Switch to schemaless open type graphs

[kornelrabczak]

Current implementation uses strictly typed subgraphs, requiring explicit CREATE NODE TABLE and CREATE REL TABLE statements before any data can be inserted (see LadybugDBConfig.initializeSchema()):

conn.query("CREATE NODE TABLE IF NOT EXISTS Entity(name STRING PRIMARY KEY, type STRING, observations STRING[])");
conn.query("CREATE REL TABLE IF NOT EXISTS RELATED_TO(FROM Entity TO Entity, relationType STRING)");

This forces a pre-defined schema. LadybugDB supports open type graphs which allow creating nodes with arbitrary labels without first defining tables:

CREATE GRAPH my_graph ANY;
USE my_graph;
-- No CREATE NODE TABLE needed!
CREATE (u:User {name: 'Alice'});
CREATE (p:Project {name: 'Archiledger', language: 'Java'});

Motivation

• Flexible entity types, users can define any entity type without schema changes
• Dynamic properties, entities can have varying properties per type
• Reduced boilerplate, no need for initializeSchema() or upfront table definitions
• Better fit for knowledge graphs, knowledge graphs are inherently schemaless; entities from different domains have different shapes
• Neo4j/GQL compatibility, open type graphs provide familiar behavior for users migrating from Neo4j

Considerations

• Open type graphs may have slightly different performance characteristics
• Decide whether to keep strict typing as an opt-in mode for users who want schema guarantees