QueryRouter design is not clear

Not sure if this is an issue or not.

The design intention of the QueryRouter database router is not clear.

https://git.astron.nl/astron-sdc/esap-api-gateway/-/blob/master/esap/query/database_router.py

From what I understand from the the DJango documentation and experimenting with different configurations. Django will try each database router listed in the DATABASE_ROUTERS array and use the first non-null response that it gets.

If a router matches a query it returns a result, the database name for db_for_read() and db_for_write(), or [True|False] for allow_relation() and allow_migrate(). If a router doesn't match a query, it should return None, allowing Django to move on to the next router in the list.

The code for QueryRouter always returns a result for db_for_read() and db_for_write() This means that any routers listed after it in the DATABASE_ROUTERS array will never be used.

https://git.astron.nl/astron-sdc/esap-api-gateway/-/blob/master/esap/esap/settings/dev.py#L24-L31

The code for QueryRouter matchs requests for batch, ida, rucio, and accounts and routes them to the corresponding database. Anything else is routed to the default database.

The result is that when adding new database router for a new app it must be added before the QueryRouter entry in DATABASE_ROUTERS or it will be ignored.

If this is by design, it would be useful to document it in the source code, and we should probably remove the unused accounts, ida, batch, and rucio routers.