Efficient OSQL union query from two classes from the same ancestor - query-optimization

Question

I have a complex class hierarchy of this form:
- A (common ancestor)
* B
- C
* D
* E
* F
- G
* H
* I
From my application I need this kind of query:
SELECT FROM D
WHERE ...
But recently I would like also to do this kind of query:
SELECT FROM A
WHERE #class IN ['D', 'I'] AND ...
My question is how much is efficient this last query, and which is the best practice to optimize it.

Answer 1

trying your query with a sample dataset I've created
SELECT FROM A
WHERE #class IN ['D', 'I']
and i'm seeing execution time from 40ms (cold cache) to 15ms (hot cache).
Using this one, instead, I see an improvement (3ms with hot cache):
select expand($c)
let $a=(SELECT FROM D),
$b=(SELECT FROM I),
$c=unionAll($a,$b)
EDIT
here is the query using WHERE and LIMIT conditions:
select from (
select expand($c)
let $a=(SELECT FROM D),
$b=(SELECT FROM I),
$c=unionAll($a,$b)
)
where value = 5 LIMIT 10
Hope it helps.
Ivan

Answer 2

I state that I don't know if I have understood the meaning of your question.
I understood you want to know if the query is faster with SELECT FROM A WHERE #class IN ['D', 'I'].
I have made a test with 5000 record for class A, 5000 for class B, 5000 for class C etc ...
With the query SELECT FROM A WHERE #class IN ['D', 'I'] limit -1
With the query SELECT EXPAND( $c ) LET $a = ( SELECT FROM D ), $b = ( SELECT FROM I ), $c = UNIONALL( $a, $b ) limit -1
The second query should be faster because it does not select all records of class A (45000) and then evaluates the clause where, but only the 5000 record of type D and 5000 of type I