panproto-schema

Protocol-indexed schema graphs.

Installation

Add the crate to a Rust 1.85 or newer project:

cargo add panproto-schema

Usage

use panproto_schema::{Protocol, SchemaBuilder};

let protocol = Protocol::default();
let schema = SchemaBuilder::new(&protocol)
    .vertex("user", "object", None)?
    .vertex("name", "string", None)?
    .edge("user", "name", "prop", Some("name"))?
    .entry("user")
    .build()?;

The default protocol is open because it has no vertex kinds or edge rules. Applications that need format-specific validation should use a protocol constructor from panproto-protocols.

Data model

Schema stores vertices, binary edges, hyper-edges, constraints, required-edge declarations, namespace identifiers, entries, variants, orderings, recursion points, usage modes, nominal-identity metadata, spans, and expression-backed enrichments. It also stores adjacency indices for outgoing, incoming, and between-vertex lookup.

Protocol names the schema and instance theories and specifies edge rules, object kinds, constraint sorts, composition metadata, and feature flags. It is configuration, not a global built-in lookup. Built-in protocol constructors live in panproto-protocols.

SchemaBuilder::vertex checks duplicate IDs and known vertex kinds. edge checks endpoints and configured edge rules. Constraints added through the builder are checked only when the caller runs validate. build rejects an empty schema and unknown entry vertices, then constructs the indices.

Pushout construction

schema_pushout(left, right, overlap) closes the declared vertex and edge pairs into an equivalence relation, builds the quotient schema, and returns morphisms from both inputs. It validates references in SchemaOverlap. The return value contains the implemented quotient and morphisms, not a proof of the universal property.

API reference

Span here is a record stored in a schema. It is distinct from the schema-span search result in panproto-mig.

License

MIT