Condensed representations of patterns offer an elegant way to represent solution sets compactly, while minimizing the redundancy and the number of patterns. This approach has been mainly developed in the context of the frequency measure and there are very few works addressing other measures. We propose a generic framework based on constraint programming to efficiently mine adequate condensed representations of patterns w.r.t. a set of measures. For this, we introduce a new global constraint with a complete polynomial filtering. We show how this constraint can be exploited in association with Pareto dominance constraints to mine skypatterns and conceptual clustering. Experiments performed on standard datasets show the efficiency of our approach and its significant advantages over existing approaches.