Problem combining select, order by, rownum (top-N) and for update

Question

Hello,
i have serious problems with this.
-- drop table testtable;
create table testTable (id number(10,0) primary key, usage number(10,10));
-- delete from testtable;
insert into testtable values (11, 0.5);
insert into testtable values (10, 0.3);
insert into testtable values (12, 0.3);
insert into testtable values (9, 0.3);
insert into testtable values (8, 0.9);
insert into testtable values (3, 0.0);
insert into testtable values (2, 0.02);
insert into testtable values (1, 0.05);
insert into testtable values (7, 0.7);
insert into testtable values (6, 0.4);
insert into testtable values (5, 0.2);
insert into testtable values (4, 0.1);
select * from testtable;
-- without FOR UPDATE
select * from (
select tt.id id_, tt.*
from testtable tt
where tt.usage > 0.1
order by tt.usage desc, tt.id desc
where rownum <= 10;
--> WORKS
-- without ORDER BY
select * from (
select tt.id id_, tt.*
from testtable tt
where tt.usage > 0.1
where rownum <= 10
for update of id_;
--> WORKS
-- without WHERE ROWNUM <= 10
select * from (
select tt.id id_, tt.*
from testtable tt
where tt.usage > 0.1
order by tt.usage desc, tt.id desc
for update of id_;
--> WORKS
-- But what i need is this:
select * from (
select tt.id id_, tt.*
from testtable tt
where tt.usage > 0.1
order by tt.usage desc, tt.id desc
where rownum <= 10
for update;
--> ORA-02014: cannot select FOR UPDATE from view with DISTINCT, GROUP BY, etc., SQL State: 42000, Error Code: 2014
select * from (
select tt.id id_, tt.*
from testtable tt
where tt.usage > 0.1
order by tt.usage desc, tt.id desc
where rownum <= 10
for update of id_;
--> ORA-02014: cannot select FOR UPDATE from view with DISTINCT, GROUP BY, etc., SQL State: 42000, Error Code: 2014
I have tried every single solution i could come up with.
But nothing worked.
My latest idea is to include a comment in the query and set up an ON SELECT trigger which evaluates the comment and enforeces the lock.
But i'm not sure if this is even possible.
I cannot split the statement into two because i need the lock immediately when the wanted rows are selected.
One major criteria for the rows is the order by. Without it i get a random set of rows.
And the rownum <= 10 is also needed because i don't want to lock the whole table but only the few needed rows.
I tried row_number() over (order by ...) but this is considdered a window/group-function which disallows the for update as well as the order by.
During these tests i noticed, that when using the row_number()-function the resultset is ordered automatically (without an additional order by clause).
But this doesn't help anyway.
I tried using piped functions to wrap the select to apply the rownum manually by cursor skip, but this doesn't work either. First of all i wasn't able to wrap the query the way i imagined and second the lock would be applied to the whole resultset anyway but only the reduced rows would be returned.
I heared about LOCK-hints from other DBs, is there anything similar here?
Any other solution??
btw. it has to be high-performance after all.
Greetings Finomosec;

Answer 1

No, not perfect.
This is the expected result (ordered by usage desc, id desc):
ID     USAGE
8     0.9
7     0.7
11     0.5
6     0.4
12     0.3
10     0.3
9     0.3
5     0.2
This ist the one produced by your statement:
ID     USAGE
5     0.2
6     0.4
7     0.7
8     0.9
9     0.3
10     0.3
11     0.5
12     0.3
Use limit 5 in your tests, and you will also notice that the following doesn't work either:
select * from testtable ww where ww.id in
select id from
select tt.id, tt.usage
from testtable tt
where tt.usage > 0.1
where rownum <= 5
order by usage desc, id desc
for update;
It's because the order is not applied to the result.
But the following modification works (at least principally):
select * from testtable ww where ww.id in
select id from
select tt.id, tt.usage
from testtable tt
where tt.usage > 0.1
order by usage desc, id desc
where rownum <= 5
order by usage desc, id desc
for update;
Thr problem here is:
I need to expand the following Statement to the above form by (Java-) Code:
-- statement-A
select tt.id, tt.usage
from testtable tt
where tt.usage > 0.1
order by usage desc, id desc;
The main problem is:
The order by clause needs to be duplicated.
The problem here is, to identify it (if present) and NOT any order by in inner selects.
I am using Hibernate and to implement this i have to modify the OracleDialect to solve this problem. I get the statement-A (see above) and have to apply the limit.
Isn't there any other solution??
Greetings Finomosec;