DB editor does not add new classes if they are empty

[jkiviluo]

Commit will not add new empty (multi-dimensional?) classes. Apparently there needs to be some data there.

[jkiviluo]

With data included, I got this error (and commit apparently failed. EDIT: apparently succeeded despite the message):

DBAPIError while inserting entity_class items: ('UNIQUE constraint failed: entity_class_dimension.entity_class_id, entity_class_dimension.dimension_id, entity_class_dimension.position',)

[PiispaH]

Might be related to #2293 and #2296

[jkiviluo]

Right. And the data wasn't actually committed. I closed and opened the DB editor, but that's not enough (apparently stuff remains in the memory). Only closing and starting Toolbox revealed that new entity class and it's data were not in the DB.

[manuelma]

What are steps to reproduce this? I just added and committed an empty multi-D class with no problem?

[jkiviluo]

It works for me too, if I start from an empty DB. But fails on the attached one. Of course, maybe it's due to corrupt DB that wouldn't happen anymore, so not sure if should be debugged (and I would just need to reimplement my DB).

generic_energy_data_specification.sqlite.txt

[jkiviluo]

I tried to look into this, since it's stopping me from progressing. It seems like the referential integrity between entity_class and entity_class_dimension tables is not working. I have rows in entity_class_dimension that have entity_class_ids not found in entity_class (since the database has seen deletion of entity_classes). This causes the error message when trying to insert records that would re-use the entity_class_id numbers.

I don't know where to fix this... Probably somewhere in the database definition, but I leave it to you @manuelma. I'll try to fix my database manually meanwhile.

[manuelma]

My hope is @PiispaH already fixed this with spine-tools/Spine-Database-API@7fa4b1e

It is somewhat clear to me how things would get messy the way you @jkiviluo describe, prior to the changes introduced by @PiispaH in the above commit.

Should we try and save the corrupted DBs? there might be a way...

[jkiviluo]

No need on my part - I've already remade the DB. So, unless someone requests that, let's not waste time trying to do that. Presumably there are so many ways things can go wrong that it can be futile.