Question

I would like help with mysql query that refers to the cell above.
for example, in the following table:

primary key(id)     day             count           percentage change
1                  monday             1               0  
2                  tuesday            2              (1-0)*100%=100%  
3                  wednesday          5              (2-1)*100%=100%  
4                  thursday           9              (5-2)*100%=300%  
5                  friday             27             (9-5)*100%=400%  

The percentage change results are based on the results of the previous two days of the count column. Is there a way to incorporate the primary key(id) to refer to the cells “above”?

Answer 1

There is not an easy way. You need to use a self join.

select t.*,
       (coalesce(t_1.count, 0) - coalesce(t_2.count, 0)) * 100.0
from t left outer join
     t t_1
     on t.id = t_1.id + 1 left outer join
     t t_2
     on t.id = t_2.id + 2

The left outer join makes sure all the original rows stay, even when there are not preceding ids.

This works because the ids are sequential. If the ids are not sequential, and the counts are monotonically increasing, you can do this with correlated subqueries:

select t.*,
       (coalesce((select max(`count`) as val
                  from table1 t_1
                  where t_1.`count` < t.`count`
                 ), 0)
       ) -
       (coalesce((select max(`count`)
                  from table1 t_2
                  where t_2.`count` < (select max(`count`) from table1 t_1 where t_1.`count` < t.`count`)
                 ), 0)
       )
from table1 t

Note: this will not work correctly if two values are the same in a row. For that, you would need to use the id instead:

select t.*,
       (coalesce((select max(`count`) as val
                  from table1 t_1
                  where t_1.`id` < t.`id`
                 ), 0)
       ) -
       (coalesce((select max(`count`)
                  from table1 t_2
                  where t_2.`id` < (select max(`id`) from table1 t_1 where t_1.`id` < t.`id`)
                 ), 0)
       )
from table1 t

If the counts are not increasing, then you have to get the ids, and join in the values again. What fun! Here is the code:

select t.*,
       coalesce(t1.count, 0) - coalesce(t2.count, 0)
from (select t.*,
             (select max(`id`) as id1 from table1 t_1 where t_1.`id` < t.`id`
             ) as id1,
             (select max(`count`) from table1 t_2
              where t_2.`id` < (select max(`id`) from table1 t_1 where t_1.`id` < t.`id`)
             ) id2
      from table1 t
     ) t left outer join
     table1 t1
     on t.id1 = t1.id left outer join
     table1 t2
     on t.id2 = t2.id