Inductive Synthesis of Structurally Recursive Functional Programs from Non-recursive Expressions

We present a novel approach to synthesizing recursive functional programs from input-output examples. Synthesizing function is challenging because subexpressions should be constructed while the target has not been fully defined yet. address this challenge by using new technique we call block-based pruning. A block refers recursion- and conditional-free expression (i.e., straight-line code) that yields an output particular input. first synthesize as many blocks possible for each example, then explore space of programs, pruning candidates are inconsistent with blocks. Our method based on efficient version learning, thereby effectively dealing possibly enormous number In addition, uses sampled behaviors library functions enable goal-directed search program library. have implemented our in system called Trio evaluated it synthesis tasks prior work tasks. experiments show outperforms solution 98% benchmarks benchmark suite.