Confused by JOINs - data missing in result

问题

This thread is related with my older one Combine fields from different rows on condition.

I adjusted the query I got from this thread to meet some other requirements.

SELECT 
    a.Date,
    a.orderid AS AZNr,
    a.Typ,
    ROUND(a.Fees, 2) AS Fees,
    ROUND(b.Shipping, 2) AS Shipping,
    ROUND(c.Price, 2) AS Price,
    d.DeliveryLand
FROM
    (SELECT 
        posteddate AS Date,
            transactiontype AS Typ,
            orderid,
            SUM(amount) AS Fees
    FROM
        report
    WHERE
        amounttype = 'ItemFees'
    GROUP BY orderid) a
        LEFT JOIN
    (SELECT 
        orderid, SUM(amount) AS Shipping
    FROM
        report
    WHERE
        amountdescription = 'Shipping'
    GROUP BY orderid) b ON a.orderid = b.orderid
        LEFT JOIN
    (SELECT 
        orderid, SUM(amount) AS Price
    FROM
        report
    WHERE
        amountdescription = 'Principal'
    GROUP BY orderid) c ON b.orderid = c.orderid
        LEFT JOIN
    (SELECT 
        DeliveryLand, ExternalOrderId
    FROM
        orders) d ON c.orderid = d.ExternalOrderId
ORDER BY Date DESC

I had to do a LEFT JOIN on the last table to get the DeliveryLand, but not every item from the report-table has one entry in the orders-table.

After I did some calculating I found that this query does not return all entries. There should be 25 entries with "typ" = "Refund", but I only get 20 of them. Some orders might contain entries that other orders (identified by orderid) might not have.

I did another query to sum up the fees, the shipping costs and the price just for the refunds:

SELECT 
    SUM(ROUND(a.Fees, 2)) AS Fees,
    SUM(ROUND(b.Shipping, 2)) AS Shipping,
    SUM(ROUND(c.Price, 2)) AS Price
FROM
    (SELECT 
        orderid, SUM(amount) AS Fees
    FROM
        report
    WHERE
        amounttype = 'ItemFees'
            AND transactiontype = 'Refund'
    GROUP BY orderid) a
        LEFT JOIN
    (SELECT 
        orderid, SUM(amount) AS Shipping
    FROM
        report
    WHERE
        amountdescription = 'Shipping'
            AND transactiontype = 'Refund'
    GROUP BY orderid) b ON a.orderid = b.orderid
        LEFT JOIN
    (SELECT 
        orderid, SUM(amount) AS Price
    FROM
        report
    WHERE
        amountdescription = 'Principal'
            AND transactiontype = 'Refund'
    GROUP BY orderid) c ON b.orderid = c.orderid

The first two results, fees and shipping costs, are summed up correctly (I got the original data for comparison), but the last one, the price, isn't correct, its too much. I guess there is some data getting truncated by the LEFT JOIN, but I can't figure out why and where, especially this query works perfectly fine when I sum up the same colums for "transactiontype" = "Order".

I don't know why there is some data truncated or missing. Can somebody help me with these confusing JOINs I do in both queries? If you need more information, please ask.

Thank you in advance!

€dited query:

  SELECT 
    posteddate AS Date,
    transactiontype AS Typ,
    report.orderid AS AZNr,
    ROUND(SUM((amounttype = 'ItemFees') * amount),
            2) AS Fees,
    ROUND(SUM((amountdescription = 'Shipping') * amount),
            2) AS Shipping,
    ROUND(SUM((amountdescription = 'Principal') * amount),
            2) AS Price,
    orders.DeliveryLand,
    articles.ItemVAT AS VAT
FROM
    report
        LEFT JOIN
    orders ON report.orderid = orders.ExternalOrderID
        LEFT JOIN
    articles ON report.sku = articles.ItemID
GROUP BY report.orderid , transactiontype

回答1:

I'd start by simplifying the query, it can be written without JOINs. I'd also question why you ROUND before the SUM which will take longer and be less precise.

I'd do:

  SELECT orderid,
         ROUND(SUM((amountdescription='ItemFees')*amount), 2)) AS Fees,
         ROUND(SUM((amountdescription='Shipping')*amount), 2)) AS Shipping,
         ROUND(SUM((amountdescription='Principal')*amount), 2)) AS Price
    FROM report
   WHERE transactiontype='Refund'
GROUP BY orderid

As an explanation (amountdescription='ItemFees') returns 1 if true and 0 if not, thus the amounts are only summed that match each condition specified.. you can use the longer CASE if you prefer:

SUM(CASE WHEN *condition* THEN amount ELSE 0 END) 

You may be losing some data from the original query as it relies on all the subtables linking up.. if there are no Fees then an order's Shipping and Price would not have been returned.

UPDATE

Don't forget if you have two transactiontypes on the same orderid, they will all count as the same order and will both be included in the same sum.. to get the transaction types separately use:

  SELECT orderid, transactiontype,
         ROUND(SUM((amountdescription='ItemFees')*amount), 2)) AS Fees,
         ROUND(SUM((amountdescription='Shipping')*amount), 2)) AS Shipping,
         ROUND(SUM((amountdescription='Principal')*amount), 2)) AS Price
    FROM report
GROUP BY orderid, transactiontype

来源：https://stackoverflow.com/questions/26128296/confused-by-joins-data-missing-in-result