How to count unique set values in MySQL - php

Question

I would appreciate your input to help me count unique values for a SET type in MySql. I have a column named "features" defined as a SET field as follows:
CREATE TABLE cars (features SET('power steering', 'power locks', 'satellite radio', 'power windows', 'sat nav', 'turbo'));
As I fill this table, since the features are not mutually exclusive, I will get records which include a combination of 2 or more of these features. For example:
Car 1 has power steering and power windows, but none of the remaining features.
Car 2 has all features.
Car 3 has all features, except sat nav and turbo.
What I want to do is to get a list of all single listed features in the table, including the count of records associated to each in a similar fashion as a SELECT statement using a GROUP BY clause. So, following with the example above, I should be able to get the following result:
features |count
---------------+------
power steering | 3 //All cars have this feature
power locks | 2 //Only cars 2 and 3 have it
satellite radio| 2 //Only cars 2 and 3 have it
power windows | 3
sat nav | 1 //only car 2 has it
turbo | 1 //only car 2 has it
I have tried using the following query with the expectation of obtaining the aforementioned result:
SELECT features, COUNT(features) FROM cars GROUP BY features;
However, instead of what I was expecting, I got the count of each of the existing feature combinations:
features |count
------------------------------------------------+--------
power steering, power windows | 1 //i.e. only 1 car has
| //only these 2 features
| //(car 1 in this example)
|
------------------------------------------------+-------
power steering, power locks, satellite radio, |
power windows, sat nav, turbo | 1
------------------------------------------------+-------
power steering, power locks, satellite radio, |
power windows | 1
So, the question is: Is there a way of obtaining the count of each single feature, as shown in the first table, using one single MySQL query? I could do it by executing one query for each feature, but I'm sure there must be a way of avoiding such hassle. Someone might as well suggest using a different table for the features and joining, but it is not possible at this point without heavily impacting the rest of the project. Thanks in advance!

Answer 1

SELECT set_list.features, COUNT(cars.features) FROM
(SELECT TRIM("'" FROM SUBSTRING_INDEX(SUBSTRING_INDEX(
(SELECT TRIM(')' FROM SUBSTR(column_type, 5)) FROM information_schema.columns
WHERE table_name = 'cars' AND column_name = 'features'),
',', #r:=#r+1), ',', -1)) AS features
FROM (SELECT #r:=0) deriv1,
(SELECT ID FROM information_schema.COLLATIONS) deriv2
HAVING #r <=
(SELECT LENGTH(column_type) - LENGTH(REPLACE(column_type, ',', ''))
FROM information_schema.columns
WHERE table_name = 'cars' AND column_name = 'features')) set_list
LEFT OUTER JOIN cars
ON FIND_IN_SET(set_list.features, cars.features) > 0
GROUP BY set_list.features
Adapted from:
MySQL: Query for list of available options for SET
My query takes the SQL at the above post as the basis, to get a list of the available column values. All of the indented SQL is that one query, if you execute it alone you'll get the list, and I create a result set from it which I call "set_list". I just copied that query as is, but it is basically doing a lot of string manipulation to get the list - as Mike Brant suggested, the code would be far simpler (but maybe just not as dynamic) if you put the list into another table, and just joined that.
I then join set_list back against the cars table, joining each item from set_list against the rows in cars that contain that feature - FIND_IN_SET(). It's an outer join, so if anything from the set list isn't represented, it will be there with a count of zero.

Answer 2

Typically, we use the FIND_IN_SET function.
You could use a query like this to return the specified result:
SELECT f.feature
, COUNT(1)
FROM ( SELECT 'power steering' AS feature
UNION ALL SELECT 'power locks'
UNION ALL SELECT 'satellite radio'
UNION ALL SELECT 'power windows'
UNION ALL SELECT 'sat nav'
UNION ALL SELECT 'turbo'
) f
JOIN cars c
ON FIND_IN_SET(f.feature,c.features)>0
GROUP BY f.feature
ORDER BY f.feature
You could omit >0 and get the same result. This query omits "zero counts": rows with a "feature" that doesn't appear for any car. To get those, you could use an outer join (add the LEFT keyword before JOIN, and rather than COUNT(1) in the SELECT list, COUNT(expr) where expr is a column from cars that is NOT NULL, or some other expression that will be non-NULL when a matching row is found, and NULL when a matching row is not found.