panproto-gat

The math engine that defines what a valid schema language looks like.

What it does

Every schema language (JSON Schema, OpenAPI, Protobuf, SQL DDL) has its own rules: which field types are allowed, how nesting works, what counts as a required vs. optional relationship. This crate lets you describe those rules formally as a "theory": a named set of sorts (the categories of thing that exist), operations (how those things relate), and equations (rules every instance must satisfy). Think of a theory as a type system for your schema language itself.

Once you have theories for two schema languages, this crate can compute their colimit, which is the smallest theory that contains both. That is the foundation for cross-protocol translation: if you can describe OpenAPI and JSON Schema as theories, combining them tells you exactly which concepts they share and which are unique to each. Morphisms (structure-preserving maps between theories) then let you carry data, schemas, and migrations across that boundary without losing information.

The crate also provides endofunctor helpers for decomposing a morphism into a sequence of named, reusable steps. Those steps become the building blocks of protolenses: bidirectional schema transforms that can be composed, inverted, and version-controlled.

Quick example

use panproto_gat::{Theory, Sort, Operation, colimit};

let mut graph = Theory::new("SimpleGraph");
graph.add_sort(Sort::new("V"));
graph.add_sort(Sort::new("E"));
graph.add_op(Operation::new("src", vec!["E"], "V"));
graph.add_op(Operation::new("tgt", vec!["E"], "V"));

// Combine with a second theory; shared sorts are merged automatically.
let result = colimit(&graph, &other_theory, &morphism)?;

API overview

License

MIT