Published

Canada (CA) build pipeline

DataHub
DataHub
0

Postal Codes Dataset for Canada, CA including name of the city, town, or place, various administrative divisions and alternative city names.

Canada (CA) build pipeline

Clean-room rebuild — GeoNames + StatCan open cartographic boundaries only. Must never read, join against, or validate itself with StatCan's Postal Code Conversion File (PCCF). PCCF is licence-gated (Canada Post / StatCan 92-154-X, via DLI) and its End-Use Licence forbids redistribution, deriving a commercial product from it, and publishing more than a 1% extract. Clause 3 bites on derivation, so PCCF may not sit open beside the build as a "just checking" reference either — a spatial join tuned until it agrees with PCCF is a PCCF derivative. fetch_sources.py's own docstring states this; integrity_checker.py and package.py hard-fail the build if admin_name3/admin_code3 (PCCF's old signature field) is ever non-empty.

An earlier, PCCF-dependent build (scripts/ca/pccf_parse.py, build_cduid.py, admin2_enrich.py — still committed at the repo root, not here) was superseded by this one on 2026-08-10/11. This is the pipeline that is actually live, confirmed by diffing the object sizes of the currently-published ca/ca.csv and ca/consolidated/latest/ca.csv in R2 against this pipeline's local output (both match exactly; see BUILD_LOG.md, sourced alongside this folder, for the fetch SHA-256s and join-correctness checks against 6 known cities).

One separate, real problem found 2026-08-19, not fixed by this pipeline change: the raw, licence-gated PCCF file itself (ca/pccf/canada_pccf.csv, ~65.7MB) is still sitting in the same R2 bucket this paid product ships from, dated before the clean-room rebuild. publish_r2.py has an --include-pccf flag built specifically to delete it — it was apparently never invoked. This needs a human decision (the deletion is irreversible), not an automated fix.

Run order

python3 fetch_sources.py       # GeoNames CA_full + StatCan CD/CSD cartographic boundaries only
python3 spatial_join_cd.py     # attach Census Division
python3 spatial_join_csd.py    # attach Census Subdivision
python3 build_base.py
python3 integrity_checker.py   # hard-fails if PCCF signature detected
python3 package.py
python3 publish_r2.py          # dry run by default; --apply to publish

Sources

  • GeoNames CA_full — CC BY 4.0
  • StatCan CD/CSD 2021 cartographic boundary files — Open Government Licence – Canada

Verification (2026-08-19)

899,779 rows. admin_name2 99.98% filled (293/293 census divisions), admin_name3 99.96% filled (3,724/5,161 census subdivisions) — more complete than the superseded PCCF build (94.2%/79.3%). R2 object byte-size match confirmed for both ca/ca.csv and ca/consolidated/latest/ca.csv; a full SHA-256 download comparison was started but not completed in the same session — size match across two independently timestamped uploads is strong evidence, not yet cryptographic proof.