# Crawled PDFs are large (325 MB for four venue-years) and re-fetchable: the
# crawler is idempotent, so `stopslop crawl <venue> <year>` rebuilds them from the
# ids recorded in stats.json. The measurements are what matter and they ARE
# committed -- stats.json is the artifact a report can be audited against.
corpus/**/*.pdf
# Measurements are generated artifacts and live in the package, under
# src/stopslop/data/<extractor>/. corpus/ is now purely a PDF cache.
corpus/
__pycache__/

# Build artifacts
*.egg-info/
dist/
build/
.venv*/
.pytest_cache/

# Belt and braces against the one accident that would hurt: a crawl pointed at
# the packaged tree. Only stats.json is ever meant to live there, and
# tests/test_packaging.py fails if a PDF does. This stops it reaching a commit.
src/stopslop/data/**/*.pdf

# Per-extractor remeasure logs
remeasure_*.log
