I have this complex query which produces 3744 rows in about 50ms.
SELECT
srl.event_id as eid
, srl.race_num as rnum
, bts.boat_id as bid_id
, srl.series_year as yr
, srl.id as id
, IFNULL(rfi.fleet,fleet_def) as flt_old,flt_match,s.series_id as sid
, s.series_year as syr
,IFNULL(ovr_pts,POINTS('4',IFNULL(ovr_place,place),num_start)) as points
FROM
(SELECT en1.boat_id,en1.boat_name,MAX(fleet) as fleet_def FROM entries en1
JOIN series_race_list srl1 ON srl1.event_id=en1.event_id
AND srl1.series_year=en1.race_year
LEFT JOIN entries_race er1 ON en1.boat_id= er1.boat_id
AND srl1.event_id=en1.event_id
AND srl1.series_year =en1.race_year
WHERE srl1.series_id ='3' AND srl1.series_year ='2012'
AND en1.entry_deleted='N'
GROUP BY boat_id) bts
JOIN series_race_list srl LEFT JOIN series as s ON s.series_id=srl.series_id
AND s.series_year =srl.series_year
LEFT JOIN entries as en ON srl.event_id=en.event_id
AND srl.series_year =en.race_year AND bts.boat_id =en.boat_id
LEFT JOIN entries_race er ON er.race_id= srl.event_id AND er.race_num=srl.race_num
AND er.yr = srl.series_year AND bts.boat_id =er.boat_id
LEFT JOIN event_race_info as eri ON eri.race_id= srl.event_id
AND eri.race_num=srl.race_num AND eri.yr = srl.series_year
ANd er.line=eri.line AND status REGEXP 'prelim|final'
LEFT JOIN race_results as rr ON srl.event_id=rr.race_id
AND srl.race_num= rr.race_num AND srl.series_year =rr.yr
AND bts.boat_id= rr.boat_id AND checked_in='Y'
LEFT JOIN race_fleet_info as rfi ON rfi.race_id= srl.event_id
AND rfi.yr=srl.series_year AND srl.race_num= rfi.race_num
AND rfi.fleet=rr.flt AND complete='Y'
LEFT JOIN series_pts_override as spo ON srl.id =spo.id AND en.boat_id =spo.bid
WHERE s.series_id ='3' AND s.series_year ='2012' AND approved ='Y'
Sorry for the length. As I said this query executes in around 50ms. Now I want to use this data and perform queries on this 3744 row result. As soon I as wrap this with a query like
SELECT eid FROM(
......previous query here.....
) data
The execution time goes from 50 ms to 2.5 sec Ouch!
I tried creating a temporary table, That was the same. (Actually this is my preferred approach since I will need to do a few different queries on this results set.
Reading on this site I don't think this is a correlated sub query but is seems to be acting like one.
Seems like the act of creating a alias table is my issues, since the sub query has a derived table alias and the temp table is obviously a table.
How can I get access to these 3744 rows of data with out this time penalty?
If it would help I can figure out how to post the Explains.
Explain for the longer query:
id select_type table type possible_keys key key_len ref rows Extra
1 PRIMARY <derived2> ALL NULL NULL NULL NULL 3744
2 DERIVED s const PRIMARY PRIMARY 5 1
2 DERIVED srl ref series_id,series_id_2 series_id 5 16 Using where
2 DERIVED <derived3> ALL NULL NULL NULL NULL 208 Using join buffer
2 DERIVED en eq_ref PRIMARY,event_id,event_id_2 PRIMARY 9 race_reg_test.srl.event_id,bts.boat_id 1 Using index
2 DERIVED er ref PRIMARY,boat_id,boat_id_2 boat_id_2 5 bts.boat_id 5
2 DERIVED eri eq_ref PRIMARY PRIMARY 13 race_reg_test.srl.event_id,race_reg_test.srl.race_... 1
2 DERIVED rr ref PRIMARY,boat_id boat_id 4 bts.boat_id 9
2 DERIVED rfi eq_ref PRIMARY PRIMARY 31 race_reg_test.srl.event_id,race_reg_test.srl.race_... 1
2 DERIVED spo ref PRIMARY PRIMARY 8 race_reg_test.srl.id,race_reg_test.en.boat_id 1
3 DERIVED srl1 ref series_id,series_id_2 series_id 5 16 Using index; Using temporary; Using filesort
3 DERIVED en1 ref PRIMARY,event_id,event_id_2 PRIMARY 5 race_reg_test.srl1.event_id 11 Using where
3 DERIVED er1 ref boat_id,boat_id_2 boat_id 4 race_reg_test.en1.boat_id 9 Using index
You said you tried creating a temporary table, I am not sure if by that you mean a View or not.
I would create a View with that query and then perform any queries necessary on the View.
CREATE VIEW massive_query_view AS
SELECT
srl.event_id as eid
, srl.race_num as rnum
, bts.boat_id as bid_id
, srl.series_year as yr
, srl.id as id
, IFNULL(rfi.fleet,fleet_def) as flt_old,flt_match,s.series_id as sid
, s.series_year as syr
,IFNULL(ovr_pts,POINTS('4',IFNULL(ovr_place,place),num_start)) as points
FROM
(SELECT en1.boat_id,en1.boat_name,MAX(fleet) as fleet_def FROM entries en1
JOIN series_race_list srl1 ON srl1.event_id=en1.event_id
AND srl1.series_year=en1.race_year
LEFT JOIN entries_race er1 ON en1.boat_id= er1.boat_id
AND srl1.event_id=en1.event_id
AND srl1.series_year =en1.race_year
WHERE srl1.series_id ='3' AND srl1.series_year ='2012'
AND en1.entry_deleted='N'
GROUP BY boat_id) bts
JOIN series_race_list srl LEFT JOIN series as s ON s.series_id=srl.series_id
AND s.series_year =srl.series_year
LEFT JOIN entries as en ON srl.event_id=en.event_id
AND srl.series_year =en.race_year AND bts.boat_id =en.boat_id
LEFT JOIN entries_race er ON er.race_id= srl.event_id AND er.race_num=srl.race_num
AND er.yr = srl.series_year AND bts.boat_id =er.boat_id
LEFT JOIN event_race_info as eri ON eri.race_id= srl.event_id
AND eri.race_num=srl.race_num AND eri.yr = srl.series_year
ANd er.line=eri.line AND status REGEXP 'prelim|final'
LEFT JOIN race_results as rr ON srl.event_id=rr.race_id
AND srl.race_num= rr.race_num AND srl.series_year =rr.yr
AND bts.boat_id= rr.boat_id AND checked_in='Y'
LEFT JOIN race_fleet_info as rfi ON rfi.race_id= srl.event_id
AND rfi.yr=srl.series_year AND srl.race_num= rfi.race_num
AND rfi.fleet=rr.flt AND complete='Y'
LEFT JOIN series_pts_override as spo ON srl.id =spo.id AND en.boat_id =spo.bid
WHERE s.series_id ='3' AND s.series_year ='2012' AND approved ='Y'
Then, you can perform queries on the View.
SELECT * FROM massive_query_view;
Hope that speeds things up. Another thing you can do is check your indexes. Indexes make where clauses faster but inserts slower. For more information, view the MySQL documentation on how MySQL uses indexes: http://dev.mysql.com/doc/refman/5.0/en/mysql-indexes.html .
A few things, but the biggest one I see is in your original query... at the point of
boat_id) bts JOIN series_race_list srl LEFT JOIN series as s
You have no "ON" condition between bts and srl which will result in a Cartesian result and probably a big killer to you. For every record in bts, its creating an entry in srl, then from that product joining to series. From srl to series is ok as it is joined on apparent valid criteria / keys.
Next, you have a few fields that are not alias.field, such as max(fleet) in inner-most query that aliases out to "bts". In addition, why the MAX(fleet) if its grouped by the boat ID which I would interpret as a primary key and would be unique... would a boat ever change it's fleet? If so, is this accurate? If you have a table of fleets (also having its own auto-sequence ID), and a boat changes ownership/sponsor ship (whatever) to a pre-existing fleet from say... fleet 93 to a new who already had an ID on file of 47 where even though 47 was the newest relationship, but and older pre-existing ID... is that what you really want? MAX()?
Additional fields for no alias.field: ovr_pts and ovr_place, place in the field list (and what is the POINTS() function... status
at the regular expression, checked_in at race results, and complete at race fleet info, and finally approved in the final where clause. Minor, but could be helpful for index optimizing.
Lastly, your query has the WHERE clause on specific "s.series_id... and s.series_year..." yet you have a LEFT-JOIN earlier in the query. This basically cancels out the left-join component of it and turns it into an implied INNER JOIN since you are not allowing for NULL as a valid option of inclusion.
After some clarification, I might even suggest altering the query around some, but the biggest thing I see was from the start... no "ON" condition joining bts and the series_rate_list table.
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.