crdt-graph 0.3.1

An op-based 2P2P-Graph CRDT implementation in Rust.
Documentation

crdt-graph

Crates.io Docs.rs License GitHub

An op-based 2P2P-Graph CRDT implementation in Rust.

Based on the 2P2P-Graph specification from Shapiro et al., "A comprehensive study of Convergent and Commutative Replicated Data Types" (2011).

Overview

A 2P2P-Graph (Two-Phase Two-Phase Graph) is a conflict-free replicated data type that models a directed graph with add and remove operations for both vertices and edges. Each replica maintains four sets:

Set Description
V_A Vertices added
V_R Vertices removed
E_A Edges added
E_R Edges removed

Updates follow the op-based CRDT model with two phases:

  • atSource — Precondition checks executed only on the originating replica.
  • downstream — State mutations applied on all replicas (including the source).

Features

  • Three built-in graph variants — ID-only (simple), binary payload (bytes), and string payload (string).
  • FlatBuffers serialization — Compact binary encoding with 16-byte inline UUID structs (zero-copy).
  • UUID v7 identifiers — Time-ordered, globally unique IDs via uuid::Uuid.
  • petgraph integration — Convert CRDT state to a standard DiGraph for graph algorithms.
  • Convenience traitsDefault, PartialEq, Eq, Hash, From conversions on all operation types.
  • State inspectionvertex_count(), edge_count(), is_empty(), vertices(), edges() iterators.

Quick Start

Using the built-in simple types (UUID-based, no payload):

use crdt_graph::types::simple::{self, AddVertex, AddEdge, Graph};
use crdt_graph::types::{RemoveEdge, RemoveVertex};
use crdt_graph::Uuid;

# fn main() {
let mut replica_a = Graph::new();
let mut replica_b = Graph::new();

let v1 = Uuid::now_v7();
let v2 = Uuid::now_v7();

// Replica A: add vertices
let op1 = replica_a.prepare(AddVertex { id: v1 }.into()).unwrap();
let op2 = replica_a.prepare(AddVertex { id: v2 }.into()).unwrap();

// Broadcast to Replica B
replica_b.apply_downstream(op1).unwrap();
replica_b.apply_downstream(op2).unwrap();

// Replica B: add an edge
let e1 = Uuid::now_v7();
let op3 = replica_b.prepare(AddEdge { id: e1, source: v1, target: v2 }.into()).unwrap();
replica_a.apply_downstream(op3).unwrap();

// Both replicas have converged
assert_eq!(replica_a.vertex_count(), 2);
assert_eq!(replica_a.edge_count(), 1);
assert_eq!(replica_b.vertex_count(), 2);
assert_eq!(replica_b.edge_count(), 1);
# }

Custom Types

You can also define your own types by implementing the required traits:

use crdt_graph::{
    TwoPTwoPAddEdge, TwoPTwoPAddVertex, TwoPTwoPGraph, TwoPTwoPId,
    TwoPTwoPRemoveEdge, TwoPTwoPRemoveVertex, UpdateOperation,
};

type Id = u64;

#[derive(Clone, Debug)]
struct VA { id: Id }
impl TwoPTwoPId<Id> for VA { fn id(&self) -> &Id { &self.id } }
impl TwoPTwoPAddVertex<Id> for VA {}

#[derive(Clone, Debug)]
struct VR { id: Id, add_vertex_id: Id }
impl TwoPTwoPId<Id> for VR { fn id(&self) -> &Id { &self.id } }
impl TwoPTwoPRemoveVertex<Id> for VR { fn add_vertex_id(&self) -> &Id { &self.add_vertex_id } }

#[derive(Clone, Debug)]
struct EA { id: Id, source: Id, target: Id }
impl TwoPTwoPId<Id> for EA { fn id(&self) -> &Id { &self.id } }
impl TwoPTwoPAddEdge<Id> for EA {
    fn source(&self) -> &Id { &self.source }
    fn target(&self) -> &Id { &self.target }
}

#[derive(Clone, Debug)]
struct ER { id: Id, add_edge_id: Id }
impl TwoPTwoPId<Id> for ER { fn id(&self) -> &Id { &self.id } }
impl TwoPTwoPRemoveEdge<Id> for ER { fn add_edge_id(&self) -> &Id { &self.add_edge_id } }

# fn main() {
let mut graph: TwoPTwoPGraph<VA, VR, EA, ER, Id> = TwoPTwoPGraph::new();
graph.prepare(UpdateOperation::AddVertex(VA { id: 1 })).unwrap();
graph.prepare(UpdateOperation::AddVertex(VA { id: 2 })).unwrap();
graph.prepare(UpdateOperation::AddEdge(EA { id: 10, source: 1, target: 2 })).unwrap();
assert_eq!(graph.vertex_count(), 2);
assert_eq!(graph.edge_count(), 1);
# }

Built-in Graph Variants

Module Payload FlatBuffers File ID
types::simple None "CRDT"
types::bytes Option<Vec<u8>> "CRD2"
types::string Option<String> "CRD3"

Each variant provides: AddVertex, AddEdge, Graph (type alias), Operation (type alias).
RemoveVertex and RemoveEdge are shared across all variants (crdt_graph::types::{RemoveVertex, RemoveEdge}).

FlatBuffers Serialization

use crdt_graph::types::simple::{AddVertex, Graph};
use crdt_graph::flatbuffers::simple as fb;
use crdt_graph::Uuid;

# fn main() {
let mut graph = Graph::new();
let v1 = Uuid::now_v7();
let op = graph.prepare(AddVertex { id: v1 }.into()).unwrap();

// Encode
let bytes = fb::encode_operation(&op);

// Decode
let decoded = fb::decode_operation(&bytes).unwrap();
# }

Batch encoding/decoding is also supported via encode_operation_log() / decode_operation_log().

API

Method Description
prepare(op) Executes atSource checks, applies locally, and returns the operation to broadcast.
apply_downstream(op) Applies an operation received from a remote replica.
update_operation(op) Convenience wrapper around prepare that discards the return value.
lookup_vertex(id) Returns true if the vertex is in V_A \ V_R.
vertex_count() Number of active (non-removed) vertices.
edge_count() Number of active (non-removed) edges.
is_empty() true if no active vertices or edges.
vertices() Iterator over active vertices.
edges() Iterator over active edges.
generate_petgraph() Converts the current state into a petgraph::DiGraph.

Preconditions

Operation atSource downstream
addVertex(w)
addEdge(u, v) lookup(u) ∧ lookup(v)
removeVertex(w) lookup(w), no active edges addVertex(w) delivered
removeEdge(u, v) lookup((u, v)) addEdge(u, v) delivered