5

For example imagine a table as below

select accountid, createdon from account

73C56F61-5FF1-E111-A4F8-005056977FBC    2012-08-28 22:26:47.000
A7C56F61-5FF1-E111-A4F8-005056977FBC    2012-08-28 22:26:48.000
B7C56F61-5FF1-E111-A4F8-005056977FBC    2012-08-28 22:26:48.000
FBC56F61-5FF1-E111-A4F8-005056977FBC    2012-08-28 22:26:49.000
CBC66F61-5FF1-E111-A4F8-005056977FBC    2012-08-28 22:26:54.000
87C66F61-5FF1-E111-A4F8-005056977FBC    2012-08-28 22:26:53.000
53C76F61-5FF1-E111-A4F8-005056977FBC    2012-08-28 22:26:55.000
87C76F61-5FF1-E111-A4F8-005056977FBC    2012-08-28 22:26:56.000
2ED89924-5CFC-E111-A4F8-005056977FBC    2012-09-11 22:01:51.000
C0D79924-5CFC-E111-A4F8-005056977FBC    2012-09-11 22:01:49.000

then in January 2012 the count was 10 accounts with this query 

select count(*) from account

Let's say I have 5 new accounts in February 2012, then querying count(*) in February 2012 returns 15 accounts. If I have 10 new accounts in March 2012, then then querying count(*) in March 2012 returns 35 accounts in total.

I'd like to get the below results

2012-january-10 accounts
2012-febrary-15 accounts
2012-march-  35 accounts

creation date in table account is date,

but if I do this query

select year(createdon), month(createdon), count(*) quantity
from accounts
group by year(createdon), month(createdon)

I get the below result instead:

2012-january-10 accounts
2012-febrary-5 accounts
2012-march-  20 accounts

then how can I get the first result with a simple query? not loops

4

1 回答 1

4

你需要一个滚动总数。在 SQL Server 中有几种方法可以做到这一点（请参阅Calculate a Running Total in SQL Server），但对于 SQL 2005，最简单（不是最快）的方法是自连接

with cte as (
    select
        datename(year, createdon) + '-' + datename(month, createdon) as name,
        convert(nvarchar(6), createdon, 112) as grp,
        count(*) as cnt
    from account
    group by
        datename(month, createdon), datename(year, createdon),
        convert(nvarchar(6), createdon, 112)
)
select
    c1.name, sum(c2.cnt) as cnt
from cte as c1
    inner join cte as c2 on c2.grp <= c1.grp
group by c1.name

sql fiddle demo

于 2013-08-26T17:04:52.803 回答