Fix: deduplicate localized reply/forward subject prefixes

Replying to a reply produced "Re: Re: foo" (and German used the English "Re:"/"Fwd:" instead of "AW:"/"WG:"). Four code paths built reply/forward subjects and only one deduplicated - and only for the English prefix, so cross-locale threads accumulated chains. New lib/subject-prefix.ts strips any leading run of reply/forward markers across ~35 tokens from all supported languages (plus Outlook Re[2]: and Eudora Re*2: counters), then prepends the locale-appropriate prefix. All four call sites (composer getInitialSubject, the two page.tsx sites, and the three pro-tab handlers) now use buildReplySubject/buildForwardSubject. German prefix corrected to AW:/WG:.
2026-05-30 15:47:53 +02:00
parent 5a70cf95e0
commit 4c1d0931a1
5 changed files with 130 additions and 12 deletions
@@ -0,0 +1,117 @@
+// Reply / forward subject prefix handling.
+//
+// Real-world email subjects accumulate prefixes across clients and languages:
+// "Re: AW: WG: Fwd: Re: foo". The deduplication regex needs to know ALL
+// commonly-used reply/forward markers - not just the current locale's, since
+// inbound messages may come from any locale. Failing to strip a foreign-locale
+// prefix means the user's locale prefix gets *added on top* and the subject
+// chain keeps growing.
+//
+// Sources: de-facto conventions in Outlook / Thunderbird / Apple Mail per
+// language. Includes a handful of legacy short-forms (R:, Fw:) that some
+// mobile clients still emit.
+
+const REPLY_TOKENS = [
+  "Re",      // English, Italian, French (also generic ISO)
+  "RE",      // Outlook variant
+  "AW",      // German (Antwort)
+  "Antw",    // German verbose
+  "Sv",      // Danish / Swedish / Norwegian (Svar)
+  "Yn",      // Turkish (Yanit)
+  "Yanit",   // Turkish verbose
+  "Odp",     // Polish (Odpowiedz)
+  "Ответ",   // Russian
+  "Resp",    // Spanish/Portuguese variant
+  "Vá",      // Hungarian
+  "回复",    // Chinese
+  "回覆",    // Chinese traditional
+  "답장",    // Korean
+  "R",       // Italian short form
+];
+
+const FORWARD_TOKENS = [
+  "Fwd",     // English standard
+  "Fw",      // English short / Polish / German short
+  "WG",      // German (Weitergeleitet)
+  "Tr",      // French (Transfert)
+  "Vs",      // Danish (Videresend)
+  "Enc",     // Portuguese (Encaminhar)
+  "ENC",     // Portuguese caps
+  "Rv",      // Spanish (Reenviar)
+  "RV",      // Spanish caps
+  "Rvf",     // Spanish variant
+  "Inol",    // Italian (Inoltro)
+  "I",       // Italian short
+  "PD",      // Polish (Przekazane Dalej)
+  "PR",      // Czech (Preposlat)
+  "İlt",     // Turkish (Ilet)
+  "Ilt",     // Turkish ASCII
+  "Пересл",  // Russian (Peresylka)
+  "Пер",     // Russian short
+  "转发",    // Chinese
+  "轉寄",    // Chinese traditional
+  "전달",    // Korean
+];
+
+// Match a single prefix token + optional [N] counter (Outlook) or *N (Eudora)
+// + colon + whitespace. Case-insensitive. The non-capturing groups keep the
+// regex composable for stripping multiple prefixes in a row.
+function buildPrefixRegex(tokens: string[]): RegExp {
+  // Escape regex specials in tokens (none currently, but be defensive)
+  const escaped = tokens.map((t) => t.replace(/[.*+?^${}()|[\]\\]/g, "\\$&"));
+  // Sort by length DESC so longer tokens (e.g. "Пересл") win over their
+  // shorter prefixes (e.g. "Пер") during alternation matching.
+  escaped.sort((a, b) => b.length - a.length);
+  return new RegExp(
+    `^\\s*(?:${escaped.join("|")})(?:\\[\\d+\\]|\\*\\d*)?\\s*:\\s*`,
+    "i",
+  );
+}
+
+const ANY_PREFIX_RE = buildPrefixRegex([...REPLY_TOKENS, ...FORWARD_TOKENS]);
+
+/**
+ * Strip any leading sequence of reply/forward prefixes (across languages) from
+ * a subject line. Idempotent and safe for empty input.
+ *
+ * Examples:
+ *   stripSubjectPrefixes("Re: AW: WG: foo")        === "foo"
+ *   stripSubjectPrefixes("Re[2]: foo")             === "foo"
+ *   stripSubjectPrefixes("RE: Re: foo")            === "foo"
+ *   stripSubjectPrefixes("foo")                    === "foo"
+ *   stripSubjectPrefixes("")                       === ""
+ */
+export function stripSubjectPrefixes(subject: string | undefined | null): string {
+  if (!subject) return "";
+  let s = subject;
+  // Bounded loop: in practice you never see more than ~10 prefixes; the bound
+  // protects against pathological input. Each iteration must consume input.
+  for (let i = 0; i < 20; i++) {
+    const next = s.replace(ANY_PREFIX_RE, "");
+    if (next === s) break;
+    s = next;
+  }
+  return s;
+}
+
+/**
+ * Build a reply subject with the given locale-aware prefix. Strips any
+ * pre-existing prefixes (in any language) first so chains don't accumulate.
+ *
+ *   buildReplySubject("AW: WG: foo", "Re:")   === "Re: foo"
+ *   buildReplySubject("foo", "AW:")           === "AW: foo"
+ *   buildReplySubject("", "AW:")              === "AW:"
+ */
+export function buildReplySubject(subject: string | undefined | null, prefix: string): string {
+  const stripped = stripSubjectPrefixes(subject);
+  return stripped ? `${prefix} ${stripped}` : prefix;
+}
+
+/**
+ * Build a forward subject. Same logic as buildReplySubject but conceptually
+ * separate for clarity at the call site.
+ */
+export function buildForwardSubject(subject: string | undefined | null, prefix: string): string {
+  const stripped = stripSubjectPrefixes(subject);
+  return stripped ? `${prefix} ${stripped}` : prefix;
+}