fix(hub): one entry per file in the Search view's Videos grid

A library shared by two groups arrived in the cross-group Search view as
two entries per file: every film was two poster cards, every episode was
listed twice in the season list under the synopsis. Inside one group
this cannot happen — GroupIndex is keyed by blake3 — so the duplication
was the Search page's own, from concatenating N independently keyed
indexes.

source-merge.js folds entries on the content hash and resolves one
source per *unit* (a film, a whole show), so a season does not scatter
across two nodes. A group hosted by the reader's own node wins; failing
that the pick is a hash of the unit key and the reader's id, stable
across renders and reloads — a source that changed mid-stream would tear
down the connection under a film that is playing — and spread across
readers and units.

The units come from video-app.js's own groupVideoEntries rather than a
second copy of its keys here. Only the Videos view is wired up so far;
Music, Photos, failover and the "N sources" badge are phases 5-8 of
docs/refactoring-search.md.

Every test was checked against the fix removed. That is how the first
version of "a unit's files share its source" turned out to prove
nothing: with every episode in every group, per-file and per-unit
picking give the same answer, so it passed against a per-file
implementation. It now uses a unit whose files have unequal sources.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AbwJDbNTkiRUh7HTWEoyss

1 files changed, 160 insertions, 0 deletions

diff --git a/packages/meshbay-hub/src/meshbay_hub/static/source-merge.js b/packages/meshbay-hub/src/meshbay_hub/static/source-merge.js
new file mode 100644
index 0000000..ccc4cfd
--- /dev/null
+++ b/packages/meshbay-hub/src/meshbay_hub/static/source-merge.js
@@ -0,0 +1,160 @@
+/**
+ * One file, one entry — merging the same content announced by several groups.
+ *
+ * Only the cross-group Search view needs this. Inside a single group the case
+ * cannot arise: `GroupIndex` is keyed by blake3, so the same bytes at two
+ * paths are already one entry. The Search page is what creates duplicates, by
+ * concatenating N independently keyed indexes into one list — a directory
+ * shared by two groups then shows every film twice, every episode twice, every
+ * track twice. See docs/refactoring-search.md.
+ *
+ * Two rules decide everything here:
+ *
+ *  - **Identity is the content hash.** `IndexEntry.id` is blake3 of the file,
+ *    so two entries sharing an id are the same file whatever group announced
+ *    it and whatever path it sits at.
+ *  - **A source is chosen per unit, not per file.** A unit is a film, a whole
+ *    show, a whole album — whatever the reader thinks of as one thing. Picking
+ *    per file would scatter a season's episodes across two nodes, opening two
+ *    connections and two metadata lookups for one show.
+ *
+ * No imports, and there must be none: the whole module is executed standalone
+ * by `test_search_source_merge.py`, which is the only evidence the merge has.
+ */
+
+// FNV-1a, 32 bits. Spreads the source choice deterministically across readers
+// — nothing here is security-bearing, and a cryptographic hash would buy
+// nothing a multiplication does not.
+function stableHash(s) {
+  let h = 0x811c9dc5;
+  for (let i = 0; i < s.length; i++) {
+    h ^= s.charCodeAt(i);
+    h = Math.imul(h, 0x01000193);
+  }
+  return h >>> 0;
+}
+
+function _byGroupId(a, b) {
+  return String(a.groupId).localeCompare(String(b.groupId));
+}
+
+function _source(e) {
+  return { groupId: e.groupId, groupName: e.groupName, groupOwner: e.groupOwner };
+}
+
+/**
+ * Which of `sources` serves this unit.
+ *
+ * A group hosted by the reader's own node wins outright — for someone browsing
+ * their own libraries that is the whole of the reported case, and it makes the
+ * answer predictable where it matters most. Failing that the pick is a hash of
+ * the unit key and a per-reader salt: stable for one reader across renders and
+ * reloads (a source that changed mid-stream would tear down the connection
+ * under a film that is playing), different from one reader to the next, which
+ * is what spreading the load across sources was asking for.
+ *
+ * `isDown` marks a source whose connection has failed. Every source being down
+ * still returns one: a tile that fails to load is better than a film that
+ * vanished from the grid.
+ */
+function pickSource(sources, unitKey, opts) {
+  if (!sources || !sources.length) return null;
+  const { salt = '', isLocal, isDown } = opts || {};
+  const sorted = [...sources].sort(_byGroupId);
+  const up = isDown ? sorted.filter((s) => !isDown(s.groupId)) : sorted;
+  const live = up.length ? up : sorted;
+  const local = isLocal ? live.filter((s) => isLocal(s.groupId)) : [];
+  const pool = local.length ? local : live;
+  return pool[stableHash(`${unitKey}\u0000${salt}`) % pool.length];
+}
+
+/**
+ * Merge the entries of each unit by content hash and resolve one source each.
+ *
+ * `units` is `[{ key, entries }]`, and the caller builds it by running the
+ * application's *own* grouping function over the un-merged list — never a
+ * second copy of that logic, which would keep agreeing right up until one of
+ * them changed. Two lists carrying the same key are one unit, which is what
+ * lets a caller emit a naive `{key: 'movie:<id>', entries: [e]}` per copy of a
+ * film and still get one card.
+ *
+ * Every field of a merged entry comes from the chosen source's own entry, and
+ * nothing is back-filled from another source. A `thumb_hash` or a
+ * `display_title` only one node computed is only fetchable over *that* node's
+ * connection, so borrowing it would produce a poster request the chosen
+ * transport cannot answer — merging two records field by field is the obvious
+ * thing to write here and it is wrong.
+ *
+ * The result carries `_sources`: every group that has the file, sorted by
+ * group id. That is what the badge counts, and it is the only trace left that
+ * more than one group was involved.
+ */
+function mergeUnitEntries(units, opts) {
+  const byKey = new Map();
+  for (const u of units || []) {
+    if (!byKey.has(u.key)) byKey.set(u.key, []);
+    byKey.get(u.key).push(...(u.entries || []));
+  }
+
+  const merged = [];
+  for (const [key, entries] of byKey) {
+    // id -> one entry per group announcing it, first seen wins within a group
+    const byId = new Map();
+    for (const e of entries) {
+      if (!byId.has(e.id)) byId.set(e.id, []);
+      const copies = byId.get(e.id);
+      if (!copies.some((c) => c.groupId === e.groupId)) copies.push(e);
+    }
+
+    const unitSources = [];
+    const seen = new Set();
+    for (const copies of byId.values()) {
+      for (const e of copies) {
+        if (seen.has(e.groupId)) continue;
+        seen.add(e.groupId);
+        unitSources.push(_source(e));
+      }
+    }
+    const unitSource = pickSource(unitSources, key, opts);
+
+    for (const copies of byId.values()) {
+      const sources = copies.map(_source).sort(_byGroupId);
+      let chosen = unitSource
+        && copies.find((e) => e.groupId === unitSource.groupId);
+      if (!chosen) {
+        // A file the unit's source does not hold — an episode only one of the
+        // groups has. Resolved over its own sources by the same rule and the
+        // same unit key, so every such file in the unit lands on the same
+        // fallback instead of scattering one per file.
+        const alt = pickSource(sources, key, opts);
+        chosen = (alt && copies.find((e) => e.groupId === alt.groupId)) || copies[0];
+      }
+      merged.push({ ...chosen, _sources: sources });
+    }
+  }
+  return merged;
+}
+
+/**
+ * What the group badge should say for one entry.
+ *
+ * Returns `{ count, name, groupId }` rather than a rendered string: this module
+ * is executed standalone under node by its test, so it holds no reference to
+ * `i18n.js`. A caller renders `name` (a link to `groupId`) when `count` is 1,
+ * and a plural of `count` otherwise — which group was picked is deliberately
+ * never shown.
+ *
+ * An entry with no `_sources` at all is the single-group Group page, where
+ * there is one source by construction and no badge is drawn.
+ */
+function sourceLabel(entry) {
+  const sources = (entry && entry._sources) || [];
+  if (sources.length > 1) return { count: sources.length, name: '', groupId: null };
+  return {
+    count: 1,
+    name: (entry && entry.groupName) || '',
+    groupId: (entry && entry.groupId) || null,
+  };
+}
+
+export { pickSource, mergeUnitEntries, sourceLabel, stableHash };