new models with default values cannot be tracked properly

[feanor07]

As far as I see; new models always return true for didChange attr/relation after the following commit fe3a1a3#diff-d1f63f06d3656a6e2da38e3a5fee9865R406

I need to start tracking of a newly created model after setting some default attributes on it; but calling model.startTrack() (I am not using auto mode) has no effect due to the above commit I mentioned. I cannot find anyways to start tracking of a new model after setting some initial values on it; so that modelChanges() do not return the properties with those initial values if they are unchanged. Is there anyway I can achieve what I want? I am open to any suggestion. thx.

[danielspaniel]

I had to do what you are wanting to do a long time ago, and I think what I did was manually track the diff. For example if you create a model and set some defaults, you could then serialize the model ( or toJSON ) to save the current state of the new model. Then later you would compare the state to that original state you saved.
It is not too horrible but it does not use change tracker.

New models are tricky in ember-data because of how ember-data internally tracks changes.

[ianpetzer]

I have run into the same requirement of being able to track changes after a model has been created and set up with default values, so I created a variation of the isDirty, hasDirtyRelations and hasDirtyAttributes computed properties which track any changes made to a model after an arbitrary point in time by adding hasNewChanges, hasDirtyRelations and hasNewDirtyAttributes computed properties.

This is my WIP. I call trackDirtinessFromNow once my model is ready to be tracked.
It appears you can't track new dirty relations.. Dirty relationships can only be created from object creation

  export function trackDirtinessFromNow(model) {
    const relations = [];
    const relationsObserver = [];
    const attrs = [];
    const existingChanges = model.modelChanges();

    model.eachRelationship((name, descriptor) => {
      if (descriptor.kind === 'hasMany') {
        relations.push(descriptor.key);
        if (descriptor.options.async) {
          relationsObserver.push(descriptor.key + '.content.@each.id');
        } else {
          relationsObserver.push(descriptor.key + '.@each.id');
        }
      } else {
        relations.push(descriptor.key);
        relationsObserver.push(descriptor.key + '.content');
      }
    });

    model.eachAttribute(name => {
      return attrs.push(name);
    });

    const hasDirtyRelations = function() {
      const changed = model.modelChanges();
      return !!attrs.find((key) => {
        // attrribute is now changed but wasnt in existing set, or was already changed in existing set but now has different value
        return (changed[key] && !existingChanges[key]) || (changed[key] && existingChanges[key] && existingChanges[key][1]!==changed[key][1])
      });
    };

    const hasNewDirtyAttributes = function() {
      const changed = model.modelChanges();
      return !!attrs.find((key) => {
        // attrribute is now changed but wasnt in existing set, or was already changed in existing set but now has different value
        return (changed[key] && !existingChanges[key]) || (changed[key] && existingChanges[key] && existingChanges[key][1]!==changed[key][1])
      });
    };

    const hasNewChanges = function() {
      return model.get('hasNewDirtyAttributes') || model.get('hasDirtyRelations');
    };

    Ember.defineProperty(
      model,
      'hasNewDirtyAttributes',
      Ember.computed.apply(Ember, attrs.concat([hasNewDirtyAttributes]))
    );

    Ember.defineProperty(
      model,
      'hasDirtyRelations',
      Ember.computed.apply(Ember, relationsObserver.concat([hasDirtyRelations]))
    );

    Ember.defineProperty(
      model,
      'hasNewChanges',
      Ember.computed.apply(Ember, ['hasNewDirtyAttributes', 'hasDirtyRelations', hasNewChanges])
    );

  }

[danielspaniel]

looks like extreme hacking to me .. it now adds a few new "hasNew" to the already dirty properties, which would be confusing the heck out of people. I would say , maybe add this to your branch and use it , or think of another way ? maybe ?