Published

Germany (DE) build pipeline

DataHub
DataHub
0

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

Germany (DE) build pipeline

Rebuilt from the original GeoNames DE export rather than the copy already sitting in R2 (that copy was produced by scripts/geonames/geonames_parse_to_R2.py with pandas and no dtype=str, which float-coerces numeric-looking admin codes and strips punctuation from compound names — same defect class documented for PT).

Run order

python3 fetch_source.py        # GeoNames DE.txt
python3 build_base.py
python3 recover_alt_names.py
python3 resolve_duplicates.py
python3 integrity_checker.py
python3 package.py
python3 publish_r2.py          # written this session — did not exist before. dry run by default.

Sources

  • GeoNames DE postal export — CC BY 4.0

Verification (2026-08-19)

Reproducible, integrity check PASS (0 duplicates by its own key), not byte-identical — 23,297 rows rebuilt vs 23,293 live on R2, a documented and understood +4-row gap, not an unexplained mismatch:

  • The rebuild fixes two real bugs present in the live R2 copy: float-coerced admin codes losing leading zeros, and dropped hyphens in compound place names (e.g. "Rehburg-Loccum", "Ostprignitz-Ruppin"). This is a quality improvement, not a regression.
  • resolve_duplicates.py's hand-verified rule table (sources/duplicate_resolutions.csv, 4 rows) is keyed to the old, buggy place-name text from the live copy. Once the two bugs above are fixed upstream of it, none of its 4 rules match anymore, so it resolves 0 groups instead of 4.
  • Before a real publish, update duplicate_resolutions.csv to match the corrected place-name text if exact row-count parity with the current live file matters; until then, publish_r2.py's dry-run plan will show a +4 row diff and should not be run with --apply without that follow-up.