Question

我已尝试过多种方法，但无法获得正确的结果集。我不愿意在代码和处理中恢复处理这个问题。进行多次查询，因为我觉得那里有一个相当简单的解决方案。

包含流程信息的2个表。

过程：

ID (key)
PARENT_ID
ROOT_ID

过程数据

ID (key)
PROCESS_ID (key in process)
type
col1        
col2

我的数据是这样的;

过程

id, root_id, parent_id
100, 1, 10
101, 1, 10
102, 1, 10
10, 1,  1

过程数据

id,process_id,type,col1, col2
201,100,X,NULL,NULL
202,100,X,NULL,NULL
203,100,TYPE1,My Data for Type1,NULL
301,100,X,NULL,NULL
302,100,X,NULL,NULL
303,100,TYPE1,My Data for Type1,NULL
401,100,X,NULL,NULL
402,100,X,NULL,NULL
403,100,TYPE1,My Data for Type1,NULL
503,10,X,NULL,NULL
503,10,TYPE2,NULL,My data for type2

期望的结果

root_id,parent_id,process_id,col1,col2
1,100,203,My data for type1,My data for type2
1,101,303,My data for type1,My data for type2
1,102,403,My data for type1,My data for type2

您可以将其视为流程层次结构;

ROOT PROCESS
TOP_LEVEL PROCESS type=TYPE2 (has a label that applies to all it's children)
    (it's possible that between the type1 and type2 header above there is another process_id -> parent_id iteration needed - can ignore for now)
    PROCESS type=type1  (has data about this instance of the process)
    PROCESS type=type1  (has data about this instance of the process)
    PROCESS type=type1  (has data about this instance of the process)

TOP_LEVEL PROCESS type=TYPE2 (has a label that applies to all it's children)
    PROCESS type=type1  (has data about this instance of the process)
    PROCESS type=type1  (has data about this instance of the process)
    PROCESS type=type1  (has data about this instance of the process)

我最接近的是;

with processes as (     -- this gets me all the process_id's for the root process - there will be lots of these with different types
                select  id
                        ,parent_id
                from process as p 
                where root_process_id='1'
), processdatainfo as (             -- this should get the type 1 child processes
        select  pd.id
                ,pd.process_id
                ,col1
        from processdata as pd 
        where pd.process_id in ( select id from processes )
        and pd.type = 'TYPE1'
), parentprocesses as (             -- this should get the type 2 parents that have the label for the children
        select  process_id
                ,col2
        from db2inst1.processlog
        where process_id in ( select id from processes )
        and type = 'TYPE2'
)
select *                            -- and the magic.....well doesn't happen....
from processes 
    inner join processdatainfo 
        on (processes.id = processdatainfo.process_id)
    left join parentprocesses
        on (parentprocesses.process_id = processes.parent_id)

在我的结果中，col2始终为NULL 虽然这是我猜的一件事，但我只获得了我预期的3行。我已经使用子查询使用嵌套选择来回转，但这种方法似乎最接近。我已经查看了递归查询，但这个问题似乎并不合适，因为虽然我确实需要处理树，但我需要为结果获取一行而不是每行递归一行。

同样递归看起来很困难 - 它需要在到达type2进程时停止递归，而不是一直到root_id 我无法使用存储过程，因为在这种情况下我无法更新服务器 - 我的应用程序是它的客户端。

虽然表格很大 - 介于60M到140M之间，但对根进程ID进行过滤会将process_id列表降低到10-20，这就是为什么我认为上述WITH会有所帮助。

我认为我需要的是;

我的root进程的process_id列表。这限制了结果集得到我的类型列表1和他们的parentid＆＃39;
从中获取type2的列表，其中parentid是type2记录的processid

听起来很简单......

加入子查询或递归？

0 个答案: