CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

ArffFiles is a header-only C++ library for reading ARFF (Attribute-Relation File Format) files and converting them into STL vectors. The library handles both numeric and categorical features, automatically factorizing categorical attributes.

Build System

This project uses CMake with Conan for package management:

CMake: Primary build system (requires CMake 3.20+)
Conan: Package management for dependencies
Makefile: Convenience wrapper for common tasks

Common Development Commands

Building and Testing

# Build and run tests (recommended)
make build && make test

# Alternative manual build process
mkdir build_debug
cmake -S . -B build_debug -D CMAKE_BUILD_TYPE=Debug -D ENABLE_TESTING=ON -D CODE_COVERAGE=ON
cmake --build build_debug -t unit_tests_arffFiles -j 16
cd build_debug/tests && ./unit_tests_arffFiles

Testing Options

# Run tests with verbose output
make test opt="-s"

# Clean test artifacts
make clean

Code Coverage

Code coverage is enabled when building with -D CODE_COVERAGE=ON and -D ENABLE_TESTING=ON. Coverage reports are generated during test runs.

Architecture

Core Components

Single Header Library: ArffFiles.hpp contains the complete implementation.

Main Class: ArffFiles

Header-only design for easy integration
Handles ARFF file parsing and data conversion
Automatically determines numeric vs categorical features
Supports flexible class attribute positioning

Key Methods

load(fileName, classLast=true): Load with class attribute at end/beginning
load(fileName, className): Load with specific named class attribute
getX(): Returns feature vectors as std::vector<std::vector<float>>
getY(): Returns labels as std::vector<int>
getNumericAttributes(): Returns feature type mapping

Data Processing Pipeline

File Parsing: Reads ARFF format, extracts attributes and data
Feature Detection: Automatically identifies numeric vs categorical attributes
Preprocessing: Handles missing values (lines with '?' are skipped)
Factorization: Converts categorical features to numeric codes
Dataset Generation: Creates final X (features) and y (labels) vectors

Dependencies

Catch2: Testing framework (fetched via CMake FetchContent)
Standard Library: Uses STL containers (vector, map, string)
C++17: Minimum required standard

Test Structure

Tests located in tests/ directory
Sample ARFF files in tests/data/
Single test executable: unit_tests_arffFiles
Uses Catch2 v3.3.2 for test framework

Conan Integration

The project includes a conanfile.py that:

Automatically extracts version from CMakeLists.txt
Packages as a header-only library
Exports only the main header file

2.9 KiB Raw Blame History