Get the latest row based on condition

user_8669616 0
Viewed 40

I have a table material -- HERE IS THE FIDDLE

+--------+-----+-------------------+----------------+-----------+          
| ID     | REV | name              | Description    | curr      |
+--------+-----+-------------------+----------------+-----------+
| 211-32 | 001 | Screw 1.0         | Used in MAT 1  | READY     |
| 211-32 | 002 | Screw 2 plus      | can be Used-32 | WITHDRAWN |
| 212-41 | 001 | Bolt H1           | Light solid    | READY     |
| 212-41 | 002 | BOLT H2+Form      | Heavy solid    | READY     |
| 101-24 | 001 | HexHead 1-A       | NOR-1          | READY     |
| 101-24 | 002 | HexHead Spl       | NOR-22         | READY     |
| 423-98 | 001 | Nut Repair spare  | NORM1          | READY     |
| 423-98 | 002 | Nut Repair Part-C | NORM2          | WITHDRAWN |
| 423-98 | 003 | Nut SP-C          | NORM2+NORM1    | NULL      |
| 654-01 | 001 | Bar               | Specific only  | WITHDRAWN |
| 654-01 | 002 | Bar rod-S         | Designed+Spe   | WITHDRAWN |
| 654-01 | 003 | Bar OPG           | Hard spec      | NULL      |
+--------+-----+-------------------+----------------+-----------+

Here each ID can have multiple revisions. I want to take latest revisions (i.e highest of 001,002,003 etc.,). But If the latest revision has curr as either NULL(string) or WITHDRAWN then I have take the previous revision and its corresponding value. If even that's curr is NULL or WITHDRAWN I have to again go to previous revision. If all the revision has the same issue then we can ignore it. so the expected output is

+--------+-----+------------------+---------------+-------+
| ID     | REV | name             | Description   | curr  |
+--------+-----+------------------+---------------+-------+
| 211-32 | 001 | Screw 1.0        | Used in MAT 1 | READY |
| 212-41 | 002 | BOLT H2+Form     | Heavy solid   | READY |
| 101-24 | 002 | HexHead Spl      | NOR-22        | READY |
| 423-98 | 001 | Nut Repair spare | NORM1         | READY |
+--------+-----+------------------+---------------+-------+

I have tried below code, but i'm not sure how to take previous revision.

with cte as (
select *,dense_rank() over (partition by id order by rev desc) as DR ,
lead(curr) over (partition by id order by rev desc) LEAD_CURR
from material )
select * from cte where DR = 1 and curr='READY'
union all
select * from cte where LEAD_CURR='READY' and DR=2
union all
select * from cte where LEAD_CURR='READY' and DR=3
sql-server-2012 sql-server sql
1 Answers

This sounds like filtering and then calculating the row number:

select m.*
from (select m.*,
             row_number() over (partition by id order by rev desc) as seqnum
      from material m
      where curr is not null and curr not in ( 'WITHDRAWN', 'NULL' ) 
     ) m
where seqnum = 1;

You can also do this using a correlated subquery:

select m.*
from material m
where m.rev = (select max(m2.rev)
               from material m2
               where m2.id = m.id and
                     curr is not null and curr not in ( 'WITHDRAWN', 'NULL' ) 
              );

Here is a db<>fiddle.

Note: It is quite non-traditional to store the string 'NULL' in a column, because that can easily be confused with the SQL "constant" NULL.

Also, your question specifically mentions 'WITHDRAWN' and NULL, but it does not say what other values are allowed. Obviously, the logic in the queries above might be equivalent to curr = 'READY' and you can use that. The above logic follows your description of the problem.

user_11440350
Related