rockstore 0.1.2

A lightweight Python wrapper for RocksDB using CFFI

RockStore

A lightweight Python wrapper for RocksDB using CFFI.

Overview

RockStore provides a simple, Pythonic interface to RocksDB, Facebook's persistent key-value store. It uses CFFI for efficient native library bindings and focuses on clean binary data operations.

Features

• Simple API: Easy-to-use Python interface for RocksDB operations
• Binary Operations: Direct work with bytes for maximum performance
• Context Manager: Automatic resource management with with statements
• Configurable Options: Customize compression, buffer sizes, and more
• Read-Only Mode: Open databases in read-only mode for safe concurrent access
• Cross-Platform: Works on macOS, Linux, and Windows

Installation

Prerequisites

First, install RocksDB on your system:

macOS (using Homebrew):

brew install rocksdb

Ubuntu/Debian:

sudo apt-get install librocksdb-dev

CentOS/RHEL/Fedora:

sudo yum install rocksdb-devel
# or for newer versions:
sudo dnf install rocksdb-devel

Windows:

• Download pre-built RocksDB binaries or build from source
• Ensure rocksdb.dll is in your PATH

Install RockStore

pip install rockstore

Quick Start

Basic Usage

from rockstore import RockStore

# Open a database
db = RockStore('/path/to/database')

# Store and retrieve binary data
db.put(b'key1', b'value1')
value = db.get(b'key1')
print(value)  # b'value1'

# Store and retrieve string data (encode/decode manually)
db.put('name'.encode(), 'Alice'.encode())
name = db.get('name'.encode()).decode()
print(name)  # 'Alice'

# Delete data
db.delete(b'key1')

# Clean up
db.close()

Using Context Manager (Recommended)

from rockstore import open_database

with open_database('/path/to/database') as db:
    db.put(b'hello', b'world')
    value = db.get(b'hello')
    print(value)  # b'world'
# Database is automatically closed

Getting All Data

with open_database('/path/to/database') as db:
    db.put(b'key1', b'value1')
    db.put(b'key2', b'value2')
    
    # Get all key-value pairs (warning: loads everything into memory)
    all_data = db.get_all()
    for key, value in all_data.items():
        print(f"{key} -> {value}")

Range Queries and Pagination

For large databases, use range queries with pagination instead of get_all():

with open_database('/path/to/database') as db:
    # Add sample data
    for i in range(10000):
        key = f"user:{i:06d}".encode()
        value = f"User {i}".encode()
        db.put(key, value)
    
    # Paginated access - get 1000 records at a time
    batch_size = 1000
    start_key = None
    
    while True:
        # Get next batch
        batch = db.get_range(start_key=start_key, limit=batch_size)
        if not batch:
            break
            
        print(f"Processing {len(batch)} records...")
        
        # Process the batch
        for key, value in batch.items():
            process_record(key, value)
        
        # Setup for next batch
        last_key = max(batch.keys())
        start_key = last_key + b'\x00'  # Next key after last_key
    
    # Query specific ranges
    user_data = db.get_range(
        start_key=b'user:', 
        end_key=b'user:\xFF', 
        limit=500
    )
    
    # Memory-efficient iteration (one record at a time)
    for key, value in db.iterate_range(start_key=b'user:', end_key=b'user:\xFF'):
        process_user(key, value)

Handling 10M+ Record Databases

For very large databases (10M+ records), here's how to efficiently paginate in 100K batches:

def process_large_database_in_batches(db_path, batch_size=100_000):
    """
    Process a large database (10M+ records) in manageable batches.
    This approach uses constant memory regardless of database size.
    """
    with open_database(db_path) as db:
        start_key = None
        total_processed = 0
        batch_count = 0
        
        while True:
            # Get next batch
            batch = db.get_range(start_key=start_key, limit=batch_size)
            if not batch:
                break
            
            batch_count += 1
            total_processed += len(batch)
            
            print(f"Processing batch {batch_count}: {len(batch)} records")
            print(f"Total processed so far: {total_processed}")
            
            # Process each record in the batch
            for key, value in batch.items():
                # Your processing logic here
                process_record(key, value)
            
            # Prepare for next batch
            last_key = max(batch.keys())
            start_key = last_key + b'\x00'
            
            # Optional: Add progress tracking or break conditions
            if total_processed >= 10_000_000:  # Safety limit
                break
        
        print(f"Completed! Processed {total_processed} records in {batch_count} batches")

# Even more memory-efficient approach using iterator
def stream_process_large_database(db_path):
    """
    Stream process records one at a time - ultimate memory efficiency.
    """
    with open_database(db_path) as db:
        processed = 0
        for key, value in db.iterate_range():
            process_record(key, value)
            processed += 1
            
            if processed % 100_000 == 0:
                print(f"Processed {processed} records...")

Working with Strings

# Helper functions for string encoding/decoding
def encode_string(s):
    return s.encode('utf-8')

def decode_bytes(b):
    return b.decode('utf-8')

with open_database('/path/to/database') as db:
    # Store string data
    db.put(encode_string('user:123'), encode_string('John Doe'))
    
    # Retrieve and decode
    user_data = db.get(encode_string('user:123'))
    if user_data:
        print(decode_bytes(user_data))  # 'John Doe'

Configuration Options

from rockstore import RockStore

# Create database with custom options
options = {
    'create_if_missing': True,
    'compression_type': 'lz4_compression',
    'write_buffer_size': 64 * 1024 * 1024,  # 64MB
    'max_open_files': 1000
}

db = RockStore('/path/to/database', options=options)

Available Options

• create_if_missing (bool): Create database if it doesn't exist (default: True)
• read_only (bool): Open database in read-only mode (default: False)
• compression_type (str): Compression algorithm - 'no_compression', 'snappy_compression', 'zlib_compression', 'bz2_compression', 'lz4_compression', 'lz4hc_compression', 'xpress_compression', 'zstd_compression' (default: 'snappy_compression')
• write_buffer_size (int): Write buffer size in bytes (default: 64MB)
• max_open_files (int): Maximum number of open files (default: 1000)

Per-Operation Options

# Synchronous write (forces immediate disk write)
db.put(b'key', b'value', sync=True)

# Read without caching
value = db.get(b'key', fill_cache=False)

# Synchronous delete
db.delete(b'key', sync=True)

API Reference

RockStore Class

Constructor

RockStore(path, options=None)

Methods

Binary Operations:

• put(key: bytes, value: bytes, sync: bool = False) - Store binary data
• get(key: bytes, fill_cache: bool = True) -> bytes | None - Retrieve binary data
• delete(key: bytes, sync: bool = False) - Delete binary data

Bulk Operations:

• get_all(fill_cache: bool = True) -> dict[bytes, bytes] - Get all key-value pairs (loads into memory)
• get_range(start_key: bytes = None, end_key: bytes = None, limit: int = None, fill_cache: bool = True) -> dict[bytes, bytes] - Get range of key-value pairs with pagination support
• iterate_range(start_key: bytes = None, end_key: bytes = None, fill_cache: bool = True) -> Iterator[tuple[bytes, bytes]] - Memory-efficient iterator over key-value pairs

Resource Management:

• close() - Close the database
• Context manager support (with statement)

Context Manager

open_database(path, options=None) -> RockStore

Requirements

• Python 3.8+
• CFFI >= 1.15.0
• RocksDB library installed on system

Development

Running Tests

# Install development dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Run with coverage
pytest --cov=rockstore

License

MIT License - see LICENSE file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.