使用子查询从不同表中选择计数

Question

I'm new to MySQL, and I'd like some help in setting up a MySQL query to pull some data from a few tables (~100,000 rows) in a particular output format.

This problem involves three SQL tables:

allusers : This one contains user information. The columns of interest are userid and vip

table1 and table2 contain data, but they also have a userid column, which matches the userid column in allusers .

What I'd like to do:

I'd like to create a query which searches through allusers , finds the userid of those that are VIP, and then count the number of records in each of table1 and table2 grouped by the userid . So, my desired output is:

  userid  | Count in Table1  | Count in Table2
    1     |        5         |         21
    5     |        16        |         31
    8     |        21        |         12

What I've done so far:

I've created this statement:

SELECT userid, count(1) 
FROM table1 
WHERE userid IN  (SELECT userid FROM allusers WHERE vip IS NOT NULL)
GROUP BY userid

This gets me close to what I want. But now, I want to add another column with the respective counts from table2

I also tried using joins like this:

select A.userid, count(T1.userid), count(T2.userid) from allusers A
left join table1 T1 on T1.userid = A.userid
left join table2 T2 on T2.userid = A.userid
where A.vip is not null
group by A.userid

However, this query took a very long time and I had to kill the query. I'm assuming this is because using Joins for such large tables is very inefficient.

Similar Questions

This one is looking for a similar result as I am, but doesn't need nearly as much filtering with subqueries

This one sums up the counts across tables, while I need the counts separated into columns

Could someone help me set up the query to generate the data I need?

Thanks!

Answer 1

You need to pre-aggregate first, then join, otherwise the results will not be what you expect if a user has several rows in both table1 and table2 . Besides, pre-aggregation is usually more efficient than outer aggregation in a situation such as yours.

Consider:

select a.userid, t1.cnt cnt1, t2.cnt cnt2
from allusers a
left join (select userid, count(*) cnt from table1 group by userid) t1
    on t1.userid = a.userid
left join (select userid, count(*) cnt from table2 group by userid) t2
    on t2.userid = a.userid
where a.vip is not null

Answer 2

This is a case where I would recommend correlated subqueries:

select a.userid, 
       (select count(*) from table1 t1 where t1.userid = a.userid) as cnt1,
       (select count(*) from table2 t2 where t2.userid = a.userid) as cnt2
from allusers a
where a.vip is not null;

The reason that I recommend this approach is because you are filtering the alllusers table. That means that the pre-aggregation approach may be doing additional, unnecessary work.