SQL one table aggregation

Question

For the last few days I have been attempting to find a method to pull a very important set of information from a table that contains what I call daily counts. I have a table that is setup as follows.

person|company|prod1|prod2|prod3|gen_date

Each company has more than one person, and each person can have different combinations of products that they have purchased. What I have been trying to figure out is a SQL statement that will list the number of people that have bought a particular product per company. So an output similar to this:

Comp ABC | 13 Prod1 |  3 Prod2 | 5 Prod 3
Comp DEF |  2 Prod1 | 15 Prod2 | 0 Prod 3
Comp HIJ |  0 Prod1 |  0 Prod2 | 7 Prod 3 

Currently if a person did not select a product the value being stored is NULL.

Best I have right now is 3 different statements that can produce this information if run on their own.

SELECT Count(person) as puchases, company 
FROM Sales  WHERE prod1 = '1' and gendate = '3/24/2010' 
Group BY company

Answer 1

If you just want to check whether the value is in any of the product fields then that is simply done with an OR operator:

SELECT company, COUNT(person) as purchases
FROM Sales
WHERE (prod1 = '1' OR prod2 = '1' OR prod3 = '1')
AND gendate = '3/24/2010'
GROUP BY company

This won't perform very well, however, and you'll have a hard time getting it to perform well, because your schema hasn't been normalized properly. If you can, you should fix it to something like this:

Person (PersonID, CompanyID)
Sales (PurchaseID, PersonID, ProductID, GenDate)

Then this query (and many other queries) will be a lot easier to write:

SELECT p.CompanyID, COUNT(*) AS purchases
FROM Person p
INNER JOIN Sales s
    ON s.PersonID = p.PersonID
WHERE s.ProductID = 1
AND s.GenDate = '20100324'
GROUP BY p.CompanyID

Answer 2

SELECT      company,
            SUM(COALESCE(prod1, 0)) AS total_prod1,
            SUM(COALESCE(prod2, 0)) AS total_prod2,
            SUM(COALESCE(prod3, 0)) AS total_prod3
FROM        Sales  
WHERE       gendate = '2010-03-24' 
GROUP BY    company

But you definitely should normalize you table - split it in 4:

• Company,
• Person,
• Product,
• Person_Product_Purchase (with the date of the purchase).