SQL query with EXISTS not working as I thought - mysql

Question

Hi these two SQL Queries return the same result
SELECT DISTINCT ItemID
FROM Sale INNER JOIN Department
ON Department.DepartmentID = Sale.DepartmentID
WHERE DepartmentFloor = 2
ORDER BY ItemID
SELECT DISTINCT ItemID
FROM Sale
WHERE EXISTS
(SELECT *
FROM Department
WHERE Sale.DepartmentID = Department.DepartmentID
AND DepartmentFloor = 2)
ORDER BY ItemID;
The Subquery Inside the Exists returns True So why doesnt the secod query return the equivalent of
SELECT DISTINCT ItemID
FROM Sale
Which guves a different result from the two above.

Answer 1

You are getting confused by EXISTS().. It occurs on a line by line basis, based on table correlation, not just a single true/false. This line of your subquery is your correlation clause:
Sale.DepartmentID = Department.DepartmentID
It is saying "Only show the Sale.ItemIDs where that ItemID's Sale.DepartmentID is in Department."
It achieves the same function as a join predicate, like in your first query:
FROM Sale S
JOIN Department D on S.DepartmentID = D.DepartmentID --here
Conversely, this query:
SELECT DISTINCT ItemID
FROM Sale
Has no limiting factor.
As an aside, you also further limit the results of each query with:
WHERE DepartmentFloor = 2
But I don't think that is the part that is throwing you off, I think it is the concept that a correlated subquery occurs for each record. If you were to remove your correlating clause, then the subquery would actually return true always, and you would get all results back.

Answer 2

The subquery isn't always returning true. It will evaluate for each row, joining on DepartmentID where the DepartmentFloor is 2.
SQL Fiddle
MySQL 5.6 Schema Setup:
CREATE TABLE Sale ( ItemID int, DepartmentID int ) ;
INSERT INTO Sale ( ItemID, DepartmentID )
VALUES (1,1), (2,2), (3,3), (4,1), (5,4), (6,2), (7,3), (8,4) ;
CREATE TABLE Department ( DepartmentID int, DepartmentFloor int ) ;
INSERT INTO Department ( DepartmentID, DepartmentFloor )
VALUES (1,1), (2,1), (3,2), (4,2) ;
Query 1:
SELECT *
FROM Department
WHERE DepartmentFloor = 2
Results: This lists only the Departments on DepartmentFloor 2.
| DepartmentID | DepartmentFloor |
|--------------|-----------------|
| 3 | 2 |
| 4 | 2 |
Query 2:
SELECT *
FROM Sale
Results: This lists ALL of your Sales.
| ItemID | DepartmentID |
|--------|--------------|
| 1 | 1 |
| 2 | 2 |
| 3 | 3 |
| 4 | 1 |
| 5 | 4 |
| 6 | 2 |
| 7 | 3 |
| 8 | 4 |
Query 3:
SELECT *
FROM Sale
WHERE DepartmentID IN (3,4)
Results: And this one shows what is the equivalent of you EXISTS statement. It only shows 4 rows that will match up in my data. So you'd only get back ItemIDs 3,5,7 and 8.
| ItemID | DepartmentID |
|--------|--------------|
| 3 | 3 |
| 5 | 4 |
| 7 | 3 |
| 8 | 4 |

Answer 3

because the uppper part of the query is equivalent to
SELECT DISTINCT ItemID FROM Sale where EXISTS (true)
the upper is the only query that really check the condition ..