fix(scraper): parse FS.com delivery dates + lead_time; fix units_sold K/M inflation

FS.com renders per-warehouse availability as a bare word-date on each line
("2 Stk. im DE-Lager, 8. Juli 2026 Lieferbar"). The old regex only matched a
"Lieferung:"/"erwartet:" prefix or numeric dd.mm.yyyy, so warehouse_de/global
delivery_date + backorder_estimated_date were 0% populated across 85k rows, and
lead_time_days was never written at all.

- new fs-com-parse.ts: pure, unit-tested parsers (13 node:test cases).
- extractDeliveryDate finds the bare word/numeric date per warehouse, bounded to
  its own line so a date never bleeds to a neighbouring warehouse.
- computeLeadTimeDays = min(delivery_date - today), floored at 0; threaded into
  upsertStockObservation (now writes lead_time_days) and upsertPriceObservation.
- parseGermanQty: K/M abbreviations now treat '.' as a decimal ("4.8K"=4800,
  "1.4M"=1.4M) instead of stripping it (was "1.4M"->14,000,000 — the outlier).
- units_sold sanity gate at ingest (FS_MAX_UNITS_SOLD, default 2,000,000).
- sql/122: purge impossible historical units_sold, re-document columns as collected.

packages/scraper/src/scrapers/fs-com-parse.test.ts

@@ -0,0 +1,138 @@
+/**
+ * Unit tests for the FS.com pure parsers. Run with:
+ *   npx tsx --test src/scrapers/fs-com-parse.test.ts
+ */
+import { test } from "node:test";
+import assert from "node:assert/strict";
+import {
+  parseGermanQty,
+  parseGermanDate,
+  parseGermanPrice,
+  extractDeliveryDate,
+  parseWarehouseStock,
+  computeLeadTimeDays,
+  sanitizeUnitsSold,
+} from "./fs-com-parse";
+
+// ── parseGermanQty — the K/M decimal bug ─────────────────────────────────────
+
+test("parseGermanQty: plain integers use '.'/',' as thousands", () => {
+  assert.equal(parseGermanQty("4.895"), 4895);
+  assert.equal(parseGermanQty("1.587"), 1587);
+  assert.equal(parseGermanQty("1.501"), 1501);
+  assert.equal(parseGermanQty("2"), 2);
+  assert.equal(parseGermanQty("1627"), 1627);
+});
+
+test("parseGermanQty: K/M abbreviations treat '.' as a DECIMAL (bug fix)", () => {
+  // Previously "4.8K" -> 48000 and "1.4M" -> 14000000 (the garbage outlier).
+  assert.equal(parseGermanQty("4.8K"), 4800);
+  assert.equal(parseGermanQty("210.9K"), 210900);
+  assert.equal(parseGermanQty("504.4K"), 504400);
+  assert.equal(parseGermanQty("1.4M"), 1_400_000);
+  assert.equal(parseGermanQty("1.2M"), 1_200_000);
+  assert.equal(parseGermanQty("63K"), 63000); // no decimal — unchanged
+  assert.equal(parseGermanQty("14M"), 14_000_000); // caught later by the gate
+});
+
+test("parseGermanQty: K/M with German comma decimal still works", () => {
+  assert.equal(parseGermanQty("4,8K"), 4800);
+  assert.equal(parseGermanQty("1,4M"), 1_400_000);
+});
+
+// ── parseGermanDate ──────────────────────────────────────────────────────────
+
+test("parseGermanDate: word and numeric forms", () => {
+  assert.equal(parseGermanDate("8. Juli 2026"), "2026-07-08");
+  assert.equal(parseGermanDate("17. Juli 2026"), "2026-07-17");
+  assert.equal(parseGermanDate("20 Apr., 2026"), "2026-04-20");
+  assert.equal(parseGermanDate("28. März 2026"), "2026-03-28");
+  assert.equal(parseGermanDate("20.04.2026"), "2026-04-20");
+  assert.equal(parseGermanDate("nonsense"), undefined);
+});
+
+// ── extractDeliveryDate ──────────────────────────────────────────────────────
+
+test("extractDeliveryDate: finds bare word-date, no keyword needed", () => {
+  assert.equal(extractDeliveryDate("2 Stk. im DE-Lager, 8. Juli 2026 Lieferbar"), "2026-07-08");
+  assert.equal(extractDeliveryDate("1.627 Stk. im Global-Lager, 17. Juli 2026"), "2026-07-17");
+});
+
+test("extractDeliveryDate: rejects stray/garbage years and empty", () => {
+  assert.equal(extractDeliveryDate("© 2019 FS.COM"), undefined);
+  assert.equal(extractDeliveryDate("auf Lager, sofort verfügbar"), undefined);
+});
+
+// ── parseWarehouseStock — the real screenshot block ──────────────────────────
+
+const SCREENSHOT_BLOCK = [
+  "2 Stk. im DE-Lager, 8. Juli 2026 Lieferbar",
+  "1.627 Stk. im Global-Lager, 17. Juli 2026",
+  "1 in Nachlieferung, 28. Juli 2026",
+  "4.8K verkauft",
+].join("\n");
+
+test("parseWarehouseStock: full block yields qtys + per-warehouse dates + sold", () => {
+  const s = parseWarehouseStock(SCREENSHOT_BLOCK);
+  assert.equal(s.deQty, 2);
+  assert.equal(s.deDeliveryDate, "2026-07-08");
+  assert.equal(s.globalQty, 1627);
+  assert.equal(s.globalDeliveryDate, "2026-07-17");
+  assert.equal(s.backorderQty, 1);
+  assert.equal(s.backorderDate, "2026-07-28");
+  assert.equal(s.unitsSold, 4800); // 4.8K, correctly de-abbreviated
+});
+
+test("parseWarehouseStock: a warehouse with no date does not steal a neighbour's date", () => {
+  const block = [
+    "5 Stk. im DE-Lager",              // no date — available now
+    "800 Stk. im Global-Lager, 17. Juli 2026",
+  ].join("\n");
+  const s = parseWarehouseStock(block);
+  assert.equal(s.deQty, 5);
+  assert.equal(s.deDeliveryDate, undefined); // must NOT be 2026-07-17
+  assert.equal(s.globalQty, 800);
+  assert.equal(s.globalDeliveryDate, "2026-07-17");
+});
+
+test("parseWarehouseStock: gates the 1.4M -> 14M units_sold garbage", () => {
+  const block = "12 Stk. im DE-Lager, 8. Juli 2026 Lieferbar\n14M verkauft";
+  const s = parseWarehouseStock(block);
+  assert.equal(s.deQty, 12);
+  assert.equal(s.unitsSold, undefined); // 14M > cap → dropped as unknown
+});
+
+test("parseWarehouseStock: empty when no warehouse text present", () => {
+  assert.deepEqual(parseWarehouseStock("Just a product description, nothing else."), {});
+});
+
+// ── sanitizeUnitsSold ────────────────────────────────────────────────────────
+
+test("sanitizeUnitsSold: keeps plausible, drops implausible/negative", () => {
+  assert.equal(sanitizeUnitsSold(4800), 4800);
+  assert.equal(sanitizeUnitsSold(504400), 504400);
+  assert.equal(sanitizeUnitsSold(14_000_000), undefined);
+  assert.equal(sanitizeUnitsSold(-5), undefined);
+  assert.equal(sanitizeUnitsSold(undefined), undefined);
+  assert.equal(sanitizeUnitsSold(2_500_000, 2_000_000), undefined);
+  assert.equal(sanitizeUnitsSold(1_500_000, 2_000_000), 1_500_000);
+});
+
+// ── computeLeadTimeDays ──────────────────────────────────────────────────────
+
+test("computeLeadTimeDays: min across dates, floored at 0, undefined when empty", () => {
+  const now = new Date("2026-07-04T09:00:00");
+  assert.equal(computeLeadTimeDays(["2026-07-08", "2026-07-17", "2026-07-28"], now), 4);
+  assert.equal(computeLeadTimeDays([undefined, null, "2026-07-05"], now), 1);
+  assert.equal(computeLeadTimeDays(["2026-07-01"], now), 0); // past date → 0, not negative
+  assert.equal(computeLeadTimeDays([], now), undefined);
+  assert.equal(computeLeadTimeDays([null, undefined], now), undefined);
+});
+
+// ── parseGermanPrice (unchanged behaviour, regression guard) ─────────────────
+
+test("parseGermanPrice: net EUR parsing", () => {
+  assert.equal(parseGermanPrice("42,50"), 42.5);
+  assert.equal(parseGermanPrice("1.063,02 €"), 1063.02);
+  assert.equal(parseGermanPrice("5,10 € ohne MwSt."), 5.1);
+});

packages/scraper/src/scrapers/fs-com-parse.ts

@@ -0,0 +1,290 @@
+/**
+ * FS.com parsing helpers — pure, side-effect-free, unit-tested.
+ *
+ * Extracted from fs-com.ts so the number/date/warehouse parsing can be tested
+ * in isolation (fs-com.ts pulls in Crawlee + Playwright + a pg Pool, which we
+ * do not want to load in a unit test).
+ *
+ * The FS.com German storefront (www.fs.com/de/) renders per-warehouse
+ * availability as a short line per warehouse, e.g.:
+ *
+ *   2 Stk. im DE-Lager, 8. Juli 2026 Lieferbar
+ *   1.627 Stk. im Global-Lager, 17. Juli 2026
+ *   1 in Nachlieferung, 28. Juli 2026
+ *   4.8K verkauft
+ *
+ * The quantity part was already parsed correctly; the delivery-date part was
+ * not, because the previous regex only looked for a "Lieferung:"/"erwartet:"
+ * keyword prefix or a numeric dd.mm.yyyy date — neither of which matches the
+ * bare word-date ("8. Juli 2026") that FS.com actually renders.
+ */
+
+// ── Constants ────────────────────────────────────────────────────────────────
+
+const GERMAN_MONTHS: Record<string, string> = {
+  jan: "01", feb: "02", mär: "03", mar: "03",
+  apr: "04", mai: "05", may: "05", jun: "06",
+  jul: "07", aug: "08", sep: "09", okt: "10",
+  oct: "10", nov: "11", dez: "12", dec: "12",
+};
+
+/**
+ * Upper bound for a plausible per-SKU lifetime "verkauft" (units-sold) counter.
+ * Values above this are treated as unknown (dropped), not stored. FS.com's
+ * biggest sellers are in the hundreds of thousands; multi-million values are
+ * a mis-parse (see parseGermanQty K/M bug) or vendor marketing inflation.
+ * Override with FS_MAX_UNITS_SOLD.
+ */
+export const DEFAULT_MAX_UNITS_SOLD = 2_000_000;
+
+/** How far past a warehouse line to look for its delivery date (characters). */
+const DATE_WINDOW = 200;
+
+// ── Number parsing ───────────────────────────────────────────────────────────
+
+/**
+ * Parse the mantissa of a K/M-abbreviated number ("4.8" in "4.8K").
+ *
+ * K/M abbreviations carry a small decimal, never a grouped thousands value, so
+ * a lone "." is the decimal point ("4.8K" = 4 800). German pages may instead
+ * use "," as the decimal ("4,8K"); a "." alongside a "," is then a thousands
+ * grouping and is stripped.
+ */
+function parseAbbrevMantissa(s: string): number {
+  if (s.includes(",")) return parseFloat(s.replace(/\./g, "").replace(",", "."));
+  return parseFloat(s); // "." is the decimal separator
+}
+
+/**
+ * Parse a German-formatted quantity string.
+ *   "4.895"  → 4895     (period = thousands separator in a plain integer)
+ *   "1.587"  → 1587
+ *   "4.8K"   → 4800     (period = decimal in a K/M abbreviation)
+ *   "210.9K" → 210900
+ *   "1.4M"   → 1400000
+ *   "63K"    → 63000
+ */
+export function parseGermanQty(text: string): number | undefined {
+  const t = text.trim().replace(/\s/g, "");
+  if (!t) return undefined;
+
+  const kMatch = t.match(/^([\d.,]+)[Kk]$/);
+  if (kMatch) {
+    const n = parseAbbrevMantissa(kMatch[1]);
+    return isNaN(n) ? undefined : Math.round(n * 1_000);
+  }
+
+  const mMatch = t.match(/^([\d.,]+)[Mm]$/);
+  if (mMatch) {
+    const n = parseAbbrevMantissa(mMatch[1]);
+    return isNaN(n) ? undefined : Math.round(n * 1_000_000);
+  }
+
+  // Plain integer: "." and "," are both thousands separators here.
+  const n = parseInt(t.replace(/\./g, "").replace(/,/g, ""), 10);
+  return isNaN(n) ? undefined : n;
+}
+
+/**
+ * Parse a German date to ISO "YYYY-MM-DD".
+ *   "8. Juli 2026"   → "2026-07-08"
+ *   "20 Apr., 2026"  → "2026-04-20"
+ *   "20.04.2026"     → "2026-04-20"
+ */
+export function parseGermanDate(text: string): string | undefined {
+  const numericMatch = text.match(/(\d{1,2})\.(\d{1,2})\.(\d{4})/);
+  if (numericMatch) {
+    const [, d, m, y] = numericMatch;
+    return `${y}-${m.padStart(2, "0")}-${d.padStart(2, "0")}`;
+  }
+  const wordMatch = text.match(/(\d{1,2})\.?\s+([A-Za-zÄÖÜäöüß]+)\.?,?\s*(\d{4})/);
+  if (!wordMatch) return undefined;
+  const day = wordMatch[1].padStart(2, "0");
+  const monthRaw = wordMatch[2]
+    .toLowerCase()
+    .replace(/ä/g, "a").replace(/ö/g, "o").replace(/ü/g, "u")
+    .slice(0, 3);
+  const month = GERMAN_MONTHS[monthRaw];
+  if (!month) return undefined;
+  return `${wordMatch[3]}-${month}-${day}`;
+}
+
+/**
+ * Parse a German price to a EUR float.
+ *   "42,50"    → 42.50
+ *   "1.063,02" → 1063.02
+ */
+export function parseGermanPrice(raw: string): number | undefined {
+  const cleaned = raw.replace(/[^0-9.,]/g, "").trim();
+  if (!cleaned) return undefined;
+  let normalized: string;
+  if (/\d+\.\d{3},\d{2}/.test(cleaned)) {
+    normalized = cleaned.replace(/\./g, "").replace(",", ".");
+  } else if (cleaned.includes(",")) {
+    normalized = cleaned.replace(",", ".");
+  } else {
+    normalized = cleaned;
+  }
+  const n = parseFloat(normalized);
+  return isNaN(n) || n <= 0 ? undefined : n;
+}
+
+// ── Units-sold sanity gate ───────────────────────────────────────────────────
+
+/**
+ * Clamp an implausible units-sold value to `undefined` (unknown). Returns the
+ * value unchanged when it is a finite, non-negative number at or below `cap`.
+ */
+export function sanitizeUnitsSold(
+  raw: number | undefined,
+  cap: number = DEFAULT_MAX_UNITS_SOLD
+): number | undefined {
+  if (raw === undefined) return undefined;
+  if (!Number.isFinite(raw) || raw < 0) return undefined;
+  if (raw > cap) return undefined;
+  return raw;
+}
+
+// ── Delivery-date extraction ─────────────────────────────────────────────────
+
+/** dd.mm.yyyy  OR  "8. Juli 2026" / "20 Apr., 2026" (day + month word + year). */
+const DELIVERY_DATE_RE =
+  /(\d{1,2}\.\d{1,2}\.\d{4})|(\d{1,2}\.?\s+[A-Za-zÄÖÜäöüß]{3,}\.?,?\s*\d{4})/;
+
+/**
+ * Find the first plausible calendar date inside a bounded text block (the slice
+ * that belongs to one warehouse line) and return it as ISO "YYYY-MM-DD".
+ * The date can appear with or without a "Lieferbar"/"Lieferung" keyword — FS.com
+ * renders it bare ("…DE-Lager, 8. Juli 2026 Lieferbar"), so no keyword is required.
+ */
+export function extractDeliveryDate(block: string): string | undefined {
+  const m = block.match(DELIVERY_DATE_RE);
+  if (!m) return undefined;
+  const iso = parseGermanDate(m[0]);
+  if (!iso) return undefined;
+  const year = Number(iso.slice(0, 4));
+  if (year < 2024 || year > 2100) return undefined; // reject stray/garbage years
+  return iso;
+}
+
+// ── Warehouse block parsing ──────────────────────────────────────────────────
+
+export interface WarehouseStock {
+  deQty?: number;
+  deDeliveryDate?: string;
+  globalQty?: number;
+  globalDeliveryDate?: string;
+  backorderQty?: number;
+  backorderDate?: string;
+  unitsSold?: number;
+}
+
+const DE_QTY_RE: RegExp[] = [
+  /(\d[\d.,KkMm]*)\s*Stk\.\s*im\s*DE[- ]Lager/i,
+  /(\d[\d.,KkMm]*)\s*im\s*DE[- ]?Lager/i,
+];
+const GLOBAL_QTY_RE: RegExp[] = [
+  /(\d[\d.,KkMm]*)\s*Stk\.\s*im\s*Global[- ]Lager/i,
+  /(\d[\d.,KkMm]*)\s*im\s*Global[- ]?(?:Lager|Warehouse)/i,
+  /(\d[\d.,KkMm]*)\s*in\s+Global\s+Warehouse/i,
+];
+const BACKORDER_QTY_RE: RegExp[] = [
+  /(\d[\d.,KkMm]*)\s*(?:Stk\.)?\s*in\s+Nachlieferung/i,
+  /Nachlieferung[:\s]*(\d[\d.,KkMm]*)/i,
+];
+const UNITS_SOLD_RE: RegExp[] = [
+  /(\d[\d.,KkMm]*)\s*(?:[Mm]al\s+)?[Vv]erkauft/,
+  /([\d.,KkMm]+)\+?\s*sold/i,
+];
+
+interface SectionMatch { value: string; index: number; }
+
+function firstMatch(text: string, res: RegExp[]): SectionMatch | undefined {
+  for (const re of res) {
+    const m = text.match(re);
+    if (m && m[1] != null && m.index != null) return { value: m[1], index: m.index };
+  }
+  return undefined;
+}
+
+/**
+ * Parse the full warehouse-availability block out of a product page's body
+ * text: DE-Lager, Global-Lager and Nachlieferung quantities, each with the
+ * delivery date rendered on its line, plus the "verkauft"/"sold" counter.
+ *
+ * A section's delivery date is looked for only in the window between that
+ * section's quantity and the start of the next section, so a warehouse with no
+ * date never picks up a neighbouring warehouse's date.
+ */
+export function parseWarehouseStock(
+  bodyText: string,
+  unitsSoldCap: number = DEFAULT_MAX_UNITS_SOLD
+): WarehouseStock {
+  const t = bodyText;
+  const de = firstMatch(t, DE_QTY_RE);
+  const gl = firstMatch(t, GLOBAL_QTY_RE);
+  const bo = firstMatch(t, BACKORDER_QTY_RE);
+  const sold = firstMatch(t, UNITS_SOLD_RE);
+
+  const starts = [de, gl, bo, sold]
+    .filter((s): s is SectionMatch => s !== undefined)
+    .map((s) => s.index);
+
+  const dateFor = (sec: SectionMatch | undefined): string | undefined => {
+    if (!sec) return undefined;
+    const later = starts.filter((b) => b > sec.index);
+    const end = later.length ? Math.min(sec.index + DATE_WINDOW, ...later) : sec.index + DATE_WINDOW;
+    return extractDeliveryDate(t.slice(sec.index, end));
+  };
+
+  const result: WarehouseStock = {};
+
+  if (de) {
+    const q = parseGermanQty(de.value);
+    if (q !== undefined) result.deQty = q;
+    const d = dateFor(de);
+    if (d) result.deDeliveryDate = d;
+  }
+  if (gl) {
+    const q = parseGermanQty(gl.value);
+    if (q !== undefined) result.globalQty = q;
+    const d = dateFor(gl);
+    if (d) result.globalDeliveryDate = d;
+  }
+  if (bo) {
+    const q = parseGermanQty(bo.value);
+    if (q !== undefined) result.backorderQty = q;
+    const d = dateFor(bo);
+    if (d) result.backorderDate = d;
+  }
+  if (sold) {
+    const s = sanitizeUnitsSold(parseGermanQty(sold.value), unitsSoldCap);
+    if (s !== undefined) result.unitsSold = s;
+  }
+
+  return result;
+}
+
+// ── Lead-time derivation ─────────────────────────────────────────────────────
+
+/**
+ * Derive lead-time in days as the soonest availability across the warehouses
+ * that quote a delivery date: min(date − today), floored at 0. Returns
+ * `undefined` when no date is known (never fabricate a 0).
+ */
+export function computeLeadTimeDays(
+  dates: Array<string | null | undefined>,
+  now: Date = new Date()
+): number | undefined {
+  const todayUtc = Date.UTC(now.getFullYear(), now.getMonth(), now.getDate());
+  let min: number | undefined;
+  for (const d of dates) {
+    if (!d) continue;
+    const m = d.match(/^(\d{4})-(\d{2})-(\d{2})$/);
+    if (!m) continue;
+    const dUtc = Date.UTC(Number(m[1]), Number(m[2]) - 1, Number(m[3]));
+    const days = Math.max(0, Math.round((dUtc - todayUtc) / 86_400_000));
+    if (min === undefined || days < min) min = days;
+  }
+  return min;
+}

packages/scraper/src/scrapers/fs-com.ts

@@ -63,6 +63,12 @@ import {
 } from "../utils/db";
 import { contentHash } from "../utils/hash";
 import { updateVerifiedSpecs, parseSpecTable } from "../utils/spec-updater";
+import {
+  parseGermanPrice,
+  parseWarehouseStock,
+  computeLeadTimeDays,
+  DEFAULT_MAX_UNITS_SOLD,
+} from "./fs-com-parse";
 
 // ── Constants ──────────────────────────────────────────────────────────────────
 
@@ -75,6 +81,8 @@ const FORCE_REVALIDATE = process.env["TIP_FORCE_REVALIDATE"] === "1";
 const ONLY_MISSING_IMAGES = process.env["FS_ONLY_MISSING_IMAGES"] === "1";
 const DB_DETAIL_ONLY = process.env["FS_DB_DETAIL_ONLY"] === "1";
 const URL_DISCOVERY_ONLY = process.env["FS_URL_DISCOVERY_ONLY"] === "1";
+// Drop implausible "verkauft" (units-sold) counters at ingest (see fs-com-parse).
+const FS_MAX_UNITS_SOLD = parseInt(process.env["FS_MAX_UNITS_SOLD"] ?? String(DEFAULT_MAX_UNITS_SOLD), 10);
 
 const PROXY_URLS = (process.env["PROXY_URLS"] ?? "")
   .split(",")
@@ -113,83 +121,7 @@ const DE_COOKIES = [
   { name: "country",  value: "DE",  domain: ".fs.com", path: "/" },
 ];
 
-// ── German locale parsers ──────────────────────────────────────────────────────
+// German number/date/warehouse parsers live in ./fs-com-parse (pure + unit-tested).
-
-const GERMAN_MONTHS: Record<string, string> = {
-  jan: "01", feb: "02", mär: "03", mar: "03",
-  apr: "04", mai: "05", may: "05", jun: "06",
-  jul: "07", aug: "08", sep: "09", okt: "10",
-  oct: "10", nov: "11", dez: "12", dec: "12",
-};
-
-/**
- * Parse German-formatted quantity string.
- * "4.895"  → 4895  (period = thousands separator in German)
- * "210.9K" → 210900
- * "1.2M"   → 1200000
- */
-function parseGermanQty(text: string): number | undefined {
-  const t = text.trim().replace(/\s/g, "");
-  if (!t) return undefined;
-
-  const kMatch = t.match(/^([\d.,]+)[Kk]$/);
-  if (kMatch) {
-    const n = parseFloat(kMatch[1].replace(/\./g, "").replace(",", "."));
-    return isNaN(n) ? undefined : Math.round(n * 1_000);
-  }
-
-  const mMatch = t.match(/^([\d.,]+)[Mm]$/);
-  if (mMatch) {
-    const n = parseFloat(mMatch[1].replace(/\./g, "").replace(",", "."));
-    return isNaN(n) ? undefined : Math.round(n * 1_000_000);
-  }
-
-  const n = parseInt(t.replace(/\./g, "").replace(/,/g, ""), 10);
-  return isNaN(n) ? undefined : n;
-}
-
-/**
- * Parse German date to ISO "YYYY-MM-DD".
- * "20 Apr., 2026"  → "2026-04-20"
- * "20.04.2026"     → "2026-04-20"
- */
-function parseGermanDate(text: string): string | undefined {
-  const numericMatch = text.match(/(\d{1,2})\.(\d{1,2})\.(\d{4})/);
-  if (numericMatch) {
-    const [, d, m, y] = numericMatch;
-    return `${y}-${m.padStart(2, "0")}-${d.padStart(2, "0")}`;
-  }
-  const wordMatch = text.match(/(\d{1,2})\.?\s+([A-Za-zÄÖÜäöüß]+)\.?,?\s*(\d{4})/);
-  if (!wordMatch) return undefined;
-  const day = wordMatch[1].padStart(2, "0");
-  const monthRaw = wordMatch[2]
-    .toLowerCase()
-    .replace(/ä/g, "a").replace(/ö/g, "o").replace(/ü/g, "u")
-    .slice(0, 3);
-  const month = GERMAN_MONTHS[monthRaw];
-  if (!month) return undefined;
-  return `${wordMatch[3]}-${month}-${day}`;
-}
-
-/**
- * Parse German price to EUR float.
- * "42,50"    → 42.50
- * "1.063,02" → 1063.02
- */
-function parseGermanPrice(raw: string): number | undefined {
-  const cleaned = raw.replace(/[^0-9.,]/g, "").trim();
-  if (!cleaned) return undefined;
-  let normalized: string;
-  if (/\d+\.\d{3},\d{2}/.test(cleaned)) {
-    normalized = cleaned.replace(/\./g, "").replace(",", ".");
-  } else if (cleaned.includes(",")) {
-    normalized = cleaned.replace(",", ".");
-  } else {
-    normalized = cleaned;
-  }
-  const n = parseFloat(normalized);
-  return isNaN(n) || n <= 0 ? undefined : n;
-}
 
 // ── Stock level helper ─────────────────────────────────────────────────────────
 
@@ -648,64 +580,18 @@ async function scrapeProductDetails(
         }
       }
 
-      // ── DE-Lager ───────────────────────────────────────────────────────────
+      // ── Warehouse availability (qty + per-warehouse delivery date + units sold) ──
-      let deQty: number | undefined;
+      // FS.com renders each warehouse on its own line, e.g.
-      let deDeliveryDate: string | undefined;
+      //   "2 Stk. im DE-Lager, 8. Juli 2026 Lieferbar". parseWarehouseStock
-      const deM =
+      //   extracts the bare word-date per warehouse and gates units_sold.
-        t.match(/(\d[\d.,KkMm]*)\s*Stk\.\s*im\s*DE[- ]Lager/i) ??
+      const wh = parseWarehouseStock(t, FS_MAX_UNITS_SOLD);
-        t.match(/(\d[\d.,KkMm]*)\s*im\s*DE[- ]?Lager/i);
+      const deQty = wh.deQty;
-      if (deM?.[1]) {
+      const deDeliveryDate = wh.deDeliveryDate;
-        deQty = parseGermanQty(deM[1]);
+      const globalQty = wh.globalQty;
-        const idx = t.indexOf(deM[0]);
+      const globalDeliveryDate = wh.globalDeliveryDate;
-        const ctx = t.slice(idx, idx + 300);
+      const backorderQty = wh.backorderQty;
-        const dm =
+      const backorderDate = wh.backorderDate;
-          ctx.match(/Lieferung[:\s]+([0-9]{1,2}\.?\s+\w+\.?,?\s*\d{4})/i) ??
+      const unitsSold = wh.unitsSold;
-          ctx.match(/erwartet[:\s]+([0-9]{1,2}\.?\s+\w+\.?,?\s*\d{4})/i) ??
-          ctx.match(/([0-9]{1,2}\.[0-9]{1,2}\.[0-9]{4})/);
-        if (dm?.[1]) deDeliveryDate = parseGermanDate(dm[1]);
-      }
-
-      // ── Global-Lager ───────────────────────────────────────────────────────
-      let globalQty: number | undefined;
-      let globalDeliveryDate: string | undefined;
-      const glM =
-        t.match(/(\d[\d.,KkMm]*)\s*Stk\.\s*im\s*Global[- ]Lager/i) ??
-        t.match(/(\d[\d.,KkMm]*)\s*im\s*Global[- ]?(?:Lager|Warehouse)/i) ??
-        t.match(/(\d[\d.,KkMm]*)\s*in\s+Global\s+Warehouse/i);
-      if (glM?.[1]) {
-        globalQty = parseGermanQty(glM[1]);
-        const idx = t.indexOf(glM[0]);
-        const ctx = t.slice(idx, idx + 300);
-        const dm =
-          ctx.match(/Lieferung[:\s]+([0-9]{1,2}\.?\s+\w+\.?,?\s*\d{4})/i) ??
-          ctx.match(/erwartet[:\s]+([0-9]{1,2}\.?\s+\w+\.?,?\s*\d{4})/i) ??
-          ctx.match(/([0-9]{1,2}\.[0-9]{1,2}\.[0-9]{4})/);
-        if (dm?.[1]) globalDeliveryDate = parseGermanDate(dm[1]);
-      }
-
-      // ── Nachlieferung ──────────────────────────────────────────────────────
-      let backorderQty: number | undefined;
-      let backorderDate: string | undefined;
-      const boM =
-        t.match(/(\d[\d.,KkMm]*)\s*(?:Stk\.)?\s*in\s+Nachlieferung/i) ??
-        t.match(/Nachlieferung[:\s]*(\d[\d.,KkMm]*)/i);
-      if (boM?.[1]) {
-        backorderQty = parseGermanQty(boM[1]);
-        const idx = t.indexOf(boM[0]);
-        const ctx = t.slice(idx, idx + 300);
-        const dm =
-          ctx.match(/[Ee]rwartet[:\s]+([0-9]{1,2}\.?\s+\w+\.?,?\s*\d{4})/i) ??
-          ctx.match(/Lieferung[:\s]+([0-9]{1,2}\.?\s+\w+\.?,?\s*\d{4})/i) ??
-          ctx.match(/([0-9]{1,2}\.[0-9]{1,2}\.[0-9]{4})/);
-        if (dm?.[1]) backorderDate = parseGermanDate(dm[1]);
-      }
-
-      // ── Units sold ─────────────────────────────────────────────────────────
-      let unitsSold: number | undefined;
-      const soldM =
-        t.match(/(\d[\d.,KkMm]*)\s*(?:[Mm]al\s+)?[Vv]erkauft/) ??
-        t.match(/([\d.,KkMm]+)\+?\s*sold/i);
-      if (soldM?.[1]) unitsSold = parseGermanQty(soldM[1]);
 
       // ── Part number refinement ─────────────────────────────────────────────
       let partNumber = listingPn;
@@ -1010,6 +896,12 @@ export async function scrapeFs(): Promise<void> {
 
       const stockLevel = deriveStockLevel(detail.deQty, detail.globalQty, detail.backorderQty);
       const totalQty = (detail.deQty ?? 0) + (detail.globalQty ?? 0);
+      // Soonest availability across the warehouses that quote a delivery date.
+      const leadTimeDays = computeLeadTimeDays([
+        detail.deDeliveryDate,
+        detail.globalDeliveryDate,
+        detail.backorderDate,
+      ]);
 
       if (detail.priceNet && detail.priceNet > 0) {
         const hash = contentHash({
@@ -1024,6 +916,7 @@ export async function scrapeFs(): Promise<void> {
           currency: "EUR",
           stockLevel,
           quantityAvailable: totalQty > 0 ? totalQty : undefined,
+          leadTimeDays,
           url: detail.url,
           contentHash: hash,
         });
@@ -1042,6 +935,7 @@ export async function scrapeFs(): Promise<void> {
         backorderQty: detail.backorderQty,
         backorderEstimatedDate: detail.backorderDate ?? null,
         unitsSold: detail.unitsSold,
+        leadTimeDays,
         compatibleBrands: detail.compatibleBrands,
         priceNet: detail.priceNet,
         productUrl: detail.url,

packages/scraper/src/utils/db.ts

@@ -394,6 +394,8 @@ export async function upsertStockObservation(params: {
   backorderQty?: number;
   backorderEstimatedDate?: string | null;
   unitsSold?: number;
+  /** Soonest availability in days across warehouses that quote a delivery date. */
+  leadTimeDays?: number;
   compatibleBrands?: string[];
   priceNet?: number;
   productUrl?: string;
@@ -444,7 +446,7 @@ export async function upsertStockObservation(params: {
        warehouse_de_qty, warehouse_de_delivery_date,
        warehouse_global_qty, warehouse_global_delivery_date,
        backorder_qty, backorder_estimated_date,
-       units_sold, compatible_brands, price_net, product_url,
+       units_sold, lead_time_days, compatible_brands, price_net, product_url,
        stock_confidence, price_currency, price_includes_tax, stock_vendor_ts
      ) VALUES (
        NOW(), $1, $2,
@@ -452,8 +454,8 @@ export async function upsertStockObservation(params: {
        $5, $6::date,
        $7, $8::date,
        $9, $10::date,
-       $11, $12, $13, $14,
+       $11, $12, $13, $14, $15,
-       $15, $16, $17, $18
+       $16, $17, $18, $19
      )`,
     [
       params.transceiverId,
@@ -467,6 +469,7 @@ export async function upsertStockObservation(params: {
       params.backorderQty ?? null,
       params.backorderEstimatedDate ?? null,
       params.unitsSold ?? null,
+      params.leadTimeDays ?? null,
       params.compatibleBrands?.length ? params.compatibleBrands : null,
       params.priceNet ?? null,
       params.productUrl ?? null,

packages/scraper/tsconfig.json

@@ -15,5 +15,5 @@
     "incremental": true
   },
   "include": ["src/**/*"],
-  "exclude": ["node_modules", "dist"]
+  "exclude": ["node_modules", "dist", "src/**/*.test.ts"]
 }

sql/122-fs-com-delivery-dates-and-units-sold-gate.sql

@@ -0,0 +1,54 @@
+-- Migration 122: FS.com delivery dates + lead_time now collected; purge units_sold garbage
+-- ─────────────────────────────────────────────────────────────────────────────
+-- Follow-up to migration 121, which documented lead_time_days as NOT COLLECTED and
+-- flagged that populating it was a scraper-side task. That task is now done:
+--
+--   * fs-com.ts (via fs-com-parse.ts) parses the bare word-date FS.com renders on
+--     each warehouse line ("… im DE-Lager, 8. Juli 2026 Lieferbar") into
+--     warehouse_de_delivery_date / warehouse_global_delivery_date /
+--     backorder_estimated_date, and derives lead_time_days = min(date − today).
+--   * upsertStockObservation now writes lead_time_days.
+--
+-- Historical rows stay NULL until the next FS.com run re-scrapes each SKU; NULL
+-- therefore still means "unknown / not yet captured", not "zero lead time".
+--
+-- Second fix: units_sold was ~10× inflated for decimal-abbreviated values
+-- ("4.8K"→48000, "1.4M"→14000000) because the K/M parser stripped the decimal
+-- point. The parser is fixed AND ingest now drops values above a sanity cap.
+-- This migration removes the clearly-impossible historical values (the multi-
+-- million "outliers") so no consumer ingests them before the re-scrape. Sub-cap
+-- values that may still be mildly inflated self-correct on the next FS.com run
+-- (a changed value writes a fresh observation; consumers read the latest).
+-- ─────────────────────────────────────────────────────────────────────────────
+
+-- 1. Purge impossible units_sold (matches the ingest gate default, FS_MAX_UNITS_SOLD=2,000,000).
+UPDATE stock_observations
+SET units_sold = NULL
+WHERE units_sold IS NOT NULL
+  AND units_sold > 2000000;
+
+-- 2. Re-document the columns as collected-by-scraper (reversing migration 121's note).
+COMMENT ON COLUMN stock_observations.warehouse_de_delivery_date IS
+  'FS.com DE-Lager availability date (collected 2026-07-04+ by fs-com-parse). '
+  'NULL = warehouse in stock now, vendor did not quote a date, or SKU not yet re-scraped.';
+
+COMMENT ON COLUMN stock_observations.warehouse_global_delivery_date IS
+  'FS.com Global-Lager availability date (collected 2026-07-04+ by fs-com-parse). '
+  'NULL semantics as warehouse_de_delivery_date.';
+
+COMMENT ON COLUMN stock_observations.backorder_estimated_date IS
+  'FS.com Nachlieferung (backorder) estimated date (collected 2026-07-04+ by fs-com-parse). '
+  'NULL = no backorder date quoted or SKU not yet re-scraped.';
+
+COMMENT ON COLUMN stock_observations.lead_time_days IS
+  'Soonest availability in days = min(delivery_date − today) across warehouses that quote a '
+  'date. Collected 2026-07-04+ by the FS.com scraper; other vendors do not expose a lead time. '
+  'Treat NULL as "unknown / not captured" (do NOT default to 0).';
+
+COMMENT ON COLUMN stock_observations.units_sold IS
+  'FS.com "verkauft" lifetime sell-through counter. Parser de-abbreviates K/M correctly and '
+  'ingest drops values > 2,000,000 as implausible. FS.com only; NULL elsewhere.';
+
+INSERT INTO _migrations (filename)
+VALUES ('122-fs-com-delivery-dates-and-units-sold-gate.sql')
+ON CONFLICT (filename) DO NOTHING;