migration of nested objects in CRIS entities: improper handling of optional nested fields

[saschaszott]

In DSC5 we are using nested objects to model affilations in researcher profiles. Each affiliation consists of 4 fields: org unit (ou pointer; mandatory), role (mandatory), start date (optional), end date (optional).

Currently, we have several affilations without start date and / or end date.

For example, in DSC5 we have one RP with 2 affiliations (screen shot)

Currently, the CRIS migration procedure (Pentaho transformation) inverts the order of affilations. This is due to step Sort position in entity_migration.krt (ascending = N)

The expected migration result of the given example RP is:

Currently, DSC7 produces an invalid migration result in case of affiliations with optional fields (as in the given example):

In this example the assignment of end date is not correct.

This bug is caused by the Pentaho migration step Select values 4 which removes nested_object_id and positiondef in each row of the stream. This means that subsequent migration steps cannot determine the correct assignments of nested fields to a given affiliation (nested object).

This bug affects the migration of RPs that have at least 2 affilations with missing nested fields.

We'll provide a bugfix (adaption of the Pentaho migration).

[atarix83]

Hi @saschaszott

thanks for opening the issue, anyway this should be already resolved with https://github.com/4Science/DSpace/pull/276.

Feel free to re-open again the issue if is not working for you

[saschaszott]

Hi @atarix83 , the PR (276) you mentioned, does not fix the problem. We have integrated #276 into our code base and are able to reproduce the bug.

[saschaszott]

@atarix83 , the problem is raised in the pentaho transformation step named Select values 4. In this early transformation step the important sorting information in nested_object_id and positiondef is removed. We have fixed the Pentaho transformation (requires additional steps) locally. Let me know if you are interested in a PR.

[atarix83]

@saschaszott

yes please open a PR when you can, so we can verify. Thanks

[saschaszott]

@atarix83 , sorry for the long pause, but today I was able to reproduce the problem described above with the latest version of DSC (2023.02.02). To illustrate the problem, I'll give you an example of a nested affiliation object (with 3 entries):

In the migrated RP you'll find an incorrect state

As you can see in the metadata full view, there is an uneven number of affiliation.startDate and affiliation.endDate fields

[saschaszott]

To better illustrate the change in the entity migration transformation, I'll provide a before-after comparison of the change in entity-migration.ktr we propse:

Before

After

[saschaszott]

You can find our proposed bugfix in PR #425 .