from pathlib import Path
# Directory paths
BENCHMARK_DIR = Path(__file__).resolve().parent
DATA_DIR = BENCHMARK_DIR / "data"
CACHE_DIR = BENCHMARK_DIR / "cache"
REPOS_DIR = BENCHMARK_DIR / "repos"
RESULTS_DIR = BENCHMARK_DIR / "results"
# Subdirectory structure
RAW_DATA_DIR = DATA_DIR / "raw"
PROCESSED_DATA_DIR = DATA_DIR / "processed"
REPORTS_DIR = BENCHMARK_DIR / "reports"
# Default dataset paths (updated for new structure)
DEFAULT_MULOCBENCH_PATH = "data/raw/mulocbench_v1.jsonl"
DEFAULT_FILTERED_PATH = "data/processed/filtered.jsonl"
# Large repos (>=100MB) excluded by default to avoid slow cloning
EXCLUDED_REPOS: frozenset[str] = frozenset(
{
# >1GB
"langflow-ai/langflow", # 1.2G
# >500MB
"PaddlePaddle/PaddleOCR", # 645M
"odoo/odoo", # 603M
# >300MB
"ansible/ansible", # 437M
"pytorch/pytorch", # 383M
"deepfakes/faceswap", # 374M
"huggingface/transformers", # 339M
"python/cpython", # 329M
"All-Hands-AI/OpenHands", # 310M
# >100MB
"hacksider/Deep-Live-Cam", # 153M
"yt-dlp/yt-dlp", # 129M
"pandas-dev/pandas", # 114M
# Previously excluded (not in current dataset)
"home-assistant/core",
"kubernetes/kubernetes",
"torvalds/linux",
"chromium/chromium",
}
)
def get_benchmark_dir() -> Path:
return BENCHMARK_DIR
def get_repos_dir() -> Path:
return REPOS_DIR
def get_data_dir() -> Path:
return DATA_DIR
def get_cache_dir() -> Path:
return CACHE_DIR
def get_results_dir() -> Path:
return RESULTS_DIR
def get_raw_data_dir() -> Path:
return RAW_DATA_DIR
def get_processed_data_dir() -> Path:
return PROCESSED_DATA_DIR
def get_reports_dir() -> Path:
return REPORTS_DIR
