One challenging aspects of data quality modeling and management is to provide flexible, declarative and appropriate ways to express requirements on the quality of data. The paper presents a framework for specifying and checking constraints on data quality in RDBMS. The evaluation of the quality of data (QoD) is based on the declaration of data quality metrics that are computed and combined into so-called QoD analytic workflows. These workflows are designed as a composition of statistical methods and data mining techniques used to detect patterns of anomalies in the data sets. As metadata they are used to characterize various quantifiable dimensions of data quality (e.g., completeness, freshness, consistency, accuracy). The paper proposes a query language extension for constraining data quality when querying both data and its associated QoD metadata. Probabilistic approximate constraints are checked to determine if the quality of data is (or not) acceptable to build quality-constrained query results.