The best way is hard to say without knowing the details so the answer is "it depends"...

• If you have a small volume of data to test against a small volume of data stored, which I think is your situation, I would load both sets and use the ChangeDetector.
• If you have a small volume of data to test against a large volume of data stored, I would try to make use of database functionality and only load records where STATION_NAME and DATETIME matches to test for changes.

Small can vary but loading a large dataset for processing really can bog down you process. And if you need to check for changes every hour, and the process runs for more than an hour you need to change your strategy.

I think the ChangeDetector is not that hard to work with. When learning something new it helps to generate a small test environment. For example two datasets having one unchanged, one updated, one deleted and one updated record. Feed them both to the ChangeDetector and see if it works as expected. I attached a workspace demonstrating this.

Hi Niels,

Thank you - I have a small amount of data to check against a large volume of stored data. It's indexed, but takes a long time to check. I'm currently using a Feature_Joiner and have attached a Database Updater to the Joined port and a Writer to the Unjoined Left port.

Cheers for the sample workflow - that is a great help as I find the terminology around the ChangeDetector mightily confusing!

@nedwaterman​ These two webinars might give you some ideas on how to solve your change detection problem:

• Managing Changing Data with FME
• Managing Changing Data: Part 2

Thanks Mark - I will take a look!