Mandango implements the Unit Of Work pattern, which simply means "send all database operation at once".
Let's see an example:
// prepare database operations $author = $mandango->create('Model\Author')->setName('pablodip'); $mandango->persist($author); $category1 = $mandango->create('Model\Category')->setName('MongoDB'); $category2 = $mandango->create('Model\Category')->setName('PHP'); $mandango->persist(array($category1, $category2)); $article = $articleRepository->createQuery(array('name' => 'Mondongo'))->one(); $article ->setName('Mandango') ->setAuthor($author) ->addCategories(array($category1, $category2)) ; $mandango->persist($article); $articles = $articleRepository->createQuery(array('type' => 'foo'))->all(); $mandango->remove($articles); // send all database operations at once $mandango->flush();
Like you can see this is a really powerful way to work, because you can prepare all the database operations, and when all of them are ready, you just have to send them to the database.
MongoDB does not have transactions, so this is probably the best way you can work if you want avoid as much as possible database inconsistency because of application breaks.
And this is also the best way to get the maximum database performance, because Mandango does that the database operations are as efficient as possible.