panproto-check

Detects breaking changes between two versions of a schema.

What it does

When you change a schema that other code depends on (an API response type, a database table, a message format), some changes are safe and some are not. Adding an optional field is safe; removing a required field is not. Relaxing a maxLength constraint is safe; tightening it is not. The rules for what counts as breaking depend on the schema language: JSON Schema has different compatibility rules than Protobuf or SQL.

This crate computes a structural diff between two schema versions across 25+ change categories (vertices, edges, constraints, hyper-edges, variants, orderings, recursion points, usage modes, and more), then classifies each change as breaking or non-breaking using the rules of the specific protocol that governs those schemas. The result is a CompatReport you can render as human-readable text for a pull request comment or as JSON for a CI gate.

The classifier is used by the schema check CLI command and by the GitHub Actions workflow that panproto ships for breaking change detection.

Quick example

use panproto_check::{diff, classify, report_text};

let schema_diff = diff(&old_schema, &new_schema);
let report = classify(&schema_diff, &protocol);

if !report.breaking.is_empty() {
    eprintln!("{}", report_text(&report));
    std::process::exit(1);
}

API overview

License

MIT