Support upserts

[nielsenko]

Allows to upsert object graphs with Realm.add(x, update:true) (and Realm.addAll([x,y], update:true))

Resolves: #648

[blagoev]

There is some work in Core about this realm/realm-core#5516

[nielsenko]

There is some work in Core about this realm/realm-core#5516

This is still in draft. I suggest we update our code once it lands.

[nirinchev]

The code seems fine to me, but I think we could improve the test coverage of that feature.

Off the top of my head, some tests that are missing:

1. A test that upserting an object with a collection different from the original collection is clearing the old collection before populating it with the new items.
2. A test that upserting an object with an empty collection clears the old collection.
3. Tests that mix and match managed and unmanaged objects.
4. A test that upserting an object with a collection where the items of the collection have changes, those changes get applied to the original objects.

[nielsenko]

The code seems fine to me, but I think we could improve the test coverage of that feature.

Off the top of my head, some tests that are missing:

1. A test that upserting an object with a collection different from the original collection is clearing the old collection before populating it with the new items.

is tested in realm_test.dart#715

2. A test that upserting an object with an empty collection clears the old collection.

isn't that implied by 1) , ie. [bob, carol] updated to [dan]? But yes, I can add a test for that specifically.

3. Tests that mix and match managed and unmanaged objects.

Doesn't the realm_test.dart#659-668 cover this? The Friends are updated through a Party (that has no PK).

4. A test that upserting an object with a collection where the items of the collection have changes, those changes get applied to the original objects.

Isn't this tested in realm_test.dart#664-665?

[nirinchev]

You're right - I misread some of what was going on with the party test.

Re:
2. It is implied, but I think an empty collection is somewhat special, so it deserves an explicit test.
3. Yes, that covers some of it, but I think it's a bit simple. Looking at the code, I don't think there are bugs, but it's a good idea to guard against refactoring by devising a more complex hierarchy being upserted. Something like:

@RealmModel()
class _ObjWithPK {
  @PrimaryKey()
  late int id;

  late _ObjWithPK linkWithPK;
  late List<_ObjWithPK> listWithPK;

  late _ObjWithoutPK linkWithoutPK;
  late List<_ObjWithoutPK> listWithoutPK;
}

@RealmModel()
class _ObjWithoutPK {
  late _ObjWithPK linkWithPK;
  late List<_ObjWithPK> listWithPK;

  late _ObjWithoutPK linkWithoutPK;
  late List<_ObjWithoutPK> listWithoutPK;
}

And test that a more complex graph with cycles is getting correctly inserted. Again, I don't think there's a bug in the code, but I think I can come up with buggy implementations that pass the Party test.

[nielsenko]

I have gathered all test in one test case about friends and parties, with an easy to follow narrative.