Abstract
$\mathcal{DLR}$ is an expressive DL with n-ary relations, particularly suited for modeling database schemas and queries. Although $\mathcal{DLR}$ has constituted one of the crucial steps for applying DL technology to data management, there is one important aspect of database schemas that DLs do not capture yet, namely the notion of key. In this paper we give keys for free to those who like $\mathcal{DLR}$. In particular, the question addressed in this paper is as follows: can we add keys to $\mathcal{DLR}$ and still have EXPTIME associated reasoning techniques? Somewhat surprisingly, we answer positively to the question, by adapting the $\mathcal{DLR}$ reasoning algorithm in such a way that reasoning on a $\mathcal{DLR}$ schema with keys can be done with the same worst-case computational complexity as for the case without keys.