@dench.com/cli 3.11.6 → 3.11.7-staging.488

package/lib/campaign-csv.ts

@@ -11,6 +11,10 @@ export type RecipientTable = {
   hasHeader: boolean;
   columnLabels: string[];
   columnSamples: string[];
+  /** Where the file's quoting breaks (`scanCsv`); no one is read from it. */
+  error?: string;
+  /** Rows with filled cells past the header's last column. */
+  raggedRows?: number;
 };
 
 export type RecipientField =
@@ -65,8 +69,10 @@ export const MAX_CSV_CHARS = 5_000_000;
 
 const EMAIL_RE = /[A-Za-z0-9._%+'-]+@[A-Za-z0-9.-]+\.[A-Za-z]{2,}/;
 const NAMED_EMAIL_RE = /^(.*?)\s*<\s*([^<>\s]+@[^<>\s]+)\s*>$/;
+// A slug may be in any script (linkedin.com/in/张伟): exports don't always
+// percent-encode it, and `isLinkedinProfileUrl` accepts it.
 const LINKEDIN_URL_RE =
-  /(?<![A-Za-z0-9.-])(?:https?:\/\/)?(?:[a-z]{1,4}\.)?linkedin\.com\/in\/[A-Za-z0-9\-_.%]+\/?/i;
+  /(?<![A-Za-z0-9.-])(?:https?:\/\/)?(?:[a-z]{1,4}\.)?linkedin\.com\/in\/[\p{L}\p{N}\-_.%]+\/?/iu;
 
 const FILE_EXTENSION_TLDS = new Set([
   "png",
@@ -337,8 +343,16 @@ export function extractLinkedinUrl(raw: string): string | null {
   return isLinkedinProfileUrl(url) ? url : null;
 }
 
+// One person per profile, whatever host prefix (uk., www.) or encoding.
 function linkedinKey(url: string): string {
-  return `li:${url.toLowerCase().replace(/^https?:\/\/(www\.)?/, "")}`;
+  const slug = url.match(/linkedin\.com\/in\/([^/?#]+)/i)?.[1] ?? url;
+  let decoded = slug;
+  try {
+    decoded = decodeURIComponent(slug);
+  } catch {
+    decoded = slug;
+  }
+  return `li:${decoded.toLowerCase()}`;
 }
 
 function normalizeEmail(raw: string): string {
@@ -390,17 +404,31 @@ export function detectDelimiter(text: string): CsvDelimiter {
   return best ?? ",";
 }
 
-export function splitCsvRows(
+// A quote opens a quoted value only at the start of a cell; a quoted value
+// must close, with nothing but spaces after it. A broken quote would fold
+// the rows after it into one person's cell, so the file is refused with the
+// line. The same rules as the app's `scanCsv` (outreach-recipients.ts).
+export function scanCsv(
   text: string,
   delimiter: CsvDelimiter = ",",
-): string[][] {
+): { rows: string[][]; error?: string } {
   const rows: string[][] = [];
   let row: string[] = [];
   let cell = "";
   let inQuotes = false;
+  let afterQuote = false;
+  let line = 1;
+  let quoteLine = 1;
+  let error: string | undefined;
   const src = text.replace(/^\uFEFF/, "");
+  const endCell = () => {
+    row.push(cell);
+    cell = "";
+    afterQuote = false;
+  };
   for (let i = 0; i < src.length; i++) {
     const ch = src[i];
+    if (ch === "\n" || (ch === "\r" && src[i + 1] !== "\n")) line += 1;
     if (inQuotes) {
       if (ch === '"') {
         if (src[i + 1] === '"') {
@@ -408,28 +436,51 @@ export function splitCsvRows(
           i += 1;
         } else {
           inQuotes = false;
+          afterQuote = true;
         }
       } else {
         cell += ch;
       }
-    } else if (ch === '"') {
-      inQuotes = true;
     } else if (ch === delimiter) {
-      row.push(cell);
-      cell = "";
+      endCell();
     } else if (ch === "\n" || ch === "\r") {
-      if (ch === "\r" && src[i + 1] === "\n") i += 1;
-      row.push(cell);
-      cell = "";
+      if (ch === "\r" && src[i + 1] === "\n") {
+        i += 1;
+        line += 1;
+      }
+      endCell();
       if (row.some((value) => value.trim() !== "")) rows.push(row);
       row = [];
+    } else if (afterQuote) {
+      if (ch.trim() && !error) {
+        error = `Line ${line} has text after a closing quote. Put the whole value in quotes and try again.`;
+      }
+      if (ch.trim()) cell += ch;
+    } else if (ch === '"' && !cell.trim()) {
+      cell = "";
+      inQuotes = true;
+      quoteLine = line;
     } else {
       cell += ch;
     }
   }
-  row.push(cell);
+  if (inQuotes && !error) {
+    error = `Line ${quoteLine} opens a quote that never closes, so the rows after it can't be read. Fix the quote and try again.`;
+  }
+  endCell();
   if (row.some((value) => value.trim() !== "")) rows.push(row);
-  return rows;
+  return error ? { rows, error } : { rows };
+}
+
+export function splitCsvRows(
+  text: string,
+  delimiter: CsvDelimiter = ",",
+): string[][] {
+  return scanCsv(text, delimiter).rows;
+}
+
+function widestRow(rows: string[][], floor = 0): number {
+  return rows.reduce((width, row) => Math.max(width, row.length), floor);
 }
 
 function cellIsExactly(cell: string, value: string): boolean {
@@ -450,7 +501,7 @@ type ColumnStats = {
 };
 
 function columnStats(rows: string[][]): ColumnStats[] {
-  const width = Math.max(...rows.map((row) => row.length), 0);
+  const width = widestRow(rows);
   const stats: ColumnStats[] = Array.from({ length: width }, () => ({
     samples: 0,
     emailHits: 0,
@@ -516,21 +567,43 @@ function firstRowIsHeader(row: string[]): boolean {
 
 function rowPersonalization(
   row: string[],
-  header: string[],
+  names: readonly string[],
   skip: (col: number) => boolean,
 ): Record<string, string> {
   const bag: Record<string, string> = {};
-  header.forEach((rawName, col) => {
-    const name = rawName.trim();
-    if (!name || skip(col)) return;
+  names.forEach((name, col) => {
+    if (skip(col)) return;
     const value = row[col]?.trim() ?? "";
-    if (!value) return;
-    const key = `crm:${name}`;
-    if (!(key in bag)) bag[key] = value;
+    if (value) bag[`crm:${name}`] = value;
   });
   return bag;
 }
 
+// What each column is called as a merge value: its header, or "Column D".
+// A stored variable name holds printable ASCII only (Convex refuses any
+// other field name): accents drop ("Café" → "Cafe"), a header with nothing
+// left ("職位") falls back to its letter, and names are unique ("Interest",
+// "Interest 2"). The same rule as the app's `columnVariableNames`.
+export function columnVariableNames(table: RecipientTable): string[] {
+  const used = new Set<string>();
+  return table.columnLabels.map((_, col) => {
+    const base =
+      (table.header[col] ?? "")
+        .normalize("NFKD")
+        .replace(/[̀-ͯ]/g, "")
+        .replace(/[{}|]/g, " ")
+        .replace(/[^\x20-\x7e]/g, " ")
+        .replace(/\s+/g, " ")
+        .trim()
+        .slice(0, 100)
+        .trim() || `Column ${columnLetter(col)}`;
+    let name = base;
+    for (let n = 2; used.has(name.toLowerCase()); n += 1) name = `${base} ${n}`;
+    used.add(name.toLowerCase());
+    return name;
+  });
+}
+
 function findNameColumns(header: string[]): {
   first?: number;
   last?: number;
@@ -600,12 +673,17 @@ export function parseRecipientTable(
   csvText: string,
   delimiter: CsvDelimiter = ",",
 ): RecipientTable {
-  const rows = splitCsvRows(csvText, delimiter);
+  const { rows, error } = scanCsv(csvText, delimiter);
   const first = rows[0] ?? [];
   const hasHeader = rows.length > 0 && firstRowIsHeader(first);
   const header = hasHeader ? first.map((cell) => cell.trim()) : [];
   const dataRows = hasHeader ? rows.slice(1) : rows;
-  const width = Math.max(header.length, ...dataRows.map((row) => row.length));
+  const width = widestRow(dataRows, header.length);
+  const raggedRows = hasHeader
+    ? dataRows.filter((row) =>
+        row.slice(header.length).some((cell) => cell.trim()),
+      ).length
+    : 0;
   const columnLabels = Array.from(
     { length: width },
     (_, col) => header[col] || columnLetter(col),
@@ -617,25 +695,44 @@ export function parseRecipientTable(
         .find((cell) => cell.length > 0) ?? "";
     return sample.length > 40 ? `${sample.slice(0, 39)}…` : sample;
   });
-  return { header, rows: dataRows, hasHeader, columnLabels, columnSamples };
+  return {
+    header,
+    rows: dataRows,
+    hasHeader,
+    columnLabels,
+    columnSamples,
+    ...(error ? { error } : {}),
+    ...(raggedRows ? { raggedRows } : {}),
+  };
 }
 
+const EMAIL_HEADER_RE =
+  /^((work|business|primary|professional|personal)[\s_-]*)?e-?mail([\s_-]*address)?$/i;
+const LINKEDIN_HEADER_RE =
+  /^(linkedin([\s_-]*(profile|url|link|profile[\s_-]*(url|link)))?|(linkedin[\s_-]*)?profile[\s_-]*(url|link))$/i;
+
+// Best first: a column whose header names the channel ("Email") before one
+// that only holds more of it ("Backup contact"), then most-filled, leftmost.
 export function detectRecipientColumns(
   table: RecipientTable,
 ): DetectedRecipientColumns {
   const stats = columnStats(table.rows);
-  const rank = (hits: (stat: ColumnStats) => number) =>
+  const rank = (hits: (stat: ColumnStats) => number, header: RegExp) =>
     stats
-      .map((stat, col) => ({ col, hits: hits(stat) }))
+      .map((stat, col) => ({
+        col,
+        hits: hits(stat),
+        named: header.test(table.header[col]?.trim() ?? "") ? 1 : 0,
+      }))
       .filter((entry) => entry.hits > 0)
-      .sort((a, b) => b.hits - a.hits || a.col - b.col)
+      .sort((a, b) => b.named - a.named || b.hits - a.hits || a.col - b.col)
       .map((entry) => entry.col);
   const names = table.hasHeader ? findNameColumns(table.header) : {};
   return {
     firstName: names.first,
     lastName: names.last,
-    email: rank(emailHitsOf),
-    linkedin: rank(linkedinHitsOf),
+    email: rank(emailHitsOf, EMAIL_HEADER_RE),
+    linkedin: rank(linkedinHitsOf, LINKEDIN_HEADER_RE),
   };
 }
 
@@ -668,10 +765,12 @@ export function recipientsFromTable(
     hasHeader: table.hasHeader,
     error,
   });
+  if (table.error) return empty(table.error);
   if (table.rows.length === 0 && !table.hasHeader) {
     return empty("This file is empty.");
   }
   const stats = columnStats(table.rows);
+  const columnNames = columnVariableNames(table);
   const detected = detectRecipientColumns(table);
   const nameColumns = resolveNameColumns(detected, mapping);
   const preferred = (field: "email" | "linkedin") =>
@@ -698,7 +797,7 @@ export function recipientsFromTable(
     nameColumns.first === col ||
     nameColumns.last === col;
   const recipients: CsvRecipient[] = [];
-  const seen = new Set<string>();
+  const seen = new Map<string, CsvRecipient>();
   let noChannel = 0;
   let duplicates = 0;
   for (const row of table.rows) {
@@ -716,31 +815,68 @@ export function recipientsFromTable(
       noChannel += 1;
       continue;
     }
-    const keys = [
-      ...(address ? [normalizeEmail(address)] : []),
-      ...(linkedinUrl ? [linkedinKey(linkedinUrl)] : []),
-    ];
-    if (keys.some((key) => seen.has(key))) {
-      duplicates += 1;
-      continue;
-    }
-    for (const key of keys) seen.add(key);
+    const emailKey = address ? normalizeEmail(address) : undefined;
+    const profileKey = linkedinUrl ? linkedinKey(linkedinUrl) : undefined;
     const personalization = table.hasHeader
-      ? rowPersonalization(row, table.header, consumed)
+      ? rowPersonalization(row, columnNames, consumed)
       : undefined;
     const name =
       rowDisplayName(row, nameColumns)?.trim() ||
       extractNameFromNamedEmail(emailRaw) ||
       (linkedinUrl ? derivedNameFromLinkedinUrl(linkedinUrl) : undefined);
-    recipients.push({
+    const matches = new Set(
+      [emailKey, profileKey].flatMap((key) => {
+        const person = key ? seen.get(key) : undefined;
+        return person ? [person] : [];
+      }),
+    );
+    if (matches.size > 0) {
+      // The same person again: keep what this row adds (their LinkedIn, a
+      // name, a column that was blank), never what the first row said.
+      duplicates += 1;
+      const [existing] = matches;
+      if (matches.size > 1 || !existing) continue;
+      if (address && emailKey && !existing.email) {
+        existing.email = address;
+        seen.set(emailKey, existing);
+      }
+      if (linkedinUrl && profileKey && !existing.linkedinProfileUrl) {
+        existing.linkedinProfileUrl = linkedinUrl;
+        seen.set(profileKey, existing);
+      }
+      if (name && !existing.displayName) existing.displayName = name;
+      if (personalization && Object.keys(personalization).length > 0) {
+        existing.personalization = {
+          ...personalization,
+          ...existing.personalization,
+        };
+      }
+      continue;
+    }
+    const person: CsvRecipient = {
       ...(address ? { email: address } : {}),
       ...(linkedinUrl ? { linkedinProfileUrl: linkedinUrl } : {}),
       ...(name ? { displayName: name } : {}),
       ...(personalization && Object.keys(personalization).length > 0
         ? { personalization }
         : {}),
-    });
+    };
+    if (emailKey) seen.set(emailKey, person);
+    if (profileKey) seen.set(profileKey, person);
+    recipients.push(person);
   }
+  // A channel switched off with `--map email=none` reads nothing from its
+  // column, so the column is reported ignored, not as that channel.
+  const channelColumn = (
+    field: "email" | "linkedin",
+    index: number,
+    hits: (stat: ColumnStats) => number,
+  ) =>
+    preferred(field) === index || isChannelColumn(stats[index], hits)
+      ? mapping[field] === null
+        ? ("ignored" as const)
+        : field
+      : undefined;
   const columns: CsvColumn[] = table.columnLabels.map((label, index) => {
     const role: ColumnRole =
       nameColumns.first === index
@@ -749,22 +885,16 @@ export function recipientsFromTable(
           ? "last name"
           : nameColumns.full === index
             ? "full name"
-            : preferred("email") === index ||
-                isChannelColumn(stats[index], emailHitsOf)
-              ? "email"
-              : preferred("linkedin") === index ||
-                  isChannelColumn(stats[index], linkedinHitsOf)
-                ? "linkedin"
-                : table.hasHeader && table.header[index]?.trim()
-                  ? "merge value"
-                  : "ignored";
+            : (channelColumn("email", index, emailHitsOf) ??
+              channelColumn("linkedin", index, linkedinHitsOf) ??
+              (table.hasHeader ? "merge value" : "ignored"));
     return {
       index,
       label,
       sample: table.columnSamples[index] ?? "",
       role,
-      ...(role === "merge value" || role === "full name"
-        ? { mergeKey: `crm:${table.header[index].trim()}` }
+      ...((role === "merge value" || role === "full name") && table.hasHeader
+        ? { mergeKey: `crm:${columnNames[index]}` }
         : {}),
     };
   });

package/lib/enrichment-gateway.ts

@@ -1519,6 +1519,12 @@ function withTimeout(signal: AbortSignal | undefined, ms: number): AbortSignal {
 export type SignalBuyerBody = {
   companyDomain: string;
   titles?: string[];
+  /** Where the person is; the gateway drops people placed elsewhere. */
+  locations?: string[];
+  /** The company's industries, checked against the person's job. */
+  industries?: string[];
+  /** Kinds of company (Public, Nonprofit, …), checked on the company. */
+  companyTypes?: string[];
   withEmail?: boolean;
   limit?: number;
 };
@@ -1556,6 +1562,9 @@ export async function findSignalBuyer(
     companyDomain: body.companyDomain,
   };
   if (body.titles?.length) payload.titles = body.titles;
+  if (body.locations?.length) payload.locations = body.locations;
+  if (body.industries?.length) payload.industries = body.industries;
+  if (body.companyTypes?.length) payload.companyTypes = body.companyTypes;
   if (body.withEmail !== undefined) payload.withEmail = body.withEmail;
   if (body.limit !== undefined) payload.limit = body.limit;
   return (await callGatewayJson("/v1/enrichment/signals/buyer", {