euredact 0.3.0 → 0.3.2

package/README.md

@@ -43,7 +43,11 @@ console.log(result.detections);
 - **Custom patterns:** register your own regex patterns for domain-specific PII
 - **Checksum validation:** IBAN mod-97, Luhn (credit cards/IMEI), and 30+
   country-specific validators
-- **Priority-aware deduplication:** validated (checksum) > custom > regex-only
+- **Priority-aware deduplication:** validated (checksum, corroborated by the
+  document's country) > custom > regex-only
+- **Country self-detection:** infers a document's countries from the entities
+  that carry one, so an ambiguous value resolves without the caller naming a
+  country — and `countries` never gates what is looked for
 - **Context-aware:** keyword proximity checks and structural detection (JSON field
   names, CSV headers) for ambiguous patterns
 - **Zero runtime dependencies**
@@ -73,10 +77,16 @@ Main entry point. Detects and redacts PII in the given text.
 
 ```ts
 interface RedactOptions {
-  countries?: string[] | null; // Country codes (e.g. ["NL", "BE"]) — null loads all
-  referentialIntegrity?: boolean;      // Replace with consistent labels (default: false)
-  detectDates?: boolean;       // Include DOB/date-of-death detections (default: false)
-  cache?: boolean;             // Enable result caching (default: true)
+  countries?: string[] | null;    // Scope. Flags anything attributed elsewhere as
+                                  // outOfScope. Never gates what is looked for.
+  countryHint?: string[] | null;  // A prior only — resolves ambiguity without
+                                  // narrowing scope or flagging anything
+  context?: DocumentContext | null; // Share country evidence across the chunks
+                                  // of one document (see "Chunked documents")
+  chunkOffset?: number;           // Where this chunk starts in the document
+  referentialIntegrity?: boolean; // Replace with consistent labels (default: false)
+  detectDates?: boolean;          // Include DOB/date-of-death detections (default: false)
+  cache?: boolean;                // Enable result caching (default: true)
 }
 ```
 
@@ -158,6 +168,28 @@ interface Detection {
   source: DetectionSource;         // "rules" or "cloud"
   country: string | null;          // ISO code or null for shared/custom patterns
   confidence: string;              // Confidence level
+  countryConfidence?: number;      // How strongly the document supports `country`,
+                                   // in [0, 1]. 0 means the attribution rests on
+                                   // a checksum alone.
+  outOfScope?: boolean;            // Attributed outside the declared `countries`.
+                                   // Flagged, never dropped.
+}
+```
+
+#### `RedactResult`
+
+```ts
+interface RedactResult {
+  redactedText: string;
+  detections: Detection[];
+  source: string;
+  degraded: boolean;
+
+  // Country inference — see "Country-independent detection"
+  inferredCountries: Array<[string, number]>; // [country, confidence], strongest first
+  evidence: CountryEvidence[];                // every signal, with the span behind it
+  detectionMode: string;                      // "declared" if countries was passed,
+                                              // "inferred" otherwise
 }
 ```
 
@@ -191,20 +223,107 @@ redaction entirely, failing open with unredacted PII.
 
 ## Country-independent detection
 
-`countries` narrows which country-specific patterns run. It does **not** gate
-identifiers that carry their own evidence:
+**`countries` never gates detection.** Every pattern runs on every document,
+whatever you pass. The country you declare decides *how a match is labelled*,
+never *whether it is found*.
 
-| Type | Why it ignores `countries` |
-|---|---|
-| `BANK_ACCOUNT` (IBAN) | carries an ISO 3166 country code and a mod-97 checksum |
-| `PHONE` (with `+` prefix) | an E.164 country prefix is self-identifying |
-| `EMAIL`, `CREDIT_CARD`, `VIN`, `IMEI`, `IP_ADDRESS`, `UUID`, `SECRET` | no national form to gate on |
+This is the engine's central invariant, enforced by
+`src/__tests__/inference.ts`: no value of `countries` may change which spans are
+detected. A wrong or missing country cannot cause a miss — silent recall loss is
+invisible in testing and surfaces in a breach report, whereas a false positive
+is recoverable.
+
+It was not always so. `countries: ["BE"]` used to make a valid Dutch BSN vanish
+entirely, because the Dutch patterns were never run:
+
+```ts
+redact("Werknemer met BSN 111222333", { countries: ["BE"] });
+// before: 'Werknemer met BSN 111222333'   <- leaked
+// now:    'Werknemer met BSN [NATIONAL_ID]'
+```
+
+Entities found outside the countries you declared are **flagged, not dropped**:
+
+```ts
+const [det] = redact("BSN 111222333", { countries: ["BE"] }).detections;
+det.outOfScope; // true — detected, masked, and marked as outside your scope
+```
+
+So a Belgian IBAN in a document processed with `countries: ["AT"]` is still
+detected:
 
 ```ts
 redact("Rekening: BE68 5390 0754 7034", { countries: ["AT"] });
 // -> 'Rekening: [BANK_ACCOUNT]'
 ```
 
+### Which country a value belongs to
+
+Because every pattern runs, the same digits often match several countries'
+schemes. 36.6% of national-ID values in our corpus validate under more than one
+country's checksum, so the digits alone cannot decide it — the *document* does.
+
+The engine infers the document's countries from entities that carry their
+country in the string, then uses that to resolve the ambiguity:
+
+```ts
+redact("Bereikbaar op telefoon 0612345678, mail jan@test.nl");
+// inferredCountries: [["NL", 0.98]]   detections: [PHONE (NL), EMAIL]
+
+redact("Kontakt: 0612345678, e-mail jens@test.dk");
+// inferredCountries: [["DK", 0.98]]   detections: [NATIONAL_ID (DK), EMAIL]
+```
+
+Identical digits, different answer — `0612345678` is both a valid Dutch mobile
+number and a valid Danish CPR. Only the surrounding document distinguishes them.
+
+Every inference is auditable: `result.evidence` lists each signal, its weight,
+and the span that produced it, in document order.
+
+| Signal | Weight (log-odds) | Measured reliability |
+|---|---:|---|
+| `e164Prefix` | 4.00 (capped) | 41,402 / 41,402 |
+| `bicCountry` | 4.00 (capped) | 2,588 / 2,588 |
+| `emailTld` | 4.00 (capped) | 97,865 / 98,949 |
+| `vatPrefix` | 2.84 | 19,022 / 20,136 |
+| `ibanPrefix` | 1.94 | 110,572 / 126,428 |
+
+Weights are derived from the corpus, not chosen by hand, and are kept identical
+to the Python SDK's. The IBAN prefix being weakest is real: a Belgian IBAN in a
+Dutch invoice is ordinary, so an account's country is only weak evidence about
+the document's.
+
+Confidences are per-country and do **not** sum to 1 — document countries are not
+mutually exclusive. A Belgian supplier invoicing a German customer is genuinely
+both.
+
+## Chunked documents
+
+A long document is usually redacted in pieces. Each piece is scanned
+independently, so a chunk carrying no country signal of its own infers nothing —
+even when page 1 identified the document beyond doubt.
+
+```ts
+import { DocumentContext } from "euredact";
+
+const ctx = new DocumentContext();
+let offset = 0;
+for (const page of pages) {
+  const result = redact(page, { context: ctx, chunkOffset: offset });
+  offset += page.length;
+}
+
+// page 1: "Factuur — IBAN NL91ABNA0417164300, info@jansen.nl"
+// page 7: "Telefoon 0612345678"  -> PHONE (NL), not NATIONAL_ID (DK)
+```
+
+`chunkOffset` rebases spans recorded in the context so they point into the whole
+document; returned detections stay relative to the chunk you passed in. Caching
+is disabled automatically while a context is in use, because the result then
+depends on evidence the text alone does not determine.
+
+Reuse a context only for chunks of the **same** document.
+
 ## BIC detection
 
 BIC is the only bank identifier here with **no check digit**, and characters
@@ -332,18 +451,35 @@ redact("identifier: xK9mPqR7vLnW2bFjY8cGhT4sDfAeU6iO").redactedText;
 
 ## Country Hints
 
-When you know which countries appear in your data, pass them explicitly:
+Two options tell the engine about country, and neither restricts what is looked
+for:
+
+| Option | Meaning |
+|---|---|
+| `countries` | **Scope.** Resolves ambiguity, and flags anything attributed elsewhere as `outOfScope`. |
+| `countryHint` | **Prior only.** Resolves ambiguity without narrowing scope or flagging anything. |
+
+```ts
+// You know this batch is Swedish, but don't want foreign PII marked out of scope:
+redact(text, { countryHint: ["SE"] });
+
+// You want anything non-Swedish flagged for review:
+redact(text, { countries: ["SE"] });
+```
+
+Declaring a country helps where a value is genuinely ambiguous and the document
+carries no other signal:
 
 ```ts
-const result = redact(text, { countries: ["NL", "BE"] });
+redact("Telefon: 0708787668", { countryHint: ["SE"] }).detections[0];
+// PHONE / SE — without the hint this is a valid Danish CPR and nothing says otherwise
 ```
 
-This restricts pattern matching to Dutch and Belgian rules (plus shared patterns
-and custom patterns). Benefits:
+`result.detectionMode` reports which happened: `"declared"` if you passed
+`countries`, `"inferred"` otherwise.
 
-- **Fewer false positives.** A 9-digit number that passes the Dutch BSN checksum
-  won't also be tested against unrelated country patterns.
-- **Faster.** Fewer patterns to compile and scan.
+Passing neither is safe, and is the right default for mixed-origin data: the
+engine infers what it can and reports it in `result.inferredCountries`.
 
 ## Referential Integrity
 
@@ -372,13 +508,17 @@ Input text
     |
     v
 [Pass 2a: Validation] -- Checksum validators (mod-97, Luhn, entropy, ...)
-    |                     Failed-validation spans become suppression zones
+    |                     Failed spans are recorded, per entity type
+    v
+[Evidence]     -- Which countries does this document belong to? From IBAN
+    |              prefixes, +CC codes, VAT prefixes, BIC, email ccTLDs
     v
 [Pass 2b: Suppression] -- Remove false positives (currency amounts, units,
-    |                      references, overlapping matches in suppression zones)
+    |                      references). Failed checksums demote same-type
+    |                      matches rather than deleting them
     v
-[Deduplication] -- Priority-aware: validated > custom > regex-only
-    |               Within same tier, longer span wins
+[Deduplication] -- Priority-aware, country-evidence-weighted
+    |               Longer span outranks declared country
     v
 [Replacement] -- Right-to-left substitution with [ENTITY_TYPE] labels
     |              or labels
@@ -389,16 +529,60 @@ RedactResult
 ### Suppression Zones
 
 When a regex matches a pattern that has a checksum validator but the checksum
-fails, the matched span becomes a "suppression zone." Any purely regex-based
-detection fully contained within that zone is suppressed as a false positive.
+fails, the span is recorded. A validator-less match of **the same entity type**
+contained in it is *demoted* below every other candidate — not deleted. A failed
+checksum is evidence against that type, not against the span, and demotion can
+never silence a detection the way deletion could.
 
 ### Deduplication Priority
 
-1. **Validated patterns** (checksum passes) -- highest priority
-2. **Custom patterns** (registered via `addCustomPattern()`)
-3. **Regex-only patterns** (no validator) -- lowest priority
+When multiple patterns match overlapping spans, the engine resolves conflicts
+using a priority system:
+
+| Tier | What |
+|---:|---|
+| 3 | **Validated** — a checksum validator passes *and* the document corroborates its country |
+| 2 | **Custom patterns** registered via `addCustomPattern()` |
+| 1 | **Regex-only**, and validated patterns whose country the document does not corroborate |
+| 0 | **Postal codes** — a bare digit run, the weakest evidence in the engine |
+| -1 | **Demoted** — a validator-less match inside a failed checksum of *its own type* |
+
+Within a tier, ranking is: longer span, then stronger country evidence, then
+whether the country was declared. **Span length outranks country** deliberately:
+preferring the declared country over the longest match truncates entities — with
+`countries: ["BE"]` the Belgian phone pattern claimed 11 of the 14 characters of
+`06 12 34 56 78` and left three digits exposed. Country can change *which*
+country is attributed, never *what* is masked.
+
+Two tiers are less obvious than they look:
+
+- **A passing checksum does not automatically win.** A weak checksum fits by
+  luck — a mod-11 scheme accepts a random number about one time in eleven — so a
+  validated candidate from a country the document shows no trace of drops to
+  tier 1. Entities that carry their own country vouch for themselves (an IBAN
+  emits evidence for its own country), so a foreign IBAN in a domestic invoice
+  keeps tier 3.
+- **Nothing is deleted for failing a checksum.** A failed checksum demotes
+  rather than removes, and only candidates of *the same entity type*: it is
+  evidence against that type, not against the span. Deleting instead removed 454
+  detections across the corpus, of which 454 overlapped real labelled PII.
+
+## Performance
+
+Measured on one core, all 31 countries loaded, `detectDates: true`. Cost scales
+with document length, so the input size is stated rather than averaged away.
+
+| Input | Latency | Throughput |
+|---|---:|---:|
+| Short record (186 chars) | 155 µs | 6,462 docs/s |
+| Real document (3,424 chars) | 1.56 ms | 643 docs/s |
 
-Within the same tier, the longer span wins.
+No optional accelerator is needed or offered. The Python SDK ships an
+`[fast]` extra (RE2 / Aho-Corasick) because CPython's regex engine is the
+bottleneck there; V8's has literal prefilters that make it unnecessary here.
+Measured on the same 611 documents, this SDK runs about 3× faster than the
+accelerated Python path, so adding a native addon — and with it the loss of
+bundler, edge-runtime and Deno compatibility — would buy nothing.
 
 ## CommonJS
 

package/dist/cjs/index.js

@@ -1,6 +1,6 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.SEED_BIC6_PREFIXES = exports.getBicRegistry = exports.setBicRegistry = exports.COUNTRY_CONFIGS = exports.EuRedact = exports.LEGACY_TYPE_ALIASES = exports.DetectionSource = exports.EntityType = void 0;
+exports.SEED_BIC6_PREFIXES = exports.getBicRegistry = exports.setBicRegistry = exports.COUNTRY_CONFIGS = exports.EuRedact = exports.DocumentContext = exports.LEGACY_TYPE_ALIASES = exports.DetectionSource = exports.EntityType = void 0;
 exports.availableCountries = availableCountries;
 exports.redact = redact;
 exports.addCustomPattern = addCustomPattern;
@@ -10,6 +10,8 @@ var types_js_1 = require("./types.js");
 Object.defineProperty(exports, "EntityType", { enumerable: true, get: function () { return types_js_1.EntityType; } });
 Object.defineProperty(exports, "DetectionSource", { enumerable: true, get: function () { return types_js_1.DetectionSource; } });
 Object.defineProperty(exports, "LEGACY_TYPE_ALIASES", { enumerable: true, get: function () { return types_js_1.LEGACY_TYPE_ALIASES; } });
+var context_js_1 = require("./rules/context.js");
+Object.defineProperty(exports, "DocumentContext", { enumerable: true, get: function () { return context_js_1.DocumentContext; } });
 var sdk_js_1 = require("./sdk.js");
 Object.defineProperty(exports, "EuRedact", { enumerable: true, get: function () { return sdk_js_1.EuRedact; } });
 var index_js_1 = require("./rules/countries/index.js");

package/dist/cjs/rules/context.js

@@ -0,0 +1,50 @@
+"use strict";
+/**
+ * Country evidence shared across the chunks of one document.
+ *
+ * A long document is usually redacted in pieces. Each piece is scanned
+ * independently, so each infers its own country — and a chunk that happens to
+ * contain no IBAN, no `+CC` number and no ccTLD infers nothing at all, even
+ * when the first page identified the document beyond doubt:
+ *
+ *     page 1   "Factuur — IBAN NL91ABNA0417164300, info@jansen.nl"
+ *     page 7   "Telefoon 0612345678"        <- Dutch mobile, or Danish CPR?
+ *
+ * Scanned alone, page 7 has nothing to go on. Sharing a context across the two
+ * carries page 1's evidence forward, and the number resolves as a Dutch phone.
+ *
+ * Spans are rebased by `chunkOffset` on the way in, so evidence collected from
+ * a chunk points at offsets in the whole document rather than the chunk.
+ *
+ * A context influences **scoring only**. It can change which national scheme is
+ * attributed to a value. It can never change which spans are found, so a
+ * context carried over from the wrong document cannot cause a miss.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.DocumentContext = void 0;
+class DocumentContext {
+    // Deduped on the same key evidence.collect() uses, so re-running a chunk — a
+    // retry, an overlapping window — cannot let one entity vote twice and
+    // outweigh the rest of the document.
+    seen = new Set();
+    items = [];
+    /** Record evidence from a chunk starting at `chunkOffset`. */
+    add(evidence, chunkOffset = 0) {
+        for (const ev of evidence) {
+            const span = [ev.span[0] + chunkOffset, ev.span[1] + chunkOffset];
+            const key = `${span[0]}:${span[1]}:${ev.source}`;
+            if (this.seen.has(key))
+                continue;
+            this.seen.add(key);
+            this.items.push(chunkOffset === 0 ? ev : { ...ev, span });
+        }
+    }
+    /** Everything recorded so far, spans rebased to the whole document. */
+    evidence() {
+        return [...this.items];
+    }
+    get size() {
+        return this.items.length;
+    }
+}
+exports.DocumentContext = DocumentContext;

package/dist/cjs/rules/countries/index.js

@@ -217,7 +217,13 @@ const DE = {
         p(types_js_1.EntityType.PHONE, String.raw `\b0[1-9]\d{1,4}[\s/\-]?\d{3,8}\b`),
         p(types_js_1.EntityType.PHONE, String.raw `\+49\s?\d{2,5}[\s/\-]?\d{3,8}\b`),
         p(types_js_1.EntityType.PASSPORT, String.raw `\b[CFGHJK][0-9A-Z]{8,9}\b`, null, "", ["Reisepass", "passport", "Passnummer", "Reisepassnummer", "Reisepass Nummer", "Reisepass-Nr", "Pass Nr"], true),
-        p(types_js_1.EntityType.LICENSE_PLATE, String.raw `\b[A-ZÄÖÜ]{1,3}[\s\-]?[A-Z]{1,2}[\s\-]?\d{1,4}[EH]?\b`),
+        // The separator after the city code is mandatory. When it was optional, a
+        // contiguous letter run split across both groups and the hyphen was
+        // consumed as the *second* separator, so any "LETTERS-DIGITS" token
+        // matched: ICD-10 parsed as city "IC" + letters "D" + "-" + "10", and
+        // ISO-9001, COVID-19, RFC-2822, POL-2025 likewise. A real plate always
+        // separates the city code: "B-AB 1234", "M-XY 99".
+        p(types_js_1.EntityType.LICENSE_PLATE, String.raw `\b[A-ZÄÖÜ]{1,3}[\s\-][A-Z]{1,2}[\s\-]?\d{1,4}[EH]?\b`),
         p(types_js_1.EntityType.POSTAL_CODE, String.raw `\b\d{5}\b`, null, "", ["PLZ", "Postleitzahl", "postal code", "postcode", "Anschrift", "Adresse", "Wohnort", "Ort", "Stadt", "wohnt", "wohnhaft", "Straße", "Str.", "Weg", "Platz", "Allee", "Ring"], true),
         p(types_js_1.EntityType.CHAMBER_OF_COMMERCE, String.raw `\bHR[AB]\s?\d{4,6}\b`),
         p(types_js_1.EntityType.SSN, String.raw `\b\d{2}[\s]?\d{6}[\s]?[A-Z][\s]?\d{3}\b`, null, "", ["Rentenversicherung", "RV-Nummer", "Sozialversicherung", "SV-Nummer", "Versicherungsnummer", "Rentenversicherungsnummer", "Sozialversicherungsnummer"], true),

package/dist/cjs/rules/deDistricts.js

@@ -0,0 +1,92 @@
+"use strict";
+/**
+ * German vehicle registration district codes (Unterscheidungszeichen).
+ *
+ * The set is closed: codes are fixed by the Fahrzeug-Zulassungsverordnung and
+ * maintained by the Kraftfahrt-Bundesamt, changing only when a district is
+ * created, merged, or a historic code is reinstated. That makes membership a
+ * usable precision signal — unlike the open-ended blocklist of standards
+ * prefixes it replaces, which only ever catches acronyms someone thought of.
+ *
+ * Used as a **tier**, not a filter: code on this list -> emit; code absent but
+ * a plate cue nearby -> emit; neither -> reject. So a missing code costs
+ * recall only where there is also no cue. Some entries are also standards
+ * prefixes — DIN is Dinslaken as well as the standards body — which is why the
+ * standards guard is kept alongside this one.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.DE_DISTRICT_CODES = void 0;
+exports.DE_DISTRICT_CODES = new Set([
+    "A", "AA", "AB", "ABG", "AC", "AE", "AH", "AIB", "AIC", "AK", "AL", "ALF",
+    "ALS", "ALZ", "AM", "AN", "ANA", "ANG", "ANK", "AP", "APD", "AR", "ARN",
+    "ART", "AS", "ASD", "ASL", "ASZ", "AT", "AU", "AUR", "AW", "AZ", "AZE",
+    "AÖ", "B", "BA", "BAD", "BAR", "BB", "BBG", "BBL", "BC", "BCH", "BD",
+    "BE", "BED", "BEI", "BEL", "BER", "BF", "BG", "BGD", "BGL", "BH", "BI",
+    "BID", "BIN", "BIR", "BIT", "BIW", "BK", "BKS", "BL", "BLB", "BLK", "BM",
+    "BN", "BNA", "BO", "BOG", "BOH", "BOR", "BOT", "BP", "BR", "BRA", "BRB",
+    "BRG", "BRI", "BRK", "BRL", "BRV", "BS", "BSB", "BSK", "BT", "BTF", "BU",
+    "BUL", "BUR", "BW", "BWL", "BYL", "BZ", "BZA", "BÖ", "BÜD", "BÜS", "BÜZ",
+    "C", "CA", "CAS", "CB", "CE", "CHA", "CHL", "CLP", "CLZ", "CO", "COC",
+    "COE", "CR", "CT", "CUX", "CW", "D", "DA", "DAH", "DAN", "DAU", "DB",
+    "DBR", "DD", "DE", "DEG", "DEL", "DG", "DGF", "DH", "DI", "DIL", "DIN",
+    "DIZ", "DKB", "DL", "DLG", "DM", "DN", "DO", "DON", "DS", "DT", "DU",
+    "DUD", "DW", "DZ", "DÜW", "E", "EA", "EB", "EBE", "EBN", "EBS", "ECK",
+    "ED", "EE", "EF", "EG", "EH", "EHI", "EI", "EIC", "EIH", "EIL", "EIN",
+    "EIS", "EL", "EM", "EMD", "EMS", "EN", "ER", "ERB", "ERH", "ERK", "ES",
+    "ESA", "ESB", "ESW", "EU", "EUT", "EW", "F", "FAL", "FB", "FD", "FDB",
+    "FDS", "FEU", "FF", "FFB", "FG", "FH", "FI", "FKB", "FL", "FLÖ", "FN",
+    "FO", "FOR", "FR", "FRG", "FRI", "FRW", "FS", "FT", "FTL", "FW", "FZ",
+    "FÜ", "FÜS", "G", "GA", "GAN", "GAP", "GC", "GD", "GDB", "GE", "GEL",
+    "GEM", "GEO", "GER", "GF", "GG", "GHA", "GHC", "GI", "GK", "GL", "GLA",
+    "GM", "GMN", "GN", "GNT", "GOA", "GOH", "GP", "GR", "GRA", "GRH", "GRI",
+    "GRM", "GRS", "GRZ", "GS", "GT", "GTH", "GUB", "GUN", "GV", "GVM", "GW",
+    "GZ", "GÖ", "GÜ", "H", "HA", "HAB", "HAL", "HAM", "HAS", "HB", "HBN",
+    "HBS", "HC", "HCH", "HD", "HDH", "HDL", "HE", "HEB", "HEF", "HEI", "HEL",
+    "HER", "HET", "HF", "HG", "HGN", "HGW", "HH", "HHM", "HI", "HIG", "HIP",
+    "HK", "HL", "HM", "HMU", "HN", "HO", "HOG", "HOH", "HOL", "HOM", "HOR",
+    "HOT", "HP", "HR", "HRO", "HS", "HSK", "HST", "HU", "HUS", "HV", "HVL",
+    "HW", "HWI", "HX", "HY", "HZ", "HÖS", "HÜN", "IGB", "IK", "IL", "ILL",
+    "IN", "IS", "IZ", "J", "JB", "JE", "JEV", "JL", "JÜL", "K", "KA", "KAR",
+    "KB", "KC", "KE", "KEH", "KEL", "KEM", "KF", "KG", "KH", "KI", "KIB",
+    "KK", "KL", "KLE", "KLT", "KLZ", "KM", "KN", "KO", "KOZ", "KR", "KRU",
+    "KS", "KT", "KU", "KUS", "KW", "KY", "KYF", "KÖN", "KÖT", "KÜN", "L",
+    "LA", "LAN", "LAT", "LAU", "LB", "LBS", "LBZ", "LC", "LD", "LDK", "LDS",
+    "LE", "LEO", "LER", "LEV", "LF", "LG", "LH", "LI", "LIB", "LIF", "LIN",
+    "LIP", "LK", "LL", "LM", "LN", "LOH", "LOS", "LP", "LR", "LS", "LSA",
+    "LSN", "LSZ", "LU", "LUK", "LWL", "LÖ", "LÖB", "LÜD", "LÜN", "M", "MA",
+    "MAB", "MAI", "MAK", "MAL", "MAR", "MB", "MC", "MD", "ME", "MED", "MEG",
+    "MEI", "MEK", "MEL", "MEP", "MER", "MES", "MET", "MG", "MGH", "MGN", "MH",
+    "MHL", "MI", "MIL", "MK", "MKK", "ML", "MM", "MN", "MO", "MOD", "MOL",
+    "MON", "MOS", "MQ", "MR", "MS", "MSP", "MST", "MT", "MTK", "MTL", "MVL",
+    "MW", "MY", "MYK", "MZ", "MZG", "MÜ", "MÜB", "MÜL", "MÜN", "MÜR", "N",
+    "NAB", "NAI", "NAU", "NB", "ND", "NDH", "NE", "NEA", "NEB", "NEC", "NEN",
+    "NES", "NEU", "NEW", "NF", "NH", "NI", "NIB", "NK", "NL", "NM", "NMB",
+    "NMS", "NOH", "NOL", "NOM", "NOR", "NP", "NR", "NRW", "NRÜ", "NT", "NU",
+    "NVP", "NW", "NWM", "NY", "NZ", "NÖ", "OA", "OAL", "OB", "OBB", "OBG",
+    "OC", "OCH", "OD", "OE", "OF", "OG", "OH", "OHA", "OHV", "OHZ", "OK",
+    "OL", "OLD", "OP", "OPR", "OR", "OS", "OSL", "OTT", "OTW", "OVI", "OVL",
+    "OVP", "OZ", "P", "PA", "PAF", "PAN", "PAR", "PB", "PCH", "PE", "PEG",
+    "PER", "PF", "PI", "PIR", "PK", "PL", "PLÖ", "PM", "PN", "PR", "PRÜ",
+    "PS", "PW", "PZ", "QFT", "QLB", "R", "RA", "RC", "RD", "RDG", "RE", "REG",
+    "REH", "REI", "RG", "RH", "RI", "RID", "RIE", "RL", "RM", "RN", "RO",
+    "ROD", "ROF", "ROH", "ROK", "ROL", "ROS", "ROT", "ROW", "RPL", "RS",
+    "RSL", "RT", "RU", "RV", "RW", "RWL", "RY", "RZ", "RÜD", "RÜG", "S", "SA",
+    "SAB", "SAD", "SAL", "SAN", "SAO", "SAW", "SB", "SBG", "SBK", "SC", "SCZ",
+    "SDH", "SDL", "SDT", "SE", "SEB", "SEE", "SEF", "SEL", "SF", "SFA", "SFB",
+    "SFT", "SG", "SGH", "SH", "SHA", "SHG", "SHK", "SHL", "SHS", "SI", "SIG",
+    "SIM", "SK", "SL", "SLE", "SLF", "SLG", "SLK", "SLN", "SLS", "SLZ", "SLÜ",
+    "SM", "SMÜ", "SN", "SNH", "SO", "SOB", "SOG", "SOK", "SOL", "SON", "SP",
+    "SPB", "SPN", "SPR", "SR", "SRB", "SRO", "ST", "STA", "STB", "STD", "STE",
+    "STH", "STL", "STO", "SU", "SUL", "SW", "SWA", "SWZ", "SY", "SZ", "SZB",
+    "SÄK", "SÖM", "SÜW", "TBB", "TE", "TF", "TG", "TGO", "THL", "THW", "TIR",
+    "TO", "TP", "TR", "TS", "TT", "TUT", "TÖL", "TÖN", "TÜ", "UE", "UEM",
+    "UER", "UFF", "UH", "UL", "UM", "UN", "USI", "V", "VAI", "VB", "VEC",
+    "VER", "VIB", "VIE", "VIT", "VK", "VL", "VOF", "VOH", "VR", "VS", "W",
+    "WA", "WAF", "WAK", "WAM", "WAN", "WAR", "WAT", "WB", "WBS", "WD", "WDA",
+    "WE", "WEB", "WEG", "WEI", "WEL", "WEM", "WEN", "WER", "WES", "WF", "WG",
+    "WHV", "WI", "WIL", "WIS", "WIT", "WIZ", "WK", "WL", "WLG", "WM", "WMS",
+    "WN", "WND", "WO", "WOB", "WOH", "WOL", "WOR", "WOS", "WR", "WRN", "WS",
+    "WSF", "WST", "WSW", "WT", "WTL", "WTM", "WUG", "WUM", "WUN", "WUR", "WW",
+    "WZ", "WZL", "WÜ", "X", "Y", "Z", "ZE", "ZEL", "ZI", "ZIG", "ZP", "ZR",
+    "ZS", "ZW", "ZZ", "ÖHR", "ÜB",
+]);