Fixes join test cases

[johnedquinn]

Description

Picked out the relevant ones from #122. See the description there.

License

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[jpschorr]

This test logically corresponds to something like:

CREATE TABLE Stores (
    id int
);

CREATE TABLE Books (
    id int,
    title varchar(255),
  	price real
);

INSERT INTO Stores (id) VALUES (5), (6), (7);
INSERT INTO Books (id, title, price) VALUES 
    (5, 'A', 5.0), (5, 'B', 2.0), (5, 'C', 7.0), (5, 'D', 9.0),
    (6, 'A', 5.0), (6, 'E', 9.50), (6, 'F', 10.0);

SELECT s.id AS id, b.title AS title 
FROM Stores AS s LEFT OUTER JOIN Books AS b ON s.id = b.id AND b.price > 9

[johnedquinn]

Here's a corresponding DB Fiddle.

[jpschorr]

This test logically corresponds to something like:

CREATE TABLE Stores (
    id int
);

CREATE TABLE Books (
    id int,
    title varchar(255),
  	price real
);

INSERT INTO Stores (id) VALUES (5), (6), (7);
INSERT INTO Books (id, title, price) VALUES 
    (5, 'A', 5.0), (5, 'B', 2.0), (5, 'C', 7.0), (5, 'D', 9.0),
    (6, 'A', 5.0), (6, 'E', 9.50), (6, 'F', 10.0);

SELECT s.id AS id, b.title AS title 
FROM Stores AS s LEFT OUTER JOIN Books AS b ON 1=1 AND s.id = b.id  
WHERE b.* IS NULL

Based on the logic that:

• <x> LEFT CROSS JOIN <y> can be modeled as <x> LEFT OUTER JOIN <y> ON 1=1
• Since the <y> is a corellated value from <x>, we add the <x>.id = <y>.id to the ON

[johnedquinn]

Here's a corresponding DB Fiddle.

[jpschorr]

This test logically corresponds to something like (I didn't rewrite the query, just added SQL tables):

CREATE TABLE t2 (
    name varchar(255),
    n int
);

CREATE TABLE t3 (
    name varchar(255),
    n int
);

INSERT INTO t2 (name, n) VALUES 
	('bb', 12), ('cc', 22), ('ee', 42);
INSERT INTO t3 (name, n) VALUES 
	('bb', 13), ('cc', 23), ('dd', 33);

SELECT * 
FROM 
		(SELECT name, n as s2_n, 2 as s2_2 FROM t2) as s2 
    LEFT JOIN 
	    (SELECT name, n as s3_n, 3 as s3_2 FROM t3) s3 ON s2.name = s3.name

[johnedquinn]

Exactly. Here's a corresponding DB Fiddle.

[jpschorr]

This test logically corresponds to something like (I didn't rewrite the query, just added SQL tables):

CREATE TABLE t2 (
    name varchar(255),
    n int
);

CREATE TABLE t3 (
    name varchar(255),
    n int
);

INSERT INTO t2 (name, n) VALUES 
	('bb', 12), ('cc', 22), ('ee', 42);
INSERT INTO t3 (name, n) VALUES 
	('bb', 13), ('cc', 23), ('dd', 33);

SELECT * 
FROM 
		(SELECT name, n as s2_n, 2 as s2_2 FROM t2) as s2 
    LEFT JOIN 
	    (SELECT name, n as s3_n, 3 as s3_2 FROM t3) s3 ON s2.name = s3.name

And here we assert the result is

<< 
    {'name': 'bb', 'n': 12, 'name':'bb', 'b':13},
    {'name': 'cc', 'n': 22, 'name':'cc', 'b':23},
    {'name': 'ee', 'n': 42, _2:null},
>>

The SQL-equivalent result might be encoded as

<< 
    {'name': 'bb', 'n': 12, 'name':'bb', 'b':13},
    {'name': 'cc', 'n': 22, 'name':'cc', 'b':23},
    {'name': 'ee', 'n': 42, 'name':null, 'n':null},
>>

[johnedquinn]

Exactly. Here's a corresponding DB Fiddle.