@mailwoman/corpus 10.0.0 → 10.2.0

package/out/lu/adapters/bd-adresses/adapter.d.ts

@@ -0,0 +1,109 @@
+/**
+ * @copyright Sister Software
+ * @license AGPL-3.0
+ * @author Teffen Ellis, et al.
+ *
+ * `bd-adresses`: the Administration du cadastre et de la topographie's BD-Adresses CSV adapter,
+ * covering Luxembourg.
+ *
+ * Input is the semicolon-separated `addresses.csv` the ACT publishes for the whole country through
+ * `data.public.lu`, beside a GeoJSON and a shapefile carrying the same records. The adapter reads
+ * `rue` as the street, `numero` as the house number, and `code_postal` with `localite` as the
+ * postcode and its postal locality.
+ *
+ * The publisher describes BD-Adresses as `un sous-ensemble des adresses figurant dans le registre
+ * national des localités et des rues`, enriched with coordinates. The row count is therefore the
+ * published subset rather than Luxembourg's address count.
+ *
+ * `data.public.lu` states `cc-zero` for the dataset, and the address-source register elects
+ * `CC0-1.0` for `lu-property-building-1` on that statement, so the adapter records that license on
+ * every row.
+ *
+ * Luxembourg writes the house number before the street. `STREET_ORDERS` records that order as
+ * `number-first`, and `COMMA_JOINED_STREET_COUNTRIES` does not list it. `formatAddressRow` therefore
+ * renders `20 Kaesfurterstrooss, L-9755 Hupperdange`.
+ *
+ * ## The file holds one street name, in no single language
+ *
+ * Luxembourg is trilingual, and the CSV declares one street column.
+ * There is no `rue_fr`, `rue_de` or `rue_lb` to choose between, so the adapter emits `rue` as the
+ * publisher wrote it.
+ *
+ * What that column holds is mixed.
+ * Measured over the 179,622 rows of the 2026-09-28 edition with a generic-word probe rather than a
+ * language classifier: 133,449 values open with a French street generic such as `Rue`, `Avenue` or
+ * `Montée`, 18,629 close with a Luxembourgish one such as `-strooss`, `-wee` or `-gaass`, and the
+ * remaining 27,544 match neither probe and include plainly Luxembourgish forms such as
+ * `Am Stengber`, `An Hiesel` and `Um Reebou`.
+ * No column marks a row's language, so `locale` reads `und-LU`.
+ * `und` is BCP-47's undetermined language subtag, and the repository already stores it in
+ * `#recipes/bare/country` and `#tools/postcode-triples`.
+ * A `fr-LU` or `lb-LU` label would assert a language for a name the publisher did not
+ * tag.
+ *
+ * ## The number comes from `numero`, and never from the street line
+ *
+ * `rue` and `numero` are separate columns and the publisher never composes them, so neither
+ * `splitStreetLine` nor `composeHouseNumber` applies here: there is no line to split and no second
+ * column to join.
+ * A street split would also be wrong in both directions.
+ * 175 `rue` values end in a digit, among them `Route Nationale 1`, `Rue du 9 août 2019` and
+ * `Cité des Sacrifiés 1940-1945`.
+ * `numero` already holds its own suffix as one token: 162,273 rows hold bare digits, 17,087 a
+ * trailing letter such as `12A`, 239 a hyphenated range such as `22-26`, eight the word `BIS` or
+ * `bis`, and 15 are empty.
+ *
+ * `id_geoportail` is the publisher's own composition of the same two values.
+ * It ends in `_<id_caclr_rue>_<numero>` on 179,607 of the 179,607 numbered rows, so the number this
+ * adapter emits agrees with the publisher's where the publisher states one.
+ * The file publishes no preformatted address line, so a rendered single line has no publisher-side
+ * counterpart to be asserted against.
+ *
+ * ## The file needs a quote-aware parser
+ *
+ * 22 rows RFC 4180-quote `rue` and double its interior quotes:
+ * `"Cité ""Pënscherbierg""";3;Wilwerwiltz;…`.
+ * Those rows hold no semicolon inside the quotes, so a naive split on `;` still yields 13 fields
+ * and passes a field-count check while handing the street to the corpus as
+ * `"Cité ""Pënscherbierg"""`.
+ * `CSVSpliterator` unescapes it to `Cité "Pënscherbierg"`.
+ * The same edition holds 0 fields with an embedded newline and 0 carriage returns in the whole
+ * file.
+ *
+ * The file opens with a UTF-8 byte-order mark, the three bytes `EF BB BF`, which the reader strips.
+ * The first record's own first key therefore reads `rue` rather than the mark followed by `rue`.
+ * That mark is why the portal's byte count exceeds by three what a reader that discards it reports.
+ *
+ * The adapter streams with `CSVSpliterator.fromAsync`.
+ * It honors `opts.limit`, `opts.signal` and `opts.country`.
+ */
+import { type CorpusAdapter } from "#types";
+/**
+ * Registry id for this adapter.
+ *
+ * Stamped into every row it emits, so a corpus record can be traced back to the dataset it came from.
+ */
+export declare const BD_ADRESSES_ADAPTER_ID = "bd-adresses";
+/**
+ * The license the address-source register elects for this publication.
+ */
+export declare const BD_ADRESSES_DEFAULT_LICENSE = "CC0-1.0";
+/**
+ * Every jurisdiction this adapter emits, checked against a caller's `--country`.
+ *
+ * BD-Adresses is one national file and Luxembourg has no dependency, so this is a single entry
+ * where `ban`, `matrikkelen` and `ryhti` each hold more than one.
+ */
+export declare const BD_ADRESSES_COUNTRIES: readonly string[];
+/**
+ * The columns `rows` indexes by name, in the order the published header declares them.
+ *
+ * A caller that downloads the file checks its header against this list.
+ */
+export declare const BD_ADRESSES_REQUIRED_COLUMNS: readonly string[];
+export declare function createBDAdressesAdapter(): CorpusAdapter;
+/**
+ * The configured adapter instance registered with the corpus builder.
+ */
+export declare const bdAdressesAdapter: CorpusAdapter;
+//# sourceMappingURL=adapter.d.ts.map

package/out/lu/adapters/bd-adresses/adapter.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"adapter.d.ts","sourceRoot":"","sources":["../../../../lib/lu/adapters/bd-adresses/adapter.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8EG;AAQH,OAAO,EAAuD,KAAK,aAAa,EAAiB,MAAM,QAAQ,CAAA;AAE/G;;;;GAIG;AACH,eAAO,MAAM,sBAAsB,gBAAgB,CAAA;AAEnD;;GAEG;AACH,eAAO,MAAM,2BAA2B,YAAY,CAAA;AAEpD;;;;;GAKG;AACH,eAAO,MAAM,qBAAqB,EAAE,SAAS,MAAM,EAAW,CAAA;AAqB9D;;;;GAIG;AACH,eAAO,MAAM,4BAA4B,EAAE,SAAS,MAAM,EAOzD,CAAA;AAED,wBAAgB,uBAAuB,IAAI,aAAa,CA6FvD;AAED;;GAEG;AACH,eAAO,MAAM,iBAAiB,eAA4B,CAAA"}

package/out/lu/adapters/bd-adresses/adapter.js

@@ -0,0 +1,196 @@
+/**
+ * @copyright Sister Software
+ * @license AGPL-3.0
+ * @author Teffen Ellis, et al.
+ *
+ * `bd-adresses`: the Administration du cadastre et de la topographie's BD-Adresses CSV adapter,
+ * covering Luxembourg.
+ *
+ * Input is the semicolon-separated `addresses.csv` the ACT publishes for the whole country through
+ * `data.public.lu`, beside a GeoJSON and a shapefile carrying the same records. The adapter reads
+ * `rue` as the street, `numero` as the house number, and `code_postal` with `localite` as the
+ * postcode and its postal locality.
+ *
+ * The publisher describes BD-Adresses as `un sous-ensemble des adresses figurant dans le registre
+ * national des localités et des rues`, enriched with coordinates. The row count is therefore the
+ * published subset rather than Luxembourg's address count.
+ *
+ * `data.public.lu` states `cc-zero` for the dataset, and the address-source register elects
+ * `CC0-1.0` for `lu-property-building-1` on that statement, so the adapter records that license on
+ * every row.
+ *
+ * Luxembourg writes the house number before the street. `STREET_ORDERS` records that order as
+ * `number-first`, and `COMMA_JOINED_STREET_COUNTRIES` does not list it. `formatAddressRow` therefore
+ * renders `20 Kaesfurterstrooss, L-9755 Hupperdange`.
+ *
+ * ## The file holds one street name, in no single language
+ *
+ * Luxembourg is trilingual, and the CSV declares one street column.
+ * There is no `rue_fr`, `rue_de` or `rue_lb` to choose between, so the adapter emits `rue` as the
+ * publisher wrote it.
+ *
+ * What that column holds is mixed.
+ * Measured over the 179,622 rows of the 2026-09-28 edition with a generic-word probe rather than a
+ * language classifier: 133,449 values open with a French street generic such as `Rue`, `Avenue` or
+ * `Montée`, 18,629 close with a Luxembourgish one such as `-strooss`, `-wee` or `-gaass`, and the
+ * remaining 27,544 match neither probe and include plainly Luxembourgish forms such as
+ * `Am Stengber`, `An Hiesel` and `Um Reebou`.
+ * No column marks a row's language, so `locale` reads `und-LU`.
+ * `und` is BCP-47's undetermined language subtag, and the repository already stores it in
+ * `#recipes/bare/country` and `#tools/postcode-triples`.
+ * A `fr-LU` or `lb-LU` label would assert a language for a name the publisher did not
+ * tag.
+ *
+ * ## The number comes from `numero`, and never from the street line
+ *
+ * `rue` and `numero` are separate columns and the publisher never composes them, so neither
+ * `splitStreetLine` nor `composeHouseNumber` applies here: there is no line to split and no second
+ * column to join.
+ * A street split would also be wrong in both directions.
+ * 175 `rue` values end in a digit, among them `Route Nationale 1`, `Rue du 9 août 2019` and
+ * `Cité des Sacrifiés 1940-1945`.
+ * `numero` already holds its own suffix as one token: 162,273 rows hold bare digits, 17,087 a
+ * trailing letter such as `12A`, 239 a hyphenated range such as `22-26`, eight the word `BIS` or
+ * `bis`, and 15 are empty.
+ *
+ * `id_geoportail` is the publisher's own composition of the same two values.
+ * It ends in `_<id_caclr_rue>_<numero>` on 179,607 of the 179,607 numbered rows, so the number this
+ * adapter emits agrees with the publisher's where the publisher states one.
+ * The file publishes no preformatted address line, so a rendered single line has no publisher-side
+ * counterpart to be asserted against.
+ *
+ * ## The file needs a quote-aware parser
+ *
+ * 22 rows RFC 4180-quote `rue` and double its interior quotes:
+ * `"Cité ""Pënscherbierg""";3;Wilwerwiltz;…`.
+ * Those rows hold no semicolon inside the quotes, so a naive split on `;` still yields 13 fields
+ * and passes a field-count check while handing the street to the corpus as
+ * `"Cité ""Pënscherbierg"""`.
+ * `CSVSpliterator` unescapes it to `Cité "Pënscherbierg"`.
+ * The same edition holds 0 fields with an embedded newline and 0 carriage returns in the whole
+ * file.
+ *
+ * The file opens with a UTF-8 byte-order mark, the three bytes `EF BB BF`, which the reader strips.
+ * The first record's own first key therefore reads `rue` rather than the mark followed by `rue`.
+ * That mark is why the portal's byte count exceeds by three what a reader that discards it reports.
+ *
+ * The adapter streams with `CSVSpliterator.fromAsync`.
+ * It honors `opts.limit`, `opts.signal` and `opts.country`.
+ */
+import { formatAddressRow } from "@mailwoman/codex/address/format";
+import { CSVSpliterator } from "spliterator";
+import { UnsupportedCountryError } from "#adapters/errors";
+import { stableSourceID } from "#adapters/source-id";
+import { SourceRegister } from "#registers";
+import { AddressRole, SurfaceOrigin } from "#types";
+/**
+ * Registry id for this adapter.
+ *
+ * Stamped into every row it emits, so a corpus record can be traced back to the dataset it came from.
+ */
+export const BD_ADRESSES_ADAPTER_ID = "bd-adresses";
+/**
+ * The license the address-source register elects for this publication.
+ */
+export const BD_ADRESSES_DEFAULT_LICENSE = "CC0-1.0";
+/**
+ * Every jurisdiction this adapter emits, checked against a caller's `--country`.
+ *
+ * BD-Adresses is one national file and Luxembourg has no dependency, so this is a single entry
+ * where `ban`, `matrikkelen` and `ryhti` each hold more than one.
+ */
+export const BD_ADRESSES_COUNTRIES = ["LU"];
+/**
+ * The columns `rows` indexes by name, in the order the published header declares them.
+ *
+ * A caller that downloads the file checks its header against this list.
+ */
+export const BD_ADRESSES_REQUIRED_COLUMNS = [
+    "rue",
+    "numero",
+    "localite",
+    "code_postal",
+    "id_caclr_rue",
+    "id_geoportail",
+];
+export function createBDAdressesAdapter() {
+    return {
+        id: BD_ADRESSES_ADAPTER_ID,
+        defaultLicense: BD_ADRESSES_DEFAULT_LICENSE,
+        addressRole: AddressRole.Premise,
+        register: SourceRegister.BDAdresses,
+        surface: SurfaceOrigin.Rendered,
+        description: "BD-Adresses (Administration du cadastre et de la topographie): street-level addresses for Luxembourg.",
+        async *rows(opts) {
+            if (opts.country && !BD_ADRESSES_COUNTRIES.includes(opts.country)) {
+                throw new UnsupportedCountryError(BD_ADRESSES_ADAPTER_ID, BD_ADRESSES_COUNTRIES, opts.country);
+            }
+            const rows = CSVSpliterator.fromAsync(opts.inputPath, {
+                normalizeKeys: false,
+                columnDelimiter: ";",
+            });
+            let emitted = 0;
+            let unnamed = 0;
+            try {
+                for await (const record of rows) {
+                    if (opts.signal?.aborted)
+                        break;
+                    if (opts.limit !== undefined && emitted >= opts.limit)
+                        break;
+                    const street = (record.rue ?? "").trim();
+                    const house = (record.numero ?? "").trim();
+                    const postcode = (record.code_postal ?? "").trim();
+                    const locality = (record.localite ?? "").trim();
+                    if (!street) {
+                        unnamed += 1;
+                        continue;
+                    }
+                    if (!postcode && !locality)
+                        continue;
+                    const components = {};
+                    // 15 of the 179,622 rows publish an empty `numero`, among them `Al Géidgen`
+                    // in Goedange and `Jaanshaff` in Walferdange.
+                    // These describe a place, so they keep their street and postcode
+                    // rather than being assigned a number.
+                    if (house) {
+                        components.house_number = house;
+                    }
+                    components.street = street;
+                    if (postcode) {
+                        components.postcode = postcode;
+                    }
+                    if (locality) {
+                        components.locality = locality;
+                    }
+                    const rendered = formatAddressRow(components, "LU", { singleLine: true });
+                    if (!rendered)
+                        continue;
+                    const { raw, components: aligned } = rendered;
+                    const seed = (record.id_geoportail ?? "").trim();
+                    const sourceID = seed ? `${BD_ADRESSES_ADAPTER_ID}-${seed}` : stableSourceID(BD_ADRESSES_ADAPTER_ID, aligned);
+                    yield {
+                        raw,
+                        components: aligned,
+                        country: "LU",
+                        locale: "und-LU",
+                        source: BD_ADRESSES_ADAPTER_ID,
+                        source_id: sourceID,
+                        corpus_version: "",
+                        license: BD_ADRESSES_DEFAULT_LICENSE,
+                    };
+                    emitted++;
+                }
+            }
+            finally {
+                if (unnamed > 0) {
+                    process.stderr.write(`  bd-adresses: ${unnamed} rows publish an empty rue and were dropped, ${emitted} kept\n`);
+                }
+            }
+        },
+    };
+}
+/**
+ * The configured adapter instance registered with the corpus builder.
+ */
+export const bdAdressesAdapter = createBDAdressesAdapter();
+//# sourceMappingURL=adapter.js.map

package/out/lu/adapters/bd-adresses/adapter.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"adapter.js","sourceRoot":"","sources":["../../../../lib/lu/adapters/bd-adresses/adapter.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8EG;AAEH,OAAO,EAAE,gBAAgB,EAAE,MAAM,iCAAiC,CAAA;AAClE,OAAO,EAAE,cAAc,EAAE,MAAM,aAAa,CAAA;AAE5C,OAAO,EAAE,uBAAuB,EAAE,MAAM,kBAAkB,CAAA;AAC1D,OAAO,EAAE,cAAc,EAAE,MAAM,qBAAqB,CAAA;AACpD,OAAO,EAAE,cAAc,EAAE,MAAM,YAAY,CAAA;AAC3C,OAAO,EAAE,WAAW,EAA8D,aAAa,EAAE,MAAM,QAAQ,CAAA;AAE/G;;;;GAIG;AACH,MAAM,CAAC,MAAM,sBAAsB,GAAG,aAAa,CAAA;AAEnD;;GAEG;AACH,MAAM,CAAC,MAAM,2BAA2B,GAAG,SAAS,CAAA;AAEpD;;;;;GAKG;AACH,MAAM,CAAC,MAAM,qBAAqB,GAAsB,CAAC,IAAI,CAAC,CAAA;AAqB9D;;;;GAIG;AACH,MAAM,CAAC,MAAM,4BAA4B,GAAsB;IAC9D,KAAK;IACL,QAAQ;IACR,UAAU;IACV,aAAa;IACb,cAAc;IACd,eAAe;CACf,CAAA;AAED,MAAM,UAAU,uBAAuB;IACtC,OAAO;QACN,EAAE,EAAE,sBAAsB;QAC1B,cAAc,EAAE,2BAA2B;QAC3C,WAAW,EAAE,WAAW,CAAC,OAAO;QAChC,QAAQ,EAAE,cAAc,CAAC,UAAU;QACnC,OAAO,EAAE,aAAa,CAAC,QAAQ;QAC/B,WAAW,EACV,uGAAuG;QAExG,KAAK,CAAC,CAAC,IAAI,CAAC,IAAoB;YAC/B,IAAI,IAAI,CAAC,OAAO,IAAI,CAAC,qBAAqB,CAAC,QAAQ,CAAC,IAAI,CAAC,OAAO,CAAC,EAAE,CAAC;gBACnE,MAAM,IAAI,uBAAuB,CAAC,sBAAsB,EAAE,qBAAqB,EAAE,IAAI,CAAC,OAAO,CAAC,CAAA;YAC/F,CAAC;YAED,MAAM,IAAI,GAAG,cAAc,CAAC,SAAS,CAAC,IAAI,CAAC,SAAS,EAAE;gBACrD,aAAa,EAAE,KAAK;gBACpB,eAAe,EAAE,GAAG;aACpB,CAAC,CAAA;YAEF,IAAI,OAAO,GAAG,CAAC,CAAA;YACf,IAAI,OAAO,GAAG,CAAC,CAAA;YAEf,IAAI,CAAC;gBACJ,IAAI,KAAK,EAAE,MAAM,MAAM,IAAI,IAAoC,EAAE,CAAC;oBACjE,IAAI,IAAI,CAAC,MAAM,EAAE,OAAO;wBAAE,MAAK;oBAE/B,IAAI,IAAI,CAAC,KAAK,KAAK,SAAS,IAAI,OAAO,IAAI,IAAI,CAAC,KAAK;wBAAE,MAAK;oBAE5D,MAAM,MAAM,GAAG,CAAC,MAAM,CAAC,GAAG,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAA;oBACxC,MAAM,KAAK,GAAG,CAAC,MAAM,CAAC,MAAM,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAA;oBAC1C,MAAM,QAAQ,GAAG,CAAC,MAAM,CAAC,WAAW,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAA;oBAClD,MAAM,QAAQ,GAAG,CAAC,MAAM,CAAC,QAAQ,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAA;oBAE/C,IAAI,CAAC,MAAM,EAAE,CAAC;wBACb,OAAO,IAAI,CAAC,CAAA;wBAEZ,SAAQ;oBACT,CAAC;oBAED,IAAI,CAAC,QAAQ,IAAI,CAAC,QAAQ;wBAAE,SAAQ;oBAEpC,MAAM,UAAU,GAA+B,EAAE,CAAA;oBAEjD,4EAA4E;oBAC5E,8CAA8C;oBAC9C,iEAAiE;oBACjE,uCAAuC;oBACvC,IAAI,KAAK,EAAE,CAAC;wBACX,UAAU,CAAC,YAAY,GAAG,KAAK,CAAA;oBAChC,CAAC;oBAED,UAAU,CAAC,MAAM,GAAG,MAAM,CAAA;oBAE1B,IAAI,QAAQ,EAAE,CAAC;wBACd,UAAU,CAAC,QAAQ,GAAG,QAAQ,CAAA;oBAC/B,CAAC;oBAED,IAAI,QAAQ,EAAE,CAAC;wBACd,UAAU,CAAC,QAAQ,GAAG,QAAQ,CAAA;oBAC/B,CAAC;oBAED,MAAM,QAAQ,GAAG,gBAAgB,CAAC,UAAU,EAAE,IAAI,EAAE,EAAE,UAAU,EAAE,IAAI,EAAE,CAAC,CAAA;oBAEzE,IAAI,CAAC,QAAQ;wBAAE,SAAQ;oBAEvB,MAAM,EAAE,GAAG,EAAE,UAAU,EAAE,OAAO,EAAE,GAAG,QAAQ,CAAA;oBAC7C,MAAM,IAAI,GAAG,CAAC,MAAM,CAAC,aAAa,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAA;oBAEhD,MAAM,QAAQ,GAAG,IAAI,CAAC,CAAC,CAAC,GAAG,sBAAsB,IAAI,IAAI,EAAE,CAAC,CAAC,CAAC,cAAc,CAAC,sBAAsB,EAAE,OAAO,CAAC,CAAA;oBAE7G,MAAM;wBACL,GAAG;wBACH,UAAU,EAAE,OAAO;wBACnB,OAAO,EAAE,IAAI;wBACb,MAAM,EAAE,QAAQ;wBAChB,MAAM,EAAE,sBAAsB;wBAC9B,SAAS,EAAE,QAAQ;wBACnB,cAAc,EAAE,EAAE;wBAClB,OAAO,EAAE,2BAA2B;qBACpC,CAAA;oBAED,OAAO,EAAE,CAAA;gBACV,CAAC;YACF,CAAC;oBAAS,CAAC;gBACV,IAAI,OAAO,GAAG,CAAC,EAAE,CAAC;oBACjB,OAAO,CAAC,MAAM,CAAC,KAAK,CACnB,kBAAkB,OAAO,gDAAgD,OAAO,SAAS,CACzF,CAAA;gBACF,CAAC;YACF,CAAC;QACF,CAAC;KACD,CAAA;AACF,CAAC;AAED;;GAEG;AACH,MAAM,CAAC,MAAM,iBAAiB,GAAG,uBAAuB,EAAE,CAAA"}

package/out/lu/tools/fetch/bd-adresses.d.ts

@@ -0,0 +1,165 @@
+/**
+ * @copyright Sister Software
+ * @license AGPL-3.0
+ * @author Teffen Ellis, et al.
+ *
+ * Download the Administration du cadastre et de la topographie's BD-Adresses CSV for the
+ * `bd-adresses` adapter.
+ *
+ * The acquisition belongs under a `tools/` root. `@mailwoman/corpus` is one of the
+ * `TOOLING_PACKAGES` in `dependency-cruiser.config.mjs`, which keep their tooling under `lib/`, so
+ * this module sits beside `#fi/tools/fetch/ryhti` rather than in an `sdk/` root the workspace does
+ * not declare.
+ *
+ * `data.public.lu` states `cc-zero` for this dataset. The address-source register elects
+ * `CC0-1.0` for it under `lu-property-building-1`. CC0 asks for no attribution, and crediting the ACT
+ * remains good practice, so the manifest records the publisher beside the license.
+ *
+ * Four measured properties decide what this module does:
+ *
+ * 1. **The download URL holds the edition's own timestamp.** The dataset is republished weekly
+ *    and the CSV resource's `url` reads
+ *    `https://download.data.public.lu/resources/adresses-georeferencees-bd-adresses/20260928-023119/addresses.csv`,
+ *    where the path segment is the publication time. A hardcoded URL therefore pins one edition and
+ *    404s on the next. The module reads {@linkcode LU_BD_ADRESSES_DATASET_URL} and takes the `url`
+ *    of the resource whose `format` is `csv`. The resource's own `latest`,
+ *    `https://data.public.lu/fr/datasets/r/<resource id>`, answers http 302 to the same dated path,
+ *    and the dated path is the one downloaded, so the bytes and the checksum recorded beside them
+ *    belong to one edition.
+ * 2. **The publisher states an md5 for the file.** The resource records
+ *    `checksum: {type: "md5", value: "…"}` and a `filesize`, measured as
+ *    `c2df2f3c18b846144bb5d5dc57fc14ca` and 28,186,970 on the 2026-09-28 edition. The transfer is
+ *    checked against both, so a truncated body is a reported failure rather than a short CSV that
+ *    parses. A publisher that one day states a digest of another type is reported rather than
+ *    skipped silently.
+ * 3. **The file opens with a UTF-8 byte-order mark**, `EF BB BF`. The portal's `filesize` counts it.
+ *    A reader that discards the mark therefore reports three bytes fewer than the portal does.
+ *    The module writes the body to disk unaltered and checks the digest over the bytes as delivered.
+ * 4. **The re-run check is the stated checksum against the manifest.** The dataset API answer is
+ *    roughly 15 KB, so comparing it costs one small request instead of 28 MB.
+ *
+ * The CSV is the only one of the three bulk files this reads. The GeoJSON is 80,836,067 bytes and
+ * the shapefile 19,681,289 for the same records, and `#lu/adapters/bd-adresses/adapter` hands
+ * `opts.inputPath` to `CSVSpliterator.fromAsync`.
+ */
+import { APIClient } from "@mailwoman/core/api";
+import { PathBuilder, type PathBuilderLike } from "path-ts";
+import type { BaseFetchOptions, FetchSummary, SourceManifest } from "#tools/fetch/download";
+/**
+ * The dataset record that states the current edition of each bulk file.
+ */
+export declare const LU_BD_ADRESSES_DATASET_URL = "https://data.public.lu/api/1/datasets/adresses-georeferencees-bd-adresses/";
+/**
+ * The file the adapter reads.
+ *
+ * It is the publisher's own name for the resource.
+ */
+export declare const LU_BD_ADRESSES_CSV_FILENAME = "addresses.csv";
+/**
+ * The publisher the manifest credits.
+ */
+export declare const LU_BD_ADRESSES_ATTRIBUTION = "Administration du cadastre et de la topographie";
+/**
+ * The semicolon the publisher delimits with.
+ */
+export declare const LU_BD_ADRESSES_DELIMITER = ";";
+/**
+ * One bulk file as the dataset record describes it.
+ */
+export interface BDAdressesResource {
+    /**
+     * The dated download URL.
+     * It identifies one edition.
+     */
+    url: string;
+    /**
+     * The publisher's stated byte count, `null` when the record omits it.
+     */
+    filesize: number | null;
+    /**
+     * The publisher's stated md5, `null` when the record states a digest of another type or none.
+     */
+    md5: string | null;
+    /**
+     * The resource's own last-modified time, as the record states it.
+     */
+    lastModified: string | null;
+}
+/**
+ * The manifest this module writes beside the CSV.
+ */
+export interface BDAdressesManifest extends SourceManifest {
+    license: string;
+    attribution: string;
+    dataset_url: string;
+    /**
+     * The md5 the dataset record stated.
+     * The downloaded bytes were checked against it.
+     */
+    publisher_md5: string | null;
+    /**
+     * The byte count the dataset record stated, beside the `bytes` that arrived.
+     */
+    publisher_filesize: number | null;
+    last_modified: string | null;
+    columns: readonly string[];
+}
+/**
+ * The shape the dataset API answers with, narrowed to the fields this module reads.
+ */
+interface DatasetRecord {
+    resources?: readonly {
+        format?: string;
+        url?: string;
+        filesize?: number;
+        checksum?: {
+            type?: string;
+            value?: string;
+        };
+        last_modified?: string;
+    }[];
+}
+/**
+ * The CSV resource of a dataset record.
+ *
+ * @throws When the record holds no `csv` resource with a URL, because the publisher
+ * having moved the file is a reported failure rather than an empty transfer.
+ */
+export declare function readBDAdressesResource(record: DatasetRecord): BDAdressesResource;
+export interface DownloadBDAdressesOptions {
+    outputDir: PathBuilderLike;
+    /**
+     * Download although the manifest already records this edition's checksum.
+     */
+    force?: boolean;
+    retries?: number;
+    retryDelayMs?: number;
+    signal?: AbortSignal;
+    report?: (line: string) => void;
+}
+/**
+ * Download the current edition of `addresses.csv` into `options.outputDir`.
+ *
+ * `client` is the caller's, so a test supplies its own and {@linkcode fetchBDAdresses}
+ * decides the retry policy.
+ */
+export declare function downloadBDAdresses(client: Pick<APIClient, "fetch">, options: DownloadBDAdressesOptions): Promise<FetchSummary>;
+export interface FetchBDAdressesOptions extends BaseFetchOptions {
+    force?: boolean;
+    retries?: number;
+    signal?: AbortSignal;
+}
+/**
+ * Download BD-Adresses into `<outRoot>/bd-adresses/`.
+ *
+ * The registry entry point.
+ * `#lu/adapters/bd-adresses/adapter` is pointed at {@linkcode bdAdressesInputPath},
+ * the `addresses.csv` inside that directory, rather than at the directory itself.
+ */
+export declare function fetchBDAdresses(options: FetchBDAdressesOptions, report?: (line: string) => void): Promise<FetchSummary>;
+/**
+ * The file `#lu/adapters/bd-adresses/adapter` reads, under a fetch run's `outRoot`.
+ */
+export declare function bdAdressesInputPath(outRoot: PathBuilder): PathBuilder;
+export {};
+//# sourceMappingURL=bd-adresses.d.ts.map

package/out/lu/tools/fetch/bd-adresses.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"bd-adresses.d.ts","sourceRoot":"","sources":["../../../../lib/lu/tools/fetch/bd-adresses.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2CG;AAEH,OAAO,EAAE,SAAS,EAAE,MAAM,qBAAqB,CAAA;AAK/C,OAAO,EAAE,WAAW,EAAE,KAAK,eAAe,EAAE,MAAM,SAAS,CAAA;AAO3D,OAAO,KAAK,EAAE,gBAAgB,EAAE,YAAY,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAA;AAI3F;;GAEG;AACH,eAAO,MAAM,0BAA0B,+EAA+E,CAAA;AAOtH;;;;GAIG;AACH,eAAO,MAAM,2BAA2B,kBAAkB,CAAA;AAE1D;;GAEG;AACH,eAAO,MAAM,0BAA0B,oDAAoD,CAAA;AAE3F;;GAEG;AACH,eAAO,MAAM,wBAAwB,MAAM,CAAA;AAE3C;;GAEG;AACH,MAAM,WAAW,kBAAkB;IAClC;;;OAGG;IACH,GAAG,EAAE,MAAM,CAAA;IACX;;OAEG;IACH,QAAQ,EAAE,MAAM,GAAG,IAAI,CAAA;IACvB;;OAEG;IACH,GAAG,EAAE,MAAM,GAAG,IAAI,CAAA;IAClB;;OAEG;IACH,YAAY,EAAE,MAAM,GAAG,IAAI,CAAA;CAC3B;AAED;;GAEG;AACH,MAAM,WAAW,kBAAmB,SAAQ,cAAc;IACzD,OAAO,EAAE,MAAM,CAAA;IACf,WAAW,EAAE,MAAM,CAAA;IACnB,WAAW,EAAE,MAAM,CAAA;IACnB;;;OAGG;IACH,aAAa,EAAE,MAAM,GAAG,IAAI,CAAA;IAC5B;;OAEG;IACH,kBAAkB,EAAE,MAAM,GAAG,IAAI,CAAA;IACjC,aAAa,EAAE,MAAM,GAAG,IAAI,CAAA;IAC5B,OAAO,EAAE,SAAS,MAAM,EAAE,CAAA;CAC1B;AAED;;GAEG;AACH,UAAU,aAAa;IACtB,SAAS,CAAC,EAAE,SAAS;QACpB,MAAM,CAAC,EAAE,MAAM,CAAA;QACf,GAAG,CAAC,EAAE,MAAM,CAAA;QACZ,QAAQ,CAAC,EAAE,MAAM,CAAA;QACjB,QAAQ,CAAC,EAAE;YAAE,IAAI,CAAC,EAAE,MAAM,CAAC;YAAC,KAAK,CAAC,EAAE,MAAM,CAAA;SAAE,CAAA;QAC5C,aAAa,CAAC,EAAE,MAAM,CAAA;KACtB,EAAE,CAAA;CACH;AAED;;;;;GAKG;AACH,wBAAgB,sBAAsB,CAAC,MAAM,EAAE,aAAa,GAAG,kBAAkB,CAqBhF;AAED,MAAM,WAAW,yBAAyB;IACzC,SAAS,EAAE,eAAe,CAAA;IAC1B;;OAEG;IACH,KAAK,CAAC,EAAE,OAAO,CAAA;IACf,OAAO,CAAC,EAAE,MAAM,CAAA;IAChB,YAAY,CAAC,EAAE,MAAM,CAAA;IACrB,MAAM,CAAC,EAAE,WAAW,CAAA;IACpB,MAAM,CAAC,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,IAAI,CAAA;CAC/B;AAED;;;;;GAKG;AACH,wBAAsB,kBAAkB,CACvC,MAAM,EAAE,IAAI,CAAC,SAAS,EAAE,OAAO,CAAC,EAChC,OAAO,EAAE,yBAAyB,GAChC,OAAO,CAAC,YAAY,CAAC,CAwHvB;AAED,MAAM,WAAW,sBAAuB,SAAQ,gBAAgB;IAC/D,KAAK,CAAC,EAAE,OAAO,CAAA;IACf,OAAO,CAAC,EAAE,MAAM,CAAA;IAChB,MAAM,CAAC,EAAE,WAAW,CAAA;CACpB;AAED;;;;;;GAMG;AACH,wBAAsB,eAAe,CACpC,OAAO,EAAE,sBAAsB,EAC/B,MAAM,CAAC,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,IAAI,GAC7B,OAAO,CAAC,YAAY,CAAC,CAavB;AAED;;GAEG;AACH,wBAAgB,mBAAmB,CAAC,OAAO,EAAE,WAAW,GAAG,WAAW,CAErE"}