description: Launch TRACE and then run this script:

recreate table t1(id int);
recreate table t2(id int);
recreate table t3(id int);
recreate table t4(id int);
recreate table t5(id int);

insert into t1(id) select null from rdb$types rows 11;
insert into t2(id) select null from rdb$types rows 22;
insert into t3(id) select null from rdb$types rows 33;
insert into t4(id) select null from rdb$types rows 44;
insert into t5(id) select null from rdb$types rows 55;
commit;

set list on;

select sum(id) from t1;
select sum(id) from t2 ;
select count(id) from t3;
select sum(id) from t4 ;
select sum(id) from t5 ;
------------------------------------------------------

set count on;

select 1
from rdb$database
where 0 not in (
( select sum(id) from t1 )
, ( select sum(id) from t2 )
, 1000 / (select count( id) from t3 ) ----------------------- [1 ]
, ( select sum(id) from t4 )
, ( select sum(id) from t5 )
);
quit;

Output will be:

SUM <null>
SUM <null>
COUNT 0
SUM <null>
SUM <null>

Statement failed, SQLSTATE = 22012
arithmetic exception, numeric overflow, or string truncation
-Integer divide by zero. The code attempted to divide an integer value by an integer divisor of zero.
Records affected: 0

Consider line marked as "[ 1 ]".
Why this is evaluated ? Don't optimizer need to stop any further evaluation after get NULL in "(select sum(id) from t1) " ?

Trace shows that
1) optimizer evaluates NOT-IN list from left to right;
2) THREE tables was scanned instead on one (T1):

0 records fetched
0 ms, 79 fetch(es)

Table Natural
*******************************************
RDB$DATABASE 1
T1 11
T2 22
T3 33

PS.
This question was originally raised in discuss with dimitr, 14-JUL-2015. I remember that some optimizations related to this issue should be made in 4.0 only. But i decided to write here this because otherwise it can be forgotten.

=>

Launch TRACE and then run this script:

recreate table t1(id int);
recreate table t2(id int);
recreate table t3(id int);
recreate table t4(id int);
recreate table t5(id int);

insert into t1(id) select null from rdb$types rows 11;
insert into t2(id) select null from rdb$types rows 22;
insert into t3(id) select null from rdb$types rows 33;
insert into t4(id) select null from rdb$types rows 44;
insert into t5(id) select null from rdb$types rows 55;
commit;

set list on;

select sum(id) from t1;
select sum(id) from t2 ;
select count(id) from t3;
select sum(id) from t4 ;
select sum(id) from t5 ;
------------------------------------------------------

set count on;

select 1
from rdb$database
where 0 not in (
( select sum(id) from t1 )
, ( select sum(id) from t2 )
, 1000 / (select count( id) from t3 ) ----------------------- [1 ]
, ( select sum(id) from t4 )
, ( select sum(id) from t5 )
);
quit;

Output will be:

SUM <null>
SUM <null>
COUNT 0
SUM <null>
SUM <null>

Statement failed, SQLSTATE = 22012
arithmetic exception, numeric overflow, or string truncation
-Integer divide by zero. The code attempted to divide an integer value by an integer divisor of zero.
Records affected: 0

Consider line marked as "[ 1 ]".
Why this is evaluated ?
Does not optimizer need to stop any further evaluation after get NULL in "(select sum(id) from t1) " ?

Trace shows that
1) optimizer evaluates NOT-IN list from left to right;
2) THREE tables was scanned instead on one (T1):

0 records fetched
0 ms, 79 fetch(es)

Table Natural
*******************************************
RDB$DATABASE 1
T1 11
T2 22
T3 33

PS.
This question was originally raised in discuss with dimitr, 14-JUL-2015. I remember that some optimizations related to this issue should be made in 4.0 only. But i decided to write here this because otherwise it can be forgotten.