Principles and Practices of Large-Scale Code Analysis at Ant Group: A Data- and Logic-Oriented Approach

2026-06-17 · Source: cs.SE updates on arXiv.org · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Expert, extended

Summary

Ant Group's CodeFuse-Query is a static code analysis system designed for large-scale software development, capable of scanning over 10 billion lines of code daily and supporting more than 300 distinct tasks across 9 programming languages. It integrates Domain Optimized System Design, which includes resource optimization, data reusability, and incremental code extraction, with Logic Oriented Computation Design. The latter leverages Datalog and a two-tiered COREF schema to transform source code into data facts, enabling complex analysis tasks through the Gödel language. CodeFuse-Query demonstrates significant robustness, scalability, and efficiency, addressing challenges in large organizations with over ten thousand developers. The project is open-sourced, fostering further innovation in the field.

Key takeaway

For engineering leads managing vast, multi-language codebases, CodeFuse-Query offers a robust solution to overcome traditional static analysis limitations. You should consider adopting its data-centric, Datalog-based approach for scalable, efficient analysis, especially for tasks like change impact assessment or LLM training data preparation. Its incremental extraction and custom query language, Gödel, can significantly enhance productivity and maintainability. Explore its open-source implementation to tailor complex analysis needs.

Key insights

CodeFuse-Query redefines static code analysis as a data computation task for large-scale efficiency.

Principles

• Integrate domain-specific features into system design.
• Optimize data reuse across the entire processing chain.
• Anticipate and handle failures through redundancy.

Method

Formulate tasks in Gödel, generate optimized Datalog execution plans, then compile and execute against extracted code facts for analysis results.

In practice

• Perform rapid change impact analysis.
• Prepare and refine LLM training data.
• Generate R&D productivity metrics.

Topics

• Static Code Analysis
• Datalog
• CodeFuse-Query
• Gödel Language
• Large-Scale Software Development
• LLM Data Preparation

Code references

• codefuse-ai/CodeFuse-Query
• StyleCop/StyleCop
• python-poetry/poetry
• pytest-dev/pytest
• robinhood/faust

Best for: CTO, VP of Engineering/Data, AI Scientist, Software Engineer, Machine Learning Engineer, AI Architect

Related on AIssential

• WildCode Revisited: A Comprehensive Empirical Study on the Security of LLM-Generated Code — LLM-generated code, particularly from ChatGPT, consistently contains significant security vulnerabilities, largely unaddressed by users.
• A new native IDE approach to prevent code leakage to LLMs: Obfuscating ASTs before the API call (Verantyx) — Obfuscating code with Kanji-infused structural semantics allows LLMs to reason without exposing proprietary data.
• Data-Centric Benchmarking of Exploit Generation in LLMs: Understanding the Impact of Fine-Tuning — Data quality and structured evaluation are paramount for effective LLM-based exploit generation.
• LogCopilot: Automating Log Aggregation Analysis through Large Language Models — LLM-driven LogCopilot automates log analysis by translating natural language into LogQL queries via a hierarchical knowledge base.
• QCFuse: Query-Aware Cache Fusion via Compressed View for Efficient RAG Serving — QCFuse enables efficient, quality-preserving RAG serving by integrating query-aware cache fusion with compressed context views.
• [AINews] GLM-5.2: the top Frontend Coding model in the world, IndexShare for Speculative Decoding — Open-weight models can achieve frontier-level coding and agentic performance with efficient long-context handling.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.SE updates on arXiv.org.