Improve youtube matching

Dockerfile

@@ -4,6 +4,9 @@ WORKDIR /bot
 COPY . .
 RUN touch settings.cfg
 
+RUN apk add --no-cache gcc libc-dev && \
+    rm -rf /var/cache/apk/*
+
 RUN pip install --no-cache-dir --requirement requirements.txt
 
 RUN addgroup -g 9999 lilia

commands.py

@@ -104,15 +104,8 @@ class Commands(object):
         channel_ids = self.config['Youtube'].get('channel_ids').split(',')
         yt = Youtube(api_key)
         try:
-            for channel_id in channel_ids:
+            result = yt.find_best_match(channel_ids, query)
-                results = yt.search(channel_id, query, playlists=True, limit=1)
+        except YoutubeError:
-                if results:
-                    break
-                results = yt.search(channel_id, query, playlists=False, limit=1)
-                if results:
-                    break
-            result = results.pop(0)
-        except (YoutubeError, IndexError):
             raise CommandError('couldn\'t find anything on Youtube')
         else:
             return result

requirements.txt

@@ -1,4 +1,5 @@
 discord.py
+fuzzywuzzy[speedup]
 google-api-python-client
 irc
 python-dateutil

services/youtube.py

@@ -1,3 +1,5 @@
+import fuzzywuzzy.fuzz
+import fuzzywuzzy.process
 import googleapiclient
 import googleapiclient.discovery
 
@@ -63,3 +65,25 @@ class Youtube(object):
             return self._search(channel_id, query, playlists, limit)
         except googleapiclient.errors.HttpError as e:
             raise YoutubeError('Failed to query Youtube API: {}'.format(e))
+
+    def find_best_match(self, channel_ids, query):
+        results = []
+        for channel_id in channel_ids:
+            try:
+                results.extend(self._search(channel_id, query, playlists=True, limit=1))
+                results.extend(self._search(channel_id, query, playlists=False, limit=1))
+            except googleapiclient.errors.HttpError as e:
+                raise YoutubeError('Failed to query Youtube API: {}'.format(e))
+        if not results:
+            return None
+        tokens = [t for t in query.split('|') if not t.strip().startswith('-')] or ['']
+        matches = []
+        for token in tokens:
+            titles = {i: r['title'] for i, r in enumerate(results)}
+            descriptions = {i: r['description'] for i, r in enumerate(results)}
+            matches.append(fuzzywuzzy.process.extractOne(token, titles,
+                scorer=fuzzywuzzy.fuzz.token_sort_ratio))
+            matches.append(fuzzywuzzy.process.extractOne(token, descriptions,
+                scorer=fuzzywuzzy.fuzz.token_sort_ratio))
+        _, _, i = sorted(matches, key=lambda m: m[1], reverse=True)[0]
+        return results[i]