Understanding precision@k for train and test in recommendation systems

Question

I am currently building a recommender system and I am trying to evaluate it. However, many sources have differing methods of computing precision@k and recall@k. Let me give an e