SQL Server-替代包含聚合或子查询的表达式的函数的替代方法

Question

First of all sorry for the confusing code sample, simplifying my problem is proving tricky.

select [...]
    SUM(case when 
        (select [CustID] from [Source3] S3
            where (S3.[CustID] = S1.[StaffID1] or S3.[CustID] = S1.[StaffID2])
            and substring(datename(dw,S1.[AppDateTime]),1,3)=substring(S3.[DoW],1,3)) is not null
        then 1 else 0 end), [...]
from [Source 1] S1
left join [Source 2] S2
    on S2.[SbSpcID]=S1.[SbSpcID]
group by S1.[Condition1], S1.[Condition2]

Basically I need to grab (as just 1 field in this query) a number indicating how many times my Source 3 table contains lines that match the line of my joined table (Source 1 and 2 joined together in this example).

This join needs to be whenever the [CustID] field of Source 3 equals either the StaffID1 or StaffID2 fields of Source 1 and the day of the week of Source 3's [DoW] field matches the day of the week of the AppDateTime field of Source 1.

substring(datename(dw,S1.[AppDateTime]),1,3) returns things like 'Mon', 'Tue', etc.. which I use because the dates are stored as names ('Thurs', 'Wed') in Source 3.

I am getting the 'Cannot perform an aggregate function on an expression containing an aggregate or subquery' error. Normally to solve this I would join Source 3 to the table containing Source 2 and 1, however because I am expecting multiple matches that means that the line count would increase which would ruin the results of other (currently working fine) results of this query.

...

..

.

Help! My brain hurts.

---------------------EDIT:---------------------

Sample Data:

Source 1&2:

[Condition1]  [Condition2]  [SbSpcID]  [StaffID1]  [StaffID2]  [AppDateTime]
Hosp          Doc           xxx        Con123       NULL       2018-02-10 16:00
Hosp          Nur           xxx        NULL         Con123     2018-03-15 21:05
Clin          Doc           xxx        Con125       NULL       2018-04-12 18:30
Hosp          DIT           xxx        NULL         NULL       2018-02-25 16:01
Hosp          Reg           xxx        NULL         Con126     2018-06-30 09:45
Hosp          Doc           xxx        Con321       NULL       2018-03-11 11:55
Hosp          Nur           xxx        NULL         Con125     2018-01-01 06:29
Hosp          Doc           xxx        Con125       NULL       2018-02-01 17:00

Source 3:

[CustID]  [Dow]
Con123    Wed
Con123    Thurs
Con123    Fri
Con123    Sunday
Con125    Mon
Con125    Tues
Con126    Sat
Con321    Mon

Results:

[Condition1]  [Condition2]  [Query]
Hosp          Doc           0
Hosp          Nur           2
Clin          Doc           0
Hosp          DIT           0
Hosp          Reg           1

Though there are a lot of [Hosp] [Doc] entries they don't match the day of the week of those in Source 3. Whereas both [Hosp][Nur] entries match the day of the week so they both count.

Answer 1

You can still LEFT JOIN Source 3 to Sources 1 & 2. Your join will be weird and inefficient, since you're storing the weekdays as text, instead of their integer values.

With your sample data (I combined Sources 1 & 2 into 1 CTE):

-- sample data
WITH 
    Source_1_2(Condition1, Condition2, StaffID1, StaffID2, AppDateTime)
    AS
    (
        SELECT * FROM
        (
            VALUES
            ('Hosp',          'Doc',    'Con123',   NULL,       '2018-02-10 16:00'),
            ('Hosp',          'Nur',    NULL,       'Con123',   '2018-03-15 21:05'),
            ('Clin',          'Doc',    'Con125',   NULL,       '2018-04-12 18:30'),
            ('Hosp',          'DIT',    NULL,       NULL,       '2018-02-25 16:01'),
            ('Hosp',          'Reg',    NULL,       'Con126',   '2018-06-30 09:45'),
            ('Hosp',          'Doc',    'Con321',   NULL,       '2018-03-11 11:55'),
            ('Hosp',          'Nur',    NULL,       'Con125',   '2018-01-01 06:29'),
            ('Hosp',          'Doc',    'Con125',   NULL,       '2018-02-01 17:00')
        ) v(c1, c2, c3, c4, c5)
    )
    ,Source_3(CustID, Dow) AS
    (
        SELECT * FROM
        (
            VALUES
            ('Con123',    'Wed'),
            ('Con123',    'Thurs'),
            ('Con123',    'Fri'),
            ('Con123',    'Sunday'),
            ('Con125',    'Mon'),
            ('Con125',    'Tues'),
            ('Con126',    'Sat'),
            ('Con321',    'Mon')
        ) v(c1, c2)
    )
-- actual query
SELECT
    Source_1_2.Condition1,
    Source_1_2.Condition2,
    COUNT(Source_3.CustID) AS [Count]
FROM Source_1_2
    LEFT JOIN Source_3 ON
        (Source_3.CustID = Source_1_2.StaffID1 OR Source_3.CustID = Source_1_2.StaffID2)
        AND LEFT(Source_3.Dow, 3) = LEFT(DATENAME(WEEKDAY, Source_1_2.AppDateTime), 3)
GROUP BY 
    Source_1_2.Condition1, 
    Source_1_2.Condition2

giving you this result:

+------------+------------+-------+
| Condition1 | Condition2 | Count |
+------------+------------+-------+
| Hosp       | DIT        |     0 |
| Clin       | Doc        |     0 |
| Hosp       | Doc        |     0 |
| Hosp       | Nur        |     2 |
| Hosp       | Reg        |     1 |
+------------+------------+-------+