United Kingdom (GB) build pipeline

Self-contained — runs entirely from this repo. All paths are relative to this scripts/ folder; sources/ and data/ are siblings under datasets/gb/ and are gitignored (regenerable).

Restored 2026-08-25. These four scripts previously lived at repo root as uk/ and were deleted in commit 35a9565 ("Remove uk/ folder") with no reason recorded; the built dataset survived only as an untracked local copy. Recovered from git history (commit 1c247b5) and re-verified end to end against freshly downloaded sources — same row count, same coverage stats as the original build. Relocated here to match the current datasets/{cc}/ convention (the old uk/ layout predates it).

Run order

python3 parse_codepoint.py         # 1. OS Code-Point Open -> ../data/gb_base.csv (BNG -> WGS84)
python3 enrich_uk.py               # 2. join ONS name lookups -> ../data/gb.csv
python3 fill_place_name.py         # 3. nearest GeoNames town -> place_name (in place)
python3 data_integrity_checker.py  # 4. gate; writes ../data/consolidated_stats.json + sample

No fetch_source.py / package.py / publish_r2.py yet (unlike fr/sg) — sources are currently fetched by hand into ../sources/ and there is no packaging or R2-publish step. That is the gap left to close before this matches the fr/sg convention fully.

Inputs (`../sources/`, not committed)

File	From	Notes
`../../../codepo_gb/Data/CSV/*.csv` (120 files)	OS Code-Point Open	read directly, not copied into `sources/` — see `parse_codepoint.py`
`ons_lad.csv`	ONS — Local Authority Districts (April 2025) Names and Codes in the UK (V2)	columns `LAD25CD`/`LAD25NM`
`ons_county.csv`	ONS — Counties (December 2025) Names and Codes in EN	columns `CTY25CD`/`CTY25NM`
`ons_ward_2022.csv`	ONS — Wards (December 2022) Names and Codes in the UK
`ons_ward_2023.csv`	ONS — Wards (December 2023) Names and Codes in the UK
`ons_ward.csv`	ONS — Wards (May 2024) Names and Codes in the UK	"base" vintage
`ons_ward_2025.csv`	ONS — Wards (May 2025) Names and Codes in the UK
`GB.txt` / `GB.zip`	GeoNames GB export	`https://download.geonames.org/export/dump/GB.zip`

Ward vintages are unioned (newest wins on code collisions) so boundary-change drift in Code-Point Open's ward codes still resolves — see enrich_uk.py::load_wards().

ONS files carry a UTF-8 BOM as downloaded — strip it before running enrich_uk.py, which opens them as plain utf-8 and would otherwise get LAD25CD etc. as the first column name and silently fail every lookup on that column.

Source and licence

Three sources, all open, all commercial-resale-permitted, none is OSM/Nominatim — this build does not touch the ODbL/Nominatim gate that blocks KWT/SGP/JEY coordinate builds.

Source	Licence	Confirmed from
OS Code-Point Open	OGL v3.0	`codepo_gb/Doc/licence.txt` — the licence file shipped inside the actual download, not a web claim
ONS Open Geography (LAD/Ward/County lookups)	OGL v3.0	ons.gov.uk/methodology/geography/licences
GeoNames GB	CC BY 4.0	geonames.org/export

See ../.licensing-notes.md for the full attribution text and the compound-attribution correction (OS's licence.txt names three copyright holders — OS, Royal Mail, and National Statistics — not two; an earlier draft README here missed Royal Mail).

Verification (2026-08-25)

1,747,841 rows — matches the original (pre-deletion) build exactly.
admin_name2/admin_name3 (district/ward): 99.9% fill — matches original.
admin_county_name (two-tier England county): 31.9% fill — matches original.
place_name: 100.0% fill via GeoNames nearest-place join.
Integrity check: PASS, 0 duplicate postal codes, 865 rows with no coordinate (OS positional-quality flag 90) — matches original.

Known characteristics (not defects)

Great Britain only — Code-Point Open does not cover Northern Ireland.
865 postcodes (positional quality 90) have blank latitude/longitude/place_name.
admin_county_name/admin_county_code are populated only for two-tier England counties (31.9% of rows) — unitary authorities and the rest of the UK have no county tier.