何时计算空值的 SQL 案例

Question

I'm writing a query, data looks like this:

Data:

ID	name	date_completed	version
1	Sydney	2021-01-01	A
2	Melbourne	2021-01-5	A
3	Sydney	2021-01-10	B
4	Sydney	2021-02-01	A
5	Melbourne	2021-02-07	A
6	Melbourne	2021-02-13	A

My query:

SELECT name
         , '01-01-21' AS Date
         , (case
                when version = 'A' then 'A'
                when version = 'B' then 'B'
        end)          AS 'Type'
         , (case
                when version = 'A' then count(ID)
                when version = 'B' then count(ID)
        end)          AS 'Count'
    FROM table
    WHERE date_completed between '2021-01-01' and '2021-01-31'
    GROUP BY name, version

    UNION

SELECT name
         , '01-02-21' AS Date
         , (case
                when version = 'A' then 'A'
                when version = 'B' then 'B'
        end)          AS 'Type'
         , (case
                when version = 'A' then count(ID)
                when version = 'B' then count(ID)
        end)          AS 'Count'
    FROM table
    WHERE date_completed between '2021-02-01' and '2021-02-28'
    GROUP BY name, version

Output:

name	date	Type	count
Sydney	01-01-21	A	1
Melbourne	01-01-21	A	1
Sydney	01-01-21	B	1
Sydney	01-02-21	A	1
Melbourne	01-02-21	A	2

The issue I'm facing is that because version A/B doesn't exist for some months it will not show a row with count = 0

I'm trying to create an output like this:

name	date	Type	count
Sydney	01-01-21	A	1
Melbourne	01-01-21	A	1
Sydney	01-01-21	B	1
Melbourne	01-01-21	B	0
Sydney	01-02-21	A	1
Melbourne	01-02-21	A	2
Sydney	01-02-21	B	0
Melbourne	01-02-21	B	0

Is this something that can be done without having to create a separate table with name, date & type and using LEFT JOIN between the two tables?

Also is there a better way of dealing with querying data within a specific month without having to union multiple select statements?

Thanks

Answer 1

You need to find all combination of name - month - version and then left join to your table and then perform the count()

If you have individual table of name, version, you may use that instead of finding the distinct value from the table ( select distinct name from tbl )

with 
name_mth_ver as
(
    select name, mth_st, mth_en, version
    from
    (
        select distinct name
        from   tbl
    ) n
    cross join
    (
         select distinct
                mth_st = convert(date, dateadd(month, datediff(month, 0, date_completed), 0)),
                mth_en = convert(date, dateadd(month, datediff(month, 0, date_completed) + 1, -1))
         from   tbl
    ) m
    cross join
    (
        select distinct version
        from   tbl
    ) v
)
select n.name
       , n.mth_st AS [Date]
       , n.version AS [Type]
       , count(t.ID) AS [Count]
from   name_mth_ver n
       left join tbl t on  n.name    = t.name
                       and n.mth_st <= t.date_completed
                       and n.mth_en >= t.date_completed
                       and n.version = t.version
group by n.name, n.mth_st, n.version