使用子查询来优化SQL？

Question

I have following two tables in the bottom of the post. The siteUsers table have a type column that can be "Member" or "Owner". I am looking for a SQL that shows how many members and owners each site has. I could do something like:

select 
    sites.Url,
    (select count(*) from siteusers where type='member' and siteusers.siteid=sites.siteid) as members,
    (select count(*) from siteusers where type='owners' and siteusers.siteid=sites.siteid) as owners
    from sites
    group by sites.url

However this is extreemly slow. Is there a faster and smarter way?

Sites table:

SiteId  int Unchecked
SPOSiteId   uniqueidentifier    Checked
Url nvarchar(MAX)   Checked

SiteUsers Table:

SiteUserId  int Unchecked
Type    nvarchar(256)   Checked
Name    nvarchar(512)   Checked
Email   nvarchar(512)   Checked
Host    nvarchar(256)   Checked
SiteId  int Checked

Answer 1

One way is:

select s.Url,
       sum(case when su.type = 'member' then 1 else 0 end) as members,
       sum(case when su.type = 'owners' then 1 else 0 end) as owners,
from sites s left join
     siteusers su
     on s.siteid = su.siteid
where su.type in ('member', 'owners')
group by s.url;

Answer 2

You can do conditional aggregation with JOIN :

SELECT s.Url,
       SUM(CASE WHEN su.type ='member' THEN 1 ELSE 0 END) as members,
       SUM(CASE WHEN su.type ='owners' THEN 1 ELSE 0 END) as owners
FROM sites s INNER JOIN
     siteuser su
     ON su.siteid = s.siteid
GROUP BY s.Url;

However, you want index on sites(Url), siteuser(type) for better performance.

If the url is unique in sites , then this should be fast with the right indexes:

select s.Url,
    (select count(*) from siteusers su where su.type = 'member' and su.siteid = s.siteid) as members,
    (select count(*) from siteusers su where su.type = 'owners' and su.siteid = s.siteid) as owners
from sites s;

The right index is on siteusers(siteid, type) . Of course, this assumes that url is unique which may not be the case.

Answer 3

Use Group By

 select s.Url,su.type, count(*)

 from sites s join
 siteusers su
 on s.siteid = su.siteid

 group by s.url,su.type

Answer 4

I'd first check that proper primary keys and indexes are built:

• table Sites should have primary key SiteId
• table SiteUsers should have primary key SiteUserId
• table SiteUsers should have index on SiteId, type

then you can run following query to get results:

select s.url, t.members, t.owners
from Sites s inner join
(
  select
    SiteId, 
    sum(case when Type = 'member' then 1 else 0 end) as members,
    sum(case when Type = 'owners' then 1 else 0 end) as owners
  from SiteUsers
  group by SiteId
) t on s.SiteId = t.SiteId

this way you avoid group by varchar column

sample fiddle http://sqlfiddle.com/#!18/0b9b7/4