ast, parser: fix parse join

[wjhuang2016]

Signed-off-by: wjhuang2016 [email protected]

What problem does this PR solve?

Fix pingcap/tidb#20563
We can't parse t1 join t2 join t3 on xxx on xxx.
Because the order is: (t1 join t2 join t3 on xxx) on xxx. What we need is t1 join (t2 join t3 on xxx) on xxx

What is changed and how it works?

1. Make join straightJoin inner cross left right full natural below tableRefPriority, which is the same as MySQL.
2. Implement NewCrossJoin(), here is its logic:

// NewCrossJoin builds a cross join without `on` or `using` clause.
// If the right child is a join tree, we need to handle it differently to make the precedence get right.
// Here is the example: t1 join t2 join t3
//                 JOIN ON t2.a = t3.a
//  t1    join    /    \
//              t2      t3
// (left)         (right)
//
// We can not build it directly to:
//         JOIN
//        /    \
//       t1	   JOIN ON t2.a = t3.a
//             /   \
//            t2    t3
// The precedence would be t1 join (t2 join t3 on t2.a=t3.a), not (t1 join t2) join t3 on t2.a=t3.a
// We need to find the left-most child of the right child, and build a cross join of the left-hand side
// of the left child(t1), and the right hand side with the original left-most child of the right child(t2).
//          JOIN t2.a = t3.a
//         /    \
//       JOIN    t3
//       /  \
//      t1  t2
// Besides, if the right handle side join tree's join type is right join, we need to rewrite it to left join.
// So t1 join t2 right join t3 would be rewrite to t1 join t3 left join t2.
// If not, t1 join (t2 right join t3) would be (t1 join t2) right join t3. After rewrite the right join to left join.
// We get (t1 join t3) left join t2, the semantics is correct.

Check List

Tests

• Unit test

Code changes

Side effects

Related changes

[winoros]

lgtm

[kennytm]

would it be easier to make the JOIN operator right-associative? 🤔

[wjhuang2016]

NewCrossJoin() will do the reorganization, make it left-associative.

[kennytm]

i mean if you use %right here perhaps you don't need that NewCrossJoin() function

[wjhuang2016]

Then t1 right join t2 right join t3 would be t1 right join (t2 right join t3), it's not correct.

[kennytm]

why is that incorrect? a RIGHT JOIN b RIGHT JOIN c ON w1 ON w2 is indeed equivalent to a RIGHT JOIN (b RIGHT JOIN c ON w1) ON w2,
and the same for LEFT JOIN (a LEFT JOIN b LEFT JOIN c ON w1 ON w2 == a LEFT JOIN (b LEFT JOIN c ON w1) ON w2)

this is different from a RIGHT JOIN b ON w1 RIGHT JOIN c ON w2 which means (a RIGHT JOIN b ON w1) RIGHT JOIN c ON w2

however NATURAL JOIN needs to be left-associative (a NATURAL JOIN b NATURAL JOIN c == (a NATURAL JOIN b) NATURAL JOIN c)

[wjhuang2016]

Sorry, the example is wrong.
Consider select * from t1 join t2 right join t3 on t2.a=t3.a;, if using right-associative, we got
select * from t1 join (t2 right join t3 on t2.a=t3.a). They are not the same.

select * from t1 join t2 join t3;
mysql> select * from t1 join t2 right join t3 on t2.a=t3.a;
+------+------+------+
| a    | a    | a    |
+------+------+------+
| NULL | NULL |    3 |
+------+------+------+
1 row in set (0.00 sec)

mysql> select * from t1 join (t2 right join t3 on t2.a=t3.a);
+------+------+------+
| a    | a    | a    |
+------+------+------+
|    1 | NULL |    3 |
+------+------+------+
1 row in set (0.00 sec)

[tangenta]

LGTM

[kennytm]

/merge

[ti-srebot]

/run-all-tests