SQL How do I query a many-to-many relationship

How do I select 开发者_开发问答posts that contain a specific tag if there is a many-to-many relationship between posts and tags?

The problem I am having is that because of the where tag.name = 'xxx', only that tag is selected. I want to select all posts that have the tag specified, together with all of their tags, e.g.,

Post 1 -> tag1, tag2
Post 2 -> tag1, tag3
Post 3 -> tag2, tag3

Currently what I get is:

Post 1 -> tag2 // missing tag1
Post 3 -> tag2 // missing tag3 

---

Assuming these tables:

• Posts: id, author, date, content
• Tags: id, name
• PostTags: post_id, tag_id

The last table is often called a join table and facilitates a many-to-many relationship between Posts and Tags.

SELECT p.*
FROM posts p
JOIN posttags pt ON p.id = pt.post_id
JOIN tags t ON pt.tag_id = t.id
WHERE t.name = 'sql'

Basically, think of a many-to-many relationship as two one-to-many relationships, because that's how they're implemented in normal RDBMSs. So the above query has a one-to-many join from Posts to PostTags and another from Tags to PostTags.

The PostTags table I created has a composite primary key, being (post_id, tag_id). That combination will be unique. Many disfavour composite keys so you'll often see people creating a primary key column:

• PostTags: id, post_id, tag_id

Either method is fine. It's largely a philosophical difference.

Update: if you want to select all the posts that have a particular tag and all the tags those posts have then:

SELECT p.*
FROM posts p
JOIN posttags pt ON p.id = pt.post_id
JOIN tags t ON pt.tag_id = t.id
WHERE p.id IN
  (SELECT post_id
  FROM PostTags pt
  JOIN tags t ON pt.tag_id = t.id
  WHERE t.name = 'xyz')

Another way to do this is:

SELECT p.*
FROM posts p
JOIN posttags pt ON p.id = pt.post_id
JOIN tags t ON pt.tag_id = t.id
WHERE EXISTS
  (SELECT post_id
  FROM PostTags pt
  JOIN tags t ON pt.tag_id = t.id
  WHERE t.name = 'xyz'
  AND pt.post_id = p.id)

Which performs better will need to be tested and may vary depending on database vendor and version. A good optimizer (ie Oracle) will probably optimize them to perform the same. Others may not.

Now this will get you rows back like this:

Post 1, tag 1
Post 1, tag 2
Post 3, tag 2
Post 3, tag 3

so you'll need to combine them, preferably in application logic rather than SQL. Some RDBMSs have vendor-specific extensions for this kind of thing, like MySQL's GROUP_CONCAT() function.