Fix spot dedupe: merge same-named spots across geohash/city + name-suffix variants

The old identity dedupe keyed on `name|place` where place was a single field
(geohash OR location OR city, first non-empty). Two copies of the same spot —
one carrying a geohash, one carrying only a city — produced different keys and
both survived (the Yellow Park duplicate).

Now:
- Name is reduced to a signature that strips trailing generic words
  (skatepark/park/plaza/spot/diy/...) and leading articles, so "Yellow Park",
  "Yellow Park Skatepark" and "Yellow Park " collapse to "yellow".
- Places are compared on human-readable tokens (location/city/country) only;
  the opaque geohash is excluded. An empty place on either side is treated as a
  wildcard, so a geohash-only copy still merges with a "Bangkok" copy.
- Genuinely different spots (different name signatures, or non-empty places with
  no shared token) are still kept separate; single-word names are preserved so
  generic names don't over-collapse.

Co-authored-by: shakespeare.diy <assistant@shakespeare.diy>

src/hooks/useCVMClient.ts

@@ -224,27 +224,93 @@ function placeMatches(candidate: string, query: string): boolean {
     .some((word) => word.length >= 3 && editDistance(word, q, 1) <= 1);
 }
 
+/**
+ * Reduce a spot name to a stable dedupe signature: lowercased, punctuation
+ * stripped, and common generic suffixes ("skatepark", "park", "plaza", "spot",
+ * "diy", etc.) removed — so "Yellow Park", "Yellow Park Skatepark" and
+ * "Yellow Park " all collapse to the same "yellow" core. Kept conservative:
+ * we only strip trailing generic words, never interior ones.
+ */
+const GENERIC_SPOT_WORDS = new Set([
+  'skatepark', 'skateplaza', 'skate', 'park', 'plaza', 'spot', 'diy',
+  'bowl', 'ramp', 'the', 'a',
+]);
+
+function spotNameSignature(name: string): string {
+  const words = normalizeForDedupe(name)
+    .replace(/[^a-z0-9 ]/gi, '')
+    .split(' ')
+    .filter(Boolean);
+  // Drop trailing generic words ("yellow park skatepark" → "yellow").
+  while (words.length > 1 && GENERIC_SPOT_WORDS.has(words[words.length - 1])) {
+    words.pop();
+  }
+  // Also drop leading articles ("the yellow" → "yellow").
+  while (words.length > 1 && GENERIC_SPOT_WORDS.has(words[0])) {
+    words.shift();
+  }
+  return words.join(' ');
+}
+
 /**
  * Collapse spots that are the SAME real-world spot published as different
  * addressable events — e.g. the same place re-posted under a new `d` identifier
  * or by two different accounts. Coordinate dedupe (kind:pubkey:d) can't catch
- * these because the coordinates differ, so we also dedupe on a
- * name + location/geohash signature and keep the freshest event.
+ * these because the coordinates differ.
+ *
+ * Two spots merge when their NAME signatures match AND their places are
+ * compatible. "Compatible" is deliberately loose: a spot with a geohash and a
+ * spot with only a city name should still merge if the names match, because the
+ * same real-world spot is often posted once with precise coords and once with
+ * just "Bangkok". So we treat an EMPTY place on either side as a wildcard, and
+ * otherwise require the two place strings to share a token (e.g. both mention
+ * "bangkok"). This is what was letting Yellow Park duplicate: one copy carried a
+ * geohash, the other a city, so the old geohash-vs-city keys never matched.
  */
 function deduplicateByIdentity<T extends { tags: string[][]; created_at: number }>(spots: T[]): T[] {
-  const seen = new Map<string, T>();
+  const kept: T[] = [];
+
+  // Human-readable place tokens only (city/location/country) — NOT the geohash,
+  // which is an opaque code that would never token-match a city name.
+  const placeTokens = (e: T): Set<string> => {
+    const raw = [getTag(e, 'location'), getTag(e, 'city'), getTag(e, 'country')]
+      .map((v) => normalizeForDedupe(v))
+      .filter(Boolean)
+      .join(' ');
+    return new Set(raw.split(/[^a-z0-9]+/i).filter((t) => t.length >= 3));
+  };
+
+  const placesCompatible = (a: Set<string>, b: Set<string>): boolean => {
+    // Empty human-place on either side → can't contradict (e.g. one copy only
+    // carried a geohash). Treat as compatible so a geohash-only copy still
+    // merges with a "Bangkok" copy of the same-named spot.
+    if (a.size === 0 || b.size === 0) return true;
+    for (const t of a) if (b.has(t)) return true;
+    return false;
+  };
+
   for (const e of spots) {
-    const name = normalizeForDedupe(getTag(e, 'title') ?? getTag(e, 'name'));
-    // Prefer geohash (precise) then location text as the place signature.
-    const place = normalizeForDedupe(getTag(e, 'g') ?? getTag(e, 'location') ?? getTag(e, 'city'));
-    // No name → can't confidently match; keep it (unique key on coordinate).
-    const key = name ? `${name}|${place}` : `__unique__${buildNaddr(e as unknown as { kind: number; pubkey: string; tags: string[][] })}`;
-    const existing = seen.get(key);
-    if (!existing || e.created_at > existing.created_at) {
-      seen.set(key, e);
+    const sig = spotNameSignature(getTag(e, 'title') ?? getTag(e, 'name') ?? '');
+    // No usable name → can't confidently merge; always keep it.
+    if (!sig) {
+      kept.push(e);
+      continue;
+    }
+    const tokens = placeTokens(e);
+    const dupIdx = kept.findIndex(
+      (k) =>
+        spotNameSignature(getTag(k, 'title') ?? getTag(k, 'name') ?? '') === sig &&
+        placesCompatible(placeTokens(k), tokens)
+    );
+    if (dupIdx === -1) {
+      kept.push(e);
+    } else if (e.created_at > kept[dupIdx].created_at) {
+      // Same spot, newer copy → replace, preferring the richer place info.
+      kept[dupIdx] = e;
     }
   }
-  return Array.from(seen.values());
+
+  return kept;
 }
 
 function buildNaddr(event: { kind: number; pubkey: string; tags: string[][] }): string {