分組結果我有這樣的一個表:SQL查詢基於序列
ID Seq Amt
1 1 500
1 2 500
1 3 500
1 5 500
2 10 600
2 11 600
3 1 700
3 3 700
我想組連續的序列號爲一行是這樣的:
ID Start End TotalAmt
1 1 3 1500
1 5 5 500
2 10 11 1200
3 1 1 700
3 3 3 700
請幫助達到這個結果。
分組結果我有這樣的一個表:SQL查詢基於序列
ID Seq Amt
1 1 500
1 2 500
1 3 500
1 5 500
2 10 600
2 11 600
3 1 700
3 3 700
我想組連續的序列號爲一行是這樣的:
ID Start End TotalAmt
1 1 3 1500
1 5 5 500
2 10 11 1200
3 1 1 700
3 3 3 700
請幫助達到這個結果。
WITH numbered AS (
SELECT
ID, Seq, Amt,
SeqGroup = ROW_NUMBER() OVER (PARTITION BY ID ORDER BY Seq) - Seq
FROM atable
)
SELECT
ID,
Start = MIN(Seq),
[End] = MAX(Seq),
TotalAmt = SUM(Amt)
FROM numbered
GROUP BY ID, SeqGroup
ORDER BY ID, Start
;
+1花了我一會兒才弄清楚' - Seq'技巧,並驗證(對於我自己)不會有任何碰撞,但OP不能要求比這更好的。 – 2011-02-23 08:39:44
+1我懷疑ROW_NUMBER(或RANK,...)會幫助我,但放棄了我的(更復雜,可能更慢)的解決方案。 – 2011-02-23 08:50:52
@Lieven:當我說這樣的讚美時,請相信我,不需要任何正式的upvoting。 :)謝謝你,你很親切!順便說一句,你不應該刪除你的答案。在我看來,另一個CTE演示從來沒有太多,因爲這是一個相當棘手的技術,這是最好的例子。我也沒有注意到Martin或Quassnoi在這裏的存在。 :) – 2011-02-23 08:51:37
請嘗試以下查詢。
select id, min(seq), max(seq), sum(amt) from table group by id
哎呀,對不起,這是錯誤的查詢,因爲你需要序列
這似乎很好地工作。 @breakingRows
將包含破壞id
和seq
序列的所有行(即如果id
改變或者如果seq
不比以前的seq
多1)。使用該表格,您可以在@temp
之內選擇這樣的序列的所有行。但是我必須補充說,由於所有的子查詢,性能可能並不是那麼好,但是你需要測試以確定。
declare @temp table (id int, seq int, amt int)
insert into @temp select 1, 1, 500
insert into @temp select 1, 2, 500
insert into @temp select 1, 3, 500
insert into @temp select 1, 5, 500
insert into @temp select 2, 10, 600
insert into @temp select 2, 11, 600
insert into @temp select 3, 1, 700
insert into @temp select 3, 3, 700
declare @breakingRows table (ctr int identity(1,1), id int, seq int)
insert into @breakingRows(id, seq)
select id, seq
from @temp t1
where not exists
(select 1 from @temp t2 where t1.id = t2.id and t1.seq - 1 = t2.seq)
order by id, seq
select br.id, br.seq as start,
isnull ((select top 1 seq from @temp t2
where id < (select id from @breakingRows br2 where br.ctr = br2.ctr - 1) or
(id = (select id from @breakingRows br2 where br.ctr = br2.ctr - 1) and
seq < (select seq from @breakingRows br2 where br.ctr = br2.ctr - 1))
order by id desc, seq desc),
br.seq)
as [end],
(select SUM(amt) from @temp t1 where t1.id = br.id and
t1.seq <
isnull((select seq from @breakingRows br2 where br.ctr = br2.ctr - 1 and br.id = br2.id),
(select max(seq) + 1 from @temp)) and
t1.seq >= br.seq)
from @breakingRows br
order by id, seq
感謝您的努力。解決方案非常完美!但我可以只標記一個答案作爲最佳答案。 – Nagesh 2011-02-23 10:43:09
沒問題。 Andriy的解決方案顯然是最好的解決方案。 – 2011-02-23 12:11:37
嗯,有可能是一個更優雅的方式來做到這一點(的東西我有提示),但這裏的一種方法,將如果你使用一個版本的SQL Server接受公共表表達式工作:
use Tempdb
go
create table [Test]
(
[id] int not null,
[Seq] int not null,
[Amt] int not null
)
insert into [Test] values
(1, 1, 500),
(1, 2, 500),
(1, 3, 500),
(1, 5, 500),
(2, 10, 600),
(2, 11, 600),
(3, 1, 700),
(3, 3, 700)
;with
lower_bound as (
select *
from Test
where not exists (
select *
from Test as t1
where t1.id = Test.id and t1.Seq = Test.Seq - 1
)
),
upper_bound as (
select *
from Test
where not exists (
select *
from Test as t1
where t1.id = Test.id and t1.Seq = Test.Seq + 1
)
),
bounds as (
select id, (select MAX(seq) from lower_bound where lower_bound.id = upper_bound.id and lower_bound.Seq <= upper_bound.Seq) as LBound, Seq as Ubound
from upper_bound
)
select Test.id, LBound As [Start], UBound As [End], SUM(Amt) As TotalAmt
from Test
join bounds
on Test.id = bounds.id
and Test.Seq between bounds.LBound and bounds.Ubound
group by Test.id, LBound, UBound
drop table [Test]
非常感謝您的努力。你的解決方案工作正常。 – Nagesh 2011-02-23 11:01:33
由於舍甫琴科已經貼金溶液,這裏是使用UPDATE語句來得到一個臨時表的結果我取,只是爲了好玩。
declare @tmp table (
id int, seq int, amt money, start int, this int, total money,
primary key clustered(id, seq))
;
insert @tmp
select *, start=seq, this=seq, total=convert(money,amt)
from btable
;
declare @id int, @seq int, @start int, @amt money
update @tmp
set
@amt = total = case when id = @id and seq = @seq+1 then @amt+total else amt end,
@start = start = case when id = @id and seq = @seq+1 then @start else seq end,
@seq = this = seq,
@id = id = id
from @tmp
option (maxdop 1)
;
select id, start, max(this) [end], max(total) total
from @tmp
group by id, start
order by id, start
注:
它的工作原理,但這是一些新的SQL Server的東西給我。你已經設法發佈一些有價值的東西,只要玩得開心,謝謝。 :) – 2011-02-23 09:39:02
'選擇(maxdop 1)'有特定的原因嗎?編輯:我想我只是想通了。你用'@ tmp'本身的數據更新'@ tmp'。多線程會干擾,對嗎? – 2011-02-23 10:15:39
@Sem是的,因爲這是一個脆弱的查詢。在UPDATE查詢中沒有辦法使用ORDER BY,所以AFAIK應該能夠工作 - 但不能保證 - 如果我們按照需要的順序進行聚類,並且強制maxdop 1。**我再次強調**(並且也投了票),Andriy的答案是黃金之一。 – RichardTheKiwi 2011-02-23 10:18:24
的SQL版本服務器? (id + seq)是唯一的嗎? – RichardTheKiwi 2011-02-23 08:45:04