Conditionally joining tables - sql

Question

I want to create conditional join to a table in my T-SQL query. Tables used in this example are from Northwind database (with only one additional table ProductCategories)
Table Products and table Categories have many-to-many relationship, hence table ProductCategories comes into picture.
I need the sum of Quantity column on table OrderDetails for each of the products falling under certain category. So I have a query like one below
Select p.ProductName, Sum(od.Quantity) As Qty
From Products p
Join OrderDetails od On od.ProductID = p.ProductID
Join ProductCategories pc On pc.ProductID = p.ProductID
And pc.CategoryID = #CategoryID
Group By p.ProductName
#CategoryID is an optional parameter. So in case it's not supplied, the join to table ProductCategories will not be required
and query should look like one below
Select p.ProductName, Sum(od.Quantity) As Qty
From Products p
Join OrderDetails od On od.ProductID = p.ProductID
Group By p.ProductName
I want to achieve this without repeating the whole query with If conditions (as below)
If #CategoryID Is Null
Select p.ProductName, Sum(od.Quantity) As Qty
From Products p
Join OrderDetails od On od.ProductID = p.ProductID
Group By p.ProductName
Else
Select p.ProductName, Sum(od.Quantity) As Qty
From Products p
Join OrderDetails od On od.ProductID = p.ProductID
Join ProductCategories pc On pc.ProductID = p.ProductID
And pc.CategoryID = #CategoryID
Group By p.ProductName
This is simplified version of the query which has many other tables and conditions like ProductCategories. And will require multiple multiple If conditions and repetition of the query. I have also tried dynamically generating the query. It works but query is not readable at all.
Any solutions?
Thank you.

Answer 1

Try this, if you'll use properly parametrized query - there will be no performance impact but may be gain:
Select p.ProductName, Sum(od.Quantity) As Qty
From Products p
Join OrderDetails od On od.ProductID = p.ProductID
WHERE #CategoryID IS NULL OR EXISTS (SELECT * FROM ProductCategories WHERE CategoryID = #CategoryID AND ProductID = p.ProductID)
Group By p.ProductName
Actually
in your query if there can be multiple rows in ProductCategories for one row in OrderDetails - then you get duplicates of od.Quantity in your SUM - is it an intended behavior?

Answer 2

I believe you can left join the tables vs the implicit inner join you are doing.
In an inner join it matches the key on the source table with each instance of the key on the destination table.
Each instance of a match on the destination table generates a set of rows displaying the
match types.
With an outer join it will display the source row EVEN IF there is no matching row in the other table. If there is it will do essentially the same as an inner join and you'll get a row back for each instance of match. So you're getting the data you need when it's available and not getting what data is unavailable.
Take this as an example
select * from Products
Left join ProductAndCategory on ProductAndCategory.ProductID = Products.ProductID
left join Categories on Categories.CategoryID = ProductAndCategory.CategoryID
Where I have a simple Product table with a ProductID and a ProductName a ProductAndCategory table with a ProductID and a CategoryID and a Categories table with a CategoryID and a CategoryName
It will show the rows that have categories with the joined categories and the rows that don't have categories will just show the one row with null for the values that don't exist.

Answer 3

I'm not very familiar with T-SQL so I don't know if this will cut it but in mysql you could do something like
Select p.ProductName, Sum(od.Quantity) As Qty
From Products p
Join OrderDetails od On od.ProductID = p.ProductID
Join ProductCategories pc On pc.ProductID = p.ProductID
And pc.CategoryID = if(#CategoryID is null , pc.CategoryID, #CategoryId)
Group By p.ProductName
Yes and if() still remains but just "one query", if that's what you're looking for.
At the same time, you say you were able to dynamically generate a single query. If so is the readability that much of an issue? If your generated query was more performant over what I suggested above I'd go with that. It's generated; you won' be manually tweaking/reading the result.