Embedding allows to add non-interesting classes in the embedding table

[skinkie]

2025-01-29 01:04:42,325 - script_runner - ERROR - 'PostalAddress_VersionStructure' Traceback (most recent call last):
  File "/mnt/storage/home/skinkie/Sources/references/gtfs-netex-test/epip_db_to_db.py", line 77, in <module>
    main(args.source, args.target)
  File "/mnt/storage/home/skinkie/Sources/references/gtfs-netex-test/epip_db_to_db.py", line 66, in main
    missing_class_update(source_db, target_db)
  File "/mnt/storage/home/skinkie/Sources/references/gtfs-netex-test/netexio/dbaccess.py", line 708, in missing_class_update
    embedded_classes = set(target_db.embedded())
                           ^^^^^^^^^^^^^^^^^^^^
  File "/mnt/storage/home/skinkie/Sources/references/gtfs-netex-test/netexio/database.py", line 62, in embedded
    tables = {self.get_class_by_name(table) for table, in cur.fetchall() if table[0].isupper()} # TODO: Remove other classes from default namespace!
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/storage/home/skinkie/Sources/references/gtfs-netex-test/netexio/database.py", line 62, in <setcomp>
    tables = {self.get_class_by_name(table) for table, in cur.fetchall() if table[0].isupper()} # TODO: Remove other classes from default namespace!
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/storage/home/skinkie/Sources/references/gtfs-netex-test/netexio/database.py", line 66, in get_class_by_name
    return self.serializer.name_object[name]
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^
KeyError: 'PostalAddress_VersionStructure'

[ue71603]

sorry. didn't see that. As not assigned to me.

[skinkie]

It still resolves into the question: the PostalAddress has an id-attribute in their dataset, lacks version, is never referenced. Our embedding code triggers on the fact it has an id. Because of the lack of any enforcement of id's being required or not (that part was not merged NeTEx-CEN/NeTEx#471 ) need to do something better. I think the approach of selecting the class by being interesting or not would make sense. I am testing now if that could work (also with respect to the DuckDB-UDF).

[ue71603]

Good idea. i did things like that in other places already.

[skinkie]

Solved.