@bytebrand/export-cars 1.26.47 → 1.26.48

package/src/transformers/marktde.transformer.ts

@@ -1,255 +1,255 @@
-import { get, toNumber, toString } from 'lodash';
-import { DataSchemes } from '@bytebrand/car-schema';
-
-import { OpenAiImageSelection } from './openai-image-fields';
-import { envkvCo2Value, envkvFuelConsumptionValue } from './google-feed-envkv';
-
-/**
- * markt.de feed (ADV-328) on top of the legacy mobile.de-style CSV row that MobileTransformer builds
- * from the `mobile_yescar` mapping sheet (ACF-4370). Only the columns below are replaced; every other
- * column, the column order and the row format stay as they were. Pure functions, no I/O.
- *
- * Column indexes are 0-based (the sheet's importField): 1 = B, 10 = K, ... 402 = OM.
- */
-export const MARKTDE_COLUMNS = {
-  vehicleId: '1', // B  stable auto.de vehicle id (metaData.vehicleId)
-  price: '10', // K  price (net for VAT-deductible cars, see getMarktdePriceColumns)
-  vatFlag: '11', // L  0 = VAT is added on top of K, 1 = no VAT shown separately
-  vin: '14', // O  VIN (optional)
-  description: '25', // Z  description (mobile.de markup)
-  images: '26', // AA image URLs, `;` separated
-  vatRate: '29', // AD VAT rate
-  fuelCombined: '98', // CU combined fuel consumption, decimal comma
-  co2: '99', // CV CO2 g/km
-  listingUrl: '402', // OM clickout link to the auto.de vehicle page
-} as const;
-
-// Clickout parameters (ADV-330, same scheme as AutoUncle ADV-324 / meinestadt ADV-327): the partner in
-// `source`, so the reporting views, the CRM source and GA4 (utm_*) all see "marktde". Legacy and utm
-// pairs carry identical values; paid / free is a contract attribute and is not encoded in the medium.
-// Pending markt.de's OK that a structured clickout field is allowed at all.
-export const MARKTDE_LINK_PARAMS =
-  'source=marktde&medium=listing&campaign=listing&utm_source=marktde&utm_medium=listing&utm_campaign=listing';
-
-export const getMarktdeListingUrl = (domain: string, carId: string): string =>
-  `https://www.${domain}/search/vehicle/${carId}?${MARKTDE_LINK_PARAMS}`;
-
-export enum MarktdeExclusion {
-  duplicateId = 'duplicate_id',
-  filtered = 'filtered',
-  noId = 'no_id',
-  noVehicleId = 'no_vehicle_id',
-  notAvailable = 'reserved_or_sold',
-  noPrice = 'no_valid_price',
-  noImage = 'no_valid_image',
-}
-
-// Online cars flagged like this are not sellable stock; the partner gets the file without them.
-const NOT_AVAILABLE = ['selector_offerAvailability_reserved', 'selector_offerAvailability_sold'];
-
-const positiveAmount = (value: unknown): number | null => {
-  if (value === null || value === undefined || value === '' || typeof value === 'boolean') {
-    return null;
-  }
-  const amount = toNumber(value);
-  return Number.isFinite(amount) && amount > 0 ? amount : null;
-};
-
-// Cheap checks before any image probe. null = the car can be exported.
-export const getMarktdeExclusion = (car: DataSchemes.Car): MarktdeExclusion | null => {
-  if (!toString(get(car, '_id', '')).trim()) {
-    return MarktdeExclusion.noId;
-  }
-  if (!toString(get(car, 'metaData.vehicleId', '')).trim()) {
-    return MarktdeExclusion.noVehicleId;
-  }
-  if (NOT_AVAILABLE.includes(toString(get(car, 'offer.offerAvailability', '')))) {
-    return MarktdeExclusion.notAvailable;
-  }
-  if (positiveAmount(get(car, 'price.currentSalesPrice', null)) === null) {
-    return MarktdeExclusion.noPrice;
-  }
-  return null;
-};
-
-// ISO 3779: 17 characters, digits and capital letters without I, O, Q.
-const VIN_PATTERN = /^[A-HJ-NPR-Z0-9]{17}$/;
-
-/**
- * The car's own VIN when it is a well-formed 17-character VIN, otherwise empty. The legacy exporter
- * sent marketing.fakeVIN or a random generated VIN for markt.de when the VIN was missing — never again.
- */
-export const getMarktdeVin = (car: DataSchemes.Car): string => {
-  const vin = toString(get(car, 'metaData.vin', '')).trim().toUpperCase();
-  return VIN_PATTERN.test(vin) ? vin : '';
-};
-
-export interface MarktdePriceColumns {
-  price: string;
-  vatFlag: string;
-  vatRate: string;
-  deductible: boolean;
-  // vatRate was missing / 0 on a VAT-deductible car and 19 was used
-  rateDefaulted: boolean;
-}
-
-/**
- * K / L / AD with the semantics of the legacy feed (ACF-4370): price.currentSalesPrice is the gross
- * consumer price. VAT-deductible car: K = net, L = 0, AD = rate, so gross = K * (1 + AD/100) again.
- * Otherwise (margin scheme / private / exempt): K = the gross price as it is, L = 1, AD = 0 — no VAT
- * is ever added on top of an already gross price. Unlike the legacy code the net price and AD always
- * use the same rate (legacy: a null vatRate left K gross while AD said 19, i.e. +19 % on display).
- */
-export const getMarktdePriceColumns = (car: DataSchemes.Car): MarktdePriceColumns | null => {
-  const raw = get(car, 'price.currentSalesPrice', null);
-  const gross = positiveAmount(raw);
-  if (gross === null) {
-    return null;
-  }
-  const deductible = Boolean(get(car, 'price.vatDeductible', false));
-  if (!deductible) {
-    return { price: String(raw), vatFlag: '1', vatRate: '0', deductible, rateDefaulted: false };
-  }
-  const storedRate = positiveAmount(get(car, 'price.vatRate', null));
-  const rate = storedRate ?? 19;
-  const net = (gross * 100) / (100 + rate);
-  return { price: net.toFixed(2), vatFlag: '0', vatRate: String(rate), deductible, rateDefaulted: storedRate === null };
-};
-
-/**
- * The markt.de image list from the shared validated clean selection (ADV-336, same rules as the
- * OpenAI / meinestadt / AutoUncle feeds): valid 3/4 hero first, then every other valid image. Real
- * photos only — no marketing overlays, no placeholders. No valid hero => [] (car is skipped).
- */
-export const toMarktdeImages = (selection: OpenAiImageSelection, maxImages = 0): string[] => {
-  if (!selection?.fields) {
-    return [];
-  }
-  const urls = [selection.fields.image_url, ...selection.fields.additional_image_urls];
-  return maxImages > 0 ? urls.slice(0, maxImages) : urls;
-};
-
-export interface DescriptionRemovals {
-  url: number;
-  email: number;
-  phone: number;
-  address: number;
-}
-
-// mobile.de description markup: `\\` is a line break, `**…**` bold. Removed tokens never eat it.
-const URL_PATTERN = /\b(?:https?:\/\/|www\.)[^\s\\"'<>]+/gi;
-const EMAIL_PATTERN = /[A-Z0-9._%+-]+@[A-Z0-9.-]+\.[A-Z]{2,}/gi;
-// Bare domains of other websites ("autohaus-muster.de"). Not the seller's own name: auto.de is the
-// brand and the legal entity (auto.de GmbH, the commercial identification the terms require), so
-// "auto.de", "Auto.de", "auto.de-ID" stay; only a link to it (scheme / www.) goes.
-const DOMAIN_PATTERN = /(?<![\w.-])(?!auto\.de\b)[a-z0-9][a-z0-9-]*(?:\.[a-z0-9-]+)*\.(?:de|com|net|org|eu|info|at|ch|biz|shop|online)\b(?:\/[^\s\\"'<>]*)?/gi;
-// Phone candidates: start with +, 00, 0 or (0, then digits and the usual separators; label optional.
-const PHONE_PATTERN = /(?:\b(?:Tel(?:efon|efax)?|Fon|Fax|SMS|Mobil(?:funk)?|Handy|WhatsApp|Hotline)\.?\s*:?\s*)?(?:\+|\b00|\(0|\b0)[\d \t/().-]{6,}\d/gi;
-const DATE_PATTERN = /^\d{1,2}\.\d{1,2}\.\d{2,4}$/;
-// Street + house number ("Max-Planck-Straße 19", "Hauptstr. 5a") and postal code + town ("06796 Brehna").
-const STREET_PATTERN = /\b[A-ZÄÖÜ][\wäöüß.-]*(?:[Ss]traße|[Ss]trasse|[Ss]tr\.|[Ww]eg|[Aa]llee|[Pp]latz|[Rr]ing|[Gg]asse|[Dd]amm|[Cc]haussee|[Uu]fer)\s*\d{1,4}(?![.,]?\d)\s?[a-z]?\b(?![.,]\d)/g;
-const POSTCODE_TOWN_PATTERN = /\b\d{5}\s+(?!(?:Euro|EUR|Kilometer|Km|KM|km|Stück|Kilo)\b)[A-ZÄÖÜ][a-zäöüß]+(?:[ -][A-ZÄÖÜ][a-zäöüß]+)?/g;
-// Marks a removed token until the tidy-up; never left in the output.
-const CUT = '\u0000';
-// Labels left behind right before a removed value ("Tel.: ", "E-Mail:").
-const DANGLING_LABEL = /\b(?:Tel(?:efon|efax)?|Fon|Fax|SMS|Mobil(?:funk)?|Handy|WhatsApp|Hotline|E-?Mail|Mail|Web(?:seite)?|Internet|Homepage|Adresse|Anschrift)\.?[ \t]*:?[ \t]*(?=\u0000)/gi;
-
-const digitCount = (value: string): number => (value.match(/\d/g) || []).length;
-
-/**
- * markt.de's terms (nutzungsbedingungen, checked 08.10.2026) forbid phone numbers, physical addresses,
- * website links and QR codes in the listing description. The dealer's free-text header / footer of the
- * mobile.de description often carries exactly that, so it is removed from the text; the vehicle facts
- * stay. Heuristic by nature: the run report counts every removal for review.
- */
-export const sanitizeMarktdeDescription = (text: string): { text: string; removed: DescriptionRemovals } => {
-  const removed: DescriptionRemovals = { url: 0, email: 0, phone: 0, address: 0 };
-  if (!text) {
-    return { text, removed };
-  }
-  let out = text
-    .replace(EMAIL_PATTERN, () => { removed.email += 1; return CUT; })
-    .replace(URL_PATTERN, () => { removed.url += 1; return CUT; })
-    .replace(DOMAIN_PATTERN, () => { removed.url += 1; return CUT; })
-    .replace(PHONE_PATTERN, (match) => {
-      const digits = digitCount(match);
-      if (digits < 8 || digits > 16 || DATE_PATTERN.test(match.trim())) {
-        return match;
-      }
-      removed.phone += 1;
-      return CUT;
-    })
-    .replace(STREET_PATTERN, () => { removed.address += 1; return CUT; })
-    .replace(POSTCODE_TOWN_PATTERN, () => { removed.address += 1; return CUT; });
-  if (removed.url + removed.email + removed.phone + removed.address === 0) {
-    return { text, removed };
-  }
-  // Tidy only where something was cut (CUT marks the spot): the label that introduced it, the spaces
-  // around it; a cut at the start / end of a line leaves no space. The rest of the text is untouched.
-  out = out
-    .replace(DANGLING_LABEL, '')
-    .replace(/[ \t]*(?:\u0000[ \t]*)+/g, '\u0001')
-    .replace(/(^|\\\\)\u0001/g, '$1')
-    .replace(/\u0001(?=\\\\|$)/g, '')
-    .replace(/\u0001/g, ' ');
-  return { text: out, removed };
-};
-
-export interface MarktdeRowContext {
-  domain: string;
-  // final, validated image URLs, hero first (non-empty: callers skip cars without images)
-  images: string[];
-}
-
-export interface MarktdeRowFacts {
-  vin: 'sent' | 'empty' | 'invalid_dropped';
-  price: MarktdePriceColumns;
-  descriptionRemovals: DescriptionRemovals | null;
-  fuelEmpty: boolean;
-  co2Empty: boolean;
-}
-
-const decimalComma = (value: number | undefined): string => (value === undefined || value === null ? '' : String(value).replace('.', ','));
-
-/**
- * Replaces the markt.de columns of a MobileTransformer row (object keyed by column index) and returns
- * what was decided, for the run report. The row is changed in place.
- */
-export const applyMarktdeColumns = (row: Record<string, unknown>, car: DataSchemes.Car, ctx: MarktdeRowContext): MarktdeRowFacts => {
-  const c = MARKTDE_COLUMNS;
-  const price = getMarktdePriceColumns(car);
-  row[c.price] = price.price;
-  row[c.vatFlag] = price.vatFlag;
-  row[c.vatRate] = price.vatRate;
-
-  const vin = getMarktdeVin(car);
-  const storedVin = toString(get(car, 'metaData.vin', '')).trim();
-  row[c.vin] = vin;
-
-  let descriptionRemovals: DescriptionRemovals | null = null;
-  const description = row[c.description];
-  if (typeof description === 'string' && description) {
-    const sanitized = sanitizeMarktdeDescription(description);
-    row[c.description] = sanitized.text;
-    descriptionRemovals = sanitized.removed;
-  }
-
-  row[c.images] = ctx.images.join(';');
-
-  // EnVKV: the same values as the Google / Meta / AutoUncle / meinestadt feeds (WLTP first, legacy
-  // only as fallback; BEV: no fuel consumption, CO2 0). Nothing usable -> empty (legacy: "undefined").
-  const fuel = envkvFuelConsumptionValue(car);
-  const co2 = envkvCo2Value(car);
-  row[c.fuelCombined] = decimalComma(fuel);
-  row[c.co2] = co2 === undefined || co2 === null ? '' : String(co2);
-
-  row[c.listingUrl] = getMarktdeListingUrl(ctx.domain, toString(get(car, '_id', '')));
-
-  return {
-    vin: vin ? 'sent' : storedVin ? 'invalid_dropped' : 'empty',
-    price,
-    descriptionRemovals,
-    fuelEmpty: row[c.fuelCombined] === '',
-    co2Empty: row[c.co2] === '',
-  };
-};
+import { get, toNumber, toString } from 'lodash';
+import { DataSchemes } from '@bytebrand/car-schema';
+
+import { OpenAiImageSelection } from './openai-image-fields';
+import { envkvCo2Value, envkvFuelConsumptionValue } from './google-feed-envkv';
+
+/**
+ * markt.de feed (ADV-328) on top of the legacy mobile.de-style CSV row that MobileTransformer builds
+ * from the `mobile_yescar` mapping sheet (ACF-4370). Only the columns below are replaced; every other
+ * column, the column order and the row format stay as they were. Pure functions, no I/O.
+ *
+ * Column indexes are 0-based (the sheet's importField): 1 = B, 10 = K, ... 402 = OM.
+ */
+export const MARKTDE_COLUMNS = {
+  vehicleId: '1', // B  stable auto.de vehicle id (metaData.vehicleId)
+  price: '10', // K  price (net for VAT-deductible cars, see getMarktdePriceColumns)
+  vatFlag: '11', // L  0 = VAT is added on top of K, 1 = no VAT shown separately
+  vin: '14', // O  VIN (optional)
+  description: '25', // Z  description (mobile.de markup)
+  images: '26', // AA image URLs, `;` separated
+  vatRate: '29', // AD VAT rate
+  fuelCombined: '98', // CU combined fuel consumption, decimal comma
+  co2: '99', // CV CO2 g/km
+  listingUrl: '402', // OM clickout link to the auto.de vehicle page
+} as const;
+
+// Clickout parameters (ADV-330, same scheme as AutoUncle ADV-324 / meinestadt ADV-327): the partner in
+// `source`, so the reporting views, the CRM source and GA4 (utm_*) all see "marktde". Legacy and utm
+// pairs carry identical values; paid / free is a contract attribute and is not encoded in the medium.
+// Pending markt.de's OK that a structured clickout field is allowed at all.
+export const MARKTDE_LINK_PARAMS =
+  'source=marktde&medium=listing&campaign=listing&utm_source=marktde&utm_medium=listing&utm_campaign=listing';
+
+export const getMarktdeListingUrl = (domain: string, carId: string): string =>
+  `https://www.${domain}/search/vehicle/${carId}?${MARKTDE_LINK_PARAMS}`;
+
+export enum MarktdeExclusion {
+  duplicateId = 'duplicate_id',
+  filtered = 'filtered',
+  noId = 'no_id',
+  noVehicleId = 'no_vehicle_id',
+  notAvailable = 'reserved_or_sold',
+  noPrice = 'no_valid_price',
+  noImage = 'no_valid_image',
+}
+
+// Online cars flagged like this are not sellable stock; the partner gets the file without them.
+const NOT_AVAILABLE = ['selector_offerAvailability_reserved', 'selector_offerAvailability_sold'];
+
+const positiveAmount = (value: unknown): number | null => {
+  if (value === null || value === undefined || value === '' || typeof value === 'boolean') {
+    return null;
+  }
+  const amount = toNumber(value);
+  return Number.isFinite(amount) && amount > 0 ? amount : null;
+};
+
+// Cheap checks before any image probe. null = the car can be exported.
+export const getMarktdeExclusion = (car: DataSchemes.Car): MarktdeExclusion | null => {
+  if (!toString(get(car, '_id', '')).trim()) {
+    return MarktdeExclusion.noId;
+  }
+  if (!toString(get(car, 'metaData.vehicleId', '')).trim()) {
+    return MarktdeExclusion.noVehicleId;
+  }
+  if (NOT_AVAILABLE.includes(toString(get(car, 'offer.offerAvailability', '')))) {
+    return MarktdeExclusion.notAvailable;
+  }
+  if (positiveAmount(get(car, 'price.currentSalesPrice', null)) === null) {
+    return MarktdeExclusion.noPrice;
+  }
+  return null;
+};
+
+// ISO 3779: 17 characters, digits and capital letters without I, O, Q.
+const VIN_PATTERN = /^[A-HJ-NPR-Z0-9]{17}$/;
+
+/**
+ * The car's own VIN when it is a well-formed 17-character VIN, otherwise empty. The legacy exporter
+ * sent marketing.fakeVIN or a random generated VIN for markt.de when the VIN was missing — never again.
+ */
+export const getMarktdeVin = (car: DataSchemes.Car): string => {
+  const vin = toString(get(car, 'metaData.vin', '')).trim().toUpperCase();
+  return VIN_PATTERN.test(vin) ? vin : '';
+};
+
+export interface MarktdePriceColumns {
+  price: string;
+  vatFlag: string;
+  vatRate: string;
+  deductible: boolean;
+  // vatRate was missing / 0 on a VAT-deductible car and 19 was used
+  rateDefaulted: boolean;
+}
+
+/**
+ * K / L / AD with the semantics of the legacy feed (ACF-4370): price.currentSalesPrice is the gross
+ * consumer price. VAT-deductible car: K = net, L = 0, AD = rate, so gross = K * (1 + AD/100) again.
+ * Otherwise (margin scheme / private / exempt): K = the gross price as it is, L = 1, AD = 0 — no VAT
+ * is ever added on top of an already gross price. Unlike the legacy code the net price and AD always
+ * use the same rate (legacy: a null vatRate left K gross while AD said 19, i.e. +19 % on display).
+ */
+export const getMarktdePriceColumns = (car: DataSchemes.Car): MarktdePriceColumns | null => {
+  const raw = get(car, 'price.currentSalesPrice', null);
+  const gross = positiveAmount(raw);
+  if (gross === null) {
+    return null;
+  }
+  const deductible = Boolean(get(car, 'price.vatDeductible', false));
+  if (!deductible) {
+    return { price: String(raw), vatFlag: '1', vatRate: '0', deductible, rateDefaulted: false };
+  }
+  const storedRate = positiveAmount(get(car, 'price.vatRate', null));
+  const rate = storedRate ?? 19;
+  const net = (gross * 100) / (100 + rate);
+  return { price: net.toFixed(2), vatFlag: '0', vatRate: String(rate), deductible, rateDefaulted: storedRate === null };
+};
+
+/**
+ * The markt.de image list from the shared validated clean selection (ADV-336, same rules as the
+ * OpenAI / meinestadt / AutoUncle feeds): valid 3/4 hero first, then every other valid image. Real
+ * photos only — no marketing overlays, no placeholders. No valid hero => [] (car is skipped).
+ */
+export const toMarktdeImages = (selection: OpenAiImageSelection, maxImages = 0): string[] => {
+  if (!selection?.fields) {
+    return [];
+  }
+  const urls = [selection.fields.image_url, ...selection.fields.additional_image_urls];
+  return maxImages > 0 ? urls.slice(0, maxImages) : urls;
+};
+
+export interface DescriptionRemovals {
+  url: number;
+  email: number;
+  phone: number;
+  address: number;
+}
+
+// mobile.de description markup: `\\` is a line break, `**…**` bold. Removed tokens never eat it.
+const URL_PATTERN = /\b(?:https?:\/\/|www\.)[^\s\\"'<>]+/gi;
+const EMAIL_PATTERN = /[A-Z0-9._%+-]+@[A-Z0-9.-]+\.[A-Z]{2,}/gi;
+// Bare domains of other websites ("autohaus-muster.de"). Not the seller's own name: auto.de is the
+// brand and the legal entity (auto.de GmbH, the commercial identification the terms require), so
+// "auto.de", "Auto.de", "auto.de-ID" stay; only a link to it (scheme / www.) goes.
+const DOMAIN_PATTERN = /(?<![\w.-])(?!auto\.de\b)[a-z0-9][a-z0-9-]*(?:\.[a-z0-9-]+)*\.(?:de|com|net|org|eu|info|at|ch|biz|shop|online)\b(?:\/[^\s\\"'<>]*)?/gi;
+// Phone candidates: start with +, 00, 0 or (0, then digits and the usual separators; label optional.
+const PHONE_PATTERN = /(?:\b(?:Tel(?:efon|efax)?|Fon|Fax|SMS|Mobil(?:funk)?|Handy|WhatsApp|Hotline)\.?\s*:?\s*)?(?:\+|\b00|\(0|\b0)[\d \t/().-]{6,}\d/gi;
+const DATE_PATTERN = /^\d{1,2}\.\d{1,2}\.\d{2,4}$/;
+// Street + house number ("Max-Planck-Straße 19", "Hauptstr. 5a") and postal code + town ("06796 Brehna").
+const STREET_PATTERN = /\b[A-ZÄÖÜ][\wäöüß.-]*(?:[Ss]traße|[Ss]trasse|[Ss]tr\.|[Ww]eg|[Aa]llee|[Pp]latz|[Rr]ing|[Gg]asse|[Dd]amm|[Cc]haussee|[Uu]fer)\s*\d{1,4}(?![.,]?\d)\s?[a-z]?\b(?![.,]\d)/g;
+const POSTCODE_TOWN_PATTERN = /\b\d{5}\s+(?!(?:Euro|EUR|Kilometer|Km|KM|km|Stück|Kilo)\b)[A-ZÄÖÜ][a-zäöüß]+(?:[ -][A-ZÄÖÜ][a-zäöüß]+)?/g;
+// Marks a removed token until the tidy-up; never left in the output.
+const CUT = '\u0000';
+// Labels left behind right before a removed value ("Tel.: ", "E-Mail:").
+const DANGLING_LABEL = /\b(?:Tel(?:efon|efax)?|Fon|Fax|SMS|Mobil(?:funk)?|Handy|WhatsApp|Hotline|E-?Mail|Mail|Web(?:seite)?|Internet|Homepage|Adresse|Anschrift)\.?[ \t]*:?[ \t]*(?=\u0000)/gi;
+
+const digitCount = (value: string): number => (value.match(/\d/g) || []).length;
+
+/**
+ * markt.de's terms (nutzungsbedingungen, checked 08.10.2026) forbid phone numbers, physical addresses,
+ * website links and QR codes in the listing description. The dealer's free-text header / footer of the
+ * mobile.de description often carries exactly that, so it is removed from the text; the vehicle facts
+ * stay. Heuristic by nature: the run report counts every removal for review.
+ */
+export const sanitizeMarktdeDescription = (text: string): { text: string; removed: DescriptionRemovals } => {
+  const removed: DescriptionRemovals = { url: 0, email: 0, phone: 0, address: 0 };
+  if (!text) {
+    return { text, removed };
+  }
+  let out = text
+    .replace(EMAIL_PATTERN, () => { removed.email += 1; return CUT; })
+    .replace(URL_PATTERN, () => { removed.url += 1; return CUT; })
+    .replace(DOMAIN_PATTERN, () => { removed.url += 1; return CUT; })
+    .replace(PHONE_PATTERN, (match) => {
+      const digits = digitCount(match);
+      if (digits < 8 || digits > 16 || DATE_PATTERN.test(match.trim())) {
+        return match;
+      }
+      removed.phone += 1;
+      return CUT;
+    })
+    .replace(STREET_PATTERN, () => { removed.address += 1; return CUT; })
+    .replace(POSTCODE_TOWN_PATTERN, () => { removed.address += 1; return CUT; });
+  if (removed.url + removed.email + removed.phone + removed.address === 0) {
+    return { text, removed };
+  }
+  // Tidy only where something was cut (CUT marks the spot): the label that introduced it, the spaces
+  // around it; a cut at the start / end of a line leaves no space. The rest of the text is untouched.
+  out = out
+    .replace(DANGLING_LABEL, '')
+    .replace(/[ \t]*(?:\u0000[ \t]*)+/g, '\u0001')
+    .replace(/(^|\\\\)\u0001/g, '$1')
+    .replace(/\u0001(?=\\\\|$)/g, '')
+    .replace(/\u0001/g, ' ');
+  return { text: out, removed };
+};
+
+export interface MarktdeRowContext {
+  domain: string;
+  // final, validated image URLs, hero first (non-empty: callers skip cars without images)
+  images: string[];
+}
+
+export interface MarktdeRowFacts {
+  vin: 'sent' | 'empty' | 'invalid_dropped';
+  price: MarktdePriceColumns;
+  descriptionRemovals: DescriptionRemovals | null;
+  fuelEmpty: boolean;
+  co2Empty: boolean;
+}
+
+const decimalComma = (value: number | undefined): string => (value === undefined || value === null ? '' : String(value).replace('.', ','));
+
+/**
+ * Replaces the markt.de columns of a MobileTransformer row (object keyed by column index) and returns
+ * what was decided, for the run report. The row is changed in place.
+ */
+export const applyMarktdeColumns = (row: Record<string, unknown>, car: DataSchemes.Car, ctx: MarktdeRowContext): MarktdeRowFacts => {
+  const c = MARKTDE_COLUMNS;
+  const price = getMarktdePriceColumns(car);
+  row[c.price] = price.price;
+  row[c.vatFlag] = price.vatFlag;
+  row[c.vatRate] = price.vatRate;
+
+  const vin = getMarktdeVin(car);
+  const storedVin = toString(get(car, 'metaData.vin', '')).trim();
+  row[c.vin] = vin;
+
+  let descriptionRemovals: DescriptionRemovals | null = null;
+  const description = row[c.description];
+  if (typeof description === 'string' && description) {
+    const sanitized = sanitizeMarktdeDescription(description);
+    row[c.description] = sanitized.text;
+    descriptionRemovals = sanitized.removed;
+  }
+
+  row[c.images] = ctx.images.join(';');
+
+  // EnVKV: the same values as the Google / Meta / AutoUncle / meinestadt feeds (WLTP first, legacy
+  // only as fallback; BEV: no fuel consumption, CO2 0). Nothing usable -> empty (legacy: "undefined").
+  const fuel = envkvFuelConsumptionValue(car);
+  const co2 = envkvCo2Value(car);
+  row[c.fuelCombined] = decimalComma(fuel);
+  row[c.co2] = co2 === undefined || co2 === null ? '' : String(co2);
+
+  row[c.listingUrl] = getMarktdeListingUrl(ctx.domain, toString(get(car, '_id', '')));
+
+  return {
+    vin: vin ? 'sent' : storedVin ? 'invalid_dropped' : 'empty',
+    price,
+    descriptionRemovals,
+    fuelEmpty: row[c.fuelCombined] === '',
+    co2Empty: row[c.co2] === '',
+  };
+};