MySQL选择连接表中多行的位置

Question

Let's say I have two tables, something like:

+-----------------------------+
|a                            |
+------+-----------+----------+
| a_id | firstname | lastname |
+------+-----------+----------+
|    1 | Tom       | Cruise   |
|    3 | Matt      | Damon    |
|    4 | Ben       | Affleck  |
|    8 | Ryan      | Gosling  |
+------+-----------+----------+

+-----------------------+
| b                     |
+------+------+---------+
| b_id | a_id | b_value |
+------+------+---------+
|    2 |    3 | one     |
|    1 |    3 | two     |
|    4 |    8 | three   |
|    8 |    1 | four    |
+------+------+---------+

I would like to be able to SELECT rows from a which meet a set of criteria from b , where there might be multiple rows.

For example:

• Get the first and last name(s) of anyone who has a b_value of one, and other b_value of two. In the example data, this refers only to Matt Damon.

• Get the first and last name(s) of anyone who has a b_value of three, or a b_value of four. In the example data, Ryan Gosling has a b_value of three, and Tom Cruise has a b_value of four, so both Ryan Gosling and Tom Cruise would be returned.

• Get the first and last name(s) of anyone that has either:

  • A b_value of one and another b_value of two.
  • A b_value of three.

In the example data, Matt Damon meets the first condition, and Ryan Gosling meets the second condition, so both are returned.

The difficulty comes from when these conditions are complex, combining multiple and/or conditions, and a variable in length.

My current attempt (solving specifically example 3 in this case), is using aggregate functions in the HAVING :

SELECT firstname, lastname 
FROM temp_a INNER JOIN temp_b ON temp_a.a_id = temp_b.a_id 
GROUP BY temp_a.a_id 
HAVING (SUM(b_value="one") AND SUM(b_value="two")) OR SUM(b_value="three");

Which gets the correct rows, but there's two immediate problems with this approach for me:

• SUM seems wrong and complicated. I'd have to find each of the conditionals and make sure each one is wrapped in a SUM , of which there might be lots.
• The b_value is indexed, and b has lots of rows. The HAVING clause doesn't use the index and so the query is noticeably slow.

Is there a different approach I should be taking to do this sort of query, bearing in mind I'm not really in direct control of the conditionals used for filtering.

Answer 1

drop table if exists a,b;
create table a( a_id int, firstname varchar(20), lastname varchar(20));
insert into a values
(    1 , 'Tom'       , 'Cruise' ),  
(    3 , 'Matt'      , 'Damon'  ),  
(    3 , 'Ben'       , 'Affleck'),  
(    8 , 'Ryan'      , 'Gosling');  

create table b( b_id int, a_id int, b_value varchar(20));
insert into b values
(    2 ,    3 , 'one'  ),   
(    1 ,    3 , 'two'  ),   
(    4 ,    8 , 'three'),   
(    8 ,    1 , 'four' );


select firstname, lastname,s.vals
from a
join
(
select b.a_id, group_concat(b_value) vals from b group by b.a_id
) s on a.a_id = s.a_id

where s.vals in ('one,two') or s.vals in ('three')

+-----------+----------+---------+
| firstname | lastname | vals    |
+-----------+----------+---------+
| Matt      | Damon    | one,two |
| Ben       | Affleck  | one,two |
| Ryan      | Gosling  | three   |
+-----------+----------+---------+
3 rows in set (0.02 sec)

Answer 2

SELECT rows from a where it has b_value of both one and two would return Matt Damon.

if you want above result then table "a" "id" column should have unique records, table a and b both should have primary key.

or you can try following query, but it will return double records for "matt daemon"

select firstname,lastname from a left join b on a.a_id=b.a_id ;

Answer 3

In dynamic SQL it would somthing like this, first I would suggest creating a variable in your application which you fill up based on the selection of filters

 - var exampleOne = "(b_value = 'one') OR (b_value = 'two')";
 - var exampleTwo = "(b_value = 'three') OR (b_value = 'four')";
 - var exampleThree = "(b_value = 'one' or b_value = 'two') OR (b_value = 'three')"

Then you can create a stored procedure with one parameter, here you pass in your variable

CREATE PROCEDURE `spNames`(IN whereFilter VARCHAR(255))
BEGIN
  SET @SqlText = CONCAT('SELECT DISTINCT a.firstname, a.lastname 
                         FROM @a AS a INNER JOIN @b AS b ON a.a_id=b.a_id
                         WHERE ', whereFilter);
  PREPARE stmt FROM @SqlText;
  EXECUTE stmt;
  DEALLOCATE PREPARE stmt
END

Then you can call this stored procedure from your code and pass in the variable

CALL spNames (exampleOne);