根据多列选择最大值列

Question

I have the following table documents

+--------+-------------+------------+----------+---------+
| doc_id | module_name | mapping_id | doc_type | version |
+--------+-------------+------------+----------+---------+
|      5 | asdf        | asd        | POI      |       1 |
|      6 | asdf        | asd        | POI      |       2 |
|      7 | asdf        | asd        | CAF      |       1 |
|      8 | asdf        | abc        | POI      |       1 |
|      9 | asdf        | abc        | ISR      |       1 |
|     10 | asdf        | abc        | ISR      |       2 |
|     11 | asdf        | xyz        | POA      |       1 |
|     12 | asdf        | xyz        | CAF      |       1 |
|     13 | asdf        | xyz        | CAF      |       2 |
|     14 | asdf        | xyz        | CAF      |       3 |
|     15 | pqrs        | asd        | POI      |       1 |
|     16 | pqrs        | asd        | POI      |       2 |
|     17 | pqrs        | asd        | CAF      |       1 |
|     18 | pqrs        | abc        | POI      |       1 |
|     19 | pqrs        | abc        | ISR      |       1 |
|     20 | pqrs        | abc        | ISR      |       2 |
|     21 | pqrs        | xyz        | POA      |       1 |
|     22 | pqrs        | xyz        | CAF      |       1 |
|     23 | pqrs        | xyz        | CAF      |       2 |
|     24 | pqrs        | xyz        | CAF      |       3 |
+--------+-------------+------------+----------+---------+

and I want to get results as:

+--------+-------------+------------+----------+---------+
| doc_id | module_name | mapping_id | doc_type | version |
+--------+-------------+------------+----------+---------+
|      6 | asdf        | asd        | POI      |       2 |
|      7 | asdf        | asd        | CAF      |       1 |
|      8 | asdf        | abc        | POI      |       1 |
|     10 | asdf        | abc        | ISR      |       2 |
|     11 | asdf        | xyz        | POA      |       1 |
|     14 | asdf        | xyz        | CAF      |       3 |
|     16 | pqrs        | asd        | POI      |       2 |
|     17 | pqrs        | asd        | CAF      |       1 |
|     18 | pqrs        | abc        | POI      |       1 |
|     20 | pqrs        | abc        | ISR      |       2 |
|     21 | pqrs        | xyz        | POA      |       1 |
|     24 | pqrs        | xyz        | CAF      |       3 |
+--------+-------------+------------+----------+---------+

where the max version number row should be returned in combination of module_name , mapping_id and doc_type .

I am unable to figure out the query. Need help.

Answer 1

In most databases, you would use row_number() . In MySQL, this is not available. One method is join and group by :

select t.*
from t join
     (select module_name, mapping_id, doc_type, max(version) as version
      from t
      group by module_name, mapping_id, doc_type
     ) tt
     using (module_name, mapping_id, doc_type, version);

if we take the OP's remarks literally, or possibly just

select t.*
from t join
     (select doc_type, max(version) as version
      from t
      group by doc_type
     ) tt
     using (doc_type, version);

Answer 2

another alternative

select *
from documents d
where not exists
  (
    select 1 from documents dv
    where dv.doc_type = d.doc_type
      and dv.version > d.version
  )

literally: "there is no bigger version number for same values"