Expand description
An embeddable property-graph engine: nodes and edges with JSON properties, held in memory, with optional write-ahead-log durability, ACID transactions, and secondary indexes that follow the graph through a change feed.
This is the engine underneath drevo, which
adds Cypher, the HTTP and Bolt servers, and the Python bindings on top. Use
drevo-core on its own when you want the graph inside your process and
nothing else.
§Example
use drevo_core::engine::GraphEngine;
use drevo_core::model::{Direction, NewEdge, NewNode, Properties};
use drevo_core::native::NativeGraph;
use drevo_core::native_fts::NativeFtsIndex;
fn task(title: &str) -> NewNode {
NewNode {
kind: "task".into(),
title: title.into(),
body: String::new(),
body_html: String::new(),
properties: Properties::default(),
}
}
let graph = NativeGraph::new(); // or NativeGraph::open_durable("graph.wal")?
let deploy = graph.create_node(task("Deploy release"))?;
let review = graph.create_node(task("Review release notes"))?;
graph.create_edge(NewEdge {
from_id: deploy.id,
to_id: review.id,
kind: "blocked_by".into(),
weight: 1.0,
properties: Properties::default(),
})?;
let blockers = graph.neighbors(deploy.id, Direction::Outgoing, Some("blocked_by"))?;
assert_eq!(blockers[0].title, "Review release notes");
// Indexes catch up with the graph by reading its change feed.
let mut fts = NativeFtsIndex::new();
fts.sync(&graph);
assert_eq!(fts.search("deploy", 1)[0].0, deploy.id);§Where to start
native::NativeGraph— the engine: create, read, update and delete nodes and edges, traverse, take snapshots, run transactions (begin) and two-phase commits.engine::GraphEngine— the trait that operations are written against.model—Node,Edgeand their create/patch inputs.- Indexes:
native_fts(BM25 full-text search),native_label_index,native_property_index. - Replication:
replica,delta,hlc,lww.
Modules§
- bm25
- Okapi BM25 scoring primitives.
- csr
- Compressed-Sparse-Row adjacency snapshot for cache-friendly, lock-free parallel scans and graph algorithms.
- delta
- Version-vector delta exchange for multi-writer convergence.
- dump
- The
drevo-json-v1dump format: a whole graph as one value. - engine
GraphEngine— the graph operations, as a trait.- error
- The error type of
drevo-core. - hlc
- Hybrid Logical Clock (HLC) — causal versioning primitive for multi-writer convergence.
- labels
- The
_labelssecondary-label convention. - lww
- Last-Writer-Wins register, map and set CRDTs for multi-writer convergence.
- model
- Data model types for drevo.
- native
NativeGraph— the graph engine.- native_
fts - In-memory BM25 full-text search over a
NativeGraph, kept current from its change feed. - native_
label_ index - An in-memory secondary-label index over a
NativeGraph, kept current from its change feed. - native_
property_ index - An in-memory property-value index over a
NativeGraph, kept current from its change feed. - replica
- WAL-shipping read replica of a
NativeGraph, for read scaling and failover. - tokenizer
- Text tokenization for full-text search: normalization plus character trigram and word extraction, with CJK-aware handling.
- value_
encoding - Canonical byte encoding of property values for index keyspaces.