@mailwoman/resolver-wof-sqlite 9.1.0 → 9.3.0

package/out/street-normalize.js

@@ -1,272 +0,0 @@
-/**
- * @copyright Sister Software
- * @license AGPL-3.0
- * @author Teffen Ellis, et al.
- *
- *   THE street normalizer for the address-point tier (#476). One function, used by BOTH the shard
- *   builder (`scripts/build-address-point-shard.ts`) and the lookup tier (`address-point.ts`) —
- *   never two implementations (the PLACETYPE_ORDER lesson: parallel copies silently corrupt).
- *
- *   Normalization contract (deliberately aggressive — both sides apply the same function, so
- *   collisions only need to be _consistent_, not linguistically perfect):
- *
- *   1. Lowercase, NFKD-fold diacritics, collapse whitespace, strip punctuation (periods, commas,
- *        apostrophes).
- *   2. Expand USPS directional abbreviations at the FIRST and LAST token position (`n` → `north`, `se` →
- *        `southeast`) — Overture sources abbreviate inconsistently.
- *   3. Canonicalize a trailing USPS street-type token via the codex suffix table to its canonical full
- *        form (`st`/`str`/`street` → `street`).
- *
- *   Numbered streets are left as digits (`5th` stays `5th`); a SPELLED ordinal before a street suffix
- *   folds to its digit form (`tenth street` → `10th street`, #723) so the grid-city ordinal
- *   cross-streets the source data spells with digits become reachable.
- */
-import { AbbreviationToDirectional, US_STREET_SUFFIX_LOOKUP } from "@mailwoman/codex/us";
-/**
- * Token count a street must exceed before its trailing pair is merged. At or below it the pair IS the whole street
- * name, and merging would leave nothing to match on.
- */
-const MIN_TOKENS_FOR_TAIL_MERGE = 3;
-/**
- * Spelled ordinal street names → their digit-ordinal form ("tenth" → "10th"), applied ONLY when a street-type suffix
- * follows (#723 admin-tail) — so the ordinal cross-streets common in grid cities ("Tenth Street", "Fifth Avenue") match
- * the shards' digit keys, WITHOUT rewriting ordinal-WORD names where the next token is not a suffix ("First National
- * Bank Rd" stays "first national …"). Digit-source shards are unaffected (a digit token isn't in this map), so the
- * existing keys need no rebuild; a future rebuild folds any spelled-source key the same way (the one-function
- * discipline).
- */
-const SPELLED_ORDINAL_TO_DIGIT = new Map([
-    ["first", "1st"],
-    ["second", "2nd"],
-    ["third", "3rd"],
-    ["fourth", "4th"],
-    ["fifth", "5th"],
-    ["sixth", "6th"],
-    ["seventh", "7th"],
-    ["eighth", "8th"],
-    ["ninth", "9th"],
-    ["tenth", "10th"],
-    ["eleventh", "11th"],
-    ["twelfth", "12th"],
-    ["thirteenth", "13th"],
-    ["fourteenth", "14th"],
-    ["fifteenth", "15th"],
-    ["sixteenth", "16th"],
-    ["seventeenth", "17th"],
-    ["eighteenth", "18th"],
-    ["nineteenth", "19th"],
-    ["twentieth", "20th"],
-    ["thirtieth", "30th"],
-    ["fortieth", "40th"],
-    ["fiftieth", "50th"],
-    ["sixtieth", "60th"],
-    ["seventieth", "70th"],
-    ["eightieth", "80th"],
-    ["ninetieth", "90th"],
-    ["hundredth", "100th"],
-]);
-/**
- * Lowercase + diacritic-fold + punctuation strip + whitespace collapse.
- */
-function fold(input) {
-    return input
-        .normalize("NFKD")
-        .replaceAll(/[̀-ͯ]/g, "")
-        .toLowerCase()
-        .replaceAll(/[.,'’]/g, "")
-        .replaceAll(/\s+/g, " ")
-        .trim();
-}
-/**
- * Normalize a street name for address-point keying. Same function at build time and lookup time — see module docstring
- * for the contract.
- */
-export function normalizeStreetForKey(street) {
-    const tokens = fold(street).split(" ");
-    if (!tokens.length)
-        return "";
-    // Spelled-ordinal street names → digit form when a street suffix follows ("Tenth Street" →
-    // "10th street", #723). Gated on the next token being a suffix so ordinal-WORD names are untouched.
-    for (let i = 0; i < tokens.length - 1; i++) {
-        const digit = SPELLED_ORDINAL_TO_DIGIT.get(tokens[i]);
-        if (digit && US_STREET_SUFFIX_LOOKUP.has(tokens[i + 1])) {
-            tokens[i] = digit;
-        }
-    }
-    // Directional expansion at the edges only ("N Main St" / "Main St N" — never interior
-    // tokens, where "W" may be an initial in a person-named street). The codex expands
-    // compounds to two words ("SE" → "SOUTH EAST"); we key on the spaceless form
-    // ("southeast"), and also merge an already-written two-token pair ("South East …").
-    const edgeDirectional = (raw) => AbbreviationToDirectional.get(raw.toUpperCase())?.toLowerCase().replace(" ", "");
-    const mergePair = (a, b) => a && b && /^(north|south)$/.test(a) && /^(east|west)$/.test(b) ? a + b : undefined;
-    const leadPair = mergePair(tokens[0], tokens[1]);
-    if (leadPair && tokens.length > 2) {
-        tokens.splice(0, 2, leadPair);
-    }
-    const first = edgeDirectional(tokens[0]);
-    if (first && tokens.length > 1) {
-        tokens[0] = first;
-    }
-    const tailPair = mergePair(tokens.at(-2), tokens.at(-1));
-    if (tailPair && tokens.length > MIN_TOKENS_FOR_TAIL_MERGE) {
-        tokens.splice(-2, 2, tailPair);
-    }
-    if (tokens.length > 2) {
-        const last = edgeDirectional(tokens.at(-1));
-        if (last) {
-            tokens[tokens.length - 1] = last;
-        }
-    }
-    // Street-type canonicalization via the codex table (lowercase keys, UPPER canonical
-    // values). The suffix is usually the last token, but sits second-to-last when a trailing
-    // directional follows ("Main St N") — check both positions, canonicalize the first hit.
-    for (const at of [tokens.length - 1, tokens.length - 2]) {
-        if (at < 1)
-            continue; // never canonicalize the only/first token ("Street Road" exists)
-        const canonical = US_STREET_SUFFIX_LOOKUP.get(tokens[at]);
-        if (canonical) {
-            tokens[at] = canonical.toLowerCase();
-            break;
-        }
-    }
-    return tokens.join(" ");
-}
-/**
- * French street-type abbreviations → canonical full form, applied per token after {@link fold}. French address types
- * LEAD the name ("Av. de…", "Bd …", "Pl. …") and "St"/"Ste" abbreviate Saint/Sainte inside names ("Rue St-Honoré" →
- * "rue saint honore"). fold() has already stripped the trailing period, so the keys are point-free ("av", "bd").
- */
-const FR_STREET_ABBREV = new Map([
-    ["av", "avenue"],
-    ["ave", "avenue"],
-    ["bd", "boulevard"],
-    ["bld", "boulevard"],
-    ["bvd", "boulevard"],
-    ["boul", "boulevard"],
-    ["pl", "place"],
-    ["imp", "impasse"],
-    ["all", "allee"],
-    ["ch", "chemin"],
-    ["che", "chemin"],
-    ["sq", "square"],
-    ["pas", "passage"],
-    ["fg", "faubourg"],
-    ["fbg", "faubourg"],
-    ["rte", "route"],
-    ["st", "saint"],
-    ["ste", "sainte"],
-    ["sts", "saints"],
-]);
-/**
- * Normalize a street name for the address-point key in a non-US locale. Same function build-side and probe-side (the
- * one-function discipline). US delegates to {@link normalizeStreetForKey}.
- *
- * - **en** — the shared English address-key pipeline (directionals + suffix aliases), currently identical to US.
- * - **fr** — fold + expand leading type abbreviations and Saint/Sainte (token map).
- * - **de** — fold + ß→ss + canonicalize the GLUED `-str(.)` suffix to `-strasse` ("Lindenstr." → "lindenstrasse",
- *   "Lindenstraße" → "lindenstrasse"); an already-full "-strasse" is left intact.
- * - **nl** — fold + canonicalize the glued `-str` suffix to `-straat` ("Kerkstr." → "kerkstraat").
- */
-export function normalizeStreetForKeyLocale(street, locale) {
-    if (locale === "us" || locale === "en")
-        return normalizeStreetForKey(street);
-    // Hyphen → space so a compound name keys the same whether the source or the query writes the
-    // hyphen ("Champs-Élysées", "St-Honoré") or a space — both sides fold identically, so this is pure
-    // robustness. It also splits a hyphenated abbreviation ("St-Honoré" → "st honore") into tokens the
-    // per-locale type/Saint map can see.
-    const tokens = fold(street).replaceAll("ß", "ss").replaceAll("-", " ").split(/\s+/).filter(Boolean);
-    if (!tokens.length)
-        return "";
-    switch (locale) {
-        case "fr":
-            for (let i = 0; i < tokens.length; i++) {
-                tokens[i] = FR_STREET_ABBREV.get(tokens[i]) ?? tokens[i];
-            }
-            break;
-        case "de":
-            for (let i = 0; i < tokens.length; i++) {
-                const t = tokens[i];
-                if (t.endsWith("str") && !t.endsWith("strasse")) {
-                    tokens[i] = t.replace(/str$/, "strasse");
-                }
-            }
-            break;
-        case "nl":
-            for (let i = 0; i < tokens.length; i++) {
-                const t = tokens[i];
-                if (t.endsWith("str") && !t.endsWith("straat")) {
-                    tokens[i] = t.replace(/str$/, "straat");
-                }
-            }
-            break;
-    }
-    return tokens.join(" ");
-}
-/**
- * Normalize a locality name for address-point keying (fold only — no street semantics).
- */
-export function normalizeLocalityForKey(locality) {
-    return fold(locality);
-}
-/**
- * Strip a trailing French arrondissement designator from a FOLDED commune key ("paris 8e arrondissement" → "paris",
- * "lyon 1er arrondissement" → "lyon", "marseille 10e arrondissement" → "marseille"). Paris, Lyon and Marseille are the
- * only French communes subdivided into _arrondissements municipaux_; a national register (BAN) names each row per
- * arrondissement, but a query names the base commune ("Place Bellecour, Lyon", never "…, Lyon 2e"). Applied on BOTH
- * sides of the #1042 street-centroid key — build-side (deriving the `locality_base` column) and query-side (folding the
- * probe commune) — so the two agree by construction (the one-function discipline). Input must already be folded
- * (lower-case, diacritic-stripped); a no-op for every other commune. Returns the input unchanged if the strip would
- * empty it.
- */
-export function stripArrondissement(localityNorm) {
-    const stripped = localityNorm.replace(/\s+\d+(?:er|e)\s+arrondissement$/, "").trim();
-    return stripped || localityNorm;
-}
-/**
- * Strip a locality QUALIFIER for a query-side fallback — when an OA locality's exact normalized name misses the
- * gazetteer's canonical name, retry with the qualifier removed. OA address data carries disambiguating qualifiers the
- * gazetteer's canonical name omits: Austrian `Kraubath/Mur` and `Hart b.Graz` → `Hart`; Swiss `Lenk im Simmental` →
- * `Lenk`, `Roche VD` → `Roche`; Danish `Odense S`, `Hurup Thy`. A FALLBACK ONLY — the exact name is tried first, and
- * the region-bbox disambiguation resolves any base-name ambiguity downstream. The candidate table is unchanged (this is
- * purely query-side); feed the result back through {@link normalizeLocalityForKey}. Returns "" when nothing was stripped
- * (no point re-probing the identical key).
- *
- * Measured (`scripts/eval/candidate-recall.ts --strip-fallback`, EU OA holdouts): recovers AT 74.1→88.2% (+14.1pp), DK
- * 91.5→96.2%, CH 90.4→92.6%; +1.3pp overall (diluted by the already-100% locales). Conservative by design — only the
- * qualifier forms above; FI/PT/SI misses are untouched.
- */
-export function stripLocalityQualifier(locality) {
-    let s = locality.trim();
-    if (s.includes("/")) {
-        s = s.split("/")[0].trim();
-    }
-    // "Kraubath/Mur", "St.Kanzian/Klopeiner See"
-    s = s.replace(/\s+[a-zà-ÿ]\.\s*\S.*$/iu, ""); // abbreviated " b.Graz" / " o.Bleiburg" / " a.d. …"
-    s = s.replace(/\s+(im|an der|ob|bei|in der|unter|vor)\s+\S.*$/iu, ""); // " im Simmental", " bei Graz"
-    s = s.replace(/\s+(S|N|E|W|V|Ø|Sø|Fyn|Thy|Sjælland|Jylland|[A-ZÅÄÖ]{2})$/u, ""); // " S", " VD", " Thy"
-    s = s.trim();
-    return s === locality.trim() ? "" : s;
-}
-/**
- * Fold numbered-route designators to a canonical key, applied AFTER {@link normalizeStreetForKey}. Sources disagree
- * systematically on how they spell a route: TIGER says `State Rte 100` / `US Hwy 5` where E911/Overture say `VT ROUTE
- * 100` / `US ROUTE 5` — the dominant street-name miss class in the #483 interpolation eval (rural addresses live on
- * routes). `us <designator> N…` folds to `us route N…`; `state <designator> N…` and `<2-letter-prefix> <designator> N…`
- * (the state abbreviation form) fold to `state route N…`. Only digit-leading route numbers fold — `State Street` and
- * friends never match.
- *
- * Used by BOTH the segment-shard builder (`scripts/build-interpolation-shard.ts`) and the interpolation lookup — same
- * one-function discipline as {@link normalizeStreetForKey}. The address-point tier (#476) does NOT apply it yet:
- * adopting it there requires a shard rebuild (noted on #483).
- *
- * A same-numbered US and state route stay DISTINCT keys (`us route 5` vs `state route 5`); only the BARE `route N` form
- * is ambiguous (designator unknown) and it stays unfolded — a bare-route query therefore misses rather than guessing a
- * designator.
- */
-export function canonicalizeRouteKey(streetNorm) {
-    const match = /^(us|state|[a-z]{2}) (?:route|rte|rt|highway|hwy) (\d.*)$/.exec(streetNorm);
-    if (!match)
-        return streetNorm;
-    return `${match[1] === "us" ? "us" : "state"} route ${match[2]}`;
-}
-//# sourceMappingURL=street-normalize.js.map

package/out/street-normalize.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"street-normalize.js","sourceRoot":"","sources":["../street-normalize.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;GAsBG;AAEH,OAAO,EAAE,yBAAyB,EAAE,uBAAuB,EAAE,MAAM,qBAAqB,CAAA;AAExF;;;GAGG;AACH,MAAM,yBAAyB,GAAG,CAAC,CAAA;AAEnC;;;;;;;GAOG;AACH,MAAM,wBAAwB,GAAG,IAAI,GAAG,CAAiB;IACxD,CAAC,OAAO,EAAE,KAAK,CAAC;IAChB,CAAC,QAAQ,EAAE,KAAK,CAAC;IACjB,CAAC,OAAO,EAAE,KAAK,CAAC;IAChB,CAAC,QAAQ,EAAE,KAAK,CAAC;IACjB,CAAC,OAAO,EAAE,KAAK,CAAC;IAChB,CAAC,OAAO,EAAE,KAAK,CAAC;IAChB,CAAC,SAAS,EAAE,KAAK,CAAC;IAClB,CAAC,QAAQ,EAAE,KAAK,CAAC;IACjB,CAAC,OAAO,EAAE,KAAK,CAAC;IAChB,CAAC,OAAO,EAAE,MAAM,CAAC;IACjB,CAAC,UAAU,EAAE,MAAM,CAAC;IACpB,CAAC,SAAS,EAAE,MAAM,CAAC;IACnB,CAAC,YAAY,EAAE,MAAM,CAAC;IACtB,CAAC,YAAY,EAAE,MAAM,CAAC;IACtB,CAAC,WAAW,EAAE,MAAM,CAAC;IACrB,CAAC,WAAW,EAAE,MAAM,CAAC;IACrB,CAAC,aAAa,EAAE,MAAM,CAAC;IACvB,CAAC,YAAY,EAAE,MAAM,CAAC;IACtB,CAAC,YAAY,EAAE,MAAM,CAAC;IACtB,CAAC,WAAW,EAAE,MAAM,CAAC;IACrB,CAAC,WAAW,EAAE,MAAM,CAAC;IACrB,CAAC,UAAU,EAAE,MAAM,CAAC;IACpB,CAAC,UAAU,EAAE,MAAM,CAAC;IACpB,CAAC,UAAU,EAAE,MAAM,CAAC;IACpB,CAAC,YAAY,EAAE,MAAM,CAAC;IACtB,CAAC,WAAW,EAAE,MAAM,CAAC;IACrB,CAAC,WAAW,EAAE,MAAM,CAAC;IACrB,CAAC,WAAW,EAAE,OAAO,CAAC;CACtB,CAAC,CAAA;AAEF;;GAEG;AACH,SAAS,IAAI,CAAC,KAAa;IAC1B,OAAO,KAAK;SACV,SAAS,CAAC,MAAM,CAAC;SACjB,UAAU,CAAC,QAAQ,EAAE,EAAE,CAAC;SACxB,WAAW,EAAE;SACb,UAAU,CAAC,SAAS,EAAE,EAAE,CAAC;SACzB,UAAU,CAAC,MAAM,EAAE,GAAG,CAAC;SACvB,IAAI,EAAE,CAAA;AACT,CAAC;AAED;;;GAGG;AACH,MAAM,UAAU,qBAAqB,CAAC,MAAc;IACnD,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,CAAA;IAEtC,IAAI,CAAC,MAAM,CAAC,MAAM;QAAE,OAAO,EAAE,CAAA;IAE7B,2FAA2F;IAC3F,oGAAoG;IACpG,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;QAC5C,MAAM,KAAK,GAAG,wBAAwB,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAE,CAAC,CAAA;QAEtD,IAAI,KAAK,IAAI,uBAAuB,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC,GAAG,CAAC,CAAE,CAAC,EAAE,CAAC;YAC1D,MAAM,CAAC,CAAC,CAAC,GAAG,KAAK,CAAA;QAClB,CAAC;IACF,CAAC;IAED,sFAAsF;IACtF,mFAAmF;IACnF,6EAA6E;IAC7E,oFAAoF;IACpF,MAAM,eAAe,GAAG,CAAC,GAAW,EAAE,EAAE,CACvC,yBAAyB,CAAC,GAAG,CAAC,GAAG,CAAC,WAAW,EAAE,CAAC,EAAE,WAAW,EAAE,CAAC,OAAO,CAAC,GAAG,EAAE,EAAE,CAAC,CAAA;IAEjF,MAAM,SAAS,GAAG,CAAC,CAAU,EAAE,CAAU,EAAE,EAAE,CAC5C,CAAC,IAAI,CAAC,IAAI,iBAAiB,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,eAAe,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,SAAS,CAAA;IAEnF,MAAM,QAAQ,GAAG,SAAS,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,CAAA;IAEhD,IAAI,QAAQ,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QACnC,MAAM,CAAC,MAAM,CAAC,CAAC,EAAE,CAAC,EAAE,QAAQ,CAAC,CAAA;IAC9B,CAAC;IAED,MAAM,KAAK,GAAG,eAAe,CAAC,MAAM,CAAC,CAAC,CAAE,CAAC,CAAA;IAEzC,IAAI,KAAK,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QAChC,MAAM,CAAC,CAAC,CAAC,GAAG,KAAK,CAAA;IAClB,CAAC;IAED,MAAM,QAAQ,GAAG,SAAS,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAA;IAExD,IAAI,QAAQ,IAAI,MAAM,CAAC,MAAM,GAAG,yBAAyB,EAAE,CAAC;QAC3D,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,QAAQ,CAAC,CAAA;IAC/B,CAAC;IAED,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QACvB,MAAM,IAAI,GAAG,eAAe,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAE,CAAC,CAAA;QAE5C,IAAI,IAAI,EAAE,CAAC;YACV,MAAM,CAAC,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG,IAAI,CAAA;QACjC,CAAC;IACF,CAAC;IAED,oFAAoF;IACpF,yFAAyF;IACzF,wFAAwF;IACxF,KAAK,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,EAAE,CAAC;QACzD,IAAI,EAAE,GAAG,CAAC;YAAE,SAAQ,CAAC,iEAAiE;QACtF,MAAM,SAAS,GAAG,uBAAuB,CAAC,GAAG,CAAC,MAAM,CAAC,EAAE,CAAE,CAAC,CAAA;QAE1D,IAAI,SAAS,EAAE,CAAC;YACf,MAAM,CAAC,EAAE,CAAC,GAAG,SAAS,CAAC,WAAW,EAAE,CAAA;YAEpC,MAAK;QACN,CAAC;IACF,CAAC;IAED,OAAO,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAA;AACxB,CAAC;AAYD;;;;GAIG;AACH,MAAM,gBAAgB,GAAG,IAAI,GAAG,CAAiB;IAChD,CAAC,IAAI,EAAE,QAAQ,CAAC;IAChB,CAAC,KAAK,EAAE,QAAQ,CAAC;IACjB,CAAC,IAAI,EAAE,WAAW,CAAC;IACnB,CAAC,KAAK,EAAE,WAAW,CAAC;IACpB,CAAC,KAAK,EAAE,WAAW,CAAC;IACpB,CAAC,MAAM,EAAE,WAAW,CAAC;IACrB,CAAC,IAAI,EAAE,OAAO,CAAC;IACf,CAAC,KAAK,EAAE,SAAS,CAAC;IAClB,CAAC,KAAK,EAAE,OAAO,CAAC;IAChB,CAAC,IAAI,EAAE,QAAQ,CAAC;IAChB,CAAC,KAAK,EAAE,QAAQ,CAAC;IACjB,CAAC,IAAI,EAAE,QAAQ,CAAC;IAChB,CAAC,KAAK,EAAE,SAAS,CAAC;IAClB,CAAC,IAAI,EAAE,UAAU,CAAC;IAClB,CAAC,KAAK,EAAE,UAAU,CAAC;IACnB,CAAC,KAAK,EAAE,OAAO,CAAC;IAChB,CAAC,IAAI,EAAE,OAAO,CAAC;IACf,CAAC,KAAK,EAAE,QAAQ,CAAC;IACjB,CAAC,KAAK,EAAE,QAAQ,CAAC;CACjB,CAAC,CAAA;AAEF;;;;;;;;;GASG;AACH,MAAM,UAAU,2BAA2B,CAAC,MAAc,EAAE,MAAoB;IAC/E,IAAI,MAAM,KAAK,IAAI,IAAI,MAAM,KAAK,IAAI;QAAE,OAAO,qBAAqB,CAAC,MAAM,CAAC,CAAA;IAE5E,6FAA6F;IAC7F,mGAAmG;IACnG,mGAAmG;IACnG,qCAAqC;IACrC,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,UAAU,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC,UAAU,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,MAAM,CAAC,OAAO,CAAC,CAAA;IAEnG,IAAI,CAAC,MAAM,CAAC,MAAM;QAAE,OAAO,EAAE,CAAA;IAE7B,QAAQ,MAAM,EAAE,CAAC;QAChB,KAAK,IAAI;YACR,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;gBACxC,MAAM,CAAC,CAAC,CAAC,GAAG,gBAAgB,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAE,CAAC,IAAI,MAAM,CAAC,CAAC,CAAE,CAAA;YAC3D,CAAC;YACD,MAAK;QACN,KAAK,IAAI;YACR,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;gBACxC,MAAM,CAAC,GAAG,MAAM,CAAC,CAAC,CAAE,CAAA;gBAEpB,IAAI,CAAC,CAAC,QAAQ,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,QAAQ,CAAC,SAAS,CAAC,EAAE,CAAC;oBACjD,MAAM,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,OAAO,CAAC,MAAM,EAAE,SAAS,CAAC,CAAA;gBACzC,CAAC;YACF,CAAC;YACD,MAAK;QACN,KAAK,IAAI;YACR,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;gBACxC,MAAM,CAAC,GAAG,MAAM,CAAC,CAAC,CAAE,CAAA;gBAEpB,IAAI,CAAC,CAAC,QAAQ,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,QAAQ,CAAC,QAAQ,CAAC,EAAE,CAAC;oBAChD,MAAM,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,OAAO,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAA;gBACxC,CAAC;YACF,CAAC;YACD,MAAK;IACP,CAAC;IAED,OAAO,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAA;AACxB,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,uBAAuB,CAAC,QAAgB;IACvD,OAAO,IAAI,CAAC,QAAQ,CAAC,CAAA;AACtB,CAAC;AAED;;;;;;;;;GASG;AACH,MAAM,UAAU,mBAAmB,CAAC,YAAoB;IACvD,MAAM,QAAQ,GAAG,YAAY,CAAC,OAAO,CAAC,kCAAkC,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,CAAA;IAEpF,OAAO,QAAQ,IAAI,YAAY,CAAA;AAChC,CAAC;AAED;;;;;;;;;;;;GAYG;AACH,MAAM,UAAU,sBAAsB,CAAC,QAAgB;IACtD,IAAI,CAAC,GAAG,QAAQ,CAAC,IAAI,EAAE,CAAA;IAEvB,IAAI,CAAC,CAAC,QAAQ,CAAC,GAAG,CAAC,EAAE,CAAC;QACrB,CAAC,GAAG,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,CAAE,CAAC,IAAI,EAAE,CAAA;IAC5B,CAAC;IAED,6CAA6C;IAC7C,CAAC,GAAG,CAAC,CAAC,OAAO,CAAC,yBAAyB,EAAE,EAAE,CAAC,CAAA,CAAC,oDAAoD;IACjG,CAAC,GAAG,CAAC,CAAC,OAAO,CAAC,kDAAkD,EAAE,EAAE,CAAC,CAAA,CAAC,+BAA+B;IACrG,CAAC,GAAG,CAAC,CAAC,OAAO,CAAC,4DAA4D,EAAE,EAAE,CAAC,CAAA,CAAC,sBAAsB;IACtG,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,CAAA;IAEZ,OAAO,CAAC,KAAK,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAA;AACtC,CAAC;AAED;;;;;;;;;;;;;;;GAeG;AACH,MAAM,UAAU,oBAAoB,CAAC,UAAkB;IACtD,MAAM,KAAK,GAAG,2DAA2D,CAAC,IAAI,CAAC,UAAU,CAAC,CAAA;IAE1F,IAAI,CAAC,KAAK;QAAE,OAAO,UAAU,CAAA;IAE7B,OAAO,GAAG,KAAK,CAAC,CAAC,CAAC,KAAK,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,OAAO,UAAU,KAAK,CAAC,CAAC,CAAC,EAAE,CAAA;AACjE,CAAC"}

package/out/street-segment-schema.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"street-segment-schema.d.ts","sourceRoot":"","sources":["../street-segment-schema.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;GAeG;AAEH,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,QAAQ,CAAA;AAEpC;;;;GAIG;AACH,MAAM,WAAW,kBAAkB;IAClC;;OAEG;IACH,WAAW,EAAE,MAAM,CAAA;IACnB;;OAEG;IACH,IAAI,EAAE,MAAM,CAAA;IACZ,OAAO,EAAE,MAAM,CAAA;IACf,KAAK,EAAE,MAAM,CAAA;IACb;;OAEG;IACH,MAAM,EAAE,MAAM,CAAA;IACd;;OAEG;IACH,MAAM,EAAE,MAAM,CAAA;IACd;;OAEG;IACH,MAAM,EAAE,MAAM,CAAA;IACd,QAAQ,EAAE,MAAM,GAAG,IAAI,CAAA;IACvB;;OAEG;IACH,WAAW,EAAE,MAAM,CAAA;IACnB;;OAEG;IACH,UAAU,EAAE,MAAM,CAAA;IAClB;;OAEG;IACH,QAAQ,EAAE,MAAM,CAAA;IAChB;;OAEG;IACH,MAAM,EAAE,MAAM,CAAA;IACd;;OAEG;IACH,OAAO,EAAE,MAAM,CAAA;CACf;AAED;;;;;;GAMG;AACH,MAAM,WAAW,oBAAoB;IACpC;;OAEG;IACH,iBAAiB,EAAE,MAAM,CAAA;IACzB;;OAEG;IACH,MAAM,EAAE,MAAM,CAAA;IACd;;OAEG;IACH,MAAM,EAAE,MAAM,CAAA;CACd;AAED;;GAEG;AACH,MAAM,WAAW,qBAAqB;IACrC,cAAc,EAAE,kBAAkB,CAAA;IAClC,kBAAkB,EAAE,oBAAoB,CAAA;CACxC;AAED;;;GAGG;AACH,eAAO,MAAM,sBAAsB,8JAczB,CAAA;AAEV;;GAEG;AACH,wBAAsB,wBAAwB,CAAC,EAAE,EAAE,MAAM,CAAC,qBAAqB,CAAC,GAAG,OAAO,CAAC,IAAI,CAAC,CAiB/F;AAED;;;;;GAKG;AACH,wBAAsB,sBAAsB,CAC3C,EAAE,EAAE,MAAM,CAAC,qBAAqB,CAAC,EACjC,GAAG,EAAE,oBAAoB,GACvB,OAAO,CAAC,IAAI,CAAC,CASf;AAED;;GAEG;AACH,wBAAsB,0BAA0B,CAAC,EAAE,EAAE,MAAM,CAAC,qBAAqB,CAAC,GAAG,OAAO,CAAC,IAAI,CAAC,CAQjG"}

package/sqlite-utils.ts

@@ -1,46 +0,0 @@
-/**
- * @copyright Sister Software
- * @license AGPL-3.0
- * @author Teffen Ellis, et al.
- *
- *   Small shared helpers for the SQLite-backed lookups.
- */
-
-import type { DatabaseSync } from "node:sqlite"
-
-/**
- * True when `name` is a table in the open database. The street-level lookups use this to degrade gracefully on an
- * empty/tableless shard — an interrupted `build-*-shard.ts`, or a stray 0-byte file (e.g. `sqlite3 <missing>.db "…"`
- * CREATES one) — rather than throwing `no such table` at construction and taking down a whole state's geocode (#568). A
- * missing table makes the lookup a no-op miss.
- */
-export function hasTable(db: DatabaseSync, name: string): boolean {
-	try {
-		const row = db.prepare("SELECT 1 FROM sqlite_master WHERE type = 'table' AND name = ? LIMIT 1").get(name)
-
-		return row !== undefined
-	} catch {
-		return false
-	}
-}
-
-/**
- * True when `table` exists in the open database AND carries `column`.
- *
- * The column-level sibling of {@link hasTable}, and it exists for the same reason one layer down: an artifact built
- * before a column was added is still a VALID artifact, and a reader that unconditionally names the new column in its
- * `SELECT` turns "this gazetteer is a build behind" into `no such column` at the first keystroke. Probe once at
- * construction and shape the query — `table_info` is a PRAGMA, so it must not sit on a per-query path.
- *
- * Note the interpolation: PRAGMA does not take bound parameters, so `table` is spliced. Every caller passes a
- * module-level constant; never pass user input.
- */
-export function hasColumn(db: DatabaseSync, table: string, column: string): boolean {
-	try {
-		const rows = db.prepare(`PRAGMA table_info(${table})`).all() as unknown as Array<{ name: string }>
-
-		return rows.some((r) => String(r.name) === column)
-	} catch {
-		return false
-	}
-}

package/street-normalize.ts

@@ -1,318 +0,0 @@
-/**
- * @copyright Sister Software
- * @license AGPL-3.0
- * @author Teffen Ellis, et al.
- *
- *   THE street normalizer for the address-point tier (#476). One function, used by BOTH the shard
- *   builder (`scripts/build-address-point-shard.ts`) and the lookup tier (`address-point.ts`) —
- *   never two implementations (the PLACETYPE_ORDER lesson: parallel copies silently corrupt).
- *
- *   Normalization contract (deliberately aggressive — both sides apply the same function, so
- *   collisions only need to be _consistent_, not linguistically perfect):
- *
- *   1. Lowercase, NFKD-fold diacritics, collapse whitespace, strip punctuation (periods, commas,
- *        apostrophes).
- *   2. Expand USPS directional abbreviations at the FIRST and LAST token position (`n` → `north`, `se` →
- *        `southeast`) — Overture sources abbreviate inconsistently.
- *   3. Canonicalize a trailing USPS street-type token via the codex suffix table to its canonical full
- *        form (`st`/`str`/`street` → `street`).
- *
- *   Numbered streets are left as digits (`5th` stays `5th`); a SPELLED ordinal before a street suffix
- *   folds to its digit form (`tenth street` → `10th street`, #723) so the grid-city ordinal
- *   cross-streets the source data spells with digits become reachable.
- */
-
-import { AbbreviationToDirectional, US_STREET_SUFFIX_LOOKUP } from "@mailwoman/codex/us"
-
-/**
- * Token count a street must exceed before its trailing pair is merged. At or below it the pair IS the whole street
- * name, and merging would leave nothing to match on.
- */
-const MIN_TOKENS_FOR_TAIL_MERGE = 3
-
-/**
- * Spelled ordinal street names → their digit-ordinal form ("tenth" → "10th"), applied ONLY when a street-type suffix
- * follows (#723 admin-tail) — so the ordinal cross-streets common in grid cities ("Tenth Street", "Fifth Avenue") match
- * the shards' digit keys, WITHOUT rewriting ordinal-WORD names where the next token is not a suffix ("First National
- * Bank Rd" stays "first national …"). Digit-source shards are unaffected (a digit token isn't in this map), so the
- * existing keys need no rebuild; a future rebuild folds any spelled-source key the same way (the one-function
- * discipline).
- */
-const SPELLED_ORDINAL_TO_DIGIT = new Map<string, string>([
-	["first", "1st"],
-	["second", "2nd"],
-	["third", "3rd"],
-	["fourth", "4th"],
-	["fifth", "5th"],
-	["sixth", "6th"],
-	["seventh", "7th"],
-	["eighth", "8th"],
-	["ninth", "9th"],
-	["tenth", "10th"],
-	["eleventh", "11th"],
-	["twelfth", "12th"],
-	["thirteenth", "13th"],
-	["fourteenth", "14th"],
-	["fifteenth", "15th"],
-	["sixteenth", "16th"],
-	["seventeenth", "17th"],
-	["eighteenth", "18th"],
-	["nineteenth", "19th"],
-	["twentieth", "20th"],
-	["thirtieth", "30th"],
-	["fortieth", "40th"],
-	["fiftieth", "50th"],
-	["sixtieth", "60th"],
-	["seventieth", "70th"],
-	["eightieth", "80th"],
-	["ninetieth", "90th"],
-	["hundredth", "100th"],
-])
-
-/**
- * Lowercase + diacritic-fold + punctuation strip + whitespace collapse.
- */
-function fold(input: string): string {
-	return input
-		.normalize("NFKD")
-		.replaceAll(/[̀-ͯ]/g, "")
-		.toLowerCase()
-		.replaceAll(/[.,'’]/g, "")
-		.replaceAll(/\s+/g, " ")
-		.trim()
-}
-
-/**
- * Normalize a street name for address-point keying. Same function at build time and lookup time — see module docstring
- * for the contract.
- */
-export function normalizeStreetForKey(street: string): string {
-	const tokens = fold(street).split(" ")
-
-	if (!tokens.length) return ""
-
-	// Spelled-ordinal street names → digit form when a street suffix follows ("Tenth Street" →
-	// "10th street", #723). Gated on the next token being a suffix so ordinal-WORD names are untouched.
-	for (let i = 0; i < tokens.length - 1; i++) {
-		const digit = SPELLED_ORDINAL_TO_DIGIT.get(tokens[i]!)
-
-		if (digit && US_STREET_SUFFIX_LOOKUP.has(tokens[i + 1]!)) {
-			tokens[i] = digit
-		}
-	}
-
-	// Directional expansion at the edges only ("N Main St" / "Main St N" — never interior
-	// tokens, where "W" may be an initial in a person-named street). The codex expands
-	// compounds to two words ("SE" → "SOUTH EAST"); we key on the spaceless form
-	// ("southeast"), and also merge an already-written two-token pair ("South East …").
-	const edgeDirectional = (raw: string) =>
-		AbbreviationToDirectional.get(raw.toUpperCase())?.toLowerCase().replace(" ", "")
-
-	const mergePair = (a?: string, b?: string) =>
-		a && b && /^(north|south)$/.test(a) && /^(east|west)$/.test(b) ? a + b : undefined
-
-	const leadPair = mergePair(tokens[0], tokens[1])
-
-	if (leadPair && tokens.length > 2) {
-		tokens.splice(0, 2, leadPair)
-	}
-
-	const first = edgeDirectional(tokens[0]!)
-
-	if (first && tokens.length > 1) {
-		tokens[0] = first
-	}
-
-	const tailPair = mergePair(tokens.at(-2), tokens.at(-1))
-
-	if (tailPair && tokens.length > MIN_TOKENS_FOR_TAIL_MERGE) {
-		tokens.splice(-2, 2, tailPair)
-	}
-
-	if (tokens.length > 2) {
-		const last = edgeDirectional(tokens.at(-1)!)
-
-		if (last) {
-			tokens[tokens.length - 1] = last
-		}
-	}
-
-	// Street-type canonicalization via the codex table (lowercase keys, UPPER canonical
-	// values). The suffix is usually the last token, but sits second-to-last when a trailing
-	// directional follows ("Main St N") — check both positions, canonicalize the first hit.
-	for (const at of [tokens.length - 1, tokens.length - 2]) {
-		if (at < 1) continue // never canonicalize the only/first token ("Street Road" exists)
-		const canonical = US_STREET_SUFFIX_LOOKUP.get(tokens[at]!)
-
-		if (canonical) {
-			tokens[at] = canonical.toLowerCase()
-
-			break
-		}
-	}
-
-	return tokens.join(" ")
-}
-
-/**
- * Street-name locale for the address-point key. The US path is the full USPS pipeline ({@link normalizeStreetForKey});
- * the international paths fold + apply a SMALL, consistent per-locale type-token canonicalization. Same discipline as
- * the US normalizer: build side and probe side call the identical function, so the key only needs to be CONSISTENT, not
- * linguistically perfect — a folded "rue du chevaleret" keys the same on both sides whether or not we reorder the
- * article, so no salient-token / multi-key index is built yet (deferred until probing shows the normalizer can't absorb
- * the false-negatives).
- */
-export type StreetLocale = "us" | "en" | "fr" | "de" | "nl"
-
-/**
- * French street-type abbreviations → canonical full form, applied per token after {@link fold}. French address types
- * LEAD the name ("Av. de…", "Bd …", "Pl. …") and "St"/"Ste" abbreviate Saint/Sainte inside names ("Rue St-Honoré" →
- * "rue saint honore"). fold() has already stripped the trailing period, so the keys are point-free ("av", "bd").
- */
-const FR_STREET_ABBREV = new Map<string, string>([
-	["av", "avenue"],
-	["ave", "avenue"],
-	["bd", "boulevard"],
-	["bld", "boulevard"],
-	["bvd", "boulevard"],
-	["boul", "boulevard"],
-	["pl", "place"],
-	["imp", "impasse"],
-	["all", "allee"],
-	["ch", "chemin"],
-	["che", "chemin"],
-	["sq", "square"],
-	["pas", "passage"],
-	["fg", "faubourg"],
-	["fbg", "faubourg"],
-	["rte", "route"],
-	["st", "saint"],
-	["ste", "sainte"],
-	["sts", "saints"],
-])
-
-/**
- * Normalize a street name for the address-point key in a non-US locale. Same function build-side and probe-side (the
- * one-function discipline). US delegates to {@link normalizeStreetForKey}.
- *
- * - **en** — the shared English address-key pipeline (directionals + suffix aliases), currently identical to US.
- * - **fr** — fold + expand leading type abbreviations and Saint/Sainte (token map).
- * - **de** — fold + ß→ss + canonicalize the GLUED `-str(.)` suffix to `-strasse` ("Lindenstr." → "lindenstrasse",
- *   "Lindenstraße" → "lindenstrasse"); an already-full "-strasse" is left intact.
- * - **nl** — fold + canonicalize the glued `-str` suffix to `-straat` ("Kerkstr." → "kerkstraat").
- */
-export function normalizeStreetForKeyLocale(street: string, locale: StreetLocale): string {
-	if (locale === "us" || locale === "en") return normalizeStreetForKey(street)
-
-	// Hyphen → space so a compound name keys the same whether the source or the query writes the
-	// hyphen ("Champs-Élysées", "St-Honoré") or a space — both sides fold identically, so this is pure
-	// robustness. It also splits a hyphenated abbreviation ("St-Honoré" → "st honore") into tokens the
-	// per-locale type/Saint map can see.
-	const tokens = fold(street).replaceAll("ß", "ss").replaceAll("-", " ").split(/\s+/).filter(Boolean)
-
-	if (!tokens.length) return ""
-
-	switch (locale) {
-		case "fr":
-			for (let i = 0; i < tokens.length; i++) {
-				tokens[i] = FR_STREET_ABBREV.get(tokens[i]!) ?? tokens[i]!
-			}
-			break
-		case "de":
-			for (let i = 0; i < tokens.length; i++) {
-				const t = tokens[i]!
-
-				if (t.endsWith("str") && !t.endsWith("strasse")) {
-					tokens[i] = t.replace(/str$/, "strasse")
-				}
-			}
-			break
-		case "nl":
-			for (let i = 0; i < tokens.length; i++) {
-				const t = tokens[i]!
-
-				if (t.endsWith("str") && !t.endsWith("straat")) {
-					tokens[i] = t.replace(/str$/, "straat")
-				}
-			}
-			break
-	}
-
-	return tokens.join(" ")
-}
-
-/**
- * Normalize a locality name for address-point keying (fold only — no street semantics).
- */
-export function normalizeLocalityForKey(locality: string): string {
-	return fold(locality)
-}
-
-/**
- * Strip a trailing French arrondissement designator from a FOLDED commune key ("paris 8e arrondissement" → "paris",
- * "lyon 1er arrondissement" → "lyon", "marseille 10e arrondissement" → "marseille"). Paris, Lyon and Marseille are the
- * only French communes subdivided into _arrondissements municipaux_; a national register (BAN) names each row per
- * arrondissement, but a query names the base commune ("Place Bellecour, Lyon", never "…, Lyon 2e"). Applied on BOTH
- * sides of the #1042 street-centroid key — build-side (deriving the `locality_base` column) and query-side (folding the
- * probe commune) — so the two agree by construction (the one-function discipline). Input must already be folded
- * (lower-case, diacritic-stripped); a no-op for every other commune. Returns the input unchanged if the strip would
- * empty it.
- */
-export function stripArrondissement(localityNorm: string): string {
-	const stripped = localityNorm.replace(/\s+\d+(?:er|e)\s+arrondissement$/, "").trim()
-
-	return stripped || localityNorm
-}
-
-/**
- * Strip a locality QUALIFIER for a query-side fallback — when an OA locality's exact normalized name misses the
- * gazetteer's canonical name, retry with the qualifier removed. OA address data carries disambiguating qualifiers the
- * gazetteer's canonical name omits: Austrian `Kraubath/Mur` and `Hart b.Graz` → `Hart`; Swiss `Lenk im Simmental` →
- * `Lenk`, `Roche VD` → `Roche`; Danish `Odense S`, `Hurup Thy`. A FALLBACK ONLY — the exact name is tried first, and
- * the region-bbox disambiguation resolves any base-name ambiguity downstream. The candidate table is unchanged (this is
- * purely query-side); feed the result back through {@link normalizeLocalityForKey}. Returns "" when nothing was stripped
- * (no point re-probing the identical key).
- *
- * Measured (`scripts/eval/candidate-recall.ts --strip-fallback`, EU OA holdouts): recovers AT 74.1→88.2% (+14.1pp), DK
- * 91.5→96.2%, CH 90.4→92.6%; +1.3pp overall (diluted by the already-100% locales). Conservative by design — only the
- * qualifier forms above; FI/PT/SI misses are untouched.
- */
-export function stripLocalityQualifier(locality: string): string {
-	let s = locality.trim()
-
-	if (s.includes("/")) {
-		s = s.split("/")[0]!.trim()
-	}
-
-	// "Kraubath/Mur", "St.Kanzian/Klopeiner See"
-	s = s.replace(/\s+[a-zà-ÿ]\.\s*\S.*$/iu, "") // abbreviated " b.Graz" / " o.Bleiburg" / " a.d. …"
-	s = s.replace(/\s+(im|an der|ob|bei|in der|unter|vor)\s+\S.*$/iu, "") // " im Simmental", " bei Graz"
-	s = s.replace(/\s+(S|N|E|W|V|Ø|Sø|Fyn|Thy|Sjælland|Jylland|[A-ZÅÄÖ]{2})$/u, "") // " S", " VD", " Thy"
-	s = s.trim()
-
-	return s === locality.trim() ? "" : s
-}
-
-/**
- * Fold numbered-route designators to a canonical key, applied AFTER {@link normalizeStreetForKey}. Sources disagree
- * systematically on how they spell a route: TIGER says `State Rte 100` / `US Hwy 5` where E911/Overture say `VT ROUTE
- * 100` / `US ROUTE 5` — the dominant street-name miss class in the #483 interpolation eval (rural addresses live on
- * routes). `us <designator> N…` folds to `us route N…`; `state <designator> N…` and `<2-letter-prefix> <designator> N…`
- * (the state abbreviation form) fold to `state route N…`. Only digit-leading route numbers fold — `State Street` and
- * friends never match.
- *
- * Used by BOTH the segment-shard builder (`scripts/build-interpolation-shard.ts`) and the interpolation lookup — same
- * one-function discipline as {@link normalizeStreetForKey}. The address-point tier (#476) does NOT apply it yet:
- * adopting it there requires a shard rebuild (noted on #483).
- *
- * A same-numbered US and state route stay DISTINCT keys (`us route 5` vs `state route 5`); only the BARE `route N` form
- * is ambiguous (designator unknown) and it stays unfolded — a bare-route query therefore misses rather than guessing a
- * designator.
- */
-export function canonicalizeRouteKey(streetNorm: string): string {
-	const match = /^(us|state|[a-z]{2}) (?:route|rte|rt|highway|hwy) (\d.*)$/.exec(streetNorm)
-
-	if (!match) return streetNorm
-
-	return `${match[1] === "us" ? "us" : "state"} route ${match[2]}`
-}