Assistance Needed: Handling Mixed _id Types (String and ObjectId) in Quarkus with MongoDB Panache

[nitinty]

Hi Team,

We are facing an issue with mapping the _id field in our existing MongoDB collections. Our collections already exist, and several of them have a mix of String and ObjectId values for the _id field.

In our previous setup with Spring, we handled this using:

@Id  
private String id;  

This approach worked seamlessly.

However, when attempting the equivalent in Quarkus:

@BsonId  
@BsonProperty("_id")  
private String id;  

We encounter the following error:

Failed to decode 'TestEntity'. Decoding 'id' errored with: readString can only be called when CurrentBSONType is STRING, not when CurrentBSONType is OBJECT_ID.

Could you please suggest a workaround or best practice to handle this scenario in Quarkus?

Thanks in advance for your guidance!

[quarkus-bot[bot]]

/cc @FroMage (panache), @loicmathieu (mongodb,panache)

[xstefank]

I talked to @nitinty offline about this first, but since I wasn't sure if this is intended behavior or not, I delegated to opening an issue.

If someone wants a reproducer, here it is - https://github.com/xstefank/quarkus-reproducers/tree/main/mongo-id.

[loicmathieu]

Hi,

This should be the correct implementation in case you want a custom ID field that is a String:

@BsonId    
private String id;  

However, it seems you field is sometimes a String and sometimes an ObjectId, I'm afraid it cannot work OOTB as we use the PojoCode wich only do 1 <-> 1 mapping of BSON type to Java type.
The only thing I can think off would be a custom code for your field.

[nitinty]

I tried this with fresh collection/entity but it didn't worked and through similar kind of error

Failed to decode 'PersonEntity'. Decoding 'id' errored with: readString can only be called when CurrentBSONType is STRING, not when CurrentBSONType is OBJECT_ID.

@MongoEntity(collection = "persons")
@Data
public class PersonEntity extends BaseAuditer {
    @BsonId
    private String id;
   
    private String name;
   
    private int age;
}


public interface PersonRepository extends PanacheMongoRepositoryBase<PersonEntity, String> {
    void savePeople(PersonEntity person) ;
    List<PersonEntity> findAll();
}


@ApplicationScoped
public class PersonRepositoryImpl implements PersonRepository {

    @Override
    public void save(PersonEntity person) {
        persist(person);
    }

    @Override
    public List<PersonEntity> findAll() {
        return listAll();
    }
}


********* Test Case ********

    @Test
    public void testMongoDatabaseResolverUsingPersonResource() {
        PersonEntity person = new PersonEntity();
        person.setAge(34);
        person.setName("Person");
        personRepository.save(person); 
        /* 
         Above code stored the below document in the mongoDB
          {
            _id: ObjectId('677f9200d532e830782d6195'),
            age: 34,
            name: 'Person'
          }
         */
        
        personRepository.findAll();  // Getting error on this line 

    }

[xstefank]

@loicmathieu They have ObjectId in Mongo, and they want to pull ID as a String. So the question is if we should be doing ObjectId -> String mapping in the background. I personally also don't think that this is aligned with relational mapping in hibernate and it might be confusing if it differs for mongo. But this is your decision.

[xstefank]

@loicmathieu @nitinty pinged me again about this today. So I gave it another try and I kind of got it working - https://github.com/xstefank/quarkus-reproducers/tree/main/mongo-id. So if @nitinty is OK with this solution, we can close this.

[nitinty]

@xstefank
Thank you for your prompt response, as always.

The solution you shared might work, but it's not considered good practice to create multiple entity classes pointing to the same collection.

[loicmathieu]

Well, people may respond that having two different type of data in a single field is not a best practice neither ;) most databases prevent that (even Mongo can if you define a schema).

As I said, if you want to map a file that can have multiple type in database, you should create your own MongoDB codec, it's not too complex and will be automatically detected by Quarkus.

MongoDB with Panache is an opinionated extension for simplifying MongoDB usage, inspired by JPA (mirror of Hibernate with Panache). Our opinionated approach is that a database field should have a consistent type across all database records and mapped to a single type in the Entity.

So either the solution form @xstefank or a custom MongoDB codec should be used to work around this issue. I didn't plan to implement any ad-hoc solution for it directly inside MongoDB with Panache.

[xstefank]

And here is the version with the custom codec if anyone ever runs into this issue again - https://github.com/xstefank/quarkus-reproducers/tree/main/mongo-id-reproducer.

[loicmathieu]

Nice, didn't saw it.
Thanks for sharing.