid)
left JOIN ox_zones AS z ON (z.affiliateid = p.affiliateid);

alter table test.helper add key(bannerid, zoneid);

But it does not return the correct results when joined with the original table.

SELECT s.ad_id, s.date_time AS date_time,SUM(s.requests) AS sum_requests,SUM(s.impressions) AS sum_views,SUM(s.clicks) AS sum_clicks,SUM(s.conversions) AS sum_conversions,
SUM(s.total_revenue) AS sum_revenue,SUM(s.total_revenue)  AS sum_spend,
SUM(s.total_basket_value) AS sum_bv,SUM(s.total_num_items) AS sum_num_items,SUM(s.total_pubrevenue) AS sum_pubrevenue
FROM ox_data_summary_ad_hourly AS s 
inner JOIN test.helper  AS d ON (d.bannerid=s.ad_id )
where d.agencyid = 3 AND d.zone_type = 'application' AND d.delivery = 8  and s.date_time>='2012-07-01 00:00:00' AND s.date_time<='2012-07-01 23:59:59' 
GROUP BY date_time

Answer 1

Firstly, I’d replace the LEFT JOINs with INNER JOINS. Since you have a WHERE clause on ox_zones, the null values created by the LEFT JOIN will drop out anyway. Outer joins are obviously expensive, and should be avoided where possible.

On a reasonable, modern RDBMS, second-guessing the query planner by creating temporary tables is often counter-productive. Are all your join conditions indexed? Are the columns referenced in your WHERE clause indexed?