How can i optimize MySQL query, event if it have index - php

Question

Here is my query which is taking 17.9397 sec time to get response:
SELECT allbar.iBarID AS iBarID,
allbar.vName AS vName,
allbar.tAddress AS tAddress,
allbar.tDescription AS tDescription,
(SELECT COUNT(*)
FROM tbl_post p
WHERE p.vBarIDs = allbar.iBarID) AS `total_post`,
allbar.bar_usbg AS bar_usbg,
allbar.bar_enhance AS bar_enhance,
(SELECT count(*)
FROM tbl_user
WHERE FIND_IN_SET(allbar.iBarID,vBarIDs)
AND (eType = 'Bartender'
OR eType = 'Bar Manager'
OR eType = 'Bar Owner')) AS countAss,
allbar.eStatus AS eStatus
FROM
(SELECT DISTINCT b.iBarID AS iBarID,
b.vName AS vName,
b.tAddress AS tAddress,
(CASE LENGTH(b.tDescription) WHEN 0 THEN '' WHEN LENGTH(b.tDescription) > 0
AND LENGTH(b.tDescription) < 50 THEN CONCAT(LEFT(b.tDescription, 50),'...') ELSE b.tDescription END) AS tDescription,
b.usbg AS bar_usbg,
b.enhance AS bar_enhance,
b.eStatus AS eStatus
FROM tbl_bar b,
tbl_user u
WHERE b.iBarID <> '-10') AS allbar
I have tried EXPLAIN, here is the result of that:
Can anyone explain me this EXPLAIN result?

Answer 1

You should totaly rewrite that query, it's complete nonsense.
In this part
(SELECT DISTINCT b.<whatever>
FROM tbl_bar b,
tbl_user u
WHERE b.iBarID <> '-10') AS allbar
what you're basically doing is connecting every row from table tbl_bar with every row from tbl_user. Then filter tbl_bar, and when everything is selected (maybe MySQL has to write everything in a temporary table before doing this) return the result set without duplicates. You don't ever want to do that. Especially when you don't even select anything from tbl_user. When there's a connection, specify it. If there's none, don't join those tables or create a connection. I don't know if or how your tables are connected, but it should look something like this:
(SELECT DISTINCT b.<whatever>
FROM tbl_bar b
JOIN tbl_user u ON b.user_id = u.id /*or whatever the connection is*/
WHERE b.iBarID <> '-10') AS allbar
Then you have this ugly subquery.
(SELECT COUNT(*)
FROM tbl_post p
WHERE p.vBarIDs = allbar.iBarID) AS `total_post`,
allbar.bar_usbg AS bar_usbg,
allbar.bar_enhance AS bar_enhance,
which is by the way dependent (see your explain output). Which means, that this subquery is executed for every row of your outer query (yes, the one with the cross join as discussed above). Instead of this subquery, join the table in the outer query and work with GROUP BY.
So far the query should look something like this:
SELECT
b.iBarID AS iBarID,
b.vName AS vName,
b.tAddress AS tAddress,
b.tDescription AS tDescription,
COUNT(*) AS `total_post`,
allbar.bar_usbg AS bar_usbg,
allbar.bar_enhance AS bar_enhance
FROM
tbl_bar b
JOIN tbl_user u ON b.user_id = u.id
JOIN tbl_post p ON p.vBarIDs = b.iBarID
WHERE b.iBarID <> '-10'
GROUP BY b.iBarID
(In fact, this is not really right. Rule is, every column in the SELECT clause should either be in the GROUP BY clause as well or have an aggregate function (like count() or max() applied to it. Otherwise a random row of each group is displayed. But this is just an example. You will have to work out the details.)
Now comes the worst part.
(SELECT count(*)
FROM tbl_user
WHERE FIND_IN_SET(allbar.iBarID,vBarIDs)
AND (eType = 'Bartender'
OR eType = 'Bar Manager'
OR eType = 'Bar Owner')) AS countAss,
allbar.eStatus AS eStatus
The use of FIND_IN_SET() suggests, that you're storing multiple values in one column. Again, you never ever want to do that. Please read this answer to Is storing a delimited list in a database column really that bad? and then redesign your database. I won't help you with this one, as this clearly is stuff for a separate question.
All this didn't really explain the EXPLAIN result. For this question, I would have to write a whole tutorial, which I won't do, since everything is in the manual, as always.