Merge pull request #472 from Nezreka/fix/extract-ids-recognize-underscore-source

Make extract_external_ids recognize all source-tagging conventions

core/library/track_identity.py

@@ -67,15 +67,23 @@ def _get(track: Any, *names: str) -> Optional[str]:
     return None
 
 
-def extract_external_ids(track: Any) -> Dict[str, str]:
+def extract_external_ids(track: Any, source_hint: Optional[str] = None) -> Dict[str, str]:
     """Pull every recognized external ID off a metadata-source track.
 
     Handles the source-source naming drift: Spotify tracks expose ``id``
     as the Spotify track ID; Deezer tracks expose ``id`` as the Deezer
     track ID; iTunes tracks may use ``trackId`` or ``id``. The disamb-
-    iguating field is ``provider`` / ``source``. Tracks coming from a
-    SoulSync internal pipeline often carry every known ID set to its
-    source-specific value — we just collect whatever's there.
+    iguating field is ``provider`` / ``source`` / ``_source``. Tracks
+    coming from a SoulSync internal pipeline often carry every known ID
+    set to its source-specific value — we just collect whatever's there.
+
+    ``source_hint`` is the caller's known answer to "where did this
+    track dict come from?" — used as a fallback when the track itself
+    doesn't carry a provider / source / _source field. Spotify and
+    iTunes return raw API responses without provider tags, so the
+    watchlist scanner passes ``get_primary_source()`` here to make sure
+    a Spotify-primary scan isn't silently no-opping just because the
+    raw API track has no provider key.
 
     Returns a dict mapping conceptual ID name → ID value. Keys present
     in ``EXTERNAL_ID_COLUMNS``. Empty dict when no IDs are available.
@@ -106,8 +114,13 @@ def extract_external_ids(track: Any) -> Dict[str, str]:
 
     # Provider field tells us which native ``id`` belongs to. Without
     # this, a Deezer track's ``id`` field would be silently ignored
-    # (we wouldn't know to map it to deezer_id).
-    provider = (_get(track, 'provider', 'source') or '').lower()
+    # (we wouldn't know to map it to deezer_id). Convention varies by
+    # client: Spotify-shaped tracks usually have no provider field,
+    # Deezer / Discogs / Hydrabase clients tag tracks with ``_source``,
+    # internal pipeline normalization may use ``source`` or ``provider``.
+    # Fall back to the caller's source_hint when the track has no
+    # provider field of its own (Spotify / iTunes raw API responses).
+    provider = (_get(track, 'provider', 'source', '_source') or source_hint or '').lower()
     native_id = _get(track, 'id')
     if native_id and provider:
         provider_to_key = {

core/watchlist_scanner.py

@@ -2077,7 +2077,16 @@ class WatchlistScanner:
                     find_provenance_by_external_id,
                 )
                 import os as _os_local
-                source_ids = extract_external_ids(track)
+                # Pass the configured primary source as a hint so the
+                # extractor can disambiguate raw Spotify / iTunes API
+                # responses that don't carry a provider / source field
+                # of their own (Deezer / Discogs / Hydrabase clients
+                # already tag tracks with _source).
+                try:
+                    _source_hint = get_primary_source()
+                except Exception:
+                    _source_hint = None
+                source_ids = extract_external_ids(track, source_hint=_source_hint)
                 if source_ids:
                     matched = find_library_track_by_external_id(
                         self.database,

tests/test_library_track_identity.py

@@ -99,11 +99,43 @@ class TestExtractExternalIdsFromProviderField:
         track = {'source': 'deezer', 'id': 'dz1', 'name': 'Hello'}
         assert extract_external_ids(track) == {'deezer_id': 'dz1'}
 
+    def test_underscore_source_field_treated_same_as_provider(self):
+        """Deezer / Discogs / Hydrabase clients tag normalized tracks
+        with ``_source`` (underscore prefix). Must be recognized as the
+        provider disambiguator — otherwise watchlist scans against those
+        sources silently miss the ID match and fall through to fuzzy."""
+        track = {'_source': 'deezer', 'id': 'dz1', 'name': 'Hello'}
+        assert extract_external_ids(track) == {'deezer_id': 'dz1'}
+
+    def test_underscore_source_hydrabase(self):
+        track = {'_source': 'hydrabase', 'id': 'hyd-soul-1', 'name': 'Hello'}
+        assert extract_external_ids(track) == {'soul_id': 'hyd-soul-1'}
+
     def test_native_id_without_provider_is_ignored(self):
         """Without a provider field we can't tell which source 'id' belongs to."""
         track = {'id': 'unknown', 'name': 'Hello'}
         assert extract_external_ids(track) == {}
 
+    def test_source_hint_disambiguates_when_track_has_no_provider(self):
+        """Spotify and iTunes raw API responses don't carry a provider /
+        source / _source field. The watchlist scanner passes the
+        configured primary source as a hint so the extractor can map
+        the track's native ``id`` field correctly."""
+        track = {'id': 'sp1', 'name': 'Hello'}  # No provider field
+        assert extract_external_ids(track, source_hint='spotify') == {'spotify_id': 'sp1'}
+        assert extract_external_ids(track, source_hint='itunes') == {'itunes_id': 'sp1'}
+        assert extract_external_ids(track, source_hint='deezer') == {'deezer_id': 'sp1'}
+
+    def test_source_hint_does_not_override_track_provider(self):
+        """If the track already carries an explicit provider, that wins
+        over the hint — defensive against the hint being wrong."""
+        track = {'_source': 'deezer', 'id': 'dz1', 'name': 'Hello'}
+        assert extract_external_ids(track, source_hint='spotify') == {'deezer_id': 'dz1'}
+
+    def test_source_hint_none_is_ignored(self):
+        track = {'id': 'unknown', 'name': 'Hello'}
+        assert extract_external_ids(track, source_hint=None) == {}
+
 
 class TestExtractExternalIdsMixedAndDefensive:
     def test_track_with_multiple_provider_specific_fields(self):