Recommender systems solve the problem of information overload by efficiently utilizing huge quantities of data and trying its best to predict potential preference which aims at a certain user. They are widely applied in numerous fields. However, hardly can we see a code recommender system for programmers though it is desperately expected. Raw data of the code are not so convenient to handle for the difference in structure and lack of relevance. Fortunately, in real world, there are abundant data affiliated to the code, such as context, tags, social relations of users and view histories. In this paper, we firstly formulate a new task of code recommendation. Then, we propose a hybrid linear algorithm for recommending source codes, in which we maximize the utility of multivariate heterogeneous auxiliary data with code. Experiments on the dataset from Code Review Community show that our proposed method works for the new code recommendation task. Our system is hopefully designed to be adaptive to new source of heterogeneous information, and hopefully performs better with more significant data and new inspired components.
