PersistentList[] of `EmbeddedPersistentObject` yields a new instance

[Goutte]

Querying a PersistentList made of EmbeddedPersistentObject always yields a new instance. This is confusing. Is this wanted ?
How can I get around this ? Should I wrap the players getter with memoization/caching logic ? Shouldn't Objectory take care of this ?

Here's a stripped-down test case :

/// A player is per-game, unlike the user, so I embed it.
class Player extends EmbeddedPersistentObject {
  Player ();
}

class Game extends PersistentObject {

  // There may be more than two players
  List<Player> get players => getPersistentList(Player, 'players');

  // ... but the first two are black and white.
  Player get black => players[0];
  Player get white => players[1];

  Game(){
    players.add(new Player());
    players.add(new Player());
  }

}

g = new Game();
expect(g.black, same(g.black)); // <-- FAILS

I'm probably missing something here (I'm very much new to Objectory).

[Goutte]

I hacked around PersistentList :

class PersistentList<E> extends ListBase<E>{

  // ... properties

  Map _cache = new Map();       // #hack

  // ... entrails of the class

  // #hack
  void operator[]=(int index, E value){
    _list[index] = internValue(value);
    _cache[_list[index]] = value;
    setDirty(null);
  }

  // #hack
  E operator[](int index) {
    if (!_cache.containsKey(_list[index])) {
      _cache[_list[index]] = valueConverter.convertValue(_list[index]);
    }
    return _cache[_list[index]];
  }

  // ...

}

I'm unsure of the consequences of this, but it solves the expectation of my previous comment ; what do you think ?

[vadimtsushko]

Hi Antoine.

IIRC there was some hidding problems with memoization of PersistentList. On the other hand current scheme with constantly returning new PersisentList did not work well with Polymer events for example.
So it should be changed. I'll try to look at it at again as soon as will have time.

[vadimtsushko]

Ooops, I've misread your issue. I believe I did not think about that scenario yet. Maybe identity for embedded persistent objects should depend on identity of their data map.

[vadimtsushko]

Just published new version. I believe it shoud solve the problem for your scenario

[Goutte]

I don't even have to take your word for it, as you pushed a test-case ! I'll test and merge later on (i polluted my fork some more with preSave and postFind hooks useful respectively for user-controlled serialization and hydratation.).

Thanks !