Postgres GROUP BY 查看日期范围

Question

我有一个 table 具有 "Code" 值更改的历史记录。每个月，此 table 都会获得一条新记录，其中包含指定月份的 "Code" 的新值。

+----------+------------+------------+------+
| Employee | FromDate   |  ToDate    | Code |
+----------+------------+------------+------+
| Employee | 01/07/2016 | 31/07/2016 |    4 |
| Employee | 01/06/2016 | 30/06/2016 |    2 |
| Employee | 01/05/2016 | 31/05/2016 |    2 |
| Employee | 01/04/2016 | 30/04/2016 |    3 |
| Employee | 01/03/2016 | 31/03/2016 |    3 |
| Employee | 01/02/2016 | 29/02/2016 |    4 |
| Employee | 01/01/2016 | 31/01/2016 |    4 |
+----------+------------+------------+------+

每次 "Code" 更改时，我需要按此数据分组以获得新记录，并为 "From date" 取最小值，为 "To date" 取最大值。数据必须按 "FromDate" 降序排列。通过我的查询，我得到了这个结果：

+----------+------------+------------+------+
| Employee | FromDate   |  ToDate    | Code |
+----------+------------+------------+------+
| Employee | 01/05/2016 | 30/06/2016 |    2 |
| Employee | 01/03/2016 | 30/04/2016 |    3 |
| Employee | 01/01/2016 | 31/07/2016 |    4 |
+----------+------------+------------+------+

它工作正常，但如果相同的 "Code" 有更多的日期范围（请参阅第一个 table 中的 4 代码），我每个代码只有一行。我想用 2 条记录中的 4 个代码得到这个结果，因为它的周期不是连续的，但它被其他代码（3 和 2）打破了：

+----------+------------+------------+------+
| Employee | FromDate   |  ToDate    | Code |
+----------+------------+------------+------+
| Employee | 01/07/2016 | 31/07/2016 |    4 |
| Employee | 01/05/2016 | 30/06/2016 |    2 |
| Employee | 01/03/2016 | 30/04/2016 |    3 |
| Employee | 01/01/2016 | 29/02/2016 |    4 |
+----------+------------+------------+------+

我使用这个查询：

SELECT
    d."Employee",
    MIN (d."FromDate") AS "FromDate",
    MAX (d."ToDate") AS "ToDate",
    d."Code"
FROM
    (
        SELECT
            "Employees"."FromDate",
            "Employees"."ToDate",
            "Employees"."Code",
            "Employees"."Employee"
        FROM
            schema_estelspa."Employees"
        ORDER BY
            "Employees"."FromDate" DESC
    ) d
GROUP BY
    d."Code",
    d."Employee"
ORDER BY
    (MIN(d."FromDate")) DESC

有什么技巧可以得到我想要的结果吗？

日期格式为：dd/MM/yyyy

Answer 1

这里需要设置日期范围，将from_date作为按列分组的一部分。您还需要自行加入才能达到此结果。我准备在 teradata 中关注 SQL。请对您的数据库进行必要的更改（coalesc 用作 if null 表达式，您也可以使用 nvl 或 case 语句）

查询：

SELECT E.EMPLOYEE, E.CODE,COALESCE(ET1.FROMdATE,E.FROMDATE)FROM_DATE ,MAX(E.TODATE)TO_D
FROM EMP_TEST E
LEFT OUTER JOIN EMP_TEST ET1
ON E.EMPLOYEE=ET1.EMPLOYEE
AND E.CODE=ET1.CODE
AND E.FromDate=ET1.ToDate+1
GROUP BY 1,2,3
ORDER BY FROM_DATE

输出：

    Employee    Code    FROM_DATE   TO_D
1   Employee    4   1/1/2016    2/29/2016
2   Employee    2   5/1/2016    6/30/2016
3   Employee    4   7/1/2016    7/31/2016
4   Employee    3   3/1/2016    4/30/2016

Answer 2

连接点

的标准递归解决方案

在实践中，半开区间 (lower_limit <= X < upper_limit) 更容易处理
递归从任何没有 lower neigbor
相邻的片段粘在右边，构建更长的链
最终查询抑制了部分结果

注意：下面的代码不处理重叠间隔。

        -- Table
CREATE TABLE ecode
        ( employee varchar NOT NULL
        , code INTEGER NOT NULL
        , fromdate DATE NOT NULL
        , uptodate DATE NOT NULL
        );
SET datestyle = 'DMY' ;

        -- Data
INSERT INTO ecode(employee, fromdate, uptodate, code) VALUES
 ('Employee','01/07/2016','31/07/2016', 4)
, ('Employee','01/06/2016','30/06/2016', 2)
, ('Employee','01/05/2016','31/05/2016', 2)
, ('Employee','01/04/2016','30/04/2016', 3)
, ('Employee','01/03/2016','31/03/2016', 3)
, ('Employee','01/02/2016','29/02/2016', 4)
, ('Employee','01/01/2016','31/01/2016', 4)
        ;

        -- Convert to half-open interval
UPDATE ecode SET uptodate = uptodate + '1 day'::interval;
-- SELECT * FROM ecode;

WITH RECURSIVE zzz AS (
        SELECT employee, code, fromdate, uptodate
        FROM ecode e0
        WHERE NOT EXISTS ( -- first one in series
                SELECT * FROM ecode nx
                WHERE nx.employee = e0.employee
                AND nx.code = e0.code
                AND nx.uptodate = e0.fromdate
                )
        UNION ALL -- append consecutive intervals
        SELECT e1.employee, e1.code, zzz.fromdate, e1.uptodate
        FROM ecode e1
        JOIN zzz ON zzz.employee = e1.employee
        AND zzz.code = e1.code
        AND zzz.uptodate = e1.fromdate
        )
SELECT * FROM zzz
        -- suppress the partial results
WHERE NOT EXISTS (SELECT * FROM ecode nx
                WHERE nx.employee = zzz.employee
                AND nx.code = zzz.code
                AND nx.fromdate = zzz.uptodate
                )
ORDER BY employee, code, fromdate
        ;

结果：

DROP SCHEMA
CREATE SCHEMA
SET
CREATE TABLE
SET
INSERT 0 7
UPDATE 7
 employee | code |  fromdate  |  uptodate  
----------+------+------------+------------
 Employee |    2 | 2016-05-01 | 2016-07-01
 Employee |    3 | 2016-03-01 | 2016-05-01
 Employee |    4 | 2016-01-01 | 2016-03-01
 Employee |    4 | 2016-07-01 | 2016-08-01
(4 rows)

Postgres GROUP BY 查看日期范围

Postgres GROUP BY looking at dates ranges

sql

postgresql

group-by

gaps-and-islands

结果：