We present a principled theoretical framework for inferring and checking the union types, and show its work in practice on JSON data structures. The framework poses a union type inference as a learning problem from multiple examples. The categorical framework is generic and easily extensible.

中文翻译：

---

走向更完美的工会类型

我们提出了一个用于推断和检查联合类型的原则性理论框架，并展示了它在 JSON 数据结构上的实践工作。该框架将联合类型推理作为从多个示例中的学习问题提出。分类框架是通用的且易于扩展。