This paper presents an on-going work that includes semantic analysis of SQL statements in the database system DES. Following Brass&Goldberg, SQL statements, while syntactically correct, can exhibit symptoms of bad design. By warning users about such symptoms, both the learning curve of students as well as productivity of SQL practitioners can be enhanced. Some errors include inconsistent conditions, lack of correlations in joins, unused tuple variables and the like. Here, we describe the semantic checker developed for DES, which applies several techniques. In particular, it applies abstraction with CLP solving, and specific algorithms for the different kind of errors. This proposal has been implemented in the deductive system DES and available at des.sourceforge.net.