SQL query over multiple rows

Question

I have a Table which looks like this:

---------------------------
|housing_id | facility_id |
---------------------------
|    1      |      7      |
|    1      |      4      |
|    2      |      7      |
---------------------------

Now what i want to do is get all housing_ids with a facility_id of 7 AND 4. So the query should only return the housing_id 1 in this case. Database is mysql.

Answer 1

Another approach would be -

SELECT housing_id
FROM mytable
WHERE facility_id IN (4,7)
GROUP BY housing_id
HAVING COUNT(DISTINCT facility_id) = 2

UPDATE - inspired by the comment by Josvic I decided to do some more testing and thought I would include my findings.

One of the benefits of using this query is that it is easy to modify to include more facility_ids. If you want to find all housing_ids that have facility_ids 1, 3, 4 & 7 you just do -

SELECT housing_id
FROM mytable
WHERE facility_id IN (1,3,4,7)
GROUP BY housing_id
HAVING COUNT(DISTINCT facility_id) = 4

The performance of all three of these queries varies hugely based on the indexing strategy employed. I was unable to get reasonable performance, on my test dataset, from the dependant subquery version regardless of indexing used.

The self join solution provided by Tim performs very well given separate single column indices on the two columns but does not perform quite so well as the number of criteria increases.

Here are some basic stats on my test table - 500k rows - 147963 housing_ids with potential values for facility_id between 1 and 9.

Here are the indices used for running all these tests -

SHOW INDEXES FROM mytable;
+---------+------------+---------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+
| Table   | Non_unique | Key_name            | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type |
+---------+------------+---------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+
| mytable |          0 | UQ_housing_facility |            1 | housing_id  | A         |      500537 |     NULL | NULL   |      | BTREE      |
| mytable |          0 | UQ_housing_facility |            2 | facility_id | A         |      500537 |     NULL | NULL   |      | BTREE      |
| mytable |          0 | UQ_facility_housing |            1 | facility_id | A         |          12 |     NULL | NULL   |      | BTREE      |
| mytable |          0 | UQ_facility_housing |            2 | housing_id  | A         |      500537 |     NULL | NULL   |      | BTREE      |
| mytable |          1 | IX_housing          |            1 | housing_id  | A         |      500537 |     NULL | NULL   |      | BTREE      |
| mytable |          1 | IX_facility         |            1 | facility_id | A         |          12 |     NULL | NULL   |      | BTREE      |
+---------+------------+---------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+

First query tested is the dependant subquery -

SELECT SQL_NO_CACHE DISTINCT housing_id
FROM mytable
WHERE housing_id IN (SELECT housing_id FROM mytable WHERE facility_id=4)
AND housing_id IN (SELECT housing_id FROM mytable WHERE facility_id=7);

17321 rows in set (9.15 sec)

+----+--------------------+---------+-----------------+----------------------------------------------------------------+---------------------+---------+------------+--------+---------------------------------------+
| id | select_type        | table   | type            | possible_keys                                                  | key                 | key_len | ref        | rows   | Extra                                 |
+----+--------------------+---------+-----------------+----------------------------------------------------------------+---------------------+---------+------------+--------+---------------------------------------+
|  1 | PRIMARY            | mytable | range           | NULL                                                           | IX_housing          | 4       | NULL       | 500538 | Using where; Using index for group-by |
|  3 | DEPENDENT SUBQUERY | mytable | unique_subquery | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | UQ_housing_facility | 8       | func,const |      1 | Using index; Using where              |
|  2 | DEPENDENT SUBQUERY | mytable | unique_subquery | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | UQ_housing_facility | 8       | func,const |      1 | Using index; Using where              |
+----+--------------------+---------+-----------------+----------------------------------------------------------------+---------------------+---------+------------+--------+---------------------------------------+

SELECT SQL_NO_CACHE DISTINCT housing_id
FROM mytable
WHERE housing_id IN (SELECT housing_id FROM mytable WHERE facility_id=1)
AND housing_id IN (SELECT housing_id FROM mytable WHERE facility_id=3)
AND housing_id IN (SELECT housing_id FROM mytable WHERE facility_id=4)
AND housing_id IN (SELECT housing_id FROM mytable WHERE facility_id=7);

567 rows in set (9.30 sec)

+----+--------------------+---------+-----------------+----------------------------------------------------------------+---------------------+---------+------------+--------+---------------------------------------+
| id | select_type        | table   | type            | possible_keys                                                  | key                 | key_len | ref        | rows   | Extra                                 |
+----+--------------------+---------+-----------------+----------------------------------------------------------------+---------------------+---------+------------+--------+---------------------------------------+
|  1 | PRIMARY            | mytable | range           | NULL                                                           | IX_housing          | 4       | NULL       | 500538 | Using where; Using index for group-by |
|  5 | DEPENDENT SUBQUERY | mytable | unique_subquery | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | UQ_housing_facility | 8       | func,const |      1 | Using index; Using where              |
|  4 | DEPENDENT SUBQUERY | mytable | unique_subquery | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | UQ_housing_facility | 8       | func,const |      1 | Using index; Using where              |
|  3 | DEPENDENT SUBQUERY | mytable | unique_subquery | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | UQ_housing_facility | 8       | func,const |      1 | Using index; Using where              |
|  2 | DEPENDENT SUBQUERY | mytable | unique_subquery | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | UQ_housing_facility | 8       | func,const |      1 | Using index; Using where              |
+----+--------------------+---------+-----------------+----------------------------------------------------------------+---------------------+---------+------------+--------+---------------------------------------+

Next is my version using the GROUP BY ... HAVING COUNT ...

SELECT SQL_NO_CACHE housing_id
FROM mytable
WHERE facility_id IN (4,7)
GROUP BY housing_id
HAVING COUNT(DISTINCT facility_id) = 2;

17321 rows in set (0.79 sec)

+----+-------------+---------+-------+---------------------------------+-------------+---------+------+--------+------------------------------------------+
| id | select_type | table   | type  | possible_keys                   | key         | key_len | ref  | rows   | Extra                                    |
+----+-------------+---------+-------+---------------------------------+-------------+---------+------+--------+------------------------------------------+
|  1 | SIMPLE      | mytable | range | UQ_facility_housing,IX_facility | IX_facility | 4       | NULL | 198646 | Using where; Using index; Using filesort |
+----+-------------+---------+-------+---------------------------------+-------------+---------+------+--------+------------------------------------------+

SELECT SQL_NO_CACHE housing_id
FROM mytable
WHERE facility_id IN (1,3,4,7)
GROUP BY housing_id
HAVING COUNT(DISTINCT facility_id) = 4;

567 rows in set (1.25 sec)

+----+-------------+---------+-------+---------------------------------+-------------+---------+------+--------+------------------------------------------+
| id | select_type | table   | type  | possible_keys                   | key         | key_len | ref  | rows   | Extra                                    |
+----+-------------+---------+-------+---------------------------------+-------------+---------+------+--------+------------------------------------------+
|  1 | SIMPLE      | mytable | range | UQ_facility_housing,IX_facility | IX_facility | 4       | NULL | 407160 | Using where; Using index; Using filesort |
+----+-------------+---------+-------+---------------------------------+-------------+---------+------+--------+------------------------------------------+

And last but not least the self join -

SELECT SQL_NO_CACHE a.housing_id
FROM mytable a
INNER JOIN mytable b
    ON a.housing_id = b.housing_id
WHERE a.facility_id = 4 AND b.facility_id = 7;

17321 rows in set (1.37 sec)

+----+-------------+-------+--------+----------------------------------------------------------------+---------------------+---------+-------------------------+-------+-------------+
| id | select_type | table | type   | possible_keys                                                  | key                 | key_len | ref                     | rows  | Extra       |
+----+-------------+-------+--------+----------------------------------------------------------------+---------------------+---------+-------------------------+-------+-------------+
|  1 | SIMPLE      | b     | ref    | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | IX_facility         | 4       | const                   | 94598 | Using index |
|  1 | SIMPLE      | a     | eq_ref | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | UQ_housing_facility | 8       | test.b.housing_id,const |     1 | Using index |
+----+-------------+-------+--------+----------------------------------------------------------------+---------------------+---------+-------------------------+-------+-------------+

SELECT SQL_NO_CACHE a.housing_id
FROM mytable a
INNER JOIN mytable b
    ON a.housing_id = b.housing_id
INNER JOIN mytable c
    ON a.housing_id = c.housing_id
INNER JOIN mytable d
    ON a.housing_id = d.housing_id
WHERE a.facility_id = 1
AND b.facility_id = 3
AND c.facility_id = 4
AND d.facility_id = 7;

567 rows in set (1.64 sec)

+----+-------------+-------+--------+----------------------------------------------------------------+---------------------+---------+-------------------------+-------+--------------------------+
| id | select_type | table | type   | possible_keys                                                  | key                 | key_len | ref                     | rows  | Extra                    |
+----+-------------+-------+--------+----------------------------------------------------------------+---------------------+---------+-------------------------+-------+--------------------------+
|  1 | SIMPLE      | b     | ref    | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | IX_facility         | 4       | const                   | 93782 | Using index              |
|  1 | SIMPLE      | d     | eq_ref | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | UQ_housing_facility | 8       | test.b.housing_id,const |     1 | Using index              |
|  1 | SIMPLE      | c     | eq_ref | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | UQ_housing_facility | 8       | test.b.housing_id,const |     1 | Using index              |
|  1 | SIMPLE      | a     | eq_ref | UQ_housing_facility,UQ_facility_housing,IX_housing,IX_facility | UQ_housing_facility | 8       | test.d.housing_id,const |     1 | Using where; Using index |
+----+-------------+-------+--------+----------------------------------------------------------------+---------------------+---------+-------------------------+-------+--------------------------+

Answer 2

SELECT housing_id
FROM mytable
WHERE housing_id IN (SELECT housing_id FROM mytable WHERE facility_id=4)
AND housing_id IN (SELECT housing_id FROM mytable WHERE facility_id=7)

Answer 3

You can also do a self JOIN, the one which is fastest will very much depend on how much data is in your table.

SELECT a.housing_id
FROM mytable a
INNER JOIN mytable b
ON a.housing_id = b.housing_id AND a.facility_id <> b.facility_id
WHERE a.facility_id = 4 AND b.facility_id = 7