Improve a query's performance

Question

Hello
I need to make a query that will return a value when there are rows that meet a certain condition and another value when there are rows that meet another condition, the most simple way I thought of was using two queries like that:
+1. select count(*)+
from  T1
where (A = a OR B = b) and C > 0
+2. select count(*)+
from  T2
LEFT JOIN T1 ON T1.X = T2.X
where (T1.A = a OR T1.B = b) and T2.A is not null
However, this way seems to be very slow when called numerous times from the application so I thought to use this next single query:
select distinct
case when (select count(*)
from  T1
where (A = a or B = b ) and C > 0) > 0
then 1
else 0
end  ARE_THERE_ROWS_THAT_MEET_CONDITION_A,
case when (select count(*)
from  T2
LEFT JOIN T1 ON T1.X = T2.X
where (T1.A = a OR T1.B = b) and T2.A is not null)) >0
then 1
else 0
end  ARE_THERE_ROWS_THAT_MEET_CONDITION_B
from T1;
This way I can connect to the DB only once (instead of twice) but running this way did not improve the performance dramaticly.
Is there a better more efficent and more simple way (single query prefered) to perform this query?

Answer 1

Hi,
You siad :
>
1. select count(*)
from T1
where (A = a OR B = b) and C > 0
However, this way seems to be very slow when called numerous times from the application so I thought to use this next single query:
>
It's slow because:
1) you don't probably read only one row per call (you probably read the FULL table) -> post the explain plan.
2) you call a "numerous time" this query.
What it suggest as a starting point (i guess you're calling a function having a, b as parameters):
- Give us the cardinality for the columns A, B and C (how many different values did you have VS the complete volume of T1)?
- Tell us which indexes you have on T1
You can run:
SELECT table_name, num_rows, blocks, empty_blocks, avg_space, avg_row_len,
       last_analyzed
  FROM user_tables
WHERE table_name IN ('T1', 'T2')
SELECT index_name, index_type, table_name, uniqueness, blevel, leaf_blocks,
       distinct_keys, clustering_factor, last_analyzed
  FROM user_indexes
WHERE table_name IN ('T1', 'T2')Edited by: user11268895 on Jul 18, 2010 4:32 PM
Edited by: user11268895 on Jul 18, 2010 4:50 PM