A Modern C++ Fork of LMDB (Lightning Memory-Mapped Database)

FiksDataStore is a modern C++ reimplementation of the Lightning Memory-Mapped Database (LMDB), originally created by Howard Chu at Symas Corporation. This project aims to modernize the LMDB codebase while maintaining its exceptional performance characteristics and reliability.

FiksDataStore is a Btree-based database management library modeled loosely on the BerkeleyDB API, but much simplified. The entire database is exposed in a memory map, and all data fetches return data directly from the mapped memory, so no malloc's or memcpy's occur during data fetches. Key features include:

• Zero-copy architecture: Data is accessed directly from memory-mapped files
• ACID compliance: Full transactional semantics with durability guarantees
• Multi-version concurrency: Readers never block writers, writers never block readers
• Crash-proof: Copy-on-write strategy prevents corruption
• Maintenance-free: No periodic checkpointing or compaction required
• Cross-platform: Works on Windows, Linux, macOS, and other Unix systems

FiksDataStore aims to modernize LMDB by:

• Modern C++ Standards: Converting from C to modern C++ (C++17/20/23)
• Type Safety: Leveraging C++ type system for better safety
• RAII: Proper resource management with constructors/destructors
• STL Integration: Native support for STL containers and algorithms
• Template-based API: Generic programming for better performance
• Exception Safety: Modern error handling with exceptions
• Smart Pointers: Automatic memory management where appropriate
• Namespace Organization: Proper C++ namespace structure

🚧 This project is currently in early development phase 🚧

The original LMDB C code is being systematically converted to modern C++. The current codebase represents the baseline LMDB 0.9.70 implementation.

• Phase 1: Core API modernization
  • Convert core data structures to C++ classes
  • Implement RAII for resource management
  • Add proper exception handling
• Phase 2: STL integration
  • STL-compatible iterators
  • Container-like interfaces
  • Algorithm compatibility
• Phase 3: Template-based optimizations
  • Generic key/value types
  • Compile-time optimizations
  • Type-safe database handles
• Phase 4: Advanced features
  • Async I/O support
  • Custom allocators
  • Coroutine integration

• Extremely fast: Often outperforms other embedded databases
• Memory efficient: Minimal memory overhead
• Scalable: Handles databases larger than RAM efficiently
• Concurrent: Multiple readers, single writer model

• Memory-mapped I/O: Direct access to data without copying
• B+ tree storage: Efficient key-value storage and retrieval
• Transactions: Full ACID properties with nested transaction support
• Multiple databases: Single environment can host multiple named databases
• Duplicate keys: Optional support for multiple values per key
• Cursor operations: Efficient iteration and range queries

• Modern C++ compiler (GCC 9+, Clang 10+, MSVC 2019+)
• XMake 2.5+
• Git

The project uses XMake as the build system:

# Build the project
xmake

# Build and run tests
xmake build tests
xmake run tests

# Clean build artifacts
xmake clean

# Configure build options
xmake config --mode=debug    # Debug build
xmake config --mode=release  # Release build

# Show configuration
xmake show

#include "fds.h"

FDS_env *env;
FDS_dbi dbi;
FDS_val key, data;
FDS_txn *txn;

// Create environment
fds_env_create(&env);
fds_env_open(env, "./testdb", 0);

// Open database
fds_txn_begin(env, NULL, 0, &txn);
fds_dbi_open(txn, NULL, 0, &dbi);

// Store data
key.mv_size = sizeof(int);
key.mv_data = &some_key;
data.mv_size = sizeof(some_data);
data.mv_data = &some_data;
fds_put(txn, dbi, &key, &data, 0);

fds_txn_commit(txn);
fds_env_close(env);

#include <fiksdatastore/database.hpp>

// Planned modern C++ API
fiksdatastore::Environment env("./testdb");
auto db = env.open_database();

// RAII transaction
{
    auto txn = env.begin_transaction();
    db.put(txn, 42, "Hello, World!");
    txn.commit();
}

// STL-style iteration
for (const auto& [key, value] : db) {
    std::cout << key << ": " << value << std::endl;
}

The project includes the original LMDB test suite:

# Build and run all tests
xmake build tests
xmake run tests

# Run individual tests
xmake run mtest
xmake run mtest2
xmake run mtest3
# ... etc

• Original LMDB Documentation
• API Reference
• Getting Started Guide

Contributions are welcome! This project is in active development and we're looking for:

• C++ modernization expertise
• Performance optimization
• Testing and validation
• Documentation improvements

• Follow modern C++ best practices
• Maintain backward compatibility during transition
• Comprehensive testing for all changes
• Clear documentation for new APIs
• Adhere to Design Constraints - Mandatory design standards for the project
• Follow Coding Conventions - Mandatory coding style and formatting standards

FiksDataStore maintains the same licensing as the original LMDB:

• OpenLDAP Public License 2.8 - See LICENSE file
• Copyright 2011-2021 Howard Chu, Symas Corp.
• Additional modernization work: Copyright 2025 FiksDataStore contributors

• Howard Chu and Symas Corporation for creating and maintaining LMDB
• The OpenLDAP Project for the foundational work
• Martin Hedenfalk for the original btree implementation
• The LMDB community for years of testing and feedback

• Original LMDB - The source project

• Issues: GitHub Issues
• Discussions: GitHub Discussions

Note: This is a fork and modernization effort. For production use of the stable, battle-tested LMDB, please use the original LMDB project.