Merge pull request #135 from metabrainz/integration-fixes

Integration fixes

docs/lb_radio.rst

@@ -35,6 +35,7 @@ The LB Radio supports the following entities:
 #. **playlist**: Use a ListenBrainz playlist as a source of recordings. (mode also does not apply to playlists)
 #. **stats**: Use a ListenBrainz user's statistics as a source of recordings.
 #. **recs**: Use a ListenBrainz user's recommended recordings as a source of recordings.
+#. **country**: Select recordings from artists who are from the given country.
 
 Options
 -------
@@ -165,21 +166,21 @@ If LB-radio does not find your artist, you can specify an artist using an Artist
 
 ::
 
-  artist:8f6bd1e4-fbe1-4f50-aa9b-94c450ec0f11
+  artist:(8f6bd1e4-fbe1-4f50-aa9b-94c450ec0f11)
 
 LB-radio also supports MusicBrainz collections as sources:
 
 ::
 
-  collection:8be1a919-a386-45f3-8cc2-0d9249b02aa4
+  collection:(8be1a919-a386-45f3-8cc2-0d9249b02aa4)
 
 Will select random recordings from a MusicBrainz recording collection -- the modes wont have any affect on collections, since
 collections have no inherent ranking that could be used to select recordings according to mode. :(
 
 
 ::
 
-  playlist:8be1a919-a386-45f3-8cc2-0d9249b02aa4
+  playlist:(8be1a919-a386-45f3-8cc2-0d9249b02aa4)
 
 Will select random recordings from a ListenBrainz playlist -- the modes wont have any affect on collections, since
 plylists have no inherent ranking that could be used to select recordings according to mode. :(
@@ -198,6 +199,14 @@ Will select random recordings from the ListenBrainz user lucifer recordings stat
 
 Will select random recordings from the ListenBrainz user mr_monkey's recommended recordings that mr_monkey hasn't listened to.
 
+::
+
+  country:(Mali)
+
+Will select random recordings from artists who are from the given country. While this features generally represents music from
+that selected country, some artists leave their home country and don't perform music representative of their country, so 
+this element may not always be 100% on point. But it can still create some very interesting playlists!
+
 
 More complex examples
 ---------------------

tests/test_parser.py

@@ -8,19 +8,21 @@ class TestParser(unittest.TestCase):
 
     def test_basic_entities(self):
         pp = PromptParser()
-        r = pp.parse("artist:57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")
+        r = pp.parse("artist:(57baa3c6-ee43-4db3-9e6a-50bbc9792ee4)")
         assert r[0] == {"entity": "artist", "values": [UUID("57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")], "weight": 1, "opts": []}
 
-        r = pp.parse("artist:57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")
+        r = pp.parse("artist:(57baa3c6-ee43-4db3-9e6a-50bbc9792ee4)")
         assert r[0] == {"entity": "artist", "values": [UUID("57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")], "weight": 1, "opts": []}
 
         self.assertRaises(ParseError, pp.parse, "wrong:57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")
+        self.assertRaises(ParseError, pp.parse, "artist:57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")
 
         r = pp.parse("artist:(the knife)")
         assert r[0] == {"entity": "artist", "values": ["the knife"], "weight": 1, "opts": []}
 
         self.assertRaises(ParseError, pp.parse, "artist:u2:nosim")
         self.assertRaises(ParseError, pp.parse, "artists:u2:nosim")
+        self.assertRaises(ParseError, pp.parse, "country:andorra")
 
     def test_tags(self):
         pp = PromptParser()
@@ -52,6 +54,9 @@ def test_tags(self):
         r = pp.parse("tag:(モーニング娘。)")
         assert r[0] == {"entity": "tag", "values": ["モーニング娘。"], "weight": 1, "opts": []}
 
+        r = pp.parse("tag:(57baa3c6-ee43-4db3-9e6a-50bbc9792ee4)")
+        assert r[0] == {"entity": "tag", "values": ["57baa3c6-ee43-4db3-9e6a-50bbc9792ee4"], "weight": 1, "opts": []}
+
     def test_tag_errors(self):
         pp = PromptParser()
         self.assertRaises(ParseError, pp.parse, "t:(abstract rock blues):bork")
@@ -73,25 +78,25 @@ def test_shortcuts(self):
 
     def test_compound(self):
         pp = PromptParser()
-        r = pp.parse('artist:05319f96-e409-4199-b94f-3cabe7cc188a:2 tag:(downtempo):1 tag:(trip hop, abstract):2')
+        r = pp.parse('artist:(05319f96-e409-4199-b94f-3cabe7cc188a):2 tag:(downtempo):1 tag:(trip hop, abstract):2')
         assert r[0] == {"entity": "artist", "values": [UUID("05319f96-e409-4199-b94f-3cabe7cc188a")], "weight": 2, "opts": []}
         assert r[1] == {"entity": "tag", "values": ["downtempo"], "weight": 1, "opts": []}
         assert r[2] == {"entity": "tag", "values": ["trip hop", "abstract"], "weight": 2, "opts": []}
 
     def test_weights(self):
         pp = PromptParser()
-        r = pp.parse("artist:57baa3c6-ee43-4db3-9e6a-50bbc9792ee4:1 artist:f54ba4c6-12dd-4358-9136-c64ad89420c5:2")
+        r = pp.parse("artist:(57baa3c6-ee43-4db3-9e6a-50bbc9792ee4):1 artist:(f54ba4c6-12dd-4358-9136-c64ad89420c5):2")
         assert r[0] == {"entity": "artist", "values": [UUID("57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")], "weight": 1, "opts": []}
         assert r[1] == {"entity": "artist", "values": [UUID("f54ba4c6-12dd-4358-9136-c64ad89420c5")], "weight": 2, "opts": []}
 
         self.assertRaises(ParseError, pp.parse,
                           "artist:57baa3c6-ee43-4db3-9e6a-50bbc9792ee4:1 artist:f54ba4c6-12dd-4358-9136-c64ad89420c5:fussy")
         self.assertRaises(ParseError, pp.parse, "artist:57baa3c6-ee43-4db3-9e6a-50bbc9792ee4:1 artist:f54ba4c6-12dd-4358-9136-c64ad89420c5:.5")
 
-        r = pp.parse("artist:portishead::easy")
+        r = pp.parse("artist:(portishead)::easy")
         assert r[0] == {"entity": "artist", "values": ["portishead"], "weight": 1, "opts": ["easy"]}
 
-        r = pp.parse("artist:57baa3c6-ee43-4db3-9e6a-50bbc9792ee4::easy")
+        r = pp.parse("artist:(57baa3c6-ee43-4db3-9e6a-50bbc9792ee4)::easy")
         assert r[0] == {"entity": "artist", "values": [UUID("57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")], "weight": 1, "opts": ["easy"]}
 
     def test_opts(self):
@@ -112,13 +117,13 @@ def test_parens(self):
 
     def test_collection_playlist(self):
         pp = PromptParser()
-        r = pp.parse("collection:57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")
+        r = pp.parse("collection:(57baa3c6-ee43-4db3-9e6a-50bbc9792ee4)")
         assert r[0] == {"entity": "collection", "values": [UUID("57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")], "weight": 1, "opts": []}
 
-        r = pp.parse("playlist:57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")
+        r = pp.parse("playlist:(57baa3c6-ee43-4db3-9e6a-50bbc9792ee4)")
         assert r[0] == {"entity": "playlist", "values": [UUID("57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")], "weight": 1, "opts": []}
 
-        r = pp.parse("playlist:57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")
+        r = pp.parse("playlist:(57baa3c6-ee43-4db3-9e6a-50bbc9792ee4)")
         assert r[0] == {"entity": "playlist", "values": [UUID("57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")], "weight": 1, "opts": []}
 
     def test_stats(self):
@@ -156,3 +161,11 @@ def test_recs(self):
 
         r = pp.parse("recs:(rob zombie)")
         assert r[0] == {"entity": "recs", "values": ["rob zombie"], "weight": 1, "opts": []}
+
+    def test_country(self):
+        pp = PromptParser()
+        r = pp.parse("country:(57baa3c6-ee43-4db3-9e6a-50bbc9792ee4)")
+        assert r[0] == {"entity": "country", "values": [UUID("57baa3c6-ee43-4db3-9e6a-50bbc9792ee4")], "weight": 1, "opts": []}
+
+        r = pp.parse("country:(mali)")
+        assert r[0] == {"entity": "country", "values": ["mali"], "weight": 1, "opts": []}

tests/test_plist.py

@@ -0,0 +1,25 @@
+import unittest
+
+from troi.plist import plist
+
+
+class TestSplitter(unittest.TestCase):
+
+    def test_plist(self):
+        pl = plist([0,1,2,3,4,5,6,7,8,9])
+
+        assert pl[0:10] == [0]
+        assert pl[0:20] == [0,1]
+        assert pl[50:100] == [5,6,7,8,9]
+        assert pl.uslice(0, 10) == [0]
+        assert pl.uslice(0, 20) == [0,1]
+        assert pl.uslice(50, 100) == [5,6,7,8,9]
+
+        assert pl.dslice(0, 2) == [0,1]
+
+        assert pl.random_item(50, 100) in [5,6,7,8,9]
+
+    def test_plist_unique(self):
+        pl = plist([0,1,2,3,4,5,6,7,8,9])
+        rlist = pl.random_item(count=9)
+        assert len(rlist) == len(set(rlist))

troi/cli.py

@@ -108,7 +108,11 @@ def playlist(patch, quiet, save, token, upload, args, created_for, name, desc, m
             "This is a local patch and should be invoked via the specific troi function, rather than the playlist function.")
         return None
 
-    ret = patch.generate_playlist()
+    try:
+        ret = patch.generate_playlist()
+    except RuntimeError as err:
+        logger.error(err)
+        ret = 0
 
     user_feedback = patch.user_feedback()
     if len(user_feedback) > 0:

troi/content_resolver/artist_search.py

@@ -10,8 +10,21 @@
 from troi.content_resolver.model.recording import Recording, RecordingMetadata
 from troi.content_resolver.utils import select_recordings_on_popularity
 from troi.recording_search_service import RecordingSearchByArtistService
-from troi.splitter import plist
+from troi.plist import plist
 
+OVERHYPED_SIMILAR_ARTISTS = [
+    "b10bbbfc-cf9e-42e0-be17-e2c3e1d2600d",  # The Beatles
+    "83d91898-7763-47d7-b03b-b92132375c47",  # Pink Floyd
+    "a74b1b7f-71a5-4011-9441-d0b5e4122711",  # Radiohead
+    "8bfac288-ccc5-448d-9573-c33ea2aa5c30",  # Red Hot Chili Peppers
+    "9c9f1380-2516-4fc9-a3e6-f9f61941d090",  # Muse
+    "cc197bad-dc9c-440d-a5b5-d52ba2e14234",  # Coldplay
+    "65f4f0c5-ef9e-490c-aee3-909e7ae6b2ab",  # Metallica
+    "5b11f4ce-a62d-471e-81fc-a69a8278c7da",  # Nirvana
+    "f59c5520-5f46-4d2c-b2c4-822eabf53419",  # Linkin Park
+    "cc0b7089-c08d-4c10-b6b0-873582c17fd6",  # System of a Down
+    "ebfc1398-8d96-47e3-82c3-f782abcdb13d",  # Beach boys
+]
 
 class LocalRecordingSearchByArtistService(RecordingSearchByArtistService):
     '''
@@ -21,23 +34,50 @@ class LocalRecordingSearchByArtistService(RecordingSearchByArtistService):
     def __init__(self):
         RecordingSearchByArtistService.__init__(self)
 
-    def search(self, artist_mbids, begin_percent, end_percent, num_recordings):
+    def get_similar_artists(self, artist_mbid):
+        """ Fetch similar artists, given an artist_mbid. Returns a sored plist of artists. """
+
+        r = requests.post("https://labs.api.listenbrainz.org/similar-artists/json",
+                          json=[{
+                              'artist_mbid':
+                              artist_mbid,
+                              'algorithm':
+                              "session_based_days_7500_session_300_contribution_5_threshold_10_limit_100_filter_True_skip_30"
+                          }])
+        if r.status_code != 200:
+            raise RuntimeError(f"Cannot fetch similar artists: {r.status_code} ({r.text})")
+
+        try:
+            artists = r.json()[3]["data"]
+        except IndexError:
+            return []
+
+        # Knock down super hyped artists
+        for artist in artists:
+            if artist["artist_mbid"] in OVERHYPED_SIMILAR_ARTISTS:
+                artist["score"] /= 3  # Chop!
+
+        return plist(sorted(artists, key=lambda a: a["score"], reverse=True))
+
+    def search(self, mode, artist_mbid, pop_begin, pop_end, max_recordings_per_artist, max_similar_artists):
+
         """
         Perform an artist search. Parameters:
 
-        tags - a list of artist_mbids for which to search recordings
-        begin_percent - if many recordings match the above parameters, return only
-                        recordings that have a minimum popularity percent score
-                        of begin_percent.
-        end_percent - if many recordings match the above parameters, return only
-                      recordings that have a maximum popularity percent score
-                      of end_percent.
-        num_recordings - ideally return these many recordings
-
-        If only few recordings match, the begin_percent and end_percent are
-        ignored.
+        mode: the mode used for this artist search
+        pop_begin: if many recordings match the above parameters, return only
+                       recordings that have a minimum popularity percent score
+                       of pop_begin.
+        pop_end: if many recordings match the above parameters, return only
+                     recordings that have a maximum popularity percent score
+                     of pop_end.
+        max_recordings_per_artist: The number of recordings to collect for each artist.
+        max_similar_artists: The maximum number of similar artists to select.
+
+        If only few recordings match, the pop_begin and pop_end are ignored.
         """
 
+        similar_artists = self.get_similar_artists(artist_mbid)
         query = """SELECT popularity
                         , recording_mbid
                         , artist_mbid
@@ -50,8 +90,8 @@ def search(self, artist_mbids, begin_percent, end_percent, num_recordings):
                  ORDER BY artist_mbid
                         , popularity"""
 
-        placeholders = ",".join(("?", ) * len(artist_mbids))
-        cursor = db.execute_sql(query % placeholders, params=tuple(artist_mbids))
+        placeholders = ",".join(("?", ) * len(similar_artists))
+        cursor = db.execute_sql(query % placeholders, params=tuple(similar_artists))
 
         artists = defaultdict(list)
         for rec in cursor.fetchall():
@@ -64,6 +104,6 @@ def search(self, artist_mbids, begin_percent, end_percent, num_recordings):
             })
 
         for artist in artists:
-            artists[artist] = select_recordings_on_popularity(artists[artist], begin_percent, end_percent, num_recordings)
+            artists[artist] = select_recordings_on_popularity(artists[artist], pop_begin, pop_end, num_recordings)
 
         return artists