Extensible base for creating objects that behave like a Property Graph.
The property-graph
package provides a set of utility classes to be used for implementating libraries or applications based on a data model represented as a Property Graph. A Property Graph representation is useful for working with dependency chains, nested resource references, node-based state workflows, and more.
Conceptually, a Property Graph is a labeled, directed multigraph, in which entities ("nodes") may have named relationships ("edges") with other nodes on the graph. Both nodes and edges may be assigned arbitrary key/value attributes. Beyond that, property-graph
is intended to be small and practical, rather than providing a large standard library for graph theory — if you need something more comprehensive, I'd suggest graphology
.
Typically, you'll define custom classes inheriting from the base GraphNode
. When using TypeScript, an interface should be provided defining the kinds of connections that each type of graph node allows. Then, .set
and .get
methods may be used to set key/value attributes (strings, numbers, booleans, ...), and .getRef
and .setRef
methods may be used to create edges (or relationships) to other nodes of a compatible type. All references have names, and support compile-time type-checking.
In a codebase with many distinct types of entities and relationships among them (e.g. "Client has N Projects", "Project has N Tasks"), this project can make management of entities and their relationships considerably easier than writing plain getters/setters for each case.
- Traversal: GraphEdges are tracked and can be traversed up or down
- Disposal: GraphNode disposal automatically cleans up incoming references from other nodes
- Finding dependents: Efficiently locate all GraphNodes that refer to a given GraphNode, or that have references from a given GraphNode
- Change detection: GraphNodes dispatch events when changed, which can be optionally propagated throughout the graph
- Extensibility: Operations like
.copy()
,.equals()
, and.swap(a, b)
can be implemented abstractly
Definitions:
import { GraphNode, RefSet } from 'property-graph';
interface IPerson {
name: string;
age: number;
friends: RefSet<Person>;
pet: Pet;
}
interface IPet {
type: 'dog' | 'cat';
name: string;
}
class Person extends GraphNode<IPerson> {
getDefaults(): Nullable<IPerson> {
return {name: '', age: 0, friends: new RefSet(), pet: null};
}
}
class Pet extends GraphNode<IPet> {
getDefaults(): Nullable<IPet> {
return {type: 'dog', name: ''};
}
}
Basic usage:
const graph = new Graph();
const spot = new Pet(graph)
.set('type', 'dog')
.set('name', 'Spot');
const jo = new Person(graph)
.set('name', 'Jo')
.set('age', 41)
.setRef('pet', spot);
const sam = new Person(graph)
.set('name', 'Sam')
.set('age', 45)
.addRef('friends', jo);
Lifecycles:
jo.equals(sam); // recursive equality → false
console.log(sam.listRefs('friends')); // → [jo]
jo.dispose();
console.log(sam.listRefs('friends')); // → []
Literal attributes (string, number, boolean, ...) are modified with two methods:
node.get('key'): Literal
node.set('key', value: Literal): this
References support one named connection to a single graph node of a given type:
node.getRef('key'): GraphNode
node.setRef('key', node: GraphNode): this
Reference Lists support a named list of connections to graph nodes of a given type:
node.addRef('key', node: GraphNode): this
node.removeRef('key', node: GraphNode): this
node.listRefs('key'): GraphNode[]
Reference Maps support a named map having any number of subkeys, where each subkey points to a graph node of a given type:
node.getRefMap('key', 'subkey'): GraphNode
node.setRefMap('key', 'subkey', node: GraphNode): this
node.listRefMapKeys('key'): string[]
node.listRefMapValues('key'): GraphNode[]