conda-forge-paths

Find which conda package provides a given file path

Bootstrap database

The main script is conda_forge_paths/path_to_artifacts_db.py. This is the last iteration of the other underscore-leading script. It has a few subcommands we can use to bootstrap and query the conda-forge path metadata.

To bootstrap, you need a local copy of the libcfgraph repository:

# Initialize the database; takes ~10-15min
$ python conda_forge_paths/path_to_artifacts_db.py bootstrap path/to/libcfgraph-repo/artifacts
# Enable full text search; takes ~2min
$ python conda_forge_paths/path_to_artifacts_db.py fts

This should create a ~9GB path_to_artifacts.db file. It compresses nicely with zstd:

$ ZSTD_NBTHREADS=4 ZSTD_CLEVEL=19 tar --zstd -cf path_to_artifacts.tar.zst path_to_artifacts.db

Queries

The script also has a couple of find-* subcommands:

# Find artifacts providing this exact file
$ python conda_forge_paths/path_to_artifacts_db.py find-artifacts 'bin/python'
# Find full paths given a path component
$ python conda_forge_paths/path_to_artifacts_db.py find-paths 'python'

The most recent artifact can be found with:

$ python conda_forge_paths/path_to_artifacts_db.py most-recent-artifact
# returns: cf/linux-64/llama-cpp-python-0.2.20-cuda112_habc0a91_2 1701711069.438 2023-12-04 17:31:09 UTC

This repo is also preconfigured for a datasette deployment, which offers the same query functionality:

$ datasette serve -i path_to_artifacts.db -m datasette.yml