We present a completely declarative approach to synthesizing pattern matching construct implementation based on application of relational programming, a specific form of constraint logic programming. Our approach is based on relational representations of both the high-level semantics of pattern matching and the semantics of an intermediate-level implementation language. This choice makes our approach, in principle, very scalable as we only need to modify the high-level semantics in order to synthesize the implementation of a pattern matching new feature. Our evaluation on a set of small samples, partially taken from existing literature shows, that our framework is capable of synthesizing optimal implementations quickly. Our in-depth stress evaluation on a number of artificial benchmarks, however, has shown the need for future improvements.