---
license: cc0-1.0
pretty_name: fatcat files — hash-keyed scholarly file catalog
tags: [file-hashes, bibliography, fatcat, provenance]
configs:
- config_name: default
  data_files:
  - split: train
    path: data/*.parquet
- config_name: maps
  data_files:
  - split: sha1
    path: maps/sha1.parquet
  - split: md5
    path: maps/md5.parquet
---

One row per (file x release) from the **final fatcat bulk export**
(2024-02-18, Internet Archive; fatcat.wiki is offline). Sorted by
`sha256`, so hash point-lookups read one row group:

    SELECT title, doi, wayback_url
    FROM 'hf://datasets/david-ar/fatcat-files/data/*.parquet'
    WHERE sha256 = '...';

`maps/sha1.parquet` and `maps/md5.parquet` (sorted by key) translate
weak digests to sha256 first. `wayback_url` points at the archived
bytes that actually carry the hash; DOI/arXiv/PMID identify the work;
fatcat idents remain as join keys into the raw dumps.

Built by [hashdex](https://github.com/davidar/hashdex)
`tools/fatcat_dataset_job.py`. Data: CC0 (fatcat's license).