AI for Automated Code Updates

Abstract

Most modern code bases extensively rely on external libraries to provide robust functionality out of the box. When these libraries are updated they can sometimes introduce breaking changes in the process, which require extensive developer maintenance. To mitigate this we propose to use artificial intelligence to parse the text of release notes to capture code deprecations in structured form. This, in turn, enables us to develop an IDE plugin that can automatically detect deprecated library usages in live code bases and even suggest recommended fixes. We evaluated our system on over 30 internal projects within J.P. Morgan.