一对多查询，同时基于不同的主键进行限制

Question

I have a table like this:

create table images (
    image_id serial primary key,
    user_id int references users(user_id),
    date_created timestamp with time zone
);

I then have a tag table for tags that images can have:

create table images_tags (
    images_tag_id serial primary key,
    image_id int references images(image_id),
    tag_id int references tags(tag_id)       
);

To get the results I want, I run a query like this:

select image_id,user_id,tag_id from images left join images_tags using(image_id)
where (?=-1 or user_id=?)
and (?=-1 or tag_id in (?, ?, ?, ?)) --have up to 4 tag_ids to search for
order by date_created desc limit 100;

The problem is, I want to limit based on the number of unique image_id s because my output will look like this:

{"images":[
    {"image_id":1, "tag_ids":[1, 2, 3]},
    ....
]}

Notice how I group the tag_id s into an array for output, even though the SQL returns a row for each tag_id and image_id combo.

So, when I say limit 100 , I want it to apply to 100 unique image_id s.

Answer 1

Maybe you should put one image on each row? If that works, you can do:

select image_id, user_id, string_agg(cast(tag_id as varchar(2000)), ',') as tags
from images left join
     images_tags
     using (image_id)
where (?=-1 or user_id=?) and
      (?=-1 or tag_id in (?, ?, ?, ?)) --have up to 4 tag_ids to search for
group by image_id, user_id
order by date_created desc
limit 100;

If that doesn't work, then use a CTE:

with cte as (
      select image_id, user_id, tag_id,
             dense_rank() over (order by date_created desc) as seqnum
      from images left join
           images_tags
           using (image_id)
      where (?=-1 or user_id=?) and
            (?=-1 or tag_id in (?, ?, ?, ?)) --have up to 4 tag_ids to search for
    )
select *
from cte
where seqnum <= 100
order by seqnum;

Answer 2

Select 100 qualifying images first, and then join images_tags.
Use an EXISTS semi-join to satisfy the condition on images_tags and take care to get the parentheses right.

SELECT i.*, t.tag_id
FROM  (
   SELECT i.image_id, i.user_id
   FROM   images i
   WHERE (? = -1 OR i.user_id = ?)
   AND   (? = -1 OR EXISTS (
      SELECT 1
      FROM   images_tags t
      WHERE  t.image_id = i.image_id
      AND    t.tag_id IN (?, ?, ?, ?)
      ))
   ORDER  BY i.date_created DESC
   LIMIT  100
   ) i
LEFT   JOIN images_tags t
            ON t.image_id = i.image_id
           AND (? = -1 OR t.tag_id in (?, ?, ?, ?)) -- repeat condition

This should be faster than a solution with window functions and CTEs.
Test performance with EXPLAIN ANLAYZE . As always run a couple of times to warm up cache.