Monadic joins to SQL

gistfile1.txt

t1.flatMap(a => t2.filter(b => b.s == a.s).map(b => b.s))

  SELECT t2.s FROM t1, t2 WHERE t2.s = t1.s

t1.flatMap(a => t2.map(b => b.s).take(10))

  SELECT x.s FROM t1, (SELECT * FROM t2 LIMIT 10) x

t1.flatMap(a => t2.filter(b => b.s == a.s).map(b => b.s).take(10))

  Is it possible to translate this composition to SQL?
  

This is not valid SQL:

SELECT x.s FROM t1, (SELECT * FROM t2 WHERE t2.s = t1.s LIMIT 10) x

This one is wrong. It'll take in consideration only the first 10 rows of t2, not the first 10 rows that match the filter.

SELECT x.s FROM t1, (SELECT ROWNUM r, t2.* FROM t2) x WHERE t2.s = t1.s AND x.r <= 10

I've found one alternative, but it's not generic enough to cover other similar compositions:

SELECT x.s FROM t1, (SELECT t2.* from t2 WHERE EXISTS (SELECT * FROM t1 WHERE t2.s = t1.s) LIMIT 10)

The normalized composition would be:

t1.flatMap(a => t2.filter(b => t1.filter(b.s == a.s).nonEmpty).take(10))

Example of composition that can't be normalized using this approach:

t1.flatMap(a => t2.filter(b => b.s == a.s).sortBy(b => a.s).take(10))