I was pulled into a meeting to help out with a query that needed to find some data to update a field. Basically the situation is:
1. a file is received from an external entity that has 2 pieces of information needed for the lookup, a GROUP and a UCC*
2. this file is loaded into a table and then an ETL (Extract, Transform, Load) process is called
3. the ETL needs to know a PriceType that is stored in a table that contains GROUP and VendorID
4. VendorID can be found in a query using the UCC

Ultimately what I need is the PriceType to include in my ETL process.
If I know the GROUP and UCC, I can find the VendorID that is needed to determine the correct PriceType.

So, the current process takes FOREVER – days!! It is using cursors and fetch next into the cursor and a whole lot of things that I don’t get! That is why we were having the meeting…to try and find a way to make the load run in a reasonable amount of time.

So I am listening to the conversation and my co-worker keeps talking about looping through the database to find the VendorID. And I am thinking, maybe I am not getting it, but it just seems like we can take the data we know, GROUP and UPC, and using JOINs get the VendorID. Why would we need to loop? So I asked and everyone agreed that what I was saying sounded reasonable, but weren’t really positive that what I was saying would work.

So, I rewrote this UPDATE_CURSOR statement with a query. I got it 90% there. The query was then put in a table and called as part of the ETL and I was missing a key field to join on in order to perform the update, but once that piece was included the process completed in an hour!

From 27 hours with no completion to complete in an hour! That is the power of joins and thinking in a set. If you are looping through your data set to find data in your tables it try to approach it and see if you can find a way for it to be done in a single query!!

The idea of iterating is a very programmatic approach and has it’s place, but not when you are trying to extract data from a database. You should try to think of everything you want to get and what connections have to be made in order to get that data.

Start with a small query. What is the main table that you need the most data from or has the key field that will be needed to find data in other tables? Write the query that just gets those fields from that main table. Where is the next piece of data that you need? Are there any criteria that need to be applied before JOINing into that data? Can you JOIN into the table or do you need to write a SUB-QUERY to extract just some of the data from that table? Once you get the details of the secondary information worked out, JOIN your main query to that and add the fields you need. It maybe that some of the fields you included you don’t need as output, but you do need as something to JOIN on in a later part of the query.

As you can see, building a query to get your full result set takes some time and knowledge of your database structure but it really how queries are supposed to work.

Now, once your application gets the result set returned, feel free to iterate through it as many times as you like to get data from it, but the idea of running one query to get a piece of information and looping through that data to run another query should now be contrary to all you believe as a Platypus Hunter.

*The UCC is part of the UPC that is on products. It is assigned to a manufacturer and is the first 6 digits in a UPC. Every product a company makes has a UPC that starts with their assigned UCC (more information can be found here http://www.insightu.org/hobby/guide_wd/ch3.htm)