Add support for Albania

[bk86a]

Country

Albania (AL)

Request Type

New country support

Description

Add support for Albanian postal codes (4-digit, no prefix, introduced 2006) mapping to NUTS-2024 codes. Albania is in NUTS-2024 as a candidate-country statistical region with the full hierarchy: 1 SR-1 (AL), 3 SR-2 regions (North/Central/South), 12 SR-3 qarqe (counties), and 61 LAU-1 bashki (municipalities). However, like Montenegro, GISCO does not currently publish a TERCET flat file for Albania, so a TERCET-only implementation would yield zero exact matches. A practical implementation can ingest GeoNames' AL postal codes (494 rows covering 31 of 61 bashki) via the existing extra_sources mechanism with a 12-row qarku-name → AL0xx lookup table.

See Research analysis below for the full data-source survey, format details, and the open questions before implementation can start.

Example Postal Codes

1001    Tirana (central)
1017    Tirana (Tirana 17)
3501    Korçë
9700    Sarandë (southern coast)
1700    (special: Transit)
1800    (special: EMS)

All examples 4 digits, no internal separator. No AL- prefix used by Posta Shqiptare.

Data Sources

• Eurostat — GISCO postal codes notice — same "missing" note as for Montenegro
• Eurostat — NUTS correspondence tables — Albania SR-1/2/3 hierarchy with AL0xx codes for qarqe
• INSTAT — General and regional statistics — Albanian national statistics on the 12 qarqe / 61 bashki
• GeoNames AL.zip — 494 rows with lat/lng + qarku name (CC-BY 4.0); covers all 12 qarqe and 31 of 61 bashki
• Posta Shqiptare — official lookup UI (no bulk export)
• Google libaddressinput / chromium-i18n — format \d{4}, examples 1001, 1017, 3501
• UPU AL addressing PDF

---

Research analysis (2026-04-29)

1. Postal code format (Kodi Postar)

• Format: 4 digits, no letters, no separator. Regex ^\d{4}$.
• Adopted: Council of Ministers Decision No. 676, 4 October 2006; rolled out from late 2006, operational from 2007.
• Structure: First 2 digits = district-level branch; last 2 digits = post office serving a bashki (municipality). Range observed 1000 (Tirana) to 9700 (Sarandë). Special codes 1700 (Transit) and 1800 (EMS).
• Variations: No AL- or SH- prefix is used by Posta Shqiptare or in Wikipedia/Grokipedia documentation. Nonetheless, an optional AL[\s\-.]* capture group should be allowed defensively, mirroring the existing patterns for MK/RS.
• Suggested regex (consistent with project style):

"AL": {
  "regex": "^(?:AL[\\s\\-–—.]*)?([0-9]{4})$",
  "example": "1001, AL-1001, 9700, 3501",
  "expected_digits": 4
}

2. Country prefix / addressing convention

• Posta Shqiptare (eposta.postashqiptare.al/postalcodes) lists raw 4-digit codes without prefix.
• UPU addressing template places the 4-digit postcode on its own line above the locality name (Albania UPU sheet albEn.pdf).
• Google libaddressinput (https://chromium-i18n.appspot.com/ssl-address/data/AL) returns:

{"id":"data/AL","key":"AL","name":"ALBANIA","fmt":"%N%n%O%n%A%n%Z%n%C","zip":"\\d{4}","zipex":"1001,1017,3501"}

No require, no sub_keys — Google does not enforce subdivision selection for AL.

• GeoNames uses bare 4-digit codes (no prefix).

3. GISCO TERCET availability

• The TERCET browser at https://gisco-services.ec.europa.eu/tercet/flat-files is a JavaScript SPA; the directory itself returns the SPA index (855 bytes HTML) for both missing files and unrelated paths.
• Probing the canonical filename pattern used by _guess_zip_urls_for_country() for AL across every plausible variant:
  • pc{2020..2025}_AL_NUTS-2024_v{1.0..4.0}.zip → all return the SPA fallback (Content-Type text/html, 855 B).
  • pc{2020,2021}_AL_NUTS-2021_v1.0.zip → also SPA fallback.
• Control: pc2020_MK_NUTS-2021_v1.0.zip returns a real ZIP (Content-Type application/zip, 600 B) confirming the probe distinguishes "exists" vs "missing".

Conclusion: Eurostat GISCO does not currently publish a TERCET postal-code-to-NUTS file for AL under NUTS-2024 or NUTS-2021. AL is treated like Montenegro/Bosnia in TERCET — absent.

4. NUTS / statistical regions

Albania is in NUTS-2024 as Statistical Regions:

• AL (SR1): 1 unit (whole country)
• Rajoneve statistikore (SR2): 3 units — North (Dibër, Durrës, Kukës, Lezhë, Shkodër), Central (Elbasan, Tiranë), South (Berat, Fier, Gjirokastër, Korçë, Vlorë)
• Qarqe (SR3): 12 units (administrative counties)
• Bashki (LAU): 61 units

Authoritative file: Correspondence-table-2024-NUTS-SR-EN-DE-FR.xlsx on ec.europa.eu/eurostat/web/nuts/correspondence-tables. Codes follow the AL0…/AL01…/AL011… convention used for other candidates.

5. Alternative data sources

• GeoNames AL.zip (download, 8.7 KB): 494 rows, tab-delimited, covers all 12 qarqe and 31 of 61 bashki with lat/lng. Includes county name (col 4) directly mappable to AL0-x NUTS3. CC-BY 4.0.
• Google libaddressinput: format/regex only, no postcode list.
• Posta Shqiptare lookup UI exists but no published CSV/API; would require scraping.
• Spotzi / GeoPostcodes / postcodebase.com: commercial datasets for 4-digit AL boundaries — not free.
• OSM: addr:postcode coverage exists but is sparse and inconsistent.

6. Architecture fit

Adding AL to app/settings.json countries and app/postal_patterns.json is mechanically trivial, but the existing GISCO loader will silently fail discovery (Strategy 1) and exhaust Strategy 2 guesses for AL. Two viable paths:

• A. GISCO-only (current architecture): add AL to countries, accept zero entries until Eurostat publishes a TERCET AL file. Net result: pattern validation works, lookups return not_found. Aligned with how the project handles missing data.
• B. Parallel loader (like UK/NSPL): ingest GeoNames AL.zip via extra_source_urls mechanism with a new transformer mapping qarku-name → AL0xx NUTS3 code (12-row static map). Reuses cache/DB/overwrite plumbing — no new top-level loader needed.

Option B is the lowest-effort productive path and matches the "extra source" hook already in the loader.

7. Open questions

1. Coverage threshold: is 494 GeoNames rows / 31 of 61 bashki acceptable, or should AL wait for Eurostat to publish TERCET?
2. NUTS3 code mapping authority: which AL NUTS-2024 code list should be the source of truth — the Eurostat correspondence XLSX, or the labels INSTAT publishes? (Names need normalisation: GeoNames uses Qarku i Beratit, NUTS uses Berat.)
3. Confidence tier: GeoNames-derived rows should likely map to medium or low in confidence_map (settings.json) rather than the high tier reserved for GISCO-validated codes.
4. Prefix handling: should the regex also tolerate SH- (an occasionally-seen colloquial prefix from "Shqipëria") or stay strict to AL-? like Google/UPU?

Source URLs

• https://en.wikipedia.org/wiki/Postal_codes_in_Albania
• https://eposta.postashqiptare.al/postalcodes
• https://www.upu.int/UPU/media/upu/PostalEntitiesFiles/addressingUnit/albEn.pdf
• https://chromium-i18n.appspot.com/ssl-address/data/AL
• https://gisco-services.ec.europa.eu/tercet/flat-files
• https://ec.europa.eu/eurostat/web/nuts/correspondence-tables
• https://www.instat.gov.al/en/themes/general-and-regional-statistics/regions/
• https://download.geonames.org/export/zip/AL.zip
• https://blog.opencagedata.com/post/nuts-eu-candidate-countries