A tiny, in-memory, typed directed multigraph with:

• Typed nodes (e.g., Person, Team, Project, Resource)
• Typed edges (e.g., WORKS_FOR, MANAGES, ASSIGNED_TO, DEPENDS_ON)
• Multiple relationships between the same nodes
• A minimal, Cypher-inspired pattern engine for traversal

See runnable examples in example/bin/ and the sample graph in example/lib/data.dart.

Table of Contents #

• Quick Preview
• Pattern Query Examples
• Mini-Cypher Reference
• Comparison with Cypher
• Core concepts
• Pattern syntax (mini, Cypher-inspired)
• Quick start
• Row-wise pattern results (new)
• Generic traversal utilities (new)
• JSON Serialization
• Examples index
• License

Quick Preview #

Mini-Cypher query highlighting nodes and edges in the Flutter demo.

Pattern Query Examples #

• Simple patterns: "user:User" → {alice, bob, charlie}
• Forward patterns: "user-[:MEMBER_OF]->group"
• Backward patterns: "resource<-[:CAN_ACCESS]-group<-[:MEMBER_OF]-user" → {alice, bob}
• Label filtering: "user:User{label~Admin}" → {bob}

Mini-Cypher Reference #

Graph_kit supports a subset of Cypher syntax with some extensions. Here's the complete reference:

Keywords #

Node Syntax #

variable:Type{filters}

Components:

• variable: Name for results map (required) - e.g., person, user, employee
• :Type: Filter by node type (optional) - e.g., :Person, :Team
• {filters}: Label filtering (optional)

Node Examples:

person:Person              # All Person nodes
user                       # All nodes (any type)
manager:Person{label=Bob}  # Person nodes with exact label "Bob"
admin:User{label~Admin}    # User nodes containing "Admin" in label

Edge Syntax #

-[:EDGE_TYPE]->    # Forward relationship
<-[:EDGE_TYPE]-    # Backward relationship

Edge Examples:

person-[:WORKS_FOR]->team           # Forward: person works for team
team<-[:WORKS_FOR]-person           # Backward: people who work for team
user-[:MEMBER_OF]->group-[:CAN_ACCESS]->resource   # Multi-hop

Label Filters #

Filter Examples:

person:Person{label=Alice Cooper}     # Exact name match
user:User{label~admin}               # Any admin user
team:Team{label=Engineering}         # Specific team

Complete Pattern Examples #

# Basic queries (with/without MATCH)
person:Person
MATCH person:Person

# Relationships
person:Person-[:WORKS_FOR]->team:Team
MATCH person:Person-[:MANAGES]->team-[:ASSIGNED_TO]->project

# Filtered queries
person:Person{label~Alice}-[:WORKS_FOR]->team
team:Team{label=Engineering}<-[:WORKS_FOR]-person

# Multi-hop traversal
MATCH person:Person-[:WORKS_FOR]->team-[:ASSIGNED_TO]->project

Variable Names in Results #

Query results are organized by variable names:

final results = query.match('manager:Person-[:MANAGES]->team:Team');
// Returns: {'manager': {...}, 'team': {...}}

final results2 = query.match('boss:Person-[:MANAGES]->group:Team');
// Returns: {'boss': {...}, 'group': {...}}

Same node type, different roles:

owner:Person-[:OWNS]->project<-[:ASSIGNED_TO]-team<-[:WORKS_FOR]-worker:Person
# Results: {'owner': {...}, 'worker': {...}, 'project': {...}, 'team': {...}}

Syntax Limitations #

Not Supported:

• Mixed directions in single pattern: person-[:A]->team<-[:B]-other
• Variable length paths: person-[:KNOWS*1..3]->friend
• Complex WHERE clauses: WHERE person.age > 25
• Multiple MATCH statements
• OPTIONAL MATCH

Workarounds:

• Use matchMany() for multiple patterns
• Use label filters instead of WHERE
• Use matchRows() for path-specific results

Comparison with Cypher #

Core concepts #

• Node
  • Each node has id, type, and label.
  • Example: u1, type=User, label=Mark.
• Edge types
  • Relationship labels between nodes, used to traverse (e.g., WORKS_FOR, MANAGES).
• Graph<N extends Node>
  • addNode(n), addEdge(src, edgeType, dst).
  • outNeighbors(srcId, edgeType), inNeighbors(dstId, edgeType).
• PatternQuery<N extends Node>
  • match(pattern, {startId}) – run a single chain.
  • matchMany([patterns], {startId}) – run multiple independent chains and union results by variable name.

Pattern syntax (mini, Cypher-inspired) #

• Seeding without IDs: alias:Type
  • Example: 'users:User' seeds the first segment with all nodes whose type == 'User'.
• Directional edges:
  • Outgoing: -[:EDGE]-> (uses outNeighbors)
  • Incoming: <-[:EDGE]- (uses inNeighbors)
• Variables (aliases):
  • Each segment name is a key in the returned map.
  • Example: 'users:User-[:MEMBER_OF]->group' returns keys 'users:User' and 'group'.

Quick start #

1. Build the example graph

import 'package:graph_kit/graph_kit.dart';

final g = Graph<Node>();
// Add your nodes and edges here
final pq = PatternQuery(g);

2. All users (no IDs needed)

final res = pq.match('users:User');
for (final id in res['users:User'] ?? {}) {
  final n = g.nodesById[id];
  print('$id (${n?.type}: ${n?.label})');
}

Runnable: dart run example/bin/allusers.dart

3. Users of a group (by group ID)

final res = pq.match('group-[:MEMBER_OF]<-user', startId: 'g_admins');
print(res['user']); // Set of user IDs

Runnable: dart run example/bin/group_users.dart g_admins or by label "Admins".

4. Resources a person can access through their team

final res = pq.match(
  'person-[:WORKS_FOR]->team-[:HAS_ACCESS]->resource',
  startId: 'alice',
);
print(res['resource']); // Set of resource IDs

Runnable: dart run example/bin/user_assets.dart u1

5. People who can work on a project (through team assignments)

final res = pq.match(
  'project-[:ASSIGNED_TO]<-team-[:WORKS_FOR]<-person',
  startId: 'web_app_project',
);
print(res['person']);

Design and performance #

• Traversal from a known ID (startId) is fast:
  • Each hop uses adjacency maps; cost is proportional to the edges visited.
• Seeding by type (alias:Type) does a one-time node scan to find initial seeds.
  • For small/medium graphs, this is effectively instant; indexing can be added later if needed.
• matchMany([...]) mirrors “multiple MATCH/OPTIONAL MATCH” lines in Cypher by running several independent chains from the same start and unioning results.

Row-wise pattern results (new) #

For cases where you need to know which variables co-occurred on the same matched path (e.g., which team gives access to which resource), use matchRows():

final rows = pq.matchRows(
  'person-[:WORKS_FOR]->team-[:HAS_ACCESS]->resource',
  startId: 'alice',
);
// rows: [{person: alice, team: engineering, resource: database}, ...]

// Build resource -> teams map from rows
final resourceToTeams = <String, Set<String>>{};
for (final r in rows) {
  final resource = r['resource']!;
  final team = r['team']!;
  resourceToTeams.putIfAbsent(resource, () => <String>{}).add(team);
}

You can union multiple chains while preserving row bindings with matchRowsMany([...]):

final rows = pq.matchRowsMany([
  'person-[:WORKS_FOR]->team-[:ASSIGNED_TO]->project-[:USES]->resource',
  // If people can be assigned directly too:
  'person-[:ASSIGNED_TO]->project-[:USES]->resource',
], startId: 'alice');

// Build resource -> projects mapping
final resourceToProjects = <String, Set<String>>{};
for (final r in rows) {
  final resource = r['resource'];
  final project = r['project'];
  if (resource != null && project != null) {
    resourceToProjects.putIfAbsent(resource, () => <String>{}).add(project);
  }
}

Notes:

• The first segment supports optional :Type and {label=...}/{label~...} filters for seeding.
• Intermediate segments currently match by structure (alias and edges); type/label filters may be added later if needed.

Generic traversal utilities (new) #

For BFS-style expansions and subgraph extraction with hop limits, use expandSubgraph from traversal.dart:

import 'package:graph_kit/graph_kit.dart';

final seeds = {'u1'};
final rightward = {'WORKS_FOR', 'ASSIGNED_TO', 'HAS_ACCESS'}; // your edge types

final sub = expandSubgraph(
  g,
  seeds: seeds,
  edgeTypesRightward: rightward,
  forwardHops: 3,
  backwardHops: 0,
);

print('Nodes: ' + sub.nodes.length.toString());
print('Edges: ' + sub.edges.length.toString());

JSON Serialization #

Save and load graphs to/from JSON for persistence and data exchange:

import 'dart:io';
import 'package:graph_kit/graph_kit.dart';

// Build your graph
final graph = Graph<Node>();
graph.addNode(Node(id: 'alice', type: 'User', label: 'Alice',
  properties: {'email': '[email protected]', 'active': true}));
graph.addNode(Node(id: 'team1', type: 'Team', label: 'Engineering'));
graph.addEdge('alice', 'MEMBER_OF', 'team1');

// Serialize to JSON
final json = graph.toJson();
final jsonString = graph.toJsonString(pretty: true);

// Save to file
await File('graph.json').writeAsString(jsonString);

// Load from file
final loadedJson = await File('graph.json').readAsString();
final restoredGraph = GraphSerializer.fromJsonString(loadedJson, Node.fromJson);

// Graph is fully restored - queries work immediately
final query = PatternQuery(restoredGraph);
final members = query.match('team<-[:MEMBER_OF]-user', startId: 'team1');
print(members['user']); // {alice}

Examples index #

• example/bin/allusers.dart – list all users
• example/bin/group_users.dart – users in a group (by ID or by label)
• example/bin/user_assets.dart – assets a user can connect to

License #

See LICENSE.