从 SQL Server 获取逐月、逐年、逐日的记录

Question

我正在寻找一种快速简便的 SQL 查询来逐年、逐月和逐日获取记录。

我的数据库示例：

ID - DATE        - CatID - VALUE
1  - 2013-08-06  -  32   -  243
2  - 2013-08-16  -  2    -  45
3  - 2013-08-21  -  2    -  1
4  - 2013-08-05  -  32   -  450
5  - 2013-08-05  -  32   -  449
6  - 2013-08-05  -  32   -  11
7  - 2013-08-01  -  2    -  221
8  - 2013-08-02  -  32   -  0
9  - 2013-08-02  -  32   -  0
10 - 2013-08-02  -  32   -  987
...
..

我在这张表上有超过 200 万条记录。

第一：

我想获得 CatID = 32 和当天的单次/更高记录。

结果：

1  - 2013-08-06  -  32   -  243
4  - 2013-08-05  -  32   -  450
10 - 2013-08-02  -  32   -  987

第二：

我想在 MONTH 之前得到相同的结果。

我怎样才能用 SQL SERVER 做到这一点。目前是 Sql Server 2012。

也使用 C# 5，如果您想使用，请使用 EF 6。

压倒选民，这个问题成了“热门问题”。现在请解释为什么投反对票？

Answer 1

以下假设您想要的不仅仅是每个日期/月份的最大值，还想知道该行的 id 等。

每天...

WITH
  sorted AS
(
  SELECT
    *,
    ROW_NUMBER() OVER (PARTITION BY CatID,
                                    Date
                           ORDER BY Value DESC)   AS ordinal_cat_date
  FROM
    yourTable
)
SELECT
  *
FROM
  sorted
WHERE
  ordinal_cat_date = 1
  AND catID = 32

每月...

WITH
  sorted AS
(
  SELECT
    *,
    ROW_NUMBER() OVER (PARTITION BY CatID,
                                    DATEADD(MONTH, DATEDIFF(MONTH, 0, Date), 0)
                           ORDER BY Value DESC)   AS ordinal_cat_month
  FROM
    yourTable
)
SELECT
  *
FROM
  sorted
WHERE
  ordinal_cat_month = 1
  AND catID = 32

在这两种情况下，ROW_NUMBER() 函数都会为catID, date 中的每个PARTITION 创建一个数字序列(1,2,3,4,etc)，按value 的降序排列。具有1 值的行是该分区中具有最高值的行。

公式DATEADD(MONTH, DATEDIFF(MONTH, 0, Date), 0) 只是将Date 向下舍入到当月的第一天，从而为整个月份创建分区。

如果这些行都共享相同的最高值，则每天/每月会带来多行的其他一些选项...

SELECT  -- Daily Version
  *
FROM
  yourTable
WHERE
  NOT EXISTS (SELECT *
                FROM yourTable   AS lookup
               WHERE lookup.CatID = yourTable.CatID
                 AND lookup.Value > yourTable.Value
                 AND lookup.Date  = yourTable.Date
             )
  AND CatID = 32

SELECT  -- Monthly Version
  *
FROM
  yourTable
WHERE
  NOT EXISTS (SELECT *
                FROM yourTable   AS lookup
               WHERE lookup.CatID  = yourTable.CatID
                 AND lookup.Value >  yourTable.Value
                 AND lookup.Date  >= DATEADD(MONTH,   DATEDIFF(MONTH, 0, yourTable.Date), 0)
                 AND lookup.Date  <  DATEADD(MONTH, 1+DATEDIFF(MONTH, 0, yourTable.Date), 0)
             )
  AND CatID = 32

或者可能……

SELECT  -- Daily Version
  *
FROM
  yourTable
INNER JOIN
(
  SELECT
    Date,
    CatID,
    MAX(Value)   AS max_value
  FROM
    yourTable
  GROUP BY
    CatID,
    Date
)
  AS lookup
    ON  yourTable.Date  = lookup.Date
    AND yourTable.CatID = lookup.CatID
    AND yourTable.Value = lookup.max_value
WHERE
  yourTable.CatID = 32

SELECT  -- Monthly Version
  *
FROM
  yourTable
INNER JOIN
(
  SELECT
    CatID,
    DATEADD(MONTH, DATEDIFF(MONTH, 0, yourTable.Date), 0)   AS month_start,
    MAX(Value)   AS max_value
  FROM
    yourTable
  GROUP BY
    CatID,
    DATEADD(MONTH, DATEDIFF(MONTH, 0, yourTable.Date), 0)
)
  AS lookup
    ON  yourTable.Date  >= lookup.month_start
    AND yourTable.Date  <  DATEADD(MONTH, 1, lookup.month_start)
    AND yourTable.CatID  = lookup.CatID
    AND yourTable.Value  = lookup.max_value
WHERE
  yourTable.CatID = 32