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?
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
来源:https://stackoverflow.com/questions/9763538/nested-query-in-squeel