Disable `supports_left_associative_joins_without_parens` for MySQL and Redshift by revitalkr · Pull Request #2357 · apache/datafusion-sqlparser-rs

revitalkr · 2026-05-28T05:15:57Z

…ntly applied to Snowflake) to MySQL and Redshift.

Summary

This PR extends the existing supports_left_associative_joins_without_parens = false behavior (currently applied to Snowflake) to MySQL and Redshift.

Problem

The following query is currently accepted only for Snowflake and rejected for other dialects:

SELECT DISTINCT p.product_id
FROM orders AS o
INNER JOIN customers AS c
INNER JOIN products AS p ON p.customer_id = c.customer_id
ON c.order_id = o.order_id;

However, this query executes successfully in both MySQL and Redshift.

Solution

Set supports_left_associative_joins_without_parens = false for MySQL and Redshift, enabling them to accept the same class of queries already supported for Snowflake.

Validation

Attached scripts demonstrate that:

The query is accepted by both dialects
Their behavior is compatible with the existing Snowflake parsing mode

Tests

The existing Snowflake test has been moved to the main test suite and is now applied to all dialects where supports_left_associative_joins_without_parens = false.

scripts

-- =========================================================
-- MySQL validation script
-- 1. Acceptance of deferred ON query
-- 2. Behavior consistent with Snowflake parsing mode
-- =========================================================

-- PART A: REAL QUERY

DROP TABLE IF EXISTS orders;
DROP TABLE IF EXISTS customers;
DROP TABLE IF EXISTS products;

CREATE TABLE orders (
order_id INT
);

CREATE TABLE customers (
customer_id INT,
order_id INT
);

CREATE TABLE products (
product_id INT,
customer_id INT
);

INSERT INTO orders VALUES (1), (2), (3);

INSERT INTO customers VALUES
(10,1),
(20,2),
(30,3);

INSERT INTO products VALUES
(100,10),
(200,20),
(300,30);

-- ORIGINAL
SELECT 'PART_A_ORIGINAL' AS src, p.product_id
FROM orders AS o
INNER JOIN customers AS c
INNER JOIN products AS p ON p.customer_id = c.customer_id
ON c.order_id = o.order_id
ORDER BY p.product_id;

-- CANONICAL (Snowflake-style)
SELECT 'PART_A_CANONICAL' AS src, p.product_id
FROM orders AS o
INNER JOIN (
customers AS c
INNER JOIN products AS p ON p.customer_id = c.customer_id
)
ON c.order_id = o.order_id
ORDER BY p.product_id;

-- PART B: WITNESS QUERY
-- Demonstrates observable associativity difference

DROP TABLE IF EXISTS w_orders;
DROP TABLE IF EXISTS w_customers;
DROP TABLE IF EXISTS w_products;

CREATE TABLE w_orders (
order_id INT
);

CREATE TABLE w_customers (
customer_id INT,
order_id INT
);

CREATE TABLE w_products (
product_id INT,
customer_id INT
);

INSERT INTO w_orders VALUES (1), (2);

INSERT INTO w_customers VALUES
(10,1),
(20,2);

INSERT INTO w_products VALUES
(100,10);

-- ORIGINAL
SELECT 'PART_B_ORIGINAL' AS src, o.order_id, c.customer_id, p.product_id
FROM w_orders AS o
LEFT JOIN w_customers AS c
INNER JOIN w_products AS p ON p.customer_id = c.customer_id
ON c.order_id = o.order_id
ORDER BY o.order_id, c.customer_id, p.product_id;

-- CANONICAL
SELECT 'PART_B_CANONICAL' AS src, o.order_id, c.customer_id, p.product_id
FROM w_orders AS o
LEFT JOIN (
w_customers AS c
INNER JOIN w_products AS p ON p.customer_id = c.customer_id
)
ON c.order_id = o.order_id
ORDER BY o.order_id, c.customer_id, p.product_id;

-- ALTERNATIVE (left-deep)
SELECT 'PART_B_LEFT_DEEP' AS src, o.order_id, c.customer_id, p.product_id
FROM (
w_orders AS o
LEFT JOIN w_customers AS c ON c.order_id = o.order_id
)
INNER JOIN w_products AS p ON p.customer_id = c.customer_id
ORDER BY o.order_id, c.customer_id, p.product_id;

-- =========================================================
-- Redshift validation script
-- 1. Acceptance of deferred ON query
-- 2. Behavior consistent with Snowflake parsing mode
-- =========================================================

-- PART A: REAL QUERY

DROP TABLE IF EXISTS orders;
DROP TABLE IF EXISTS customers;
DROP TABLE IF EXISTS products;

CREATE TEMP TABLE orders (
order_id INT
);

CREATE TEMP TABLE customers (
customer_id INT,
order_id INT
);

CREATE TEMP TABLE products (
product_id INT,
customer_id INT
);

INSERT INTO orders VALUES (1), (2), (3);

INSERT INTO customers VALUES
(10,1),
(20,2),
(30,3);

INSERT INTO products VALUES
(100,10),
(200,20),
(300,30);

-- ORIGINAL
SELECT 'PART_A_ORIGINAL' AS src, p.product_id
FROM orders AS o
INNER JOIN customers AS c
INNER JOIN products AS p ON p.customer_id = c.customer_id
ON c.order_id = o.order_id
ORDER BY p.product_id;

-- CANONICAL (Snowflake-style)
SELECT 'PART_A_CANONICAL' AS src, p.product_id
FROM orders AS o
INNER JOIN (
customers AS c
INNER JOIN products AS p ON p.customer_id = c.customer_id
)
ON c.order_id = o.order_id
ORDER BY p.product_id;

-- PART B: WITNESS QUERY

DROP TABLE IF EXISTS w_orders;
DROP TABLE IF EXISTS w_customers;
DROP TABLE IF EXISTS w_products;

CREATE TEMP TABLE w_orders (
order_id INT
);

CREATE TEMP TABLE w_customers (
customer_id INT,
order_id INT
);

CREATE TEMP TABLE w_products (
product_id INT,
customer_id INT
);

INSERT INTO w_orders VALUES (1), (2);

INSERT INTO w_customers VALUES
(10,1),
(20,2);

INSERT INTO w_products VALUES
(100,10);

-- ORIGINAL
SELECT 'PART_B_ORIGINAL' AS src, o.order_id, c.customer_id, p.product_id
FROM w_orders AS o
LEFT JOIN w_customers AS c
INNER JOIN w_products AS p ON p.customer_id = c.customer_id
ON c.order_id = o.order_id
ORDER BY o.order_id, c.customer_id, p.product_id;

-- CANONICAL
SELECT 'PART_B_CANONICAL' AS src, o.order_id, c.customer_id, p.product_id
FROM w_orders AS o
LEFT JOIN (
w_customers AS c
INNER JOIN w_products AS p ON p.customer_id = c.customer_id
)
ON c.order_id = o.order_id
ORDER BY o.order_id, c.customer_id, p.product_id;

-- ALTERNATIVE (left-deep)
SELECT 'PART_B_LEFT_DEEP' AS src, o.order_id, c.customer_id, p.product_id
FROM (
w_orders AS o
LEFT JOIN w_customers AS c ON c.order_id = o.order_id
)
INNER JOIN w_products AS p ON p.customer_id = c.customer_id
ORDER BY o.order_id, c.customer_id, p.product_id;
``

…h INNER JOIN in Snowflake dialect produces non-equivalent query

iffyio · 2026-05-28T10:53:58Z

    );
 }

+#[test]


does this test fail without the changes in this PR? if not it looks like we can skip it entirely, it looks similar to https://github.com/SatoriCyber/datafusion-sqlparser-rs/blob/00e1d917d439ff40e0c6a10829c491b0735b8818/tests/sqlparser_common.rs#L17445-L17465

yes, it fails without this PR.
These are different cases which behave differently. Natural join behaves different than the other joins because it can't have a condition after it.

iffyio · 2026-05-29T08:14:05Z

+fn non_left_associative_dialects() -> TestedDialects {
+    all_dialects_where(|d| !d.supports_left_associative_joins_without_parens())
+}


can we inline this into the function that uses it?

…ntly applied to Snowflake) to MySQL and Redshift.

revitalkr · 2026-05-31T12:40:35Z

Rebased this PR on top of #2355 since it depends on those changes.

iffyio · 2026-06-05T09:07:10Z

        comment: Option<String>,
    },
    /// ```sql
    /// CREATE [ OR REPLACE ] [ { TEMP | TEMPORARY | VOLATILE } ] FILE FORMAT [ IF NOT EXISTS ] <name>


the diff seems to contain unrelated (unintentional ?) changes

Fix issue apache#2353: Incorrect canonicalization of NATURAL JOIN wit…

293b38c

…h INNER JOIN in Snowflake dialect produces non-equivalent query

iffyio reviewed May 28, 2026

View reviewed changes

iffyio reviewed May 29, 2026

View reviewed changes

iffyio changed the title ~~Extend supports_left_associative_joins_without_parens behavior (curre…~~ Disable supports_left_associative_joins_without_parens for MySQL and Redshift May 29, 2026

ramnes and others added 2 commits May 31, 2026 15:30

Snowflake: Add support for CREATE FILE FORMAT (apache#2336)

0abda75

Extend supports_left_associative_joins_without_parens behavior (curre…

d60c9a3

…ntly applied to Snowflake) to MySQL and Redshift.

revitalkr force-pushed the fix/non-left-associative-mysql-redshift branch from 00e1d91 to d60c9a3 Compare May 31, 2026 12:30

revitalkr requested a review from iffyio June 4, 2026 08:08

iffyio reviewed Jun 5, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Disable `supports_left_associative_joins_without_parens` for MySQL and Redshift#2357

Disable `supports_left_associative_joins_without_parens` for MySQL and Redshift#2357
revitalkr wants to merge 3 commits into
apache:mainfrom
SatoriCyber:fix/non-left-associative-mysql-redshift

revitalkr commented May 28, 2026

Uh oh!

iffyio May 28, 2026

Uh oh!

revitalkr May 28, 2026

Uh oh!

iffyio May 29, 2026

Uh oh!

revitalkr May 31, 2026

Uh oh!

revitalkr commented May 31, 2026

Uh oh!

iffyio Jun 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

                   );
               }
+              #[test]

Conversation

revitalkr commented May 28, 2026

Summary

Problem

Solution

Validation

Tests

scripts

Uh oh!

iffyio May 28, 2026

Choose a reason for hiding this comment

Uh oh!

revitalkr May 28, 2026

Choose a reason for hiding this comment

Uh oh!

iffyio May 29, 2026

Choose a reason for hiding this comment

Uh oh!

revitalkr May 31, 2026

Choose a reason for hiding this comment

Uh oh!

revitalkr commented May 31, 2026

Uh oh!

iffyio Jun 5, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants