如何在MySQL中使用表A与自身以一对多关系联接两个表？

Question

I have two tables.

The 'Store' table is a demographics table, with fields like these:

ID, ParentID, ConsolidationType

ID is the unique identifier for a given store.

ParentID may be either NULL or contain the ID of its parent store if it has a ConsolidationType of 'Secondary'.

ConsolidationType is either NULL, 'Primary', or 'Secondary'.

The 'Sales' table has fields like these:

StoreID, SalesDate, SalesAmount

StoreID refers to an ID in the Store table.

---

I'm trying to get the total consolidated and individual sales for a given store in a single row of a single query. I've written a SQL query like the following:

SELECT 
Store.ID AS 'Store ID',
SUM(IF(YEAR(main.SalesDate) = 2012 AND QUARTER(main.SalesDate) = 1,main.SalesAmount,0)) AS 'Individual Sales',
SUM(IF(YEAR(main.SalesDate) = 2012 AND QUARTER(main.SalesDate) = 1 AND YEAR(secondaries.SalesDate) = 2012 AND QUARTER(secondaries.SalesDate) = 1,main.SalesAmount + secondaries.SalesAmount,0) AS 'Consolidated Sales'
FROM Store 
LEFT JOIN Sales AS 'main' ON Store.ID = main.StoreID 
LEFT JOIN Sales AS 'secondaries' ON Store.ParentID = secondaries.StoreID 
GROUP BY Store.ID

I can't figure out why this doesn't work as intended. What am I missing? What's wrong with my logic?

Answer 1

The problem is that your query is generating a Cartesian-product (-like) result set.

Basically, a row from 'main' is getting repeated multiple times, for each matching row from 'secondaries'.

To get the result you want, join to the sales table just one time, and match on both the StoreID and the ParentID.

To get Individual Sales , only include in the SUM rows where the StoreID matches, something like this:

SELECT Store.ID AS `Store ID`
     , SUM(IF(main.StoreID = Store.ID,
       IF(YEAR(main.SalesDate) = 2012 AND QUARTER(main.SalesDate) = 1,main.SalesAmount,0)
       ,0)) AS `Individual Sales`
     , SUM(
       IF(YEAR(main.SalesDate) = 2012 AND QUARTER(main.SalesDate) = 1,main.SalesAmount,0)
       ) AS `Consolidated Sales`
  FROM Store 
  LEFT JOIN Sales AS `main` ON main.StoreID = Store.ID OR main.StoreID = Store.ParentID
 GROUP BY Store.ID

---

My bad. (DOH!) ParentID is on the Store table, not the Sales table.

The query above does not return the specified result. (Working on it.)

I think you had the solution already...

Use the ParentID column in place of the ID column from the Store table. Where the ParentID column is NULL, we use the value from the ID column.

 
 
 
  
  SELECT IFNULL(Store.ParentID,Store.ID) AS `Store ID` , SUM(IF(main.StoreID = Store.ID, IF(YEAR(main.SalesDate) = 2012 AND QUARTER(main.SalesDate) = 1,main.SalesAmount,0) ,0)) AS `Individual Sales` , SUM( IF(YEAR(main.SalesDate) = 2012 AND QUARTER(main.SalesDate) = 1,main.SalesAmount,0) ) AS `Consolidated Sales` FROM Store s LEFT JOIN Sales AS `main` ON main.StoreID = Store.ID OR main.StoreID = Store.ParentID GROUP BY IFNULL(Store.ParentID,Store.ID)
 
  

---

This query returns the specified result set:

 SELECT t.StoreID AS `Store ID` , SUM(IF(t.source='p', IF(YEAR(t.SalesDate) = 2012 AND QUARTER(t.SalesDate) = 1,t.SalesAmount,0) ,0)) AS `Individual Sales` , SUM( IF(YEAR(t.SalesDate) = 2012 AND QUARTER(t.SalesDate) = 1,t.SalesAmount,0) ) AS `Consolodiated Sales` FROM ( SELECT 'p' AS source , a.StoreID , a.SalesDate , a.SalesAmount FROM Sales a UNION ALL SELECT 's' AS source , s.ParentID , b.SalesDate , b.SalesAmount FROM Sales b JOIN Store s ON s.ID = b.StoreID WHERE s.ParentID IS NOT NULL ) t GROUP BY t.StoreID ORDER BY t.StoreID 

This is not the most efficient, the inline view (or "derived table") is on the order of the size of the Sales table. It would be more efficient to push the predicates on the date down into the inline view, or to summarize by month, or quarter, in the derived table.

I would more likely return the "quarter" as part of the result set, to allow me to pull more than one quarter.

 SELECT t.StoreID AS `Store ID` , t.SalesQuarter , SUM(IF(t.source='p',t.SalesAmount,0)) AS `Individual Sales` , SUM(t.SalesAmount) AS `Consolodiated Sales` FROM ( SELECT 'p' AS source , a.StoreID , ADDDATE(MAKEDATE(YEAR(a.SalesDate),1), INTERVAL FLOOR(MONTH(a.SalesDate)/4) QUARTER) AS SalesQuarter , SUM(a.SalesAmount) AS SalesAmount FROM Sales a -- WHERE a.SalesDate >= '2012-01-01' -- AND a.SalesDate < '2012-04-01' GROUP BY a.StoreID, SalesQuarter UNION ALL SELECT 's' AS source , s.ParentID , ADDDATE(MAKEDATE(YEAR(b.SalesDate),1), INTERVAL FLOOR(MONTH(b.SalesDate)/4) QUARTER) AS SalesQuarter , SUM(b.SalesAmount) AS SalesAmount FROM Sales b JOIN Store s ON s.ID = b.StoreID WHERE s.ParentID IS NOT NULL -- AND a.SalesDate >= '2012-01-01' -- AND a.SalesDate < '2012-04-01' GROUP BY s.ParentID, SalesQuarter ) t GROUP BY t.StoreID, t.SalesQuarter ORDER BY t.StoreID, t.SalesQuarter 

Answer 2

I usually try to keep the filtering logic in the WHERE clause. Your original query might have too many JOINS and cases where there are no children might not give you what you expect. The IS NULL check in the WHERE below allows for stores with no children.

SELECT
  s.ID
  , SUM(s.SalesAmount) AS `Individual Sales`
  , SUM(s.SalesAmount) + SUM(c.SalesAmount) AS `Consolidated Sales`
FROM Store AS s
LEFT OUTER JOIN Store AS c ON c.ParentID = s.ID
WHERE YEAR(s.SalesDate) = 2012 AND QUARTER(s.SalesDate) = 1
  AND (c.SalesDate IS NULL OR
    (YEAR(c.SalesDate) = 2012 AND QUARTER(c.SalesDate) = 1))
GROUP BY s.ID