Effects of NULL in Aggregate Operator - SQL
Let aggop(x) be an invocation of some aggregate operator aggop in SQL, where x is an expression (usually an open expression) to be evaluated against each row of the table t determined by the context in which the invocation appears. Then aggop considers only those rows that satisfy the condition x IS NOT NULL. It follows that if aggop is EVERY or SOME and x evaluates to TRUE or FALSE for at least one row of t, then the result is either TRUE or FALSE, never UNKNOWN. However, if x evaluates to UNKNOWN for every row of t (which is true in the particular case when t is empty), then SQL's other general rule kicks in, requiring the result to be NULL, which is equivalent to UNKNOWN when it appears in the place of a BOOLEAN value.
That anomaly is to some extent compensated for, when EVERY is used in constraint declarations, by SQL's rule that a constraint is deemed to be satisfied when it evaluates to UNKNOWN. However, (SELECT SOME(TRUE) FROM (tx) AS T) is not reliable as an existence test because it evaluates to UNKNOWN if the result of tx is empty, when a constraint based on that condition would be deemed satisfied. That problem could be addressed by writing COALESCE((SELECT SOME(TRUE) FROM (tx) AS T), FALSE) or, equivalently, (SELECT SOME(TRUE) FROM (tx) AS T) IS TRUE .