An enterprise information worker is often aware of a few example tuples (but not the entire result) that should be present in the output of the query. We study the problem of discovering the minimal project join query that contains the given example tuples in its output. Efficient discovery of such queries is challenging. We propose novel algorithms to solve this problem. Our experiments on real-life datasets show that the proposed solution is significantly more efficient compared with naive adaptations of known techniques.