2 Replies Latest reply: May 24, 2012 12:15 PM by esarachik RSS

    Link tables

      I'm new to Qlik but have been developing BI solutions for many years. Could someone please give me some help on link tables. I have attached some notes. all help most appreciated.

        • Link tables

          Hi Elaine.

           

           

          My suggestion:

           

           

          Sales:

          load

          "%FK_ItemKey"&"%FK_Customer" as "%FK_ItemKeyCustomer",

          SHead_OrderSource,

          Sell-to Customer No_,

          PK_SO.No,

          SLine_OSQuantity

          from Sales.qvd

           

           

          ItemLEntry:

          load

          "%FK_ItemKey"&"%FK_Customer" as "%FK_ItemKeyCustomer",

          PK_LEntry No_

          VEntry.Valued Quantity,

          VEntry.Document Type,

          ...

          from ItemLEntry.qvd;

           

           

          after load SnowFlake....

           

           

          concatenate(Snowflake)

          load

          "%FK_ItemKey"&"%FK_Customer" as "%FK_ItemKeyCustomer",

          %FK_ItemKey,

          %FK_Customer

          from Sales.qvd;

           

           

          concatenate(Snowflake)

          load

          "%FK_ItemKey"&"%FK_Customer" as "%FK_ItemKeyCustomer",

          %FK_ItemKey,

          %FK_Customer

          from ItemLEntry.qvd;

           

           

          Sorry bad english

          • Link tables

            A key/link table is frequently required in QlikView to resolve Synthetic Key or Circular Join issues.  It can also help to tidy  up a schema that has tables all over the place.  The goal is to create a star (and sometimes snowflake) schema with dimension tables connected to a central key table.  Unlike in classical data warehousing, the central table doesn't often have the measures - they generally stay in the dimension tables.

             

            There are 3 rules for creating a key table.  The first 2 are very straightforward - the last is where you need to use your skill and judgement to create the right result.  So, here are the rules:

             

            1.  All tables should have a primary key.  If a table doesn't have one unique key, derive one using a function like autonumber or autonumberhash256.

             

            2.  Break all the existing joins by renaming foreign keys (e.g. in the Orders table, rename CustomerID to O_CustomerID).  For a "pure" star schema, there should be no QlikView Associations remaining and all the tables should be standalone.  From a pragmatic point of view, it is fine to leave some hierarchical tables associated (e.g. Product to ProductCategory) to have a more "snowflake" schema.

             

            3.  Use a mixture of Concatenate and Join to generate the Key table using the Resident data.  You will load the previously renamed foreign key with the correct name so that they connect to the right dimension table (e.g. ... O_CustomerID As CustomerID).

             

            For an example, if I have a simple structure with Customer, Calendar, Order, OrderDetail and Product.

             

            Step 1 - Customer, Product, Calendar (DateKey), and Order already have a primary key.  In OrderDetail I will create an ID from OrderID and LineNo (we will do a bit of step 2 while we are at it):

             

               ...

               AutoNumberHash256(OrderID, LineNo) As OrderDetailID,

               OrderID as OD_OrderID,   // rename order foreign key

               ProductID as OD_ProductID, // rename product foreign key

               ...

             

            Step 2 - Customer and Product are not an issue because they don't have a foreign key.  I already renamed my foreign keys in OrderDetail so I need to attend to Order:

             

               ...

               OrderID,

               CustomerID as O_CustomerID,

               DateKey as O_DateKey,

               ...

             

            Now all my links will be broken.

             

            Step 3 - Now I load my key table.  I will begin with data from the Order table:

             

               Key:

               Load

                  OrderID,

                  O_CustomerID As CustomerID,

                  O_DateKey as DateKey,

               ...

               Resident Order;

             

            Now Join in the Product and OrderDetail keys from the OrderDetail table.

             

               Join (Key) Load

                  OD_OrderID As OrderID,

                  OD_ProductID As ProductID,

                  OrderDetailID,

               ...

               Resident OrderDetail;

             

            Now I have a key table which will connect all my detail.  I can extend this by Joining or Concatenating additional tables.  For example, I could concatenate data from Purchases that also has Date and ProductID information.

             

            Any of the previously renamed foreign keys (e.g. O_CustomerID) can actually be dropped now - their information is encapsulated in the key table so that keeping them is just duplicating data.  I might choose to hang onto them for a while, just to test my relations, but best practice will be to remove them.