The framework of database repairs and consistent answers to queries is a principled approach to managing inconsistent databases. We describe the first system able to compute the consistent answers of general aggregation queries with the COUNT ($A$), COUNT (*), and SUM operators, and with or without grouping constructs. Our system uses reductions to optimization versions of Boolean satisfiability (SAT) and then leverages powerful SAT solvers. We carry out an extensive set of experiments on both synthetic and real-world data that demonstrate the usefulness and scalability of this approach.

Consistent Query Answering (CQA) is a rigorous and principled approach to answering queries posed against inconsistent databases. Computing consistent answers to a Select-Project-Join (SPJ) query or an SPJ query with aggregation operators on a given inconsistent database can be an intractable problem. We demonstrate CAvSAT, a system for CQA that leverages a set of natural reductions from a given CQA instance to boolean Satisfiability (SAT) and its optimization variants. CAvSAT is the first system that is capable of handling unions of SPJ queries with aggregation operators SUM and COUNT, and databases that are inconsistent w.r.t. key constraints, functional dependencies, and denial constraints.

Managing inconsistencies in databases is an old, but recurring, problem. An inconsistent database is a database that violates one or more integrity constraints. In the real-world, inconsistent databases arise in several different contexts, including data warehousing and information integration. The framework of database repairs and consistent query answering (CQA) is a principled way of handling inconsistencies. In this work, we propose a novel approach that has a potential to build a comprehensive and scalable CQA system. We report preliminary experimental results on a prototype CQA system CAvSAT (Consistent Answering via Satisfiability), implemented using this approach.