Creating canonical item without primary key in data hub

[former_member1113192]

Hi Experts,

I have a requirement where canonical item needs to be created without primary key. E.g. addressId is not required in Address Item. If I try to do this, I am getting following error:

com.hybris.datahub.validation.ValidationException: [ValidationFailure{FATAL, "No primary key is set for canonical item CanonicalCustomer"}] at com.hybris.datahub.validation.AbstractValidator.validate(AbstractValidator.java:32)

Is there a way to define canonical Item without primary key?

[mpern · Accepted solution]

No, there is no way, because how would you create a canonical item out of your raw item without a primary key to determine a match?

And what happens in case of an update? To use your example: how would you know which Address needs to be updated without some kind of primary key?

[former_member1113192 · Accepted solution]

So is there any way to auto generate primary key(Address id in my case) in Data Hub? If yes, How do we do that in our canonical extension xml file?

[mpern · Accepted solution]

I would highly recommend that you get some kind of ID for the address from the source system, for the reasons mentioned above. I bet they do have some kind of identifier, even if you have to be a bit creative (e.g. use the customer ID as address ID, in case a customer can only have a single address)

If that isn't an option, then you have to generate some kind of unique identifier when loading the raw items. Maybe you can write a Spel Expression to generated a GUID in the canonical XML file, or you have to provide it later in the processing with a grouping handler (never tried that).

But you may amass a lot of unnecessary data in SAP Commerce that way. Every address has a new unique id, even updates for existing addresses -> every update is a new address in SAP Commerce, if you don't have some special handling during publish / import of the data in Commerce.

In other words, just because you have the requirement that addresses don't have a primary key, doesn't mean that the requirement makes sense or is even feasible when integrating data.