SQL Server查询（更新分组？）

Question

I got the following tables to work with.(Just a sample)

thead

id      date
------------------
2399    01/01/2017
2400    18/07/2017
2300    11/11/2016
...

tstocks

id      product_id  type    stockq  result    
--------------------------------------------
2399    0000001       1        5    
2399    0000001       2       10
2399    0000002       1       15    
2399    0000002       2        2
2400    0000001       1        4
2400    0000001       2        6
2300    0000003       1        0
2300    0000003       2        5

Each product_id has 2 values, type 1 and type2. thead holds the id and the date.

I need to update result, where stockq type is 1 for each group of product_id, where the date of tproducts.id is '01/01/2017'

So it should end up with something like this:

tstocks :

id     product_id  type   stockq    result  
--------------------------------------------
2399    0000001     1        5      15
2399    0000001     2       10
2399    0000002     1       15      17
2399    0000002     2        2
2400    0000001     1        4
2400    0000001     2        6
2300    0000003     1        0
2300    0000003     2        5

I tried with the following query but it says with an more than one value on the subquery error.

update tstocks
set result = (select sum(stockq) 
              from tstocks 
              group by tstocks.product_id) 
from tstocks
inner join thead ON tstocks.id = thead.id
where tstocks.type = '1' and thead.date = '01/01/2017'

Any ideas would be appreciated.

Thanks in advance.

Answer 1

I would use an updatable CTE and window functions:

with toupdate as (
      select s.*, sum(stockq) over (partition by id, product_id) as new_result
      from tstocks
     )
update toupdate
    set result = new_result
    from toupdate join
         thead
         on toupdate.id = thead.id
    where thead.date = '2017-01-01' and toupdate.type = 1;

Note that I changed the date to a standard format and dropped the single quotes around the type. This assumes the columns have the types of date / datetime and number respectively.

You can also follow your reasoning, but with a correlated subquery:

update s
    set result = (select sum(s2.stockq)
                  from tstocks s2
                  where s2.product_id = s.product_id an s2.id = s.id
                 ) 
    from tstocks s inner join
         thead h
         on s.id = h.id
    where s.stockq = 1 and h.date = '2017-01-01';