Using REGEXP vs IN on a subquery mysql - mysql

Question

I want to use the data from table 'similar' to find results from table 'releases'
Table 'Similar' has this structure
artist similar_artist
Moodymann Theo Parrish
Moodymann Jeff Mills
Moodymann Marcellus Pittman
Moodymann Rick Wilhite
My query so far is
SELECT * FROM releases
WHERE
releases.all_artists REGEXP 'Moodymann'
OR releases.label_no_country='KDJ'
OR releases.all_artists IN (SELECT similar_artist
FROM similar
WHERE artist='Moodymann')
ORDER BY date DESC
the column 'all_artists' has records like this:
Moodymann | Theo Parrish | Rick Wade
Jeff Mills | Moodymann | Rick Wilhite
So the end query that I want will essentially be this
SELECT * FROM releases
WHERE
releases.all_artists REGEXP 'Moodymann'
OR releases.label_no_country='KDJ'
OR releases.all_artists IN ('Theo Parrish','Jeff Mills','Marcellus Pittman','Rick Wilhite')
To make matches I think I need to use REGEXP instead of IN - REGEXP returns the 'Subquery returns more than 1 row'. How can use the data returned from the subquery?
Also the query is taking a long time to run (up to 20 seconds) - is there anyway to speed this up as this is not usable in my web app.
Thanks!

Answer 1

The only way I would know of how to use REGEXP with a subquery, would be to use that subquery to produce a REGEXP string.
SELECT * FROM releases
WHERE
releases.all_artists REGEXP 'Moodymann'
OR releases.label_no_country='KDJ'
OR releases.all_artists REGEXP (
SELECT GROUP_CONCAT(similar_artist SEPARATOR '|')
FROM similar
WHERE artist='Moodymann'
GROUP BY similar_artist)
ORDER BY date DESC
The above isn't tested, is just a theory to what I might try. It's not going to be very optimal however.
update
Have since tested this and found that GROUP BY similar_artist should be GROUP BY artist
SELECT * FROM releases
WHERE
releases.all_artists REGEXP 'Moodymann'
OR releases.label_no_country='KDJ'
OR releases.all_artists REGEXP (
SELECT GROUP_CONCAT(similar_artist SEPARATOR '|')
FROM similar
WHERE artist='Moodymann'
GROUP BY artist)
ORDER BY date DESC
However, as mentioned by Pheonix you would be better off refactoring your structure to have a releases_artist table. You could then do all this work via JOINs which would be much, much faster.

Answer 2

Try this SQL
SELECT *
FROM releases
WHERE releases.all_artists LIKE '%Moodymann%'
OR releases.label_no_country='KDJ'
ORDER BY date DESC
SQL Fiddle
MySQL 5.5.30 Schema Setup:
CREATE TABLE Table1
(`artist` varchar(9), `similar_artist` varchar(17))
;
INSERT INTO Table1
(`artist`, `similar_artist`)
VALUES
('Moodymann', 'Theo Parrish'),
('Moodymann', 'Jeff Mills'),
('Moodymann', 'Marcellus Pittman'),
('Moodymann', 'Rick Wilhite')
;
create table allt(allf varchar(50));
insert into allt values('Moodymann | Theo Parrish | Rick Wade'),
('Jeff Mills | Moodymann | Rick Wilhite'),
('Jeff Mills | asdasdadasd | Rick Wilhite');
Query 1:
SELECT *
FROM allt
WHERE allt.allf LIKE '%Moodymann%'
Results:
| ALLF |
-----------------------------------------
| Moodymann | Theo Parrish | Rick Wade |
| Jeff Mills | Moodymann | Rick Wilhite |

Answer 3

You can do a join on a comma separated list (won't be fast, but might be quicker than using LIKE with a leading wild card), and you can replace your existing delimiter with a comma to allow this. Also you can use a load of UNIONs to get your list of artists to behave like a table to do a join on.
Further you can use union instead of your other WHERE clauses which might well help with allowing the use of indexes (MySQL will only use one index per table in a query, hence using OR to query on a different column forces it to not use an index for one of the columns it is checking).
As such you can do something like the following:-
SELECT releases.*
FROM releases
INNER JOIN (SELECT 'Theo Parrish' AS anArtist UNION SELECT 'Jeff Mills' UNION SELECT 'Marcellus Pittman' UNION SELECT 'Rick Wilhite') Sub1
ON FIND_IN_SET(Sub1.anArtist, REPLACE(releases.all_artists, " | ", ",")) > 0
UNION
SELECT releases.*
FROM releases
WHERE releases.label_no_country='KDJ'
However if changing the database design to split the pipe separated list of artists onto a different table is even a slight option then do that instead. It will be far quicker and will cope with far greater numbers of artists.