当 oracle 中的连接字段之一是多对 1 时，如何在连接 3 个表时根据最大日期获得准确计数？

Question

So, I have 3 tables that I am attempting to get counts for based on a groupid, and a task code. There are a few issues I am having as some of the relationships are many to one, which I think is somehow inflating my counts. I will list my 3 tables with the pertinent attributes.

task_table contains:

task_code - would like to get the counts of each one in a group id, would like to use the latest instance basedon event date.

sol_id -used to join to worktable; many sol_id to one m_id is possible

edate -need to use to get one record

cur_id - where cur_id = 1 in the where clause

worktable contains:

sol_id - used to join to task_table

m_id - used to join to grouptable

grouptable contains:

m_id

groupid- used to group the task_code to get count

I'd like the end result to look like:

group_id    task_count  task
5555        45          A
5555        4           N
5624        67          A
5624        23          O
5624        42          X

I have been attempting to run a number of queries, but the counts I am getting back do not look correct. I am concerned that it is somehow returning more than one instance of the m_id somehow? Here is the query in question:

select  c.groupid, count(c.groupid) group_count, a.task_code from task_table a
join worktable b
on a.sol_id = b.sol_id
join grouptable c
on b.m_id= c.m_id
where   a.cur_id = 1 and a.task_code is not null 
group by c.groupid, a.task_code;

If I add 'edate = (select max(edate) from task_table)' in the where clause, it returns an empty table.

I am unsure how to incorporate edate to get only the newest record that fits the criteria in the where clause. The reason I think I want to use this is because there could be more than one sol_id that is associated with a m_id, so i'd just like to include only the newest record with a cur_id in the count. Thank you for your time.

sample data

task_table

task_code  sol_id  edate   cur_id
A          23      6/7/09    1
A          24      6/4/09    1
A          23      6/10/09   0
B          45      6/2/09    1
B          42      6/3/09    1
C          34      10/8/10   0
C          83      9/10/09   1   

work table

sol_id    m_id
23        1234
24        1234
45        1832
42        1343
83        7623

group table

m_id  group_id
1234   A76
1832   Y23
1343   A76
7623   Y23

looking at these tables, the result should look like the following

group_id    task_count  task
A76       2         A            
Y23       1         C       
 

( A76 should only count sol_id 23 and 42) ( Y23 should only count sol_id 83)

Answer 1

So, there's a conflict in your requested data result. According to your own sample, A76 should have a task_count of 2: sol_id 23, which has Task A, and sol_id 42, which has Task B. It's not possible to have it return a row like you have at your example result table because it would need to group by TASK_CODE, which means losing the COUNT(task_code). Can't have it both ways.

In order to obtain only the most recent edate, I did a separate calculation to location that max(edate) by task_code, then joined it back to obtain the sol_id. If this isn't accurate for your data set, you'll need to determine another way of obtaining max(edate). This works for your sample set.

with recentTasks as (
   select task_code, max(edate) as recentDate
   from task_table m 
   where cur_id = 1
     and task_code is not null 
   group by task_code
), recentTaskWithSols as (
  select m.task_code, m.recentDate as edate, t.sol_id
  from recentTasks m 
  join task_table t on m.task_code = t.task_code AND m.recentDate = t.edate
  where t.cur_id = 1
)
select c.group_id, 
  count(a.sol_id) task_count
from group_table c
join work_table b on c.m_id = b.m_id 
join recentTaskWithSols a on b.sol_id = a.sol_id 
group by c.group_id;

gives the result:

+------------------------+
| GROUP_ID  | TASK_COUNT |
+------------------------+
|    A76    |     2      |
|    Y23    |     1      |
+-----------+------------+

Demo here .