Question

Please consider the following project table:

Project:
  table: project
  manyToMany:
    themes:
      targetEntity: Theme
      inversedBy: projects
      joinTable:
        name: project_theme
        joinColumns:
          project_id:
            referencedColumnName: id
        inverseJoinColumns:
          theme_id:
            referencedColumnName: id
    platforms:
      targetEntity: Platform
      joinTable:
        name: project_platform
        joinColumns:
          project_id:
            referencedColumnName: id
        inverseJoinColumns:
          platform_id:
            referencedColumnName: id
  manyToOne:
    client:
      targetEntity: Client

As you can see a project has three relations; themes through the project_theme join table, platforms through the project_platform table and clients through a client_id column.

I'm trying a produce a query which will find all related projects - projects with the same themes, platforms or clients - and order them by a 'score'.

For example:

Project A:
  Themes: 18, 19
  Platforms: 1, 4
  Client: 22

Find related projects to Project A..

Project D:
  Themes: 18, 19
  Platforms: 1, 4
  Client: 22
Score: 5

Project G:
  Themes: 18, 21
  Platforms: 3, 4
  Client: 22
Score: 3

Project B:
  Themes: 8, 21
  Platforms: 2, 4
  Client: 1
Score: 1

I'd really appreciate some assistant with writing a MySQL query for this. I've been struggling for a while with the following - but I'm probably miles off:

SELECT 
    `project`.*,
    GROUP_CONCAT(`project_theme`.`theme_id`) as themes,
    GROUP_CONCAT(`project_platform`.`platform_id`) as platforms,
    `project`.`client_id` as client
FROM `project`
LEFT JOIN `project_theme` ON `project`.`id` = `project_theme`.`project_id`
LEFT JOIN `project_platform` ON `project`.`id` = `project_platform`.`project_id`
GROUP BY `project`.`id`

Many thanks in advance for any help
Pete

Answer 1

Thanks for your hint Marlin, after reading Advance query. Rank most related fields in mysql and a bit of experimentation I think I might have cracked it…

SELECT 
    p.*,
    (
        IFNULL(themes.matches, 0) + IFNULL(platforms.matches, 0) + IFNULL(clients.matches, 0)
    ) as score
FROM `project` p
LEFT JOIN (
    SELECT t2.project_id, COUNT(*) as matches FROM `project_theme` t1, `project_theme` t2
    WHERE 
      t1.theme_id = t2.theme_id AND t1.project_id = 1
      GROUP BY t2.project_id
) themes ON p.id = themes.project_id
LEFT JOIN (
    SELECT f2.project_id, COUNT(*) as matches FROM `project_platform` f1, `project_platform` f2
    WHERE 
      f1.platform_id = f2.platform_id AND f1.project_id = 1
      GROUP BY f2.project_id
) platforms ON p.id = platforms.project_id
LEFT JOIN (
    SELECT p2.id, COUNT(*) as matches FROM `project` p1, `project` p2
    WHERE 
      p1.client_id = p2.client_id AND p1.id = 1
      GROUP BY p2.id
) clients ON p.id = clients.id
GROUP BY p.`id`
HAVING score > 0
ORDER BY score DESC;