PyART: Python API Recommendation in Real-Time

Abstract

This is the research artifact of the paper titled 'PyART: Python API Recommendation in Real-Time'. PyART is a real-time API recommendation tool for Python, which includes two main functions: data-flow analysis and real-time API recommendation for both incomplete and complete Python code context. Compared to classical tools, PyART has two important particularities: it is able to work on real-time recommendation scenario, and it provides data-flow analysis and API recommendation for dynamic language. Classical tools often fail to make static analysis in real-time recommendation scenario, due to the incompletion of syntax. And the dynamic features of Python language also bring challenges to type inference and API recommendation. Different from classical tools, PyART derives optimistic data-flow that is neither sound nor complete but sufficient for API recommendation and cost-effective to collect, and provides real-time API recommendations based on novel candidate collection, context analysis and feature learning techniques. The artifact evaluation experiments of PyART include three main aspects: data-flow analysis, intra-project API recommendation and across-project API recommendation. We assume users of the artifact is able to use Linux Ubuntu Operating System.