@mailwoman/corpus 9.4.0 → 10.1.0

package/out/recipes/index.js

@@ -1,91 +0,0 @@
-/**
- * @copyright Sister Software
- * @license AGPL-3.0
- * @author Teffen Ellis, et al.
- *
- *   The synthetic-corpus SLICE RECIPE registry. Each recipe ({@link CorpusRecipe}) is what one legacy
- *   `build-*-slice.mjs` script used to be; the `mailwoman corpus slice <recipe>` command dispatches
- *   to them. A new slice is a new entry here, not a new script.
- */
-import { anchorAbsorptionRecipe } from "#recipes/anchor-absorption";
-import { bareCountryRecipe } from "#recipes/bare-country";
-import { boundaryStressRecipe } from "#recipes/boundary-stress";
-import { cnOrganizationalUnitsRecipe } from "#recipes/cn-organizational-units";
-import { countryBalancedRecipe } from "#recipes/country-balanced";
-import { czPcFirstPrepositionRecipe } from "#recipes/cz-pcfirst-preposition";
-import { frAdminSplitRecipe } from "#recipes/fr-admin-split";
-import { frBareStreetRecipe } from "#recipes/fr-bare-street";
-import { frFragmentRecipe } from "#recipes/fr-fragment";
-import { frLieuditRecipe } from "#recipes/fr-lieudit";
-import { frOrderRecipe } from "#recipes/fr-order";
-import { germanRecipe } from "#recipes/german";
-import { houseVenueRecipe } from "#recipes/house-venue";
-import { intersectionRecipe } from "#recipes/intersection";
-import { localeRecipe } from "#recipes/locale";
-import { nlPostcodeRecipe } from "#recipes/nl-postcode";
-import { noFragmentRecipe } from "#recipes/no-fragment";
-import { noStreetRecipe } from "#recipes/no-street";
-import { noStreetLedRecipe } from "#recipes/no-street-led";
-import { poBoxRecipe } from "#recipes/po-box";
-import { poBoxCedexRecipe } from "#recipes/po-box-cedex";
-import { reviewedPostcodeTailRecipe } from "#recipes/reviewed-postcode-tail";
-import { siBareVillageRecipe } from "#recipes/si-bare-village";
-import { streetRecipe } from "#recipes/street";
-import { streetAffixRecipe, suffixBoundaryRecipe } from "#recipes/street-affix";
-import { streetBareRecipe } from "#recipes/street-bare";
-import { subVenueRecipe } from "#recipes/sub-venue";
-import { trailingRegionRecipe } from "#recipes/trailing-region";
-import { unitRecipe } from "#recipes/unit";
-export * from "#recipes/scaffold";
-/**
- * Every registered recipe, in display order.
- */
-const RECIPES = [
-    bareCountryRecipe,
-    trailingRegionRecipe,
-    reviewedPostcodeTailRecipe,
-    streetRecipe,
-    streetBareRecipe,
-    streetAffixRecipe,
-    suffixBoundaryRecipe,
-    noStreetRecipe,
-    houseVenueRecipe,
-    poBoxRecipe,
-    poBoxCedexRecipe,
-    unitRecipe,
-    intersectionRecipe,
-    germanRecipe,
-    cnOrganizationalUnitsRecipe,
-    localeRecipe,
-    frOrderRecipe,
-    frAdminSplitRecipe,
-    frBareStreetRecipe,
-    frFragmentRecipe,
-    frLieuditRecipe,
-    czPcFirstPrepositionRecipe,
-    nlPostcodeRecipe,
-    noStreetLedRecipe,
-    noFragmentRecipe,
-    siBareVillageRecipe,
-    countryBalancedRecipe,
-    boundaryStressRecipe,
-    anchorAbsorptionRecipe,
-    subVenueRecipe,
-];
-/**
- * Recipe name → recipe.
- */
-export const SLICE_RECIPES = new Map(RECIPES.map((r) => [r.name, r]));
-/**
- * Look up a recipe by its `<recipe>` name.
- */
-export function getSliceRecipe(name) {
-    return SLICE_RECIPES.get(name);
-}
-/**
- * All recipes (for `--list` / help).
- */
-export function listSliceRecipes() {
-    return RECIPES;
-}
-//# sourceMappingURL=index.js.map

package/out/recipes/index.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"index.js","sourceRoot":"","sources":["../../lib/recipes/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AAEH,OAAO,EAAE,sBAAsB,EAAE,MAAM,4BAA4B,CAAA;AACnE,OAAO,EAAE,iBAAiB,EAAE,MAAM,uBAAuB,CAAA;AACzD,OAAO,EAAE,oBAAoB,EAAE,MAAM,0BAA0B,CAAA;AAC/D,OAAO,EAAE,2BAA2B,EAAE,MAAM,kCAAkC,CAAA;AAC9E,OAAO,EAAE,qBAAqB,EAAE,MAAM,2BAA2B,CAAA;AACjE,OAAO,EAAE,0BAA0B,EAAE,MAAM,iCAAiC,CAAA;AAC5E,OAAO,EAAE,kBAAkB,EAAE,MAAM,yBAAyB,CAAA;AAC5D,OAAO,EAAE,kBAAkB,EAAE,MAAM,yBAAyB,CAAA;AAC5D,OAAO,EAAE,gBAAgB,EAAE,MAAM,sBAAsB,CAAA;AACvD,OAAO,EAAE,eAAe,EAAE,MAAM,qBAAqB,CAAA;AACrD,OAAO,EAAE,aAAa,EAAE,MAAM,mBAAmB,CAAA;AACjD,OAAO,EAAE,YAAY,EAAE,MAAM,iBAAiB,CAAA;AAC9C,OAAO,EAAE,gBAAgB,EAAE,MAAM,sBAAsB,CAAA;AACvD,OAAO,EAAE,kBAAkB,EAAE,MAAM,uBAAuB,CAAA;AAC1D,OAAO,EAAE,YAAY,EAAE,MAAM,iBAAiB,CAAA;AAC9C,OAAO,EAAE,gBAAgB,EAAE,MAAM,sBAAsB,CAAA;AACvD,OAAO,EAAE,gBAAgB,EAAE,MAAM,sBAAsB,CAAA;AACvD,OAAO,EAAE,cAAc,EAAE,MAAM,oBAAoB,CAAA;AACnD,OAAO,EAAE,iBAAiB,EAAE,MAAM,wBAAwB,CAAA;AAC1D,OAAO,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAA;AAC7C,OAAO,EAAE,gBAAgB,EAAE,MAAM,uBAAuB,CAAA;AACxD,OAAO,EAAE,0BAA0B,EAAE,MAAM,iCAAiC,CAAA;AAE5E,OAAO,EAAE,mBAAmB,EAAE,MAAM,0BAA0B,CAAA;AAC9D,OAAO,EAAE,YAAY,EAAE,MAAM,iBAAiB,CAAA;AAC9C,OAAO,EAAE,iBAAiB,EAAE,oBAAoB,EAAE,MAAM,uBAAuB,CAAA;AAC/E,OAAO,EAAE,gBAAgB,EAAE,MAAM,sBAAsB,CAAA;AACvD,OAAO,EAAE,cAAc,EAAE,MAAM,oBAAoB,CAAA;AACnD,OAAO,EAAE,oBAAoB,EAAE,MAAM,0BAA0B,CAAA;AAC/D,OAAO,EAAE,UAAU,EAAE,MAAM,eAAe,CAAA;AAE1C,cAAc,mBAAmB,CAAA;AAEjC;;GAEG;AACH,MAAM,OAAO,GAA4B;IACxC,iBAAiB;IACjB,oBAAoB;IACpB,0BAA0B;IAC1B,YAAY;IACZ,gBAAgB;IAChB,iBAAiB;IACjB,oBAAoB;IACpB,cAAc;IACd,gBAAgB;IAChB,WAAW;IACX,gBAAgB;IAChB,UAAU;IACV,kBAAkB;IAClB,YAAY;IACZ,2BAA2B;IAC3B,YAAY;IACZ,aAAa;IACb,kBAAkB;IAClB,kBAAkB;IAClB,gBAAgB;IAChB,eAAe;IACf,0BAA0B;IAC1B,gBAAgB;IAChB,iBAAiB;IACjB,gBAAgB;IAChB,mBAAmB;IACnB,qBAAqB;IACrB,oBAAoB;IACpB,sBAAsB;IACtB,cAAc;CACd,CAAA;AAED;;GAEG;AACH,MAAM,CAAC,MAAM,aAAa,GAAsC,IAAI,GAAG,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,CAAA;AAExG;;GAEG;AACH,MAAM,UAAU,cAAc,CAAC,IAAY;IAC1C,OAAO,aAAa,CAAC,GAAG,CAAC,IAAI,CAAC,CAAA;AAC/B,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,gBAAgB;IAC/B,OAAO,OAAO,CAAA;AACf,CAAC"}

package/out/recipes/locale.d.ts

@@ -1,136 +0,0 @@
-/**
- * @copyright Sister Software
- * @license AGPL-3.0
- * @author Teffen Ellis, et al.
- *
- *   `locale` slice recipe — the multi-locale generalization of the `german` recipe. Reads REAL
- *   OpenAddresses tuples for a `--country` (DE/FR/NL/IT/ES), renders each via
- *   {@link synthesizeLocaleRow} in BOTH orders (`--intl-fraction`, default 0.4 international / the
- *   rest country-native), aligns to BIO, and emits a labeled JSONL. Generate-mode: it STREAMS each
- *   source CSV (a streamed zip member for cached zips, plain `createReadStream` for extracted CSVs) and
- *   reservoir-samples to {@link RESERVOIR_CAP} (so FR/ES countrywide work in bounded memory), then
- *   draws `--count` rows from the pool with the passed `random`. Ported from
- *   scripts/build-locale-slice.mjs.
- *
- *   The reservoir uses its OWN seeded PRNG ({@link makeMulberry32}, per part), independent of the
- *   emit `random`, so the input sample is reproducible WITHOUT perturbing the synth/order draws.
- *
- *   SURFACE DIVERSITY (#241): two per-country shape draws ride the emit loop, sized by the
- *   2026-07-02 format-diversity audit against the `openaddresses-{es,nl,it}-sample.jsonl` observed
- *   forms. ES: the OpenCage template comma-joins the house number (`CALLE MAYOR, 12`) but all 3,000
- *   eval rows space-join (`CALLE MAYOR 12`) — {@link ES_SPACE_JOIN_FRACTION} of native rows collapse
- *   the comma. NL: OA (and the eval, 3,000/3,000) glue the postcode (`1187LM`) while the national
- *   convention spaces it (`1187 LM`) — {@link NL_GLUED_POSTCODE_FRACTION} of rows keep the glued
- *   source shape, the rest the spaced conventional one. These draws are consumed ONLY for their
- *   country, so DE/FR emit streams are unchanged for a given seed.
- */
-import type { PathBuilderLike } from "path-ts";
-import type { CorpusRecipe } from "#recipes/scaffold";
-import { type LocaleBaseTuple, type SynthesizedLocaleRow } from "#synthesizers/german";
-/**
- * One per-country OA source part: either a cached `zip` + `csv` member (streamed out of the archive) or an extracted
- * plain `path` (streamed via `createReadStream`). Both carry the standard OA header
- * (LON,LAT,NUMBER,STREET,UNIT,CITY,DISTRICT,REGION,POSTCODE,ID,HASH). An optional `region` fallback covers countries
- * whose REGION column is empty (DE — the Bundesland is implied by the per-state file).
- */
-export interface LocalePart {
-    zip?: PathBuilderLike;
-    csv?: string;
-    path?: PathBuilderLike;
-    region?: string;
-    /**
-     * NZ — the OA DISTRICT column holds the city (`Auckland`) and CITY holds the suburb (`Birkenhead`). When set, map
-     * DISTRICT→locality and CITY→dependent_locality (falling back to CITY→locality when DISTRICT is empty, ~18% of NZ
-     * rows). Without this, the default CITY→locality mapping wrongly trains the suburb as the locality.
-     */
-    districtAsLocality?: boolean;
-    /**
-     * ES pedanía part only — the header is the RAW (un-conformed) CNIG export schema (`numero`, `tipo_vial`,
-     * `nombre_via`, `poblacion`, `municipio`, `comunidad_autonoma`, `cod_postal`), NOT the standard OA
-     * NUMBER/STREET/CITY/DISTRICT/REGION/POSTCODE header every other part uses. Verified 2026-07-22 by exact- coordinate
-     * cross-check: the OA-conformed `extracted/es/countrywide.csv` collapses CITY to `municipio` and drops `poblacion`
-     * (Spain's below-municipio núcleo/pedanía name) entirely, so the pedanía signal survives ONLY in this raw export.
-     * `street` is reconstructed as `tipo_vial + " " + nombre_via` (verified byte-identical to the conformed STREET column
-     * for the same row). CITY-analog = `poblacion`, DISTRICT-analog = `municipio`.
-     */
-    cnigRaw?: boolean;
-}
-export interface LocaleCountrySource {
-    source: string;
-    parts: LocalePart[];
-    /**
-     * The `corpus_version` stamped on emitted rows. DE/FR keep the historical `0.4.0` (regenerating those slices must
-     * stay lineage-identical); ES/IT/NL are the #241 staging lineage (`v0.9.9-es-it-nl`).
-     */
-    corpusVersion: string;
-    /**
-     * An ALTERNATE part list, read instead of {@link parts} when the `--district-as-locality` override is explicitly
-     * `true` for this invocation (see `run()`). ES-only for now — the standard `parts` entry can't supply real
-     * dependent-locality signal (its OA-conformed CSV drops `poblacion`; see {@link LocalePart.cnigRaw}), so the pedanía
-     * build reads a wholly different raw source instead of flipping the standard CITY/DISTRICT columns. `undefined` for
-     * every other country — the override then just forces `districtAsLocality` on the normal `parts`, as GB/NZ already do
-     * per-part.
-     */
-    pedaniaParts?: LocalePart[];
-}
-/**
- * OA CITY-noise normalization (#241) — the documented cleaning step, derived from the 2026-07-02 FULL-STREAM audit of
- * the ES (15.6M rows), IT (13.9M), and NL (9.1M) sources (not a hand-list). Returns the cleaned city, or `null` to drop
- * the tuple.
- *
- * Cleaned classes:
- *
- * 1. DROP pseudo-localities — the ES cadastral aggregates (`Comunidad de 09076, 09150 y 09578`, `Ledanía de …`; 0.06% of
- *    ES rows): any CITY containing a comma or a ≥4-digit run is a land-register aggregate, not a renderable city.
- *    Structural, locale-safe — NL's genuine `2e Valthermond` (one digit) survives; IT/NL have zero hits.
- * 2. STRIP a trailing parenthesized 1–3-letter admin code — the NL BAG province disambiguator (`Bergen (NH)`, `Rijswijk
- *    (GLD)` → `Bergen`, `Rijswijk`; 0.13% of NL rows). The analogue of the German Kreis/region-suffix class (#241 names
- *    `Rabenau Sachs` / `Weißwasser /O.L.`): an admin-region gloss glued onto the locality value that dirties locality
- *    labels.
- *
- * Audit-verified NON-noise, deliberately NOT cleaned (a naive suffix rule would mangle real names):
- *
- * - ES/IT city-ends-with-province (`Alhama de Almería`, `GENZANO DI ROMA`; ~0.8% each): genuine toponyms whose linking
- *   `de`/`di` makes them full names, unlike the German glued-abbreviation class.
- * - ES bilingual slash names (`Laudio/Llodio`; 2.16%): official co-names — the eval expects them verbatim.
- * - IT ALL-CAPS city casing (98.79% of the source, and the eval's observed form): casing is the #829 case-augmentation
- *   change, not this slice's.
- */
-export declare function cleanCityNoise(city: string): string | null;
-/**
- * Stream real tuples out of an OA source part and reservoir-sample to {@link RESERVOIR_CAP}. Reads the CSV row-by-row —
- * `readZipEntry | CSVSpliterator` for zip parts, `createReadStream | CSVSpliterator` for extracted parts (both bounded
- * memory) — and keeps a uniform random sample (Algorithm R) seeded by `rng`, separate from the emit loop's PRNG. NO
- * global dedup (a 25M-key Set would OOM; OA rows are near-unique). The city passes through {@link cleanCityNoise}; the
- * region falls back to `part.region` when the row's REGION cell is empty (DE).
- *
- * Exported for {@link locale.test.ts} — the CSV read path (quote handling, CRLF, region fallback) has no other test.
- */
-export declare function readTuples(part: LocalePart, rng: () => number): Promise<LocaleBaseTuple[]>;
-/**
- * Country-append fraction (the fr-admin-split #728 pattern, generalized to the locale recipe): mutates `synth` in
- * place, `countryFraction` of the time appending an explicit country surface form ("United Kingdom") to `raw` + a
- * `country` component — the model relearns to emit country WHEN the token is present without over-firing it on the
- * (still-majority) country-less rows. `countryFraction <= 0` (the default) short-circuits the `random()` draw away
- * entirely — no `synth` mutation and no RNG consumption — so every existing locale's emit stream stays byte-identical
- * to before this option existed. Exported for {@link locale.test.ts}.
- */
-export declare function applyCountryAppend(synth: SynthesizedLocaleRow, country: string, countryFraction: number, random: () => number): void;
-/**
- * Merge the `--district-as-locality` CLI override onto one part. `undefined` (flag absent) returns `part` unchanged
- * (same object — no allocation, no behavior change); `true`/`false` returns a shallow copy with `districtAsLocality`
- * forced to that value for this invocation only. Exported for {@link locale.test.ts}.
- */
-export declare function applyDistrictAsLocalityOverride(part: LocalePart, override: boolean | undefined): LocalePart;
-/**
- * Pick which part list a `--country` run reads: {@link LocaleCountrySource.pedaniaParts} when the override is explicitly
- * `true` AND the country registers one (ES only, so far), else the default `parts` — unchanged for every other
- * country/override combination. Exported for {@link locale.test.ts}.
- */
-export declare function resolveLocaleParts(countrySource: LocaleCountrySource, override: boolean | undefined): LocalePart[];
-/**
- * Slice recipe registered with the corpus builder — see the file header for the parse behaviour it exists to exercise,
- * and `description` below for the surface form it generates.
- */
-export declare const localeRecipe: CorpusRecipe;
-//# sourceMappingURL=locale.d.ts.map

package/out/recipes/locale.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"locale.d.ts","sourceRoot":"","sources":["../../lib/recipes/locale.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;GAyBG;AAQH,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,SAAS,CAAA;AAI9C,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,mBAAmB,CAAA;AACrD,OAAO,EAAE,KAAK,eAAe,EAAE,KAAK,oBAAoB,EAAuB,MAAM,sBAAsB,CAAA;AAI3G;;;;;GAKG;AACH,MAAM,WAAW,UAAU;IAC1B,GAAG,CAAC,EAAE,eAAe,CAAA;IACrB,GAAG,CAAC,EAAE,MAAM,CAAA;IACZ,IAAI,CAAC,EAAE,eAAe,CAAA;IACtB,MAAM,CAAC,EAAE,MAAM,CAAA;IACf;;;;OAIG;IACH,kBAAkB,CAAC,EAAE,OAAO,CAAA;IAC5B;;;;;;;;OAQG;IACH,OAAO,CAAC,EAAE,OAAO,CAAA;CACjB;AAED,MAAM,WAAW,mBAAmB;IACnC,MAAM,EAAE,MAAM,CAAA;IACd,KAAK,EAAE,UAAU,EAAE,CAAA;IACnB;;;OAGG;IACH,aAAa,EAAE,MAAM,CAAA;IACrB;;;;;;;OAOG;IACH,YAAY,CAAC,EAAE,UAAU,EAAE,CAAA;CAC3B;AA6FD;;;;;;;;;;;;;;;;;;;;;;GAsBG;AACH,wBAAgB,cAAc,CAAC,IAAI,EAAE,MAAM,GAAG,MAAM,GAAG,IAAI,CAM1D;AAgBD;;;;;;;;GAQG;AACH,wBAAsB,UAAU,CAAC,IAAI,EAAE,UAAU,EAAE,GAAG,EAAE,MAAM,MAAM,GAAG,OAAO,CAAC,eAAe,EAAE,CAAC,CAkJhG;AAED;;;;;;;GAOG;AACH,wBAAgB,kBAAkB,CACjC,KAAK,EAAE,oBAAoB,EAC3B,OAAO,EAAE,MAAM,EACf,eAAe,EAAE,MAAM,EACvB,MAAM,EAAE,MAAM,MAAM,GAClB,IAAI,CAgBN;AAED;;;;GAIG;AACH,wBAAgB,+BAA+B,CAAC,IAAI,EAAE,UAAU,EAAE,QAAQ,EAAE,OAAO,GAAG,SAAS,GAAG,UAAU,CAE3G;AAED;;;;GAIG;AACH,wBAAgB,kBAAkB,CAAC,aAAa,EAAE,mBAAmB,EAAE,QAAQ,EAAE,OAAO,GAAG,SAAS,GAAG,UAAU,EAAE,CAElH;AAED;;;GAGG;AACH,eAAO,MAAM,YAAY,EAAE,YAmK1B,CAAA"}