compare two versions of a feature class to find difference

I think you can use a serie of ChangeDetectors for this.

First compare both featureclasses on GlobalID only to find the new records in each featureclass. Then compare the Unchanges output again on all attributes and geometry.

Use a FGDB reader and Csv writer for in and output

Alternative method

Add with Schema (Any Format) reader

Thanks koen. what is the name for Transformer NEW_in_A and New_in_B in your diagram?

In my exampe it's an attributecreator, but those transformers are irrelevant, thats just for the picture.

It's the ChangeDetectors who do the work.

There are various transformers you can use for this. The ChangeDetector is the most obvious one.

The Matcher and DuplicateFilter are others you can use, though you then need to test the value of fme_feature_type to see where each original feature came from.

Finally, if performance is important, you could try the FeatureMerger. Technically this transformer is for merging data on a common key, but obviously if there is no merge then there are no matching attributes, so it is a test of sorts and way faster transformer than the ChangeDetector.

So, lots of options, but the ChangeDetector is what I suggest to try first.

Oh - also see this blog post - tip #6 covers some of what I've said but in more detail.

I have tested the above workspace. I can get added, deleted, and attribute-modified records. But for attribute-modified records, is there a way to know which attribute or attributes have been changed for each feature? for instance, for feature with an EventID of {120oui11000}, field A changed from "12345" to "54321", field B changed from "On" to "Off", and so on.

thanks Mark for your response. do you know if there is a way in FME to capture specific attribute changes? For instance, for feature with an EventID of {120oui11000}, field A changed from "12345" to "54321", field B changed from "On" to "Off", and so on.

I think to pick up individual changes you would need to use a ChangeDetector for each attribute, so you could find out which specific attribute is different. To be honest that's not very efficient so I will file an enhancement request with our developers to ask if they can tag features with which component is different. Doesn't do much for you in the short term, but it would help in the future.

Alternatively, you could try using the AttributeSplitter before the ChangeDetector. That would give you a separate feature per attribute, so when a feature fails to match, you know which attribute it is and therefore which attribute is different. (it's like pratap's schema reader idea, just a little easier to implement).

Hope that helps.

Appreciate your response, Mark!

Hi @fmenewbie

we improved ChangeDetector in FME 2019. Now your problem can be solved with a single ChangeDetector that creates report with all changes between the original dataset and the revised dataset.

Please give FME 2019 beta a try.