Issues creating top N query for grouped datsets

Question

I am working on a database query and I am facing issues creating the right query to get required output. Below is my create table query and insert statements.

create table DETAILS (dep_name varchar(50), emp_name varchar(50), salary int);

insert into DETAILS values ('marketing', 'ravi', 50000);  
insert into DETAILS values ('marketing', 'suresh', 25000);  
insert into DETAILS values ('marketing', 'mahesh', 60000);  
insert into DETAILS values ('legal', 'raja', 40000);  
insert into DETAILS values ('legal', 'gavi', 50000);  
insert into DETAILS values ('legal', 'rana', 90000);  
insert into DETAILS values ('legal', 'jyoti', 60000);  

My requirement is to display only top 2 salaried employees from each department. Please see sample output below.

dep_name    emp_name    salary  
legal       rana        90000  
legal       jyoti       60000  
marketing   mahesh      60000  
marketing   ravi        50000

I have tried several queries but not able to crack this. Can somebody get me to frame the right query?

Answer 1

If you are using Oracle then it can be done by dense_rank() over (partition by .. order by ..)

If your RDBMS is mysql you cannot use dense_rank() you have to emulate the concept.

See this link.

Try:

select 
    dep_name,
    emp_name,
    salary
from (
    SELECT 
        @row_number:=CASE
            WHEN @dep_name = dep_name THEN @row_number + 1
            ELSE 1
        END AS num,
        @dep_name := dep_name as dep_name,
        emp_name,
        salary
    FROM DETAILS ,(SELECT @dep_name := '', @row_number := 0) as t
    ORDER BY dep_name, salary desc
) x
where num < 3
ORDER BY dep_name, salary desc;

sql fiddle demo