从5个表格中选择复杂的MYSQL

Question

I have five tables storing facilities, employees, completed units, projects, and shifts.

I need to determine the productivity (cumulative # of completes and cumulative hours) of a project at each facility, if the facility has hours logged on the project. The facility is stored in the employee table.

Here is my table structure, (reduced to necessary fields for this question ):

facilities
+------------+
| id         |
+------------+
| label      |
+------------+

employees
+------------+
| id         |
+------------+
| facility   |
+------------+

completes
+------------+
| id         |
+------------+
| project    |
+------------+
| completes  |
+------------+
| employee   |
+------------+

projects
+------------+
| id         |
+------------+
| title      |
+------------+

shifts
+------------+
| id         |
+------------+
| project    |
+------------+
| length     |
+------------+
| employee   |
+------------+

Where I am running into trouble is selecting the productivity for a facility, when it is stored with the employee, not with the shift or the # of completes. Maybe that is a design flaw? Should the shift and completes table also have a facility column? It seems redundant to have a facility column, when the employee already determines the facility.

SELECT p.title, SUM(c.completes), SUM(s.length)
FROM projects as p
LEFT JOIN completes as c ON c.project = p.id
LEFT JOIN shifts as s ON s.project = p.id
GROUP BY p.id, shifts.employee.facility;

That GROUP BY statement of course is not valid. But it demonstrates what I want to accomplish. I would also like to select the facility label in the statement. How do I approach this?

Thanks in advance for any insight you can provide!

Answer 1

This is one of the rare instances that I would find a CROSS JOIN to be the best approach:

SELECT p.title, f.label, SUM(c.completes), SUM(s.length)
FROM facilities as f        
CROSS JOIN projects as p
JOIN employees as e ON f.id = e.facility
LEFT JOIN completes as c ON c.employee = e.id AND c.project = p.id
LEFT JOIN shifts as s ON s.employee = e.id AND s.project = p.id
WHERE c.id IS NOT NULL OR s.id IS NOT NULL
GROUP BY p.id, p.title, f.id, f.label

This isn't inherently a bad design, but it does make it difficult that you can't use the typical parent-child relationships here since there are actually 2 parents on your completes and shifts tables.

An alternative approach would to use a UNION without the LEFT JOINs. This may return a faster result, depending on the amount of rows currently in your facilities and projects tables:

SELECT p.title, 
       f.label, 
       SUM(CASE aggregated.type WHEN 'completes' THEN units ELSE 0 END),
       SUM(CASE aggregated.type WHEN 'shifts' THEN units ELSE 0 END)
FROM (
    SELECT 'completes' as type, project, employee, completes as units
    FROM completes
    UNION
    SELECT 'shifts', project, employee, length
    FROM shifts ) aggregated
JOIN projects as p ON aggregated.project = p.id
JOIN employees as e ON aggregated.employee = e.id
JOIN facilities as f ON e.facility = f.id
GROUP BY p.id, p.title, f.id, f.label

As you can see from this query, the problem could be solved fairly easily if completes and shifts were stored in the same table. That might not work for you depending on your needs, but it may be something to consider as an alternative schema.