Program Sketching Using Lifted Analysis for Numerical Program Families

This work presents a novel approach for synthesizing numerical program sketches using lifted (family-based) static analysis. In particular, our leverages analysis based on abstract interpretation, which is used analyzing families with features. It takes as input the common code base, encodes all variants of family, and produces precise results in single run. The elements underlying domain are decision trees, nodes labeled linear constraints defined over features leaf belong to given single-program domain.We encode sketch family such that holes correspond possible realizations family. Then, we preform so only those satisfy assertions under inputs represent correct sketch.We have implemented an experimental synthesizer resolving C sketches. analyzer \(\texttt {\#if}\)-based families, uses domains from APRON library. An evaluation yields promising results. Moreover, provides speedups some cases against popular sketching tool Sketch can solve benchmarks cannot handle.