How can I get the intersection of n SELECTS? - mysql

Question

I have these 3 tables: data, tags and data_tag_rel.
data
id data
------------------------------------
1 A string of long data A.
2 A string of long data B.
3 A string of long data C.
4 A string of long data D.
5 A string of long data E.
6 A string of long data F.
7 A string of long data G.
tags
id tag
------------
1 gold
2 silver
3 copper
data_tag_rel
data tag
------------------
1 1
1 2
2 1
3 2
4 3
5 1
5 2
5 3
6 1
7 1
As you can see, there is data and tags, and a relationship table to determine what tags are assigned to what data. Here the data is talking about metals. In this example:
The gold tag has been assigned to 5 data strings.
The silver tag has been assigned to 3 data strings.
The copper tag has been assigned to 2 data strings.
I want to query the database and obtain an INTERSECTION of tags gold, silver and copper. Meaning I want to obtain the table_data that is assigned to all 3 tags. The result would be just 1 row from the data table, row id 5: "A string of long data E."
What query would accomplish this INTERSECTION?
So far I can get the query working querying only 1 tag:
SELECT data.id, data.data
FROM data
INNER JOIN data_tag_rel ON data.id = data_tag_rel.data
INNER JOIN tags ON data_tag_rel.tag = tags.id
WHERE tags.tag = "gold"
Thanks so much!

Answer 1

Aggregation provides one option:
SELECT d.id, d.data
FROM data d
INNER JOIN data_tag_rel dtr ON d.id = dtr.data
INNER JOIN tags t ON dtr.tag = t.id
WHERE t.tag IN ('gold', 'silver', 'copper')
GROUP BY d.id, d.data
HAVING COUNT(DISTINCT t.tag) = 3;
Demo

Answer 2

Speed is going to vary based on your data. Tim's answer is likely to be fast enough for practical purposes, but if you find it is not, you may be able to slightly improve it by not joining data until needed (the other changes here are just stylistic and unlikely to have any effect):
select d.id, d.data
from (
select dtr.data as id
from data_tag_rel dtr
where dtr.tag in (select id from tag where tag in ('gold','silver','copper')
group by dtr.data
having count(tag) = 3
) d_ids
join data d using (id)
If you have a great deal of data, doing separate joins for each tag is likely to be faster, especially if you know which tags are rare and can go from rarest to least rare:
select d.id, d.data
from data_tag_rel dtr1 on dtr
join data_tag_rel dtr2 on dtr2.data=dtr1.data and dtr2.tag=(select id from tag where tag='silver')
join data_tag_rel dtr3 on dtr3.data=dtr2.data and dtr3.tag=(select id from tag where tag='copper')
join data d on d.id=dtr3.data
where dtr1.tag=(select id from tag where tag='gold')
(both queries untested)