如何删除子查询？

Question

I have to get all count for all orders, which was passed to query, even for null. I have developed such kind of query in SQL Server 2008R2:

SELECT COUNT(UserId), b.OrderId
FROM @OrderList b
LEFT JOIN 
(
    SELECT DISTINCT pat_base.UserId, prov_list.OrderId
    FROM Users pat_base
    LEFT JOIN Assetments hlth_asmt ON hlth_asmt.UserId = pat_base.UserId
        AND hlth_asmt.Date BETWEEN @StartDate AND @EndDate
    LEFT JOIN Logs proc_log ON proc_log.UserId = pat_base.UserId 
        AND proc_log.PLDATE BETWEEN @StartDate AND @EndDate
    LEFT JOIN Encounter med_enc_log ON med_enc_log.UserId = pat_base.UserId
        AND med_enc_log.EncounterDate BETWEEN @StartDate AND @EndDate
    LEFT JOIN @OrderList prov_list ON
        hlth_asmt.OrderId = prov_list.OrderId
        OR proc_log.OrderId = prov_list.OrderId
        OR med_enc_log.OrderId = prov_list.OrderId
    WHERE hlth_asmt.UserId IS NOT NULL
        OR proc_log.UserId IS NOT NULL
    OR med_enc_log.UserId IS NOT NULL
) a ON a.OrderId=b.OrderId
GROUP BY b.OrderId

If I remove joined subquery, order is being cut with many-to-many relation. Is it possible to replace subquery with join?

PS Here is output now:

            OrderId
----------- -----------
8           10001
0           10003

Output without subquery:

            OrderId
----------- -----------
8           10001

Answer 1

you have to change

LEFT JOIN @OrderList prov_list ON

to

RIGHT JOIN @OrderList prov_list ON

if you use left join you filter out orders wich are in @OrderList but dont have expected user

so the query would be like

SELECT COUNT(UserId), prov_list.OrderId
    FROM Users pat_base
    LEFT JOIN Assetments hlth_asmt ON hlth_asmt.UserId = pat_base.UserId
        AND hlth_asmt.Date BETWEEN @StartDate AND @EndDate
    LEFT JOIN Logs proc_log ON proc_log.UserId = pat_base.UserId 
        AND proc_log.PLDATE BETWEEN @StartDate AND @EndDate
    LEFT JOIN Encounter med_enc_log ON med_enc_log.UserId = pat_base.UserId
        AND med_enc_log.EncounterDate BETWEEN @StartDate AND @EndDate
    RIGHT JOIN @OrderList prov_list ON
        hlth_asmt.OrderId = prov_list.OrderId
        OR proc_log.OrderId = prov_list.OrderId
        OR med_enc_log.OrderId = prov_list.OrderId
    WHERE hlth_asmt.UserId IS NOT NULL
        OR proc_log.UserId IS NOT NULL
    OR med_enc_log.UserId IS NOT NULL
group by prov_list.OrderId

edit

yeah, that means that those rows were filtered out in where clause

change it to

WHERE hlth_asmt.UserId IS NOT NULL
OR proc_log.UserId IS NOT NULL
OR med_enc_log.UserId IS NOT NULL
OR pat_base IS NULL

but i think what it means, is that you have inconsistent data within db

Answer 2

You can return the result from a single query, but I'd recommend not using a join. Instead, remove your outer query, and change to inner query to use a windowing function on the count. So your SELECT clause would be something similar to:

SELECT DISTINCT
   pat_base.UserId,
   prov_list.OrderId,
   COUNT(pat_base.UserId) OVER (PARTITION BY prov_list.OrderId) as UserCount