13.2.11.7 Correlated Subqueries

A correlated subquery is a subquery that contains a reference to a table that also appears in the outer query. For example:

SELECT * FROM t1
  WHERE column1 = ANY (SELECT column1 FROM t2
                       WHERE t2.column2 = t1.column2);

Notice that the subquery contains a reference to a column of t1, even though the subquery's FROM clause does not mention a table t1. So, MySQL looks outside the subquery, and finds t1 in the outer query.

Suppose that table t1 contains a row where column1 = 5 and column2 = 6; meanwhile, table t2 contains a row where column1 = 5 and column2 = 7. The simple expression ... WHERE column1 = ANY (SELECT column1 FROM t2) would be TRUE, but in this example, the WHERE clause within the subquery is FALSE (because (5,6) is not equal to (5,7)), so the expression as a whole is FALSE.

Scoping rule: MySQL evaluates from inside to outside. For example:

SELECT column1 FROM t1 AS x
  WHERE x.column1 = (SELECT column1 FROM t2 AS x
    WHERE x.column1 = (SELECT column1 FROM t3
      WHERE x.column2 = t3.column1));

In this statement, x.column2 must be a column in table t2 because SELECT column1 FROM t2 AS x ... renames t2. It is not a column in table t1 because SELECT column1 FROM t1 ... is an outer query that is farther out.

Beginning with MySQL 8.0.24, the optimizer can transform a correlated scalar subquery to a derived table. Consider the query shown here:

SELECT * FROM t1 
    WHERE ( SELECT a FROM t2 
              WHERE t2.a=t1.a ) > 0;

To avoid materializing several times for a given derived table, we can instead materialize—once—a derived table which adds a grouping on the join column from the table referenced in the inner query (t2.a) and then an outer join on the lifted predicate (t1.a = derived.a) in order to select the correct group to match up with the outer row. (If the subquery already has an explicit grouping, the extra grouping is added to the end of the grouping list.) The query previously shown can thus be rewritten like this:

SELECT t1.* FROM t1 
    LEFT OUTER JOIN
        (SELECT a, COUNT(*) AS ct FROM t2 GROUP BY a) AS derived
    ON  t1.a = derived.a 
        AND 
        REJECT_IF(
            (ct > 1),
            "ERROR 1242 (21000): Subquery returns more than 1 row"
            )
    WHERE derived.a > 0;

In the rewritten query, REJECT_IF() represents an internal function which tests a given condition (here, the comparison ct > 1) and raises a given error (in this case, ER_SUBQUERY_NO_1_ROW) if the condition is true. This reflects the cardinality check that the optimizer performs as part of evaluating the JOIN or WHERE clause, prior to evaluating any lifted predicate, which is done only if the subquery does not return more than one row.

This type of transformation can be performed, provided the following conditions are met:

• The subquery can be part of a SELECT list, WHERE condition, or HAVING condition, but cannot be part of a JOIN condition, and cannot contain a LIMIT or OFFSET clause. In addition, the subquery cannot contain any set operations such as UNION.

• The WHERE clause may contain one or more predicates, combined with AND. If the WHERE clause contains an OR clause, it cannot be transformed. At least one of the WHERE clause predicates must be eligible for transformation, and none of them may reject transformation.

• To be eligible for transformation, a WHERE clause predicate must be an equality predicate in which each operand should be a simple column reference. No other predicates—including other comparison predicates—are eligible for transformation. The predicate must employ the equality operator = for making the comparison; the null-safe ≪=> operator is not supported in this context.

• A WHERE clause predicate that contains only inner references is not eligible for transformation, since it can be evaluated before the grouping. A WHERE clause predicate that contains only outer references is eligible for transformation, even though it can be lifted up to the outer query block. This is made possible by adding a cardinality check without grouping in the derived table.

• To be eligible, a WHERE clause predicate must have one operand that contains only inner references and one operand that contains only outer references. If the predicate is not eligible due to this rule, transformation of the query is rejected.

• A correlated column can be present only in the subquery's WHERE clause (and not in the SELECT list, a JOIN or ORDER BY clause, a GROUP BY list, or a HAVING clause). Nor can there be any correlated column inside a derived table in the subquery's FROM list.

• A correlated column can not be contained in an aggregate function's list of arguments.

• A correlated column must be resolved in the query block directly containing the subquery being considered for transformation.

• A correlated column cannot be present in a nested scalar subquery in the WHERE clause.

• The subquery cannot contain any window functions, and must not contain any aggregate function which aggregates in a query block outer to the subquery. A COUNT() aggregate function, if contained in the SELECT list element of the subquery, must be at the topmost level, and cannot be part of an expression.

See also Section 13.2.11.8, “Derived Tables”.