quasar7/pre1900-catalogue

Pre-1900 English & German Catalogue

Universal catalogue of pre-1900 (or undated) English and German works: 2,344,644 works, 2,047,100 with full text (869.4GB).

• Browse & read: https://quasar7-pre1900-catalogue-site.static.hf.space (serverless — searches and reads run in your browser against this repo)
• Query here: use the SQL Console on this page (the catalogue config is the works table)

Layout

path	contents
site/catalog_flat.parquet	the catalogue: one row per work (title, author, year, language, quality tier, source, status, chars)
site/text_locator_all.parquet	work_id → which text file holds its full text
text/…	full texts, one row per book (work_qid, corpus, ref, text); text/fast/… are small re-cut files for fast single-book reads
corpus_db8/part_00..20.bin	the master DuckDB database in 8GiB pieces — cat part_*.bin > corpus.duckdb (175,734,009,856 bytes, SHA256-verified)
aux/	author database, HathiTrust catalogue data, release backups
site/catalog_app.py	the original server app (runs anywhere with the above files)

Read one book from anywhere

-- 1. find the work and its text file
SELECT l.shard FROM 'hf://datasets/quasar7/pre1900-catalogue/site/text_locator_all.parquet' l
WHERE l.work_id = 'ht:uc1.b4534619';
-- 2. fetch the text (replace /data/text/ with .../resolve/main/text/)
SELECT text FROM '<that file URL>' WHERE work_qid = 'ht:uc1.b4534619';

Texts carry page structure as whole-line markers ([[p N · folio]], [[vol …]]); strip them with the regex ^\[\[(?:p|vol)\b[^\]\n]*\]\]$\n?.

Companion repo (raw preservation: HathiTrust page-zips, METS, manifests): quasar7/pre1900-academics-corpus