C++ Implementation Reference

All C++ implementations keep single-file algorithm cores and share a thin buffer facade under algorithms/shared/cpp/include/compresskit/. There is no stateful streaming layer; encode_buffer / decode_buffer take a BufferTransform function pointer directly.

Compilation

The project is built with CMake. make build is a thin wrapper around the CMake invocation:

cmake -S . -B build && cmake --build build

Binaries are produced under build/ (e.g. build/huffman_cpp).

Recommended Flags

The CMake build enables C++17 with optimization and warnings. For debug builds, AddressSanitizer and UBSan can be enabled through the CMake configuration.

Flag	Purpose
-std=c++17	Enable C++17 features
-O2	Optimization level
-Wall -Wextra	Warnings
-fsanitize=address	AddressSanitizer (debug builds)
-fsanitize=undefined	UBSan (debug builds)

Huffman (algorithms/huffman/cpp/main.cpp)

Usage

./build/huffman_cpp encode input.bin output.huf
./build/huffman_cpp decode output.huf decoded.bin

Internal Structure

• BitWriter / BitReader — bit-level I/O classes
• Node — Huffman tree node with custom deleter
• compress_file() / decompress_file() — main encode/decode functions

File Format

Offset	Size	Field
0	4B	Magic: HFMN
4	4B	Frequency table size (always 257)
8	1028B	Frequency table (257 × uint32 LE)
1036+	Variable	Encoded bit stream

---

Arithmetic (algorithms/arithmetic/cpp/main.cpp)

Usage

./build/arithmetic_cpp encode input.bin output.aenc
./build/arithmetic_cpp decode output.aenc decoded.bin

Key Classes

• ArithmeticEncoder — state machine with low, high, pendingBits
• ArithmeticDecoder — decoder with code initialization

---

Range Coder (algorithms/range/cpp/main.cpp)

Usage

./build/rangecoder_cpp encode input.bin output.rcnc
./build/rangecoder_cpp decode output.rcnc decoded.bin

File Format

Offset	Size	Field
0	4B	Magic: RCNC
4	4B	Frequency table size
8	Variable	Frequency table
...	Variable	Byte stream (renormalized intervals)

---

RLE (algorithms/rle/cpp/main.cpp)

Usage

./build/rle_cpp encode input.bin output.rle
./build/rle_cpp decode output.rle decoded.bin

File Format

Offset	Size	Field
0	4B	Magic: RLE\x00
4+	Variable	Repeated (count: uint32 LE, value: byte) pairs

---

Common Patterns

Pattern	Description
Single-file core	Each algorithm core in one main.cpp
Shared dependencies	Uses common code from algorithms/shared/cpp/
Error handling	fprintf(stderr, ...) + exit(1)
Memory management	std::unique_ptr with custom deleters

Shared Buffer Facade

The buffer helpers live in:

• compresskit/result.hpp — StatusCode enum and Result<T> template
• compresskit/buffer_api.hpp — BufferTransform, encode_buffer, decode_buffer, file helpers
• compresskit/algorithms.hpp — per-algorithm *_encode_buffer / *_decode_buffer entry points

Status Codes

Code	Meaning
OK	Success
ERR_CORRUPT	Encoded data failed structural validation
ERR_SIZE_LIMIT	Input exceeded 4 GiB or decoded output exceeded 1 GiB

Buffer Transform

BufferTransform is a function pointer alias mapping an in-memory byte buffer to its transformed form. Each algorithm exposes one encode and one decode transform.

#include <vector>
#include "compresskit/algorithms.hpp"
#include "compresskit/buffer_api.hpp"

std::vector<uint8_t> encode(const std::vector<uint8_t>& input) {
    auto result = compresskit::encode_buffer(huffman_encode_buffer, input);
    return result.ok() ? std::move(result.value) : std::vector<uint8_t>{};
}