Question

I’m having problems with a join in MySQL. I have 2 tables, customerorder and customerorderpos. The pos table contains the items ordered – so there is always just one entry per order in the customerorder table but there may be several with the same customerorderid in the customerorderpos.

Everything works in the query except when I try to do a sum calculation on customerorder. For example sum(cart_total_complete). This is returning duplicate values (it’s counting more than once for the orders with multiple customerorderpos records).

I’m sure it’s something basic, either with the join type or how I use distinct, but I’ve been trying for hours and just can’t get it to work…

Any ideas what I’m doing wrong?
Thanks for your help!!

select concat(left(dayname(from_unixtime(customerorder.datetime)),3), ' ',
day(from_unixtime(customerorder.datetime))) as day, 
count(distinct customerorder.customerorderid) as count_totalorders, 
count(distinct customerorderpos.itemid) as count_differentitems, 
sum(customerorderpos.quantity_ordered - customerorderpos.quantity_cancelled) as quantity_ordered, 
sum(customerorderpos.itemsubtotal) as item_subtotal, 
sum(customerorderpos.pricechangetotal) as item_pricechangetotal, 
sum(customerorderpos.itemtotal) as item_total, 
sum(customerorderpos.purchase_price * (customerorderpos.quantity_ordered - customerorderpos.quantity_cancelled)) as item_purchasepricetotal, 
sum(cart_discounttotal) as total_discount, 
sum(customerorderpos.itemtotal) - sum(customerorderpos.purchase_price * (customerorderpos.quantity_ordered - customerorderpos.quantity_cancelled)) - sum(cart_discounttotal) as item_earningstotal, 
sum(cart_total_shipping) as total_shipping, 
sum(cart_total_tax) as total_tax, 
sum(cart_total_complete) as total_complete 
from customerorder inner join customerorderpos on customerorderpos.customerorderid = customerorder.customerorderid 
where customerorder.status_cancelled = 0 
group by day(from_unixtime(customerorder.datetime)) order by customerorder.datetime

Answer 1

If you first aggregate your customerorderpos table by customerorderid:

SELECT   customerorderid,
         SUM(quantity_ordered - quantity_cancelled) AS quantity_ordered,
         SUM(itemsubtotal) AS item_subtotal,
         SUM(pricechangetotal) AS item_pricechangetotal,
         SUM(itemtotal) AS item_total,
         SUM(purchase_price * (
           quantity_ordered - quantity_cancelled
         )) AS item_purchasepricetotal,
FROM     customerorderpos
GROUP BY customerorderid

You can then join the result to your customerorder table:

SELECT   FROM_UNIXTIME(customerorder.datetime, '%a %e') AS day, 
         COUNT(*)                       AS count_totalorders,
         SUM(i.quantity_ordered)        AS quantity_ordered, 
         SUM(i.item_subtotal)           AS item_subtotal,
         SUM(i.item_pricechangetotal)   AS item_pricechangetotal, 
         SUM(i.item_total)              AS item_total, 
         SUM(i.item_purchasepricetotal) AS item_purchasepricetotal, 
         SUM(o.cart_discounttotal)      AS total_discount, 
         SUM(
           i.item_total
         - i.purchasepricetotal
         - o.cart_discounttotal
         )                              AS item_earningstotal, 
         SUM(o.cart_total_shipping)     AS total_shipping, 
         SUM(o.cart_total_tax)          AS total_tax, 
         SUM(o.cart_total_complete)     AS total_complete 
FROM     customerorder o JOIN (
  SELECT   customerorderid,
           COUNT(DISTINCT itemid) AS count_differentitems,
           SUM(quantity_ordered - quantity_cancelled) AS quantity_ordered,
           SUM(itemsubtotal) AS item_subtotal,
           SUM(pricechangetotal) AS item_pricechangetotal,
           SUM(itemtotal) AS item_total,
           SUM(purchase_price * (
             quantity_ordered - quantity_cancelled
           )) AS item_purchasepricetotal,
  FROM     customerorderpos
  GROUP BY customerorderid
) i USING (customerorderid)
WHERE    o.status_cancelled = 0 
GROUP BY FROM_UNIXTIME(o.datetime, '%e')
ORDER BY o.datetime

Note that I have removed count_differentitems as it is not clear if you want to aggregate the number of distinct items in each order (which would be a simple case of extending the above) or if you want the number of distinct items across all aggregated orders (which would require an additional table join).

Note also that I have used the date format argument to FROM_UNIXTIME() in place of your manual attempt at creating a formatted date string.