Random records for a row - sql

Question

I have the following table strvals() with data. I'm looking to randomly choose 2 rows from strval() table and populate table S1 in a loop.
I want something like this
Create table s1(id primary key,strval1, strval2) as
select level,random_rec(strvals), random_rec(strvals)
from dual
connect by level<=10;
The caveat is the column strval1 has to be different THEN strval2 for each row.
Valid output
1, 'AAAA', 'BBBB'
2, 'CCCC', 'BBBB'
3, 'CCCC', 'AAAA'
Not valid
1, 'AAAA', 'AAAA'
Create table strvals(
strval varchar2(4),
constraint pk_strval primary key (strval)
);
insert into strvals
values(
'AAAA'
);
insert into strvals
values(
'BBBB'
);
insert into strvals
values(
'CCCC'
);

Answer 1

Getting a random string from a table can be tricky. One method is to use correlated subqueries -- the correlation clause ensures that the subquery is not "optimized" to run only once.
So, here is one method:
select id, strval,
(select s2.strval
from strvals s2
where s2.strval <> x.strval and x.id > 0
order by dbms_random.random fetch first 1 row only
) as strval2
from (select id,
(select strval
from strvals
where x.id > 0
order by dbms_random.random fetch first 1 row only
) as strval
from (select level as id
from dual
connect by level < 25
) x
) x;
And here is a db<>fiddle.

Answer 2

To reduce a number of slow dbms_random.value in case of many rows, I would suggest to use this technique:
with
rand as (select row_number()over(order by dbms_random.value()) n,strval from strvals)
,cnt as (select count(*) m from rand)
,generator as (select level id, ceil(dbms_random.value()*cnt.m) rnd from cnt connect by level<=10)
select generator.id, rand.strval
from generator
join rand on rand.n=generator.rnd;
As you can see, I sort the table strvals once by dbms_random.value (rand view), so each row has own random integer number, then I count them to get maximum N. And then I calculate random N for each generated row so we can join rand view using hash join.
Update: for 2 strvals: DBFiddle
with
rand as (select row_number()over(order by dbms_random.value()) n,strval from strvals)
,cnt as (select count(*) m from rand)
,generator as (
select
level id,
ceil(dbms_random.value()*cnt.m) rnd1,
ceil(dbms_random.value()*cnt.m) rnd2
from cnt connect by level<=10
)
select
generator.id,
rnd1.strval,
rnd2.strval
from generator
join rand rnd1 on rnd1.n=generator.rnd1
join rand rnd2 on rnd2.n=generator.rnd2;
Update2: now with 2 strvals: DBFiddle
with
rand as (
select
row_number()over(order by dbms_random.value()) n
,s1.strval strval1
,s2.strval strval2
from strvals s1
join strvals s2
on s1.strval!=s2.strval
)
,cnt as (select count(*) m from rand)
,generator as (
select
level id,
ceil(dbms_random.value()*cnt.m) rnd
from cnt connect by level<=10
)
select
generator.id,
rand.strval1,
rand.strval2
from generator
join rand on rand.n=generator.rnd
order by 1
;