Why are all objects updated on CSV import when data has not changed.

[Tim]

I'm trying to understand the criteria Insight uses for determining if an object is "Missing" during a CSV import.  I assume that if any attribute has changed, it's considered missing.

I have two scenarios that result in different behavior regarding missing objects.

• Object types with NO object reference attributes results in updates (remove/replace ) to only the objects with changed attributes in the CSV source.

• Object types WITH object reference attributes are considered missing and update (remove/replace ) ALL objects, even if the object references have not changed.

Here are examples of these scenarios.

City attribute as Default Text type

Using the following configurations and importing the employees.csv, I can execute the import configuration a second time with the exact same data and file.

These screen shots show the first and the second imports, and as expected, there are 6 identical objects on the second import.

City attribute as Object type

In this scenario the City object type is introduced, and the Employees City attribute is changed to object type City.

Now, after importing the objects initially, a second run with the exact same data and file causes all objects to be considered "missing". They are deleted and new issues are created.

This creates a particular problem when using the Update Missing Objects approach as all objects are considered missing, new objects created, and the existing objects have the configured field updated.

I'm hoping someone can explain why objects with object reference attributes always are considered missing on import and if there is a way to prevent it.

[Tim · Accepted solution]

This has been deemed a defect. 

[JSDCLOUD-10045] Import duplication when object type mapping has two identifiers and one is another referenced object - Create and track feature requests for Atlassian products. 

[Björn Gullander]

Hi

I tried to recreate your problem and it seems to behave differently on DC vs. Cloud. In DC it works as expected, but on Cloud I get the same problem with referenced attributes. 

In you case I would say you have two options: 

• Add an extra text attribute "City text" which you use as Identifier instead of the Object attribute. Like this:
• Only use User attribute as Identifier. The Identifiers will alone or together set a unique combination to find the correct existing object in the import. In your case you should probably use a username, userid or email as a single unique identifier. 

Björn

[Tim]

Thanks Bjorn,

The issue isn't so much the use case and the attributes I've described.  I have thousands objects with references to other object types: Manufacturer, Model, Location, etc.

Updating every object every time is not a good practice.

I think I will open a support request.

-tim