hugegraph-struct

Imported from gencau/test-practices-agent-configurations (dataset/repos/apache§hugegraph/hugegraph-struct/AGENTS.md). Install upstream with npx skills add gencau/test-practices-agent-configurations --skill hugegraph-struct. Copyright stays with the author.

AGENTS.md

This file provides guidance to an AI coding tool when working with code in this repository.

Module Overview

hugegraph-struct is a foundational data structures module that defines the core abstractions shared across HugeGraph distributed components. This module must be built before hugegraph-pd and hugegraph-store as they depend on its structure definitions.

Key Responsibilities:

• Schema element definitions (VertexLabel, EdgeLabel, PropertyKey, IndexLabel)
• Graph element structures (BaseVertex, BaseEdge, BaseProperty)
• Binary serialization/deserialization for efficient storage and RPC
• Type system definitions (HugeType enum, data types, ID strategies)
• Query abstractions (Query, ConditionQuery, IdQuery, Aggregate)
• Chinese text analyzers (multiple implementations: Jieba, IK, HanLP, etc.)
• Authentication utilities (JWT token generation, constants)

Build Commands

Building This Module

# From hugegraph-struct directory
mvn clean install -DskipTests

# Build with tests (if any exist in future)
mvn clean install

# From parent directory (hugegraph root)
mvn install -pl hugegraph-struct -am -DskipTests

Dependency Chain

This module is a critical dependency for distributed components:

# Correct build order for distributed components:
# 1. Build hugegraph-struct first
mvn install -pl hugegraph-struct -am -DskipTests

# 2. Then build PD
mvn clean package -pl hugegraph-pd -am -DskipTests

# 3. Then build Store
mvn clean package -pl hugegraph-store -am -DskipTests

Code Architecture

Package Structure

org.apache.hugegraph/
├── struct/schema/          # Schema element definitions
│   ├── SchemaElement       # Base class for all schema types
│   ├── VertexLabel         # Vertex label definitions
│   ├── EdgeLabel           # Edge label definitions
│   ├── PropertyKey         # Property key definitions
│   ├── IndexLabel          # Index label definitions
│   └── builder/            # Builder pattern implementations
├── structure/              # Graph element structures
│   ├── BaseElement         # Base class for vertices/edges
│   ├── BaseVertex          # Vertex implementation
│   ├── BaseEdge            # Edge implementation
│   ├── BaseProperty        # Property implementation
│   └── builder/            # Element builders
├── type/                   # Type system
│   ├── HugeType            # Enum for all graph types (VERTEX, EDGE, etc.)
│   ├── GraphType           # Type interface
│   ├── Namifiable          # Name-based types
│   ├── Idfiable            # ID-based types
│   └── define/             # Type definitions (DataType, IdStrategy, etc.)
├── id/                     # ID generation and management
│   ├── Id                  # ID interface
│   ├── IdGenerator         # ID generation utilities
│   ├── EdgeId              # Edge-specific ID handling
│   └── IdUtil              # ID utility methods
├── serializer/             # Binary serialization
│   ├── BytesBuffer         # Buffer for binary I/O
│   ├── BinaryElementSerializer    # Element serialization
│   └── DirectBinarySerializer     # Direct binary access
├── query/                  # Query abstractions
│   ├── Query               # Base query interface
│   ├── ConditionQuery      # Conditional queries
│   ├── IdQuery             # ID-based queries
│   ├── Condition           # Query conditions
│   └── Aggregate           # Aggregation queries
├── analyzer/               # Text analyzers (Chinese NLP)
│   ├── Analyzer            # Base analyzer interface
│   ├── AnalyzerFactory     # Factory for creating analyzers
│   ├── IKAnalyzer          # IK Chinese word segmentation
│   ├── JiebaAnalyzer       # Jieba segmentation
│   ├── HanLPAnalyzer       # HanLP NLP
│   ├── AnsjAnalyzer        # Ansj segmentation
│   ├── WordAnalyzer        # Word-based analysis
│   ├── JcsegAnalyzer       # Jcseg segmentation
│   ├── MMSeg4JAnalyzer     # MMSeg4J segmentation
│   └── SmartCNAnalyzer     # Lucene SmartCN
├── auth/                   # Authentication utilities
│   ├── TokenGenerator      # JWT token generation
│   └── AuthConstant        # Auth constants
├── backend/                # Backend abstractions
│   ├── BinaryId            # Binary ID representation
│   ├── BackendColumn       # Column abstraction
│   └── Shard               # Shard information
├── options/                # Configuration options
│   ├── CoreOptions         # Core configuration
│   └── AuthOptions         # Auth configuration
├── util/                   # Utilities
│   ├── StringEncoding      # String encoding utilities
│   ├── GraphUtils          # Graph utility methods
│   ├── LZ4Util             # LZ4 compression
│   ├── Blob                # Binary blob handling
│   └── collection/         # Collection utilities (IdSet, CollectionFactory)
└── exception/              # Exception hierarchy
    ├── HugeException       # Base exception
    ├── BackendException    # Backend errors
    ├── NotSupportException # Unsupported operations
    ├── NotFoundException   # Not found errors
    └── NotAllowException   # Permission errors

Key Architectural Concepts

1. Two-Layer Schema System

The module defines a dual schema hierarchy:

• struct.schema.*: Schema element definitions (VertexLabel, EdgeLabel, etc.) - these are metadata about the graph structure
• structure.*: Actual graph elements (BaseVertex, BaseEdge, etc.) - these are data instances

The schema layer defines the "blueprint" while the structure layer implements the "instances".

2. Type System

The HugeType enum (type/HugeType.java) defines all possible types:

• Schema types: VERTEX_LABEL, EDGE_LABEL, PROPERTY_KEY, INDEX_LABEL
• Data types: VERTEX, EDGE, PROPERTY, AGGR_PROPERTY_V, AGGR_PROPERTY_E
• Special types: META, COUNTER, TASK, OLAP, INDEX

3. ID Management

IDs are critical for distributed systems:

• Id interface provides abstraction over different ID types
• IdGenerator creates IDs based on strategy (AUTO_INCREMENT, PRIMARY_KEY, CUSTOMIZE)
• EdgeId uses special encoding: source vertex ID + edge label ID + sort values + target vertex ID
• Binary serialization optimizes ID storage

4. Binary Serialization

BytesBuffer and serializers enable:

• Efficient storage in RocksDB and other backends
• Fast gRPC message passing between PD/Store/Server
• Compact on-disk and in-memory representation

5. Query Abstraction

Query classes provide backend-agnostic query building:

• Query: Base interface with limit, offset, ordering
• ConditionQuery: Supports conditions (EQ, GT, LT, IN, CONTAINS, etc.)
• IdQuery: Direct ID-based lookups
• Aggregate: Aggregation operations (SUM, MAX, MIN, AVG)

Dependencies

Critical Dependencies

• hg-pd-client (${project.version}): PD client for metadata coordination
• hugegraph-common (${project.version}): Shared utilities
• Apache TinkerPop 3.5.1: Graph computing framework
• Guava 25.1-jre: Google utilities
• Eclipse Collections 10.4.0: High-performance collections
• fastutil 8.1.0: Fast primitive collections

Text Analysis Dependencies

Multiple Chinese NLP libraries for different use cases:

• jieba-analysis 1.0.2: Popular Chinese word segmentation
• IKAnalyzer 2012_u6: IK word segmentation
• HanLP portable-1.5.0: Natural language processing
• Ansj 5.1.6: Ansj segmentation
• Word 1.3: APDPlat word segmentation
• Jcseg 2.2.0: Jcseg segmentation
• mmseg4j-core 1.10.0: MMSeg4J segmentation
• lucene-analyzers-smartcn 7.4.0: Lucene SmartCN

Security Dependencies

• jjwt-api/impl/jackson 0.11.2: JWT token handling
• jbcrypt 0.4: Password hashing

Development Notes

When Modifying This Module

1. Understand the impact: Changes here affect hugegraph-pd, hugegraph-store, and hugegraph-server
2. Rebuild dependent modules: After modifying, rebuild PD and Store modules
3. Binary compatibility: Serialization changes require careful version migration
4. ID changes: Modifying ID generation can break existing data

Working with Schema Elements

When adding or modifying schema elements in struct/schema/:

• Extend SchemaElement base class
• Implement required interfaces (Namifiable, Typifiable)
• Add corresponding HugeType enum value if needed
• Update serialization logic in BinaryElementSerializer
• Verify schema builder patterns in struct/schema/builder/

Working with Binary Serialization

When modifying serialization:

• Changes to BytesBuffer format require version migration
• Test with all backends (RocksDB, HStore)
• Ensure backward compatibility or provide migration path
• Update both write and read paths consistently

Adding Text Analyzers

To add a new text analyzer:

1. Implement the Analyzer interface in analyzer/
2. Register in AnalyzerFactory
3. Add dependency to pom.xml
4. Test with Chinese text queries

Common Patterns

Creating Schema Elements

// Schema elements use builders
PropertyKey propertyKey = schema.propertyKey("name")
                                .asText()
                                .valueSingle()
                                .create();

ID Generation

// Generate IDs based on strategy
Id id = IdGenerator.of(value, IdType.LONG);
Id edgeId = EdgeId.parse(sourceId, direction, label, sortValues, targetId);

Binary Serialization

// Write to buffer
BytesBuffer buffer = BytesBuffer.allocate(size);
buffer.writeId(id);
buffer.writeString(name);

// Read from buffer
Id id = buffer.readId();
String name = buffer.readString();

Cross-Module References

This module is referenced by:

• hugegraph-pd: Uses schema definitions for metadata management
• hugegraph-store: Uses serialization for storage and RPC
• hugegraph-server/hugegraph-core: Uses all abstractions for graph operations
• hugegraph-server/hugegraph-api: Uses structures for REST API serialization

License and Compliance

This module follows Apache Software Foundation guidelines:

• All files must have Apache 2.0 license headers
• Third-party dependencies require license documentation in install-dist/release-docs/licenses/
• Excluded from Apache RAT: None (all source files checked)