FeatX: Editing Software by Editing Features for Repository-Level Code Evolution

2026-06-30

AI Summary

Current large language models for software evolution are code-centric and rely heavily on manual context management and iterative prompt tuning, imposing significant cognitive load and limiting efficiency. This work proposes a feature-oriented software editing approach that structures development tasks into an epic-feature hierarchy and establishes explicit feature-to-code mappings. Coupled with a three-stage evolutionary agent that automatically generates code patches and a four-panel interactive interface, the method introduces the first interactive paradigm for repository-scale code evolution driven at the feature level. Evaluated on 38 real-world feature editing tasks, it improves the F1 score for functional-level change localization by 42.6% over strong baselines, with a per-run cost of merely $0.07. User studies further demonstrate its significantly higher usability compared to ChatGPT.

Abstract

Large language models (LLMs) are increasingly used for software evolution, yet most interaction paradigms remain code-centric and require manual context management and prompt iteration. We present FeatX, a feature-oriented tool for editing software by editing features. Given an existing repository, FeatX extracts a hierarchical epic-feature structure with explicit feature-to-code mappings, then invokes a three-stage Evolution Agent to translate feature edits into code patches. The workflow is exposed through four coordinated panels. Across a controlled user study and replay experiments on 38 real-world feature-editing commits, FeatX significantly reduces cognitive load and improves usability compared with vanilla ChatGPT. It also achieves a 42.6\% relative improvement in function-level modification localization F1 over strong LLM baselines, at substantially lower cost (\$0.07 in total). The tool and collected dataset are available at https://github.com/a496263365/FeatX/tree/demo, with a demonstration video at https://youtu.be/OZqKZ4Ii-yM.

Problem

• software evolution
• code-centric interaction
• context management
• prompt iteration
• cognitive load

Innovation

• feature-oriented editing
• repository-level code evolution
• LLM-based software evolution
• feature-to-code mapping
• Evolution Agent