Join returning zeroes for a column that should have values

Question

I believe I have formatted my query incorrectly but my Google skills haven't come up with a solution that works. This is my query:

use movement;
select t1.StoreNumber, t1.saleDate, t1.departmentNumber, sum(t1.dollarsSold), sum(ifnull(t2.loss,0))
from movement.movement t1
left join knownLoss.producekl t2 ON t1.StoreNumber = t2.Store
where t1.StoreNumber = 3
AND (t1.departmentNumber = 10 OR t1.departmentNumber = 20 OR t1.departmentNumber = 27 OR t1.departmentNumber = 30 OR t1.departmentNumber = 40 OR t1.departmentNumber = 50 OR t1.departmentNumber = 51 OR t1.departmentNumber = 60 OR t1.departmentNumber = 70 OR t1.departmentNumber = 80 OR t1.departmentNumber = 81 OR t1.departmentNumber = 82 OR t1.departmentNumber = 90 OR t1.departmentNumber = 95 OR t1.departmentNumber = 96 OR t1.departmentNumber = 97)
AND t1.saleDate > date_sub(curdate(), interval 7 day)
group by t1.saleDate, t1.StoreNumber, t1.departmentNumber;

My intended output is to give a table grouped by date of sale, store, and department summing sales totals and sales losses in the final two columns

My expected output should look like this:

-------------------------------------------------------------
| Store Number | Sale Date | Dept Num | Dollars Sold | Loss |
-------------------------------------------------------------
|      1       | 2017-3-9  |    10    |    7435.26   | 0.00 |
-------------------------------------------------------------
|      1       | 2017-3-9  |    20    |    30.50     | 3.00 |
-------------------------------------------------------------
|      1       | 2017-3-10 |    10    |    2503.22   | 4.00 |
-------------------------------------------------------------
|      2       | 2017-3-11 |    10    |    5685.49   |17.50 |
-------------------------------------------------------------
|      3       | 2017-3-9  |    10    |    75.26     | 4.55 |
-------------------------------------------------------------
|      4       | 2017-3-9  |    10    |    7435.26   | 0.00 |
-------------------------------------------------------------
|      4       | 2017-3-13 |    30    |      20.45   | 1.99 |
-------------------------------------------------------------

Instead of my expected output the loss column the loss column is all zeroes.

-------------------------------------------------------------
| Store Number | Sale Date | Dept Num | Dollars Sold | Loss |
-------------------------------------------------------------
|      1       | 2017-3-9  |    10    |    7435.26   | 0.00 |
-------------------------------------------------------------
|      1       | 2017-3-9  |    20    |    30.50     | 0.00 |
-------------------------------------------------------------
|      1       | 2017-3-10 |    10    |    2503.22   | 0.00 |
-------------------------------------------------------------
|      2       | 2017-3-11 |    10    |    5685.49   | 0.00 |
-------------------------------------------------------------
|      3       | 2017-3-9  |    10    |    75.26     | 0.00 |
-------------------------------------------------------------
|      4       | 2017-3-9  |    10    |    7435.26   | 0.00 |
-------------------------------------------------------------
|      4       | 2017-3-13 |    30    |      20.45   | 0.00 |
-------------------------------------------------------------

Even though when I query t2 alone it returns results

use knownLoss;
select t2.Store, t2.Department, t2.klDate, sum(t2.loss)
from producekl t2
group by klDate, Store, Department;

So my join must be messed up but I can't figure out how. Any ideas?

Answer 1

I finally worked the issue out. Neither of the tables I was joining had unique values, therefore I had to use subqueries to produce the unique values. Here is the final query that resulted in the output I expected.

SELECT t2.StoreNumber
 , t2.departmentNumber
 , sum(t2.Sales) AS netSales
 , ifnull(sum(t3.loss),0) AS knownLoss
 , ifnull(round(avg(t3.loss / t2.Sales),4),0) AS percentLoss
from ( SELECT t1.StoreNumber
   , t1.departmentNumber
   , SUM(t1.dollarsSold) AS Sales
  FROM movement.movement t1
  WHERE t1.StoreNumber = 1
  AND t1.departmentNumber = 80
  AND t1.saleDate >= date_sub(SUBDATE(curdate(), WEEKDAY(curdate())), interval 7 day)
  AND t1.saleDate < SUBDATE(curdate(), WEEKDAY(curdate()))
  GROUP BY t1.StoreNumber
   , t1.departmentNumber) t2
   left join
   (select Store, Department, sum(loss) AS loss
    from producekl
    where Department = 80
    AND Store = 1
    AND klDate >= date_sub(SUBDATE(curdate(), WEEKDAY(curdate())), interval 7 day)
    AND klDate < SUBDATE(curdate(), WEEKDAY(curdate()))
    group by Store, Department) t3 ON t2.StoreNumber = t3.Store
GROUP BY t2.StoreNumber, t2.departmentNumber

Thanks to Strawberry for encouraging good formatting, it helped me resolve my issue faster.

Answer 2

FWIW, I find this easier to read...

I'm unfamiliar with the curdatee() function, so I substituted it for something else...

SELECT t1.StoreNumber
     , t1.saleDate
     , t1.departmentNumber
     , SUM(t1.dollarsSold)
     , SUM(IFNULL(t2.loss,0))
  FROM movement.movement t1
  LEFT
  JOIN knownLoss.producekl t2 
    ON t1.StoreNumber = t2.Store
 WHERE t1.StoreNumber = 3
   AND t1.departmentNumber IN(10,20,27,30,40,50,51,60,70,80,81,82,90,95,96,97)
   AND t1.saleDate > DATE_SUB(CURDATE(), INTERVAL 7 DAY)
 GROUP BY t1.saleDate
     , t1.StoreNumber
     , t1.departmentNumber;

Note that saledate is missing from the GROUP BY. BAD.