Select first rows with rowno

Question

Problem: I search for 20 companies with highest turnover in a table. Its a dynamic SQL (created at runtime), I can not use a hard coded fetch of 20 rows.
(Example is a simple table with companyID, turnover and a key)
I read in documentation, that first rowno is applied, then the order (to be fast and not access the whole big table). Ok.
Oracle goes same way, this is the work around:
SELECT * FROM (SELECT t1.Firmid_i AS t1_Firmid_i, SUM( t1.Amount_dc ) AS t1_Amount_dc FROM DEMO.turnover t1 GROUP BY t1.Firmid_i, t1.Productid_i ORDER BY 2 desc ) WHERE  rownum <= 20
I try this on a MaxDb :
SELECT * FROM (SELECT t1.Firmid_i, MAX( t1.Amount_dc ) FROM DEMO.turnover t1 GROUP BY t1.Firmid_i ORDER BY 2 asc) WHERE rowno <= 20 
-> an error occur (General error;-5016 POS(103) Missing delimiter: ), the order in the inner sql is not allowed
I can change the statement:
SELECT * FROM (SELECT t1.Firmid_i, MAX( t1.Amount_dc ) FROM DEMO.turnover t1 GROUP BY t1.Firmid_i) WHERE  rowno <= 20  ORDER BY 2 asc 
Now I got 20 row (20 rows sorted), but not the results I need (sort is used to late).
Does anyone knows a workaroud?
TIA
Christian

Answer 1

Hello Christian,
Unfortunately, MaxDB does not support the ORDER BY/GROUP BY condition in sub-selects to achieve such a result set limitation, nor does it support a kind of LIMIT statement
as MS-SQL does.
Anyway, to give the correct results, the query will anyway
create the complete result set, to apply the correct sorting (as the rows can only be sorted when for all rows the values are known).
There is unfortunately no way to formulate such a query
in a single SQL statement, and so you can only
fetch as much rows you need, and discard the rest,
which you said is not possible for you.
Sorry
Alexander Schröder