dyedgreen / cqlite

Embedded graph database

CQLite

An embedded graph database implemented in Rust. This is currently a pre-release. It has not been extensively tested with 'real-world work-loads', and the file-format and API are not yet stabilized.

The longer term goal is to create an in-process graph database with a stable on-disk format and support for a wide range of programming languages, providing a native Rust API, as well as a C FFI interface.

use cqlite::Graph;

let graph = Graph::open_anon()?;

let mut txn = graph.mut_txn()?;
let edge: u64 = graph.prepare(
        "
        CREATE (a:PERSON { name: 'Peter Parker' })
        CREATE (b:PERSON { name: 'Clark Kent' })
        CREATE (a) -[e:KNOWS]-> (b)
        RETURN ID(e)
        "
    )?
    .query_map(&mut txn, (), |m| m.get(0))?
    .next()
    .unwrap()?;
txn.commit()?;

let name: String = graph.prepare(
        "
        MATCH (p:PERSON) <-[e:KNOWS]- (:PERSON)
        WHERE ID(e) = $edge
        RETURN p.name
        "
    )?
    .query_map(&mut graph.txn()?, ("edge", edge), |m| m.get(0))?
    .next()
    .unwrap()?;
assert_eq!("Clark Kent", name);

Architecture Overview

Parser :: src/parser

PEG grammar and parser for a subset of the CYPHER graph query language

Query Planner :: src/planner

Transforms a parsed query ast into a logical query plan. Performs some optimizations on the query plan.

Byte-Code Interpreter :: src/runtime

Defines a simple 'byte' code (Instructions) and can execute those against a given database, as well as generate instructions for a given query plan.

Storage Backend :: src/store

Uses a disc-backed btree to provide basic storage, iteration and lockup for nodes and edges.