TrainDis is a high-performance, Redis-compatible in-memory database implemented in C++. It provides key-value storage with support for multiple data types, persistence mechanisms (AOF and RDB), and master-slave replication.

• Multiple Data Types: Supports strings, hashes, and sorted sets (zsets)
• Redis Protocol: Compatible with RESP (Redis Serialization Protocol)
• Persistence Options:
  • AOF (Append-Only File) with three modes: Always, EverySec, and None
  • RDB (Redis Database) snapshotting
• Master-Slave Replication: Support for data replication between nodes
• Thread-Safe Operations: Concurrency-safe data structures
• Configurable Settings: Load configuration from file or command line
• Memory Efficient: Optimized memory usage with expiration support

1. Server: Main server implementation using epoll for I/O multiplexing
2. KV Storage: In-memory key-value storage supporting different data types
3. AOF Manager: Handles Append-Only File persistence with various sync modes
4. RDB Manager: Manages database snapshots in RDB format
5. RESP Parser: Handles Redis Serialization Protocol for client communication
6. Config Loader: Parses configuration from files or command-line arguments

• Strings: Simple key-value pairs with expiration support
• Hashes: Key-value maps stored under a single key
• Sorted Sets (ZSets): Collections of unique members ordered by scores, implemented with skip lists

The server can be configured via command-line arguments or configuration files:

./train_set [--port <port>] [--bind <ip>] [--config <file>]

• port: Server port (default: 6379)
• bind_address: IP address to bind to (default: 127.0.0.1)
• aof.enabled: Enable AOF persistence (true/false)
• aof.mode: AOF sync mode (always/everysec/none)
• aof.dir: Directory for AOF file
• aof.filename: AOF file name
• rdb.enabled: Enable RDB persistence (true/false)
• rdb.dir: Directory for RDB file
• rdb.filename: RDB file name
• replicate.enabled: Enable replication (true/false)
• replicate.master_host: Master host for replication
• replicate.master_port: Master port for replication

• SET key value [EX seconds] - Set a key-value pair
• GET key - Get value by key
• DEL key1 [key2 ...] - Delete one or more keys
• EXISTS key - Check if key exists
• TTL key - Get time to live for a key
• EXPIRE key seconds - Set expiration time

• HSET key field value - Set field in hash
• HGET key field - Get field value from hash
• HDEL key field1 [field2 ...] - Delete fields from hash
• HEXISTS key field - Check if field exists in hash
• HLEN key - Get number of fields in hash
• HGETALL key - Get all fields and values in hash

• ZADD key score member - Add member to sorted set
• ZREM key member [member ...] - Remove members from sorted set
• ZRANGE key start stop - Get range of members by rank
• ZSCORE key member - Get score of member in sorted set

• SAVE - Save data to RDB file
• BGSAVE - Save data to RDB file in background
• BGREWRITEAOF - Rewrite AOF file in background

• PING - Ping the server
• INFO - Get server information
• FLUSHALL - Remove all keys
• KEYS * - List keys matching pattern

• C++17 compatible compiler
• CMake 3.15 or higher
• POSIX-compliant system (Linux/macOS)

mkdir build
cd build
cmake ..
make

# Using a config file
./train_set --config ../conf/always.conf

# Using command line arguments
./train_set --port 6380 --bind 127.0.0.1

The repository includes several example configuration files:

• always.conf: AOF enabled with "always" sync mode
• everysec.conf: AOF enabled with "every second" sync mode
• none.conf: AOF disabled
• master.conf: Master node configuration
• replica.conf: Replica node configuration

• Efficient AOF Implementation: Multiple sync modes for balancing performance and durability
• Batch Writes: AOF supports configurable batch sizes to improve write performance
• Skip Lists: Sorted sets implemented with skip lists for efficient range operations
• Memory Pre-allocation: AOF files can be pre-allocated to reduce fragmentation
• Background Operations: AOF rewriting and RDB saving run in background threads

TrainDis supports master-slave replication:

• Slaves can connect to masters to receive data updates
• Initial synchronization happens via RDB dump
• Incremental updates are sent as commands over the replication channel

The project includes comprehensive unit tests for all major components:

• AOF functionality tests
• RDB persistence tests
• Key-value storage tests
• Configuration loading tests
• RESP protocol parsing tests