Alias resolution polish: lazy-fire on direct-match failure + worker backfill

Two perf gaps that would have failed Cin's review:

# Gap #1: alias lookup fired unconditionally

Pre-fix in this commit, `_resolve_expected_artist_aliases` ran at
the top of every `verify_audio_file` call regardless of whether
the direct artist match would have passed. For users whose library
is mostly same-script (95% of cases), every successful verification
was paying for a wasted DB query (and possibly a wasted MB API
call for un-enriched artists).

Restructured the helper to accept a callable provider instead of a
pre-resolved list. Provider invoked LAZILY only when direct
similarity falls below `ARTIST_MATCH_THRESHOLD`. Verifier passes a
memoising thunk that resolves once across the 3 comparison sites
within one verification.

`_alias_aware_artist_sim` now accepts `aliases` as either:

- iterable of strings (used eagerly — backward compat with tests
  that already know the aliases)
- callable returning the iterable (resolved on first need within
  a verification)

Happy path (direct match passes): zero DB queries, zero MB calls.
Cross-script case: one resolution shared across 3 sites — same as
the prior contract.

# Gap #2: existing-MBID artists never got alias backfill

Worker's `_process_item` artist branch had an `existing_id` short-
circuit (line 296) that updated MBID status but skipped alias
fetch. Result: every user with an already-enriched library had
MBIDs but NULL aliases on day-one of this PR. Live MB lookup at
verify-time covered them, but at the cost of N live calls for N
artists across the library.

Added one-time backfill: when existing-MBID is found AND
`artists.aliases` for that row is empty, fetch + persist aliases.
Subsequent re-scan cycles short-circuit on the populated column —
no repeated MB calls.

New helper `_artist_aliases_empty(artist_id)` does the cheap NULL
check via direct SQL. Best-effort: defensively returns True on
errors so backfill happens (a redundant MB call is cheaper than
missing the backfill entirely).

# Tests added (9)

`test_acoustid_verification_aliases.py` (+6):
- `TestLazyAliasResolution` (3): no lookup when direct match passes,
  lookup fires only when direct fails, lookup memoised across the
  3 sites within one verification.
- `TestAliasProviderCallable` (3): iterable passed directly,
  callable resolves lazily, callable returning empty falls back to
  direct sim.

`test_artist_alias_service.py` (+3):
- `test_existing_mbid_path_backfills_aliases_when_column_empty`
- `test_existing_mbid_path_skips_backfill_when_aliases_already_set`
- `test_existing_mbid_backfill_failure_does_not_break_match`

# Verification

- 79/79 matching tests pass (+9 from prior commit)
- 2537 full suite passes (+9, +79 PR-total)
- Ruff clean
- Backward compat: every prior-commit test still passes (the
  iterable-shape API still works alongside the new callable shape)

core/acoustid_verification.py

@@ -90,7 +90,7 @@ def _similarity(a: str, b: str) -> float:
 def _alias_aware_artist_sim(
     expected_artist: str,
     actual_artist: str,
-    aliases: Optional[List[str]] = None,
+    aliases: Optional[Any] = None,
 ) -> float:
     """Best artist-similarity across (expected, *aliases) vs actual.
 
@@ -104,12 +104,39 @@ def _alias_aware_artist_sim(
     threshold checks (>= ARTIST_MATCH_THRESHOLD) keep their
     semantics. When `aliases` is None or empty, behaves identically
     to the prior raw `_similarity(expected, actual)` call.
+
+    `aliases` accepts two shapes:
+
+    - **Iterable** (list/tuple/set of strings): used directly. Used
+      by tests that already know the aliases.
+    - **Callable**: invoked LAZILY only when direct similarity
+      falls below the threshold. Lets the verifier pass a memoizing
+      thunk that resolves aliases (DB / cache / live MB) only when
+      needed. Verifications where the direct match already passes
+      never trigger the lookup chain — no wasted DB query for the
+      happy path.
     """
     from core.matching.artist_aliases import artist_names_match
+
+    direct = _similarity(expected_artist, actual_artist)
+    # Fast path — direct match already passes the threshold OR caller
+    # supplied no aliases handle. Avoids any lookup work.
+    if aliases is None:
+        return direct
+    if direct >= ARTIST_MATCH_THRESHOLD:
+        return direct
+
+    # Resolve the iterable. Callable provider invoked NOW (lazily —
+    # the caller can memoize the result across multiple invocations
+    # within one verify_audio_file call).
+    resolved = aliases() if callable(aliases) else aliases
+    if not resolved:
+        return direct
+
     _matched, score = artist_names_match(
         expected_artist,
         actual_artist,
-        aliases=aliases,
+        aliases=resolved,
         threshold=ARTIST_MATCH_THRESHOLD,
         similarity=_similarity,
     )
@@ -120,17 +147,22 @@ def _find_best_title_artist_match(
     recordings: List[Dict[str, Any]],
     expected_title: str,
     expected_artist: str,
-    expected_artist_aliases: Optional[List[str]] = None,
+    expected_artist_aliases: Optional[Any] = None,
 ) -> Tuple[Optional[Dict], float, float]:
     """
     Find the AcoustID recording that best matches expected title/artist.
 
     Issue #442 — `expected_artist_aliases` (when supplied) is the
     list of alternate spellings for `expected_artist` (Japanese
-    kanji, Cyrillic, etc.). Each recording's artist is scored
-    against (expected, *aliases) and the best score wins. When the
-    list is empty or omitted, behavior is identical to the prior
-    raw similarity comparison.
+    kanji, Cyrillic, etc.). Accepts either:
+
+    - An iterable of alias strings (used eagerly), or
+    - A callable returning the list (resolved lazily — only fires
+      when at least one recording fails direct artist similarity).
+
+    Each recording's artist is scored against (expected, *aliases)
+    and the best score wins. When the list is empty/omitted/None,
+    behavior is identical to the prior raw similarity comparison.
 
     Returns:
         (best_recording, title_similarity, artist_similarity)
@@ -371,23 +403,33 @@ class AcoustIDVerification:
             # Enrich recordings that are missing title/artist via MusicBrainz lookup
             recordings = _enrich_recordings_from_musicbrainz(recordings)
 
-            # Issue #442 — resolve alternate-spelling aliases for the
-            # expected artist ONCE. Multi-tier resolution (library DB
-            # → cache → live MB), cached per artist name so 100
-            # quarantine candidates with the same artist don't trigger
-            # 100 MB API calls. Empty list on any failure → verifier
-            # falls back to prior direct-similarity behaviour.
-            expected_artist_aliases = _resolve_expected_artist_aliases(expected_artist_name)
-            if expected_artist_aliases:
-                logger.debug(
-                    "Resolved %d aliases for expected artist '%s'",
-                    len(expected_artist_aliases), expected_artist_name,
-                )
+            # Issue #442 — alias resolution is LAZY. We pass a memoising
+            # thunk to the artist-comparison sites; it only fires the
+            # multi-tier lookup (library DB → cache → live MB) when
+            # direct artist similarity falls below threshold. Verifications
+            # where the direct match already passes (the common case for
+            # same-script artist names) never trigger any lookup work,
+            # so the fix doesn't add a per-verification DB query for the
+            # happy path. When the thunk DOES fire, the result is cached
+            # in the closure so the 3 comparison sites within one
+            # verification share a single resolution pass.
+            _alias_cache: Dict[str, Any] = {}
+
+            def _aliases_provider() -> List[str]:
+                if 'value' not in _alias_cache:
+                    resolved = _resolve_expected_artist_aliases(expected_artist_name)
+                    _alias_cache['value'] = resolved
+                    if resolved:
+                        logger.debug(
+                            "Resolved %d aliases for expected artist '%s'",
+                            len(resolved), expected_artist_name,
+                        )
+                return _alias_cache['value']
 
             # Step 4: Find best title/artist match among AcoustID results
             best_rec, title_sim, artist_sim = _find_best_title_artist_match(
                 recordings, expected_track_name, expected_artist_name,
-                expected_artist_aliases=expected_artist_aliases,
+                expected_artist_aliases=_aliases_provider,
             )
 
             if not best_rec:
@@ -449,7 +491,7 @@ class AcoustIDVerification:
                 for rec in recordings:
                     rec_artist = rec.get('artist', '')
                     if _alias_aware_artist_sim(
-                        expected_artist_name, rec_artist, expected_artist_aliases,
+                        expected_artist_name, rec_artist, _aliases_provider,
                     ) >= ARTIST_MATCH_THRESHOLD:
                         msg = (
                             f"Audio verified: found '{expected_track_name}' by '{expected_artist_name}' "
@@ -499,7 +541,7 @@ class AcoustIDVerification:
                     continue
                 if (_similarity(expected_track_name, t) >= TITLE_MATCH_THRESHOLD and
                         _alias_aware_artist_sim(
-                            expected_artist_name, a, expected_artist_aliases,
+                            expected_artist_name, a, _aliases_provider,
                         ) >= ARTIST_MATCH_THRESHOLD):
                     msg = (
                         f"Audio verified: found '{t}' by '{a}' in AcoustID results "

core/musicbrainz_worker.py

@@ -283,6 +283,31 @@ class MusicBrainzWorker:
             if conn:
                 conn.close()
 
+    def _artist_aliases_empty(self, artist_id: Any) -> bool:
+        """Check if `artists.aliases` for this row is NULL or empty.
+
+        Used by the existing-MBID backfill path to skip the MB call
+        when aliases are already populated (re-scan cycles after
+        backfill complete should be no-ops). Defensive: returns True
+        on any error so the backfill attempt happens — a redundant MB
+        call is cheaper than missing the backfill entirely.
+        """
+        conn = None
+        try:
+            conn = self.db._get_connection()
+            cursor = conn.cursor()
+            cursor.execute("SELECT aliases FROM artists WHERE id = ? LIMIT 1", (artist_id,))
+            row = cursor.fetchone()
+            if not row:
+                return False  # Row doesn't exist — nothing to backfill
+            value = row[0]
+            return value is None or value == '' or value == '[]'
+        except Exception:
+            return True
+        finally:
+            if conn:
+                conn.close()
+
     def _process_item(self, item: Dict[str, Any]):
         """Process a single item (artist, album, or track)"""
         try:
@@ -301,6 +326,27 @@ class MusicBrainzWorker:
                 try:
                     if item_type == 'artist':
                         self.mb_service.update_artist_mbid(item_id, existing_id, 'matched')
+                        # Issue #442 — one-time backfill for artists
+                        # enriched before alias support landed. Users with
+                        # pre-existing libraries on day-one of this PR have
+                        # MBIDs but NULL aliases. Fetch ONLY when the
+                        # column is empty so re-scan cycles after backfill
+                        # don't re-query MB. Best-effort: failures are
+                        # logged at debug, don't regress the match outcome.
+                        try:
+                            if self._artist_aliases_empty(item_id):
+                                aliases = self.mb_service.fetch_artist_aliases(existing_id)
+                                if aliases:
+                                    self.mb_service.update_artist_aliases(item_id, aliases)
+                                    logger.debug(
+                                        "Backfilled %d aliases for artist '%s'",
+                                        len(aliases), item_name,
+                                    )
+                        except Exception as backfill_err:
+                            logger.debug(
+                                "Alias backfill failed for artist '%s': %s",
+                                item_name, backfill_err,
+                            )
                     elif item_type == 'album':
                         self.mb_service.update_album_mbid(item_id, existing_id, 'matched')
                     elif item_type == 'track':

tests/matching/test_acoustid_verification_aliases.py

@@ -274,3 +274,128 @@ class TestAliasLookupCalledOncePerVerify:
         verifier.verify_audio_file('/fake/path.mp3', 'X', 'Hiroyuki Sawano')
 
         assert fake_service.lookup_artist_aliases.call_count == 1
+
+
+# ---------------------------------------------------------------------------
+# Lazy alias resolution — happy path skips MB lookup entirely
+# ---------------------------------------------------------------------------
+
+
+class TestLazyAliasResolution:
+    """Issue #442 perf followup: alias lookup should ONLY fire when
+    the direct artist comparison fails. Verifications where artist
+    names already match (the 95% common case for same-script
+    libraries) must NOT trigger the lookup chain — no wasted DB
+    query, no wasted MB call."""
+
+    def test_no_lookup_when_direct_artist_match_passes(self, stubbed_verifier):
+        """Exact-match Latin-script artist passes verification with
+        zero alias lookups — no DB query, no MB call. Same-script
+        libraries (the 95% common case) inherit zero perf cost from
+        this PR."""
+        verifier, fake_service = stubbed_verifier
+
+        verifier.acoustid_client.fingerprint_and_lookup.return_value = {
+            'best_score': 0.95,
+            'recordings': [
+                {'title': 'Dirty White Boy', 'artist': 'Foreigner'},
+            ],
+        }
+
+        result, _ = verifier.verify_audio_file(
+            '/fake/path.mp3', 'Dirty White Boy', 'Foreigner',
+        )
+
+        assert result == VerificationResult.PASS
+        # Critical — alias lookup must NOT have been called for the
+        # happy path. Otherwise every successful verification adds a
+        # DB query for nothing.
+        fake_service.lookup_artist_aliases.assert_not_called()
+
+    def test_lookup_fires_only_when_direct_artist_match_fails(self, stubbed_verifier):
+        """Cross-script case where direct sim is 0% → lookup fires
+        as expected."""
+        verifier, fake_service = stubbed_verifier
+
+        verifier.acoustid_client.fingerprint_and_lookup.return_value = {
+            'best_score': 0.95,
+            'recordings': [
+                {'title': 'YAMANAIAME', 'artist': '澤野弘之'},
+            ],
+        }
+        fake_service.lookup_artist_aliases.return_value = ['澤野弘之']
+
+        result, _ = verifier.verify_audio_file(
+            '/fake/path.mp3', 'YAMANAIAME', 'Hiroyuki Sawano',
+        )
+
+        assert result == VerificationResult.PASS
+        # Lookup fired BECAUSE direct match would have failed
+        fake_service.lookup_artist_aliases.assert_called_once()
+
+    def test_lookup_memoised_across_three_comparison_sites(self, stubbed_verifier):
+        """When lookup DOES fire, the result must be reused across
+        the three artist-comparison sites in the verifier (best-match
+        scoring, secondary scan, fallback scan). One resolution per
+        verification — not three."""
+        verifier, fake_service = stubbed_verifier
+
+        # Force a code path that hits multiple sites: title matches
+        # several recordings but the best-match's artist sim is below
+        # threshold (forces secondary scan path).
+        verifier.acoustid_client.fingerprint_and_lookup.return_value = {
+            'best_score': 0.95,
+            'recordings': [
+                {'title': 'X', 'artist': 'Different Latin Artist'},  # 0 alias hit
+                {'title': 'X', 'artist': '澤野弘之'},                 # alias hit
+            ],
+        }
+        fake_service.lookup_artist_aliases.return_value = ['澤野弘之']
+
+        verifier.verify_audio_file('/fake/path.mp3', 'X', 'Hiroyuki Sawano')
+
+        # Memoised — one resolution shared across all sites
+        assert fake_service.lookup_artist_aliases.call_count == 1
+
+
+# ---------------------------------------------------------------------------
+# Provider-callable contract on the helper
+# ---------------------------------------------------------------------------
+
+
+class TestAliasProviderCallable:
+    """Pin the dual-shape contract on `_alias_aware_artist_sim`:
+    accepts an iterable OR a callable. Callable resolves lazily."""
+
+    def test_iterable_passed_directly(self):
+        """Plain list — used as-is, no lazy semantics."""
+        score = _alias_aware_artist_sim(
+            'Hiroyuki Sawano', '澤野弘之', aliases=['澤野弘之'],
+        )
+        assert score == 1.0
+
+    def test_callable_resolves_lazily_only_when_direct_fails(self):
+        """Callable provider — invoked ONLY when direct sim falls
+        below threshold."""
+        call_count = [0]
+
+        def provider():
+            call_count[0] += 1
+            return ['澤野弘之']
+
+        # Direct match passes → provider NOT called
+        _alias_aware_artist_sim('Foreigner', 'Foreigner', aliases=provider)
+        assert call_count[0] == 0
+
+        # Direct match fails → provider IS called
+        _alias_aware_artist_sim('Hiroyuki Sawano', '澤野弘之', aliases=provider)
+        assert call_count[0] == 1
+
+    def test_callable_returning_empty_list_falls_back_to_direct(self):
+        """Provider returns empty (e.g. MB had no aliases) →
+        score = direct sim, no error."""
+        score = _alias_aware_artist_sim(
+            'Hiroyuki Sawano', '澤野弘之', aliases=lambda: [],
+        )
+        # ~0 because direct cross-script comparison fails
+        assert score < 0.1

tests/matching/test_artist_alias_service.py

@@ -305,6 +305,79 @@ class TestWorkerAliasEnrichment:
         worker.mb_service.fetch_artist_aliases.assert_not_called()
         worker.mb_service.update_artist_aliases.assert_not_called()
 
+    def test_existing_mbid_path_backfills_aliases_when_column_empty(self, temp_db):
+        """Issue #442 perf followup: existing-MBID short-circuit path
+        was skipping alias enrichment entirely. Users with libraries
+        enriched BEFORE this PR shipped have MBIDs but NULL aliases.
+        Worker should fetch aliases on the existing-id path too —
+        one-time backfill on first re-scan post-deploy."""
+        from core.musicbrainz_worker import MusicBrainzWorker
+        artist_id = _seed_artist(temp_db, 'Hiroyuki Sawano')
+
+        worker = MusicBrainzWorker.__new__(MusicBrainzWorker)
+        worker.database = temp_db
+        worker.db = temp_db  # _artist_aliases_empty uses self.db
+        worker.mb_service = MagicMock()
+        worker.mb_service.fetch_artist_aliases.return_value = ['澤野弘之', 'SawanoHiroyuki']
+        worker.stats = {'matched': 0, 'not_found': 0, 'errors': 0}
+        # Existing MBID path
+        worker._get_existing_id = MagicMock(return_value='mb-existing-id')
+
+        worker._process_item({'type': 'artist', 'id': artist_id, 'name': 'Hiroyuki Sawano'})
+
+        # MBID was preserved
+        worker.mb_service.update_artist_mbid.assert_called_once_with(
+            artist_id, 'mb-existing-id', 'matched',
+        )
+        # Aliases backfilled
+        worker.mb_service.fetch_artist_aliases.assert_called_once_with('mb-existing-id')
+        worker.mb_service.update_artist_aliases.assert_called_once_with(
+            artist_id, ['澤野弘之', 'SawanoHiroyuki'],
+        )
+
+    def test_existing_mbid_path_skips_backfill_when_aliases_already_set(self, temp_db):
+        """If aliases are already populated, don't re-fetch — re-scan
+        cycles after backfill complete should be no-ops."""
+        from core.musicbrainz_worker import MusicBrainzWorker
+        artist_id = _seed_artist(
+            temp_db, 'X', aliases=json.dumps(['existing-alias']),
+        )
+
+        worker = MusicBrainzWorker.__new__(MusicBrainzWorker)
+        worker.database = temp_db
+        worker.db = temp_db
+        worker.mb_service = MagicMock()
+        worker.stats = {'matched': 0, 'not_found': 0, 'errors': 0}
+        worker._get_existing_id = MagicMock(return_value='mb-x')
+
+        worker._process_item({'type': 'artist', 'id': artist_id, 'name': 'X'})
+
+        # No alias work — column already populated
+        worker.mb_service.fetch_artist_aliases.assert_not_called()
+        worker.mb_service.update_artist_aliases.assert_not_called()
+
+    def test_existing_mbid_backfill_failure_does_not_break_match(self, temp_db):
+        """Backfill is best-effort — failure to fetch aliases must
+        NOT prevent the MBID-preservation update from happening."""
+        from core.musicbrainz_worker import MusicBrainzWorker
+        artist_id = _seed_artist(temp_db, 'X')
+
+        worker = MusicBrainzWorker.__new__(MusicBrainzWorker)
+        worker.database = temp_db
+        worker.db = temp_db
+        worker.mb_service = MagicMock()
+        worker.mb_service.fetch_artist_aliases.side_effect = Exception("MB down")
+        worker.stats = {'matched': 0, 'not_found': 0, 'errors': 0}
+        worker._get_existing_id = MagicMock(return_value='mb-x')
+
+        # Should NOT raise
+        worker._process_item({'type': 'artist', 'id': artist_id, 'name': 'X'})
+
+        # MBID still preserved despite alias backfill failure
+        worker.mb_service.update_artist_mbid.assert_called_once_with(
+            artist_id, 'mb-x', 'matched',
+        )
+
     def test_alias_fetch_failure_does_not_break_match(self, temp_db):
         """If alias fetch raises (network error, malformed response,
         whatever), the artist match still gets recorded — alias