0

我想验证从多个 SCD type2 表派生的视图。因此，我必须根据两个（或更多）表中可用的日期组合创建一个日期范围。请参见下面的示例。

表格1：

hid     h1name            h1_strt_dttm                  h1_end_dttm

5        value5          2/17/2008  18:19:50    2/16/2009 23:59:59:59

5       value5-upd1     2/17/2009  00:00:00    4/11/2011 23:59:59

5       value5-upd2     4/12/2011  00:00:00    3/5/2012 23:59:59

5       value5-upd3     3/6/2012   00:00:00    11/11/2012 23:59:59

5       value5-upd4     11/12/2012 00:00:00    12/31/9999 23:59:59

表 2：

hid  h2name        h2_strt_dttm           h2_end_dttm

5   name5          3/7/2008 8:14:44     7/17/2010 23:59:59

5  name5-upd1     7/18/2010 0:00:00     1/1/2011 23:59:59

5  name5-upd2     1/2/2011 0:00:00     3/5/2013 23:59:59

5  name5-upd3     3/6/2013 0:00:00     12/31/9999 23:59:59

结果集：

    id   strt_dttm               end_dttm            h1name      h2name

    5   2/17/2008 18:19:50    3/7/2008 8:14:43       value5       null

    5   3/7/2008 8:14:44        2/16/2009 23:59:59   value5       name5

    5   2/17/2009 00:00:00    7/17/2010 23:59:59   value5-upd1    name5

    5   7/18/2010 00:00:00    1/1/2011 23:59:59     value5-upd1   name5-upd1

    5   1/2/2011 00:00:00      4/11/2011 23:59:59   value5-upd1   name5-upd2

    5   4/12/2011 00:00:00    3/5/2012 23:59:59     value5-upd2   name5-upd2

    5   3/6/2012 00:00:00      11/11/2012 23:59:59   value5-upd3  name5-upd2

    5   11/12/2012 00:00:00  3/5/2013 23:59:59       value5-upd4  name5-upd2

    5   3/6/2013 00:00:00      1/1/9999 00:00:00     value5-upd4  name5-upd3

实现这一目标的步骤：我的方法：

1) 取两个（或更多）表中所有可用的 STRT_DTTM 字段值的并集。

2) 从步骤 1 中可用的上述日期分发/重新创建日期范围。（作为开始日期和结束日期）

3）然后尝试在新创建的日期范围内识别 CDC 字段（h1name 和 h2name ）并将它们放入结果集中。查询：

步骤1：

sel h1.hid hid , h1_strt_dttm dtm 

from  HIST1_DAR h1

union

sel h2.hid hid ,h2_strt_dttm dtm

from  HIST2_DAR h2

第2步：

/***************  Step 2 ********************/

WITH ALL_DATE_CTE (hid,dttm,strt_dttm,end_dttm) AS (

sel hid, 

dtm+INTERVAL '0.000001' SECOND  as dttm ,

dttm-INTERVAL '0.000001' SECOND as strt_dttm,

coalesce(min(strt_dttm-INTERVAL '1' SECOND) over (partition by hid 

order by dttm  ROWS BETWEEN 1 following 

AND 1 following ), cast('9999-01-01 00:00:00.000000' as timestamp(6)) ) as end_dttm

from 

(

/***************  Step 1 ********************/ 

sel h1.hid hid , h1_strt_dttm dtm 

from  DBNAME.HIST1_DAR h1

union 

sel h2.hid hid ,h2_strt_dttm dtm

from  DBNAME.HIST2_DAR h2

/***************  Step 1 ********************/  ) a

order by 1,2  )

/**************** Step 2********************/

第 3 步：

/**************** Step 3********************/

sel

h1.hid id,

cte.strt_dttm,

cte.end_dttm,

h1.h1name, 

h1_strt_dttm , h1_end_dttm,

h2.h2name,

h2_strt_dttm,h2_end_dttm

from DBNAME.HIST1_DAR h1

inner join ALL_DATE_CTE cte

on h1.hid = cte.hid

and cte.dttm between h1_strt_dttm and h1_end_dttm

inner join DBNAME.HIST2_DAR h2

on cte.hid = h2.hid

and cte.dttm between h2_strt_dttm and h2_end_dttm

where h1.hid = 5;

/**************** Step 3********************/

现在问题出在第 2 步：我无法在子查询和 CTE 中使用 order by。如果没有订购开始日期，我无法正确列出日期范围。

任何人都可以指导我寻求解决方案或其他简单的方法。请注意，由于我是 QA，因此我没有对大多数数据库的创建表访问权限。因此，任何使用 select 的建议都会更受重视。我已经包含了用于创建表的 DDL 和 DML。

create table DBNAME. hist1_dar ( hid integer , h1name varchar(20),  h1_strt_dttm timestamp(6),  h1_end_dttm timestamp(6));

create table DBNAME. hist2_dar ( hid integer , h2name varchar(20),  h2_strt_dttm timestamp(6),  h2_end_dttm timestamp(6));

ins  DBNAME. hist1_dar (1, 'value1' , '1987-07-02 08:30:00.000000',  '1989-12-02 23:59:59.000000');

ins  DBNAME. hist1_dar (2, 'value2' , '1997-09-12 05:20:10.000000' ,  '1999-12-12 23:59:59.000000');

ins  DBNAME. hist1_dar (3, 'value3' , '2000-06-30 07:10:50.000000',  '2001-02-10 23:59:59.000000');

ins  DBNAME. hist1_dar (4, 'value4' , '2006-05-23 23:32:13.000000',  '2007-07-03 23:59:59.000000');

ins  DBNAME. hist1_dar (5, 'value5' , '2009-02-17 18:19:50.000000',  '2011-04-11 11:09:59.000000' );


ins  DBNAME. hist1_dar (1, 'value1-upd1' , '1990-01-01 01:01:01.000000',  '9999-12-31 23:59:59.000000');

ins  DBNAME. hist1_dar (2, 'value2-upd1' , '2000-01-01 01:01:01.000000' ,  '9999-12-31 23:59:59.000000');

ins  DBNAME. hist1_dar (3, 'value3-upd1' ,  '2001-02-11 01:01:01.000000',  '2002-07-12 23:59:59.000000');

ins  DBNAME. hist1_dar (3, 'value3-upd2' ,  '2002-07-13 01:01:01.000000',  '9999-12-31 23:59:59.000000');

ins  DBNAME. hist1_dar (4, 'value4-upd1' ,  '2007-07-03 01:01:01.000000' ,  '2007-10-17 23:59:59.000000' );

ins  DBNAME. hist1_dar (4, 'value4-upd2' ,  '2007-10-18 01:01:01.000000' ,  '2007-12-23 23:59:59.000000' );

ins  DBNAME. hist1_dar (4, 'value4-upd3' ,  '2007-12-24 01:01:01.000000' , '9999-12-31 23:59:59.000000');

ins  DBNAME. hist1_dar (5, 'value5-upd1' , '2009-02-17 01:01:01.000000',  '2011-04-11 23:59:59.000000' );

ins  DBNAME. hist1_dar (5, 'value5-upd2' ,  '2011-04-11 01:01:01.000000', '2012-03-05 23:59:59.000000' );

ins  DBNAME. hist1_dar (5, 'value5-upd3' ,   '2012-03-05 01:01:01.000000', '2012-11-11 23:59:59.000000'  );

ins  DBNAME. hist1_dar (5, 'value5-upd4' ,   '2012-11-12 01:01:01.000000' , '9999-12-31 23:59:59.000000');


ins  DBNAME. hist2_dar (1, 'name1' , '1988-08-12 18:20:10.000000',  '1990-03-02 23:59:59.000000');

ins  DBNAME. hist2_dar (2, 'name2' ,  '1993-05-19 15:12:30.000000', '1999-11-01 23:59:59.000000');

ins  DBNAME. hist2_dar (3, 'name3' , '2002-09-20 17:19:57.000000',  '2003-04-10 23:59:59.000000');

ins  DBNAME. hist2_dar (4, 'name4' , '2004-04-18 13:38:23.000000' , '2005-07-28 23:59:59.000000');

ins  DBNAME. hist2_dar (5, 'name5' , '2008-03-07 08:14:44.000000',  '2010-07-17 23:59:59.000000');



ins  DBNAME. hist2_dar (1, 'name1-upd1' ,  '1990-03-03 01:01:01.000000',  '9999-12-31 23:59:59.000000');

ins  DBNAME. hist2_dar (2, 'name2-upd1' ,   '1999-11-02 01:01:01.000000',  '9999-12-31 23:59:59.000000');

ins  DBNAME. hist2_dar (3, 'name3-upd1' ,  '2003-04-11 01:01:01.000000' ,  '9999-12-31 23:59:59.000000');

ins  DBNAME. hist2_dar (4, 'name4-upd1' , '2005-07-28 21:28:56.000000' ,  '2008-08-19 23:59:59.000000' );

ins  DBNAME. hist2_dar (4, 'name4-upd2' ,  '2008-08-20 01:01:01.000000' ,  '2007-12-23 23:59:59.000000' );

ins  DBNAME. hist2_dar (5, 'name5-upd1' , '2010-07-18 01:01:01.000000',  '2011-01-01 23:59:59.000000' );

ins  DBNAME. hist2_dar (5, 'name5-upd2' ,  '2011-01-02 01:01:01.000000', '2013-03-05 23:59:59.000000' );

ins  DBNAME. hist2_dar (5, 'name5-upd3' ,   '2013-03-06 01:01:01.000000',  '9999-12-31 23:59:59.000000'  );
4

1 回答 1

0

您的 INSERT 与您的示例不匹配。

如果您的第 2 步返回正确的数据，您可以简单地添加 LAST_VALUE：

WITH cte (
    hid
    ,strt_dttm
    ,end_dttm
    ,h1name
    ,h2name
    )
AS (
    SELECT hid
        ,dttm
        ,coalesce(min(dttm - INTERVAL '1' SECOND) OVER (
                PARTITION BY hid ORDER BY dttm ROWS BETWEEN 1 following
                        AND 1 following
                ), TIMESTAMP '9999-01-01 00:00:00') AS end_dttm
        ,h1name
        ,h2name
    FROM (
        /***************  Step 1 ********************/
        SELECT h1.hid hid
            ,h1_strt_dttm dttm
            ,h1name
            ,cast(NULL AS VARCHAR(20)) AS h2name
        FROM HIST1_DAR h1

        UNION ALL

        SELECT h2.hid hid
            ,h2_strt_dttm dttm
            ,cast(NULL AS VARCHAR(20))
            ,h2name
        FROM HIST2_DAR h2
            /***************  Step 1 ********************/
        ) a
        --order by 1,2 
    )
SELECT hid
    ,strt_dttm
    ,end_dttm
    ,last_value(h1name ignore nulls) OVER (
        PARTITION BY hid ORDER BY strt_dttm
        )
    ,last_value(h2name ignore nulls) OVER (
        PARTITION BY hid ORDER BY strt_dttm
        )
FROM cte
于 2015-08-31T17:36:30.133 回答