aggregate_sum and EmbeddedDocumentListField

I had to manually do an “$unwind” to obtain a result, otherwise the result always returned a “0".

How can I develop an “aggregate sum” over a document with a “EmbeddedDocumentListField” or a “List” in order to obtain that result?. Is “$unwind” the correct / only way to do it?

class WordUsed(mongoengine.EmbeddedDocument):
    word_id = mongoengine.IntField()
    name = mongoengine.StringField()
    hits = mongoengine.IntField()
    errors = mongoengine.IntField()


class LogGames(mongoengine.EmbeddedDocument):
    game_id = mongoengine.IntField()
    used_words = mongoengine.EmbeddedDocumentListField(WordUsed)
    used_phrases = mongoengine.EmbeddedDocumentListField(PhraseUsed)


class LogSession(mongoengine.Document):
    session_date = mongoengine.DateTimeField()
    log_games = mongoengine.EmbeddedDocumentListField(LogGames)
    used_words = mongoengine.EmbeddedDocumentListField(WordUsed)
    used_phrases = mongoengine.EmbeddedDocumentListField(PhraseUsed)

Original code:

>>> sessions = LogSession.objects.all()

>>> sessions.sum("used_words.hits")
0
>>> sessions.aggregate_sum("used_words.hits")
0

My manual code:

>>> result = LogSession.objects.aggregate({"$unwind": "$used_words"}, {'$group': {'_id': 'total', 'total': {'$sum': '$used_words.hits'}}})
{'_id': 'total', 'total': 45.0}

No repository file or test is named. Reproduce the provided LogSession example and compare aggregate_sum("used_words.hits") with the manual aggregation using $unwind and $group. Done means establishing whether the zero result is a bug or supported limitation, with a regression test or clarified documented behavior.