Datasets:

kltng
/

tgaz-sqlite

license: other
license_name: chgis-tgaz
license_link: https://chgis.fairbank.fas.harvard.edu/
pretty_name: TGAZ (Temporal Gazetteer) SQLite
language:
  - zh
  - en
  - bo
tags:
  - history
  - gazetteer
  - china
  - historical-gis
  - sqlite
size_categories:
  - 10K<n<100K

TGAZ-SQLite

The Temporal Gazetteer (TGAZ) of the China Historical GIS (CHGIS) as one SQLite file. It has 82,117 historical placenames from 763 BCE to 1911 CE, with 245,042 spellings (simplified and traditional Chinese, pinyin, Tibetan), jurisdiction links over time, and historical notes.

Files

File	What it is
`tgaz_20260325.zip`	the database (`tgaz.db`, 228 MB unzipped, 62 MB zipped)
`latest.json`	the current release: file name, SHA-256 checksums, download URL

Download

The easiest way is the setup command of the tgaz-sqlite skill:

python3 scripts/query_tgaz.py setup     # downloads, checks the checksum, unzips

It puts the file at ~/.local/share/sino-authorities/tgaz.db. By hand:

curl -L -o tgaz.zip https://huggingface.co/datasets/kltng/tgaz-sqlite/resolve/main/tgaz_20260325.zip
unzip tgaz.zip      # gives tgaz.db

Source and credit

Data: China Historical GIS (Harvard University and Fudan University), with records from the Tibetan Buddhist Resource Center (3,587) and the Historical Gazetteer of Russia (761).
Converted from the MySQL dump in fccs-dci/containerized_tgaz with the scripts in kltng/tgaz-sqlite. The live API is at https://chgis.hudci.org/tgw/.
The upstream data has no explicit licence file. Credit CHGIS when you use it, and check the CHGIS site for terms of use. The conversion is by Kwok-leong Tang.

Known data problems

About 6,000 rows (mostly POLYGON records, hvd_90xxx IDs) have longitude 0.00000. On a map they land in Africa. The skill's --good-coords flag drops them.
251 rows end before they begin (e.g. hvd_112536).
mv_pn_srch.parent_vn is the earliest parent, not the parent in a given year. Use part_of for a given year.