Gigapi Metadata provides a high-performance indexing system for managing metadata about data files (typically Parquet files) organized in time-partitioned structures. It supports efficient querying, merging operations, and provides both local JSON file storage and distributed Redis storage backends.

• Dual Storage Backends: JSON file-based storage for local deployments and Redis for distributed systems
• Time-Partitioned Data: Optimized for date/hour partitioned data structures 1
• Merge Planning: Intelligent merge planning for data consolidation across different layers 2
• Async Operations: Promise-based asynchronous operations for better performance 3
• Efficient Querying: Time-range and folder-based querying capabilities 4

go get github.com/gigapi/metadata

The fundamental data structure representing metadata about a single data file: 5

For local file-based storage, suitable for single-node deployments: 6

For distributed deployments with Redis backend: 7

Before using the library, initialize merge configurations which define merge behavior across different iterations: 8

Example configuration:

import "github.com/gigapi/metadata"

// Configure merge settings: [timeout_sec, max_size_bytes, iteration_id]
metadata.MergeConfigurations = [][3]int64{
    {10, 10 * 1024 * 1024, 1}, // 10s timeout, 10MB max size, iteration 1
    {30, 50 * 1024 * 1024, 2}, // 30s timeout, 50MB max size, iteration 2
}

// Create a JSON-based table index
tableIndex := metadata.NewJSONIndex("/data/root", "my_database", "my_table")

// Add metadata entries
entries := []*metadata.IndexEntry{
    {
        Database:  "my_database",
        Table:     "my_table", 
        Path:      "date=2024-01-15/hour=14/file1.parquet",
        SizeBytes: 1000000,
        MinTime:   1705327200000000000, // nanoseconds
        MaxTime:   1705327800000000000,
    },
}

// Batch operation (async)
promise := tableIndex.Batch(entries, nil)
result, err := promise.Get()

// Query with time range
options := metadata.QueryOptions{
    After:  time.Now().Add(-24 * time.Hour),
    Before: time.Now(),
}

entries, err := tableIndex.GetQuerier().Query(options)

// Get merge plan
planner := tableIndex.GetMergePlanner()
plan, err := planner.GetMergePlan("layer1", 1)

if plan != nil {
    // Execute merge (external process)
    // ...
    
    // Mark merge as complete
    err = planner.EndMerge(plan)
}

The main interface for table-level operations: 10

For database-level operations: 11

The system expects data organized in the following structure:

/root/
  ├── database1/
  │   ├── table1/
  │   │   ├── date=2024-01-15/
  │   │   │   ├── hour=00/
  │   │   │   ├── hour=01/
  │   │   │   └── ...
  │   │   └── date=2024-01-16/
  │   └── table2/
  └── database2/

For Redis backend, use standard Redis connection URLs:

• redis://localhost:6379/0 - Standard Redis
• rediss://user:pass@host:6380/1 - Redis with TLS 12

All operations return errors through the Promise interface or standard Go error handling. The library uses async operations for better performance in high-throughput scenarios.

Both JSON and Redis implementations are thread-safe and can be used concurrently across multiple goroutines.

Run tests with a local Redis instance:

# Start Redis
docker run -d -p 6379:6379 redis:alpine

# Run tests  
go test ./...

This project is licensed under the Apache License 2.0. 13

1. Fork the repository
2. Create a feature branch
3. Add tests for new functionality
4. Ensure all tests pass
5. Submit a pull request

• The library is optimized for time-series data workloads with frequent writes and time-range queries
• Redis backend is recommended for distributed deployments and high-throughput scenarios
• JSON backend is suitable for single-node deployments and development environments
• Merge operations are designed to be executed by external processes, with the library managing the planning and coordination
• All time values are stored as Unix nanoseconds for high precision temporal operations