Question

Short version: How do I write this query in squeel?

SELECT OneTable.*, my_count
FROM   OneTable JOIN (
  SELECT DISTINCT one_id, count(*) AS my_count
  FROM   AnotherTable
  GROUP BY one_id
) counts
ON OneTable.id=counts.one_id

---

Long version: rocket_tag is a gem that adds simple tagging to models. It adds a method tagged_with. Supposing my model is User, with an id and name, I could invoke User.tagged_with ['admin','sales']. Internally it uses this squeel code:

select{count(~id).as(tags_count)}
.select("#{self.table_name}.*").
joins{tags}.
where{tags.name.in(my{tags_list})}.
group{~id}

Which generates this query:

SELECT count(users.id) AS tags_count, users.*
  FROM users INNER JOIN taggings
    ON taggings.taggable_id = users.id
   AND taggings.taggable_type = 'User'
  INNER JOIN tags
    ON tags.id = taggings.tag_id
  WHERE tags.name IN ('admin','sales')
  GROUP BY users.id

Some RDBMSs are happy with this, but postgres complains:

ERROR: column "users.name" must appear in the GROUP BY
clause or be used in an aggregate function

I believe a more agreeable way to write the query would be:

SELECT users.*, tags_count FROM users INNER JOIN (
  SELECT DISTINCT taggable_id, count(*) AS tags_count
    FROM taggings INNER JOIN tags
      ON tags.id = taggings.tag_id
    WHERE tags.name IN ('admin','sales')
    GROUP BY taggable_id
  ) tag_counts
  ON users.id = tag_counts.taggable_id

Is there any way to express this using squeel?

Answer 1

I wouldn’t know about Squeel, but the error you see could be fixed by upgrading PostgreSQL.

Some RDBMSs are happy with this, but postgres complains:

ERROR: column “users.name” must appear in the GROUP BY clause or be
used in an aggregate function

Starting with PostgreSQL 9.1, once you list a primary key in the GROUP BY you can skip additional columns for this table and still use them in the SELECT list. The release notes for version 9.1 tell us:

Allow non-GROUP BY columns in the query target list when the primary
key is specified in the GROUP BY clause

---

BTW, your alternative query can be simplified, an additional DISTINCT would be redundant.

SELECT o.*, c.my_count
FROM   onetable o
JOIN (
  SELECT one_id, count(*) AS my_count
  FROM   anothertable
  GROUP  BY one_id
) c ON o.id = counts.one_id