Nitrite/Java: Object must implement Mappable

[neroux]

I just managed to get my project to compile with Nitrite, however I immediately came across the following exception

org.dizitart.no2.exceptions.ObjectMappingException: object must implements Mappable

Does that mean every object that is supposed to be stored via an ObjectRepository needs to manage its serialisation manually? The Mappable interface would actually suggest that

https://github.com/nitrite/nitrite-java/blob/8298a0d65d539264660b8c64f0fda915ae5d9e0f/nitrite/src/main/java/org/dizitart/no2/mapper/Mappable.java#L22-L23

However, considering that I seem to be able to save a document with an embedded custom class without issues, I would assume that should not be the case and Nitrate still handles serialisation on its own, right? What would be the issue then here? Did I miss anything?

Thanks.

For the sake of good order, here the trace

org.dizitart.no2.exceptions.ObjectMappingException: object must implements Mappable
    at org.dizitart.no2.mapper.MappableMapper.convertToDocument(MappableMapper.java:82)
    at org.dizitart.no2.mapper.MappableMapper.convert(MappableMapper.java:51)
    at org.dizitart.no2.repository.RepositoryOperations.toDocument(RepositoryOperations.java:103)
    at org.dizitart.no2.repository.DefaultObjectRepository.update(DefaultObjectRepository.java:106)
    at org.dizitart.no2.repository.DefaultObjectRepository.update(DefaultObjectRepository.java:100)

[anidotnet]

Not every object should implement Mappable, if you use jackson module to serialize the object.

// initialization using builder
Nitrite db = Nitrite.builder()
        .loadModule(storeModule)
        .loadModule(new JacksonMapperModule())  // this enables jackson serialization of object without using Mappable
        .openOrCreate("user", "password");

You need to include nitrite-jackson-mapper in your dependency for this.

If you don't want to include jackson dependency in your project, you should implement Mappable. And also, mappable implementation is faster than jackson serialization for obvious reason.

[neroux]

All right, so do I understand correctly that v4 makes a lot of things optional which came out of the box in v3?

I just tried it by adding Jackson and it seemed to do the trick, though, the odd thing is, aforementioned

able to save a document with an embedded custom class without issues

still works as long as the class implements Serializable. Why would Nitrate use different serialisations for collections and repositories?

[anidotnet]

Repository is just a typed collection, underlying data structure of the repository is just a collection of documents. You need a mechanism to convert a POJO to a document and that's where Mappable or jackson serialization comes in.

Once your data is in a document inside a collection, its up to the storage engine how it stores the document into a file, for example - mvstore storage uses java serialization, rocksdb storage uses kryo to convert objects into byte array. I hope this answers your query.

[neroux]

Fair enough, makes sense. Thanks.