Serilization of Core Data objects inorder to Post/Put to server

[SAJK111]

As the title says... Does this library support serilization of Core Data objects to JSON inorder to Post/Put to a server, like RestKit?

[cnstoll]

Hi James,

Good question. Presently, it does not, but it will in version 2.0.0, which is being worked on now.

Serialization is actually a somewhat complex problem, since its hard to know exactly how one may want an object to be serialized. The solution now is to go with using the keyPaths that get used to populate the model originally.

You can check out the 2.0.0 draft of the serializer here:

https://github.com/mutualmobile/MMRecord/blob/2.0.0/Source/MMRecord/MMRecordSerializer.h

One major question is how to deal with relationships. Should the relationships be serialized at all? Should they only be serialized one level deep, or many levels deep? Its kind of a tough problem to solve. I'd love to hear what your expectations are.

Thanks,

• Conrad

[SAJK111]

Hi Conrad
Sorry for my late response. If all levels could be serialized it would be perfect, there is just one bottleneck with recursive relationship.

[ParmeshwarBayappu]

My thoughts on serialization of relationships:
Serialization of relationships should depend on dependence of the destination entity on the source entity.

• When the destination entity in the relationship is owned by the source entity the whole destination entity should be included. For such an entity delete rule for the relationship will/should be 'Cascade'. Ex: Shape entity has a relation (ordered to many) to the Point entity.
• When the destination entity in the relationship is not owned by the source entity (i.e. destination object can exist even when the source entity is deleted) then only the identification information of the destination object should be included. For such an entity the delete rule for the relationship will/should be ’Nullify’. Ex: Employee entity has a relation (to one) to the Department entity.

To provide option to override that default behavior we could allow specifying this explicitly in the user info (say Serialization_IncludeWholeEntity).

There may be a few occasions where the user may want to specify explicitly which entities to include wholly in the generated JSON. This could be provided as a set of entity names/classes - and serve to override the default behavior.

For those that really want all objects in relationships to be serialized wholly we could have flag to provide this override behavior.

During the generation process we should keep a dictionary of the objects that are already being serialized - to prevent endless recursion. When we encounter in a relationship an object that has already been serialized, then only the identification information of the destination object should be included.

BTW, any plans for active development on 2.0.

[cnstoll]

Hola,

There aren't any current development plans for 2.0. I wrote a little more on that topic here:

#120

I'll leave this discussion open though as a reference to 2.0.0 in case development on it ever resumes.

Thanks,

• Conrad