[jkrenge]

Summary

Expand postal geocoding from ten to 29 countries for return-network location analysis, consolidating source PRs #95–123 into one review. Refresh ten datasets and add nineteen. Retain source/license documentation, generators, tests, and all source commits.

country + postcode
  → country-specific normalization
  → compact country JSON, coordinates rounded to four significant digits
  → [longitude, latitude] strings, or undefined

All 603,127 geocode records keep their postal codes and row order. Coordinate values are rounded to four significant digits (not four decimal places); unnecessary trailing zeros and JSON indentation are removed. Runtime schemas and return types are unchanged. Run npm run geocode:compact after any country generator; tests enforce the distributed precision and compact encoding. Source fixtures and generator calculations retain their original precision.

Country/source manifest and limits. Brazil, India, Russia, and Turkey remain excluded because source-quality blockers are unresolved.

Evidence

• All 176 combined tests pass, including rounding boundaries, invalid coordinates, idempotence, and every distributed file's encoding/precision.
• Lint, formatting, data validation, and diff checks pass. GitHub CI passes for static checks and Node.js 24/26. Validation retains only the existing AUT/BEL region-name warnings.
• Before compaction, the integration audit verified all 29 source heads, unchanged country assets, and 604,260 normalization inputs against the source branches. Compaction preserves postal codes, row order, and all non-coordinate fields across all 603,127 records.
• Finland and UK generators use Node.js built-in modules. Their CLI outputs reproduce all 6,020 existing compact coordinates from the downloaded source data. Tests cover duplicate weighting, rounding, invalid source rows, polygon holes, and disconnected areas; no distributed data changed in this conversion.
• Three cross-country tests now use the refreshed German dataset; country lookup expectations reflect four significant digits.
• Measured with npm pack --ignore-scripts --json, same packaging rules and no publication. Decimal MB; package figures exclude separately installed dependencies:

Size	Base: 10 countries	Expanded: 29 countries	Compact: 29 countries
Geocode JSON	18.15 MB	51.49 MB	34.20 MB
npm download	2.78 MB	5.03 MB	3.76 MB
Unpacked package	26.56 MB	60.10 MB	42.82 MB

Compact npm download is 1.27 MB (25.3%) smaller than the expanded version, and 0.98 MB (35.2%) larger than base. Unpacked increase over base: 16.26 MB (61.2%).

Merge Danger

Door: two-way

Blast Radius: geocoding

Four-significant-digit rounding moves stored points by a median 0.424 km, 95th percentile 3.884 km, and maximum 5.547 km. These are displacements from the previous stored points, not source-accuracy estimates. Two rounded endpoints can compound distance error.

Source replacement also changes legacy lookup coverage, particularly France, Belgium, Germany, and US Pacific codes. Granularity varies by country. Runtime lookup remains a linear scan; production performance was not measured.

No master merge, package publication, deployment, or production tracking-data change is included.

[ilin-andrey]

did you just unlock a new level?

[ilin-andrey]

why python? is not it a js package?

[ilin-andrey]

same question, it is a nodejs repo, why is it in python?