获取行数据的计数

Question

There's a table variable (I'll write it as a regular table here)

CREATE TABLE TEST (memberid int, producttype varchar(7))

This table has hundreds of thousands of rows, but for this example I've added a lot less

Insert into test values(1,'book')
Insert into test values(1,'clothes')
Insert into test values(2,'book')
Insert into test values(3,'book')
Insert into test values(4,'clothes')
Insert into test values(5,'book')
Insert into test values(5,'clothes')
Insert into test values(6,'book')
Insert into test values(7,'book')

I need to get:

• the memberids that have 'book' only
• the memberids that have 'clothes' only
• the memberids that have both 'book' & 'clothes'

eg

Member     Book      Clothes      Both
  1          0          0           1
  2          1          0           0
  3          1          0           0
  4          0          1           0
  5          0          0           1
  6          1          0           0 
  7          1          0           0

I had managed to get it to work with sub-queries, but because of the size of the table it could take over 2 minutes to run.

I would appreciate if anyone knows a better way to achieve this?

Answer 1

One method uses conditional aggregation:

select 
    memberid,
    case when max(producttype) = 'book' then 1 else 0 end book,
    case when min(producttype) = 'clothes' then 1 else 0 end clothes,
    case when min(producttype) <> max(producttype) then 1 else 0 end both
from test
group by memberid

This works because there are only two possible producttype s. If you actually have more, then you need some expressions that are more complicated (and possibly more efficient), such as:

case when count(*) = sum(case when producttype = 'book' then 1 end)
    then 1
    else 0
end book

Answer 2

Use a CTE to get if each member has book and/or clothes :

with cte as (
  select memberid,
    count(distinct case when producttype = 'book' then 1 end) book_flag,
    count(distinct case when producttype = 'clothes' then 1 end) clothes_flag
  from test 
  group by memberid
)
select memberid,
  case when book_flag > clothes_flag then 1 else 0 end book,
  case when clothes_flag > book_flag then 1 else 0 end clothes,
  book_flag * clothes_flag both
from cte

See the demo .
Results:

> memberid | book | clothes | both
> -------: | ---: | ------: | ---:
>        1 |    0 |       0 |    1
>        2 |    1 |       0 |    0
>        3 |    1 |       0 |    0
>        4 |    0 |       1 |    0
>        5 |    0 |       0 |    1
>        6 |    1 |       0 |    0
>        7 |    1 |       0 |    0

Answer 3

A table variable with hundreds of thousands of rows is going to be problematic for you.

If you check your query plan, you'll likely see that the optimizer expects that table variable to only contain one row.

Changing the structure to a local temp table, and perhaps adding an index to producttype , should significantly improve the performance of the query even before you optimize your code.

CREATE TABLE #TEST (memberid int, producttype varchar(7));

CREATE NONCLUSTERED INDEX tempTest ON #TEST(producttype);