slow limit -N queries without order by ts queries #2330
Description
Describe the bug
Looks like in latest release queries that use limit -N without explicit order by issued against tables with designated timestamp traverse partitions from first to last . This can make a big difference for tables with many partitions that get queried rarely enough to not be present in os cache .
To reproduce
create table test ( ts timestamp, d double, f float, i int, l long, s1 string, s2 string, sy symbol index) timestamp(ts) partition by day;
insert into test
SELECT cast( (x*2500000) as timestamp), x, x, cast(x as int), x, 's1' || (x%10), 's2' || (x%10), 'sy' || (x%10)
from long_sequence(100000000);
selecT count(*) from test limit -5;
--observe many 'open partition' messages in logs
--after restart run
selecT * from test order by ts desc limit 5;
--check that there's only 1 'open partition' messageExpected Behavior
No response
Environment
- **QuestDB version**: 6.4.4-snapshot
- **OS**: Windows
- **Browser**: ChromeAdditional context
No response