将子查询关联到 select 个最早条目的 ID

Question

我目前正在使用这个：

IF OBJECT_ID('tempdb..#Temp') IS NOT NULL DROP TABLE #Temp

CREATE TABLE #Temp
(
    SomeId INT,
    UtcDateTime DATETIME2
)

INSERT INTO #Temp
    SELECT 1, DATETIME2FROMPARTS(2015,1,1,1,1,1,0,0)
        UNION 
    SELECT 1, DATETIME2FROMPARTS(2015,1,1,2,1,1,0,0)
        UNION 
    SELECT 2, DATETIME2FROMPARTS(2015,1,1,3,1,1,0,0)
        UNION 
    SELECT 2, DATETIME2FROMPARTS(2015,1,12,4,1,1,0,0)
        UNION 
    SELECT 2, DATETIME2FROMPARTS(2015,1,12,5,1,1,0,0)
        UNION 
    SELECT 3, DATETIME2FROMPARTS(2015,1,12,5,1,1,0,0)

SELECT * FROM #Temp ORDER BY UtcDateTime ASC

SELECT 
    *
FROM #Temp AS O1
WHERE UtcDateTime =
(
    SELECT MIN(UtcDateTime) FROM #Temp AS O2 WHERE O1.SomeId = O2.SomeId 
)
AND SomeId =
(
    SELECT MAX(SomeId) FROM #Temp AS O2 WHERE O1.SomeId = O2.SomeId AND O1.UtcDateTime = O2.UtcDateTime
)

目的是 select 每个条目的第一次出现都具有基于 UtcDateTime 的不同 SomeId。换句话说，我在这些行之后：

SomeId  UtcDateTime
1   2015-01-01 01:01:01.0000000
2   2015-01-01 03:01:01.0000000
3   2015-01-12 05:01:01.0000000

上述关联子查询的做法是否正确？

Answer 1

Select 没有具有相同 ID 和较早时间戳的行：

SELECT SomeId, UtcDateTime
FROM Temp AS T1
WHERE NOT EXIST (
    SELECT 1 FROM Temp AS T2
    WHERE T1.SomeId = T2.SomeId
    AND T2.UtcDateTime < T1.UtcDateTime)

Answer 2

两次自连接（提到 3 次相同 table）- 在我看来太多了

1)

SELECT *
FROM #Temp AS O1
WHERE not exists
    ( select 1 from #Temp O2 
      where O2.SomeId = O1.SomeId and O2.UtcDateTime < O1.UtcDateTime)

2)

SELECT *
FROM #Temp AS O1
CROSS APPLY
(
  SELECT TOP 1 O2.UtcDateTime
  FROM #Temp as O2
  WHERE O2.SomeId = O1.SomeId
  ORDER BY O2.UtcDateTime ASC
) O2
WHERE O1.UtcDateTime = O2.UtcDateTime

2.1) 带 group by、min 和 join 的子查询（apply 很可能会导致使用循环连接的计划，而 group by first and join after - to hash join）

3) 示例案例

SELECT O1.id, MIN(O1.UtcDateTime)
FROM #Temp AS O1
GROUP BY O1.id

4) 对于小 tables；请注意，row_number 没有搜索，因此您 "disabling" 所有索引并请求额外的内存来填充新列

SELECT *
FROM
(
  SELECT O1.*,
    ROW_NUMBER() OVER(PARTITION BY O1.SomeID ORDER BY O1.UtcDateTime) as row_no
  FROM #Temp as O1
) O2
WHERE O2.row_no = 1

Answer 3

您还可以使用 ROW_NUMBER 按 [UtcDateTime] 排序并按 [SomeId] 分区以获得第一次出现。

SELECT  *
FROM    (
            SELECT  *,
                    ROW_NUMBER () OVER (PARTITION BY SomeId ORDER BY [UtcDateTime]) RN 
            FROM    #Temp
        ) t
WHERE   Rn = 1

Answer 4

你的没有错，但你可以试试这个（更简单的方法）：

SELECT SomeId, MIN(UTCDateTime) AS UTCDateTime FROM #Temp GROUP BY SomeId

每个 'SomeId'

将捕获第一次出现

将子查询关联到 select 个最早条目的 ID

correlated sub query to select ids of earliest entry

tsql

sql-server

sql-server-2014