Effective Iterative Program Synthesis with Knowledge Searched from Internet

Abstract

This paper presents the ongoing work of studying the iterative program synthesis based on knowledge searched from the Internet, which can fairly reduce the scale of program space and improve the efficiency of synthesis. First, we implement a tool named Args(api Recommendation via General Search) to obtain the API knowledge from the Internet. Second, we propose an iterative method that incrementally constructs the program space to quickly approach the target program. The initial experimental result shows the effectiveness of our work.