unicode in str MediaFile fields (#311, closes #15)

beets/mediafile.py

@@ -114,7 +114,12 @@ def _safe_cast(out_type, val):
         if val is None:
             return u''
         else:
-            return unicode(val)
+            if isinstance(val, str):
+                return val.decode('utf-8', errors='ignore')
+            elif isinstance(val, unicode):
+                return val
+            else:
+                return unicode(val)
 
     elif out_type == float:
         if val is None:

docs/changelog.rst

@@ -8,6 +8,8 @@ Changelog
   included with beets, making it easy to fetch **song lyrics**.
 * Fix a bug in the ``rewrite`` plugin that broke the use of multiple rules for
   a single field.
+* Fix a crash with non-ASCII characters in bytestring metadata fields (e.g.,
+  MusicBrainz IDs).
 
 
 1.0b12 (January 16, 2012)

test/test_mediafile.py

@@ -106,6 +106,11 @@ class InvalidValueToleranceTest(unittest.TestCase):
     def test_safe_cast_negative_string_to_float(self):
         self.assertAlmostEqual(_sc(float, '-1.234'), -1.234)
 
+    def test_safe_cast_special_chars_to_unicode(self):
+        us = _sc(unicode, 'caf\xc3\xa9')
+        self.assertTrue(isinstance(us, unicode))
+        self.assertTrue(us.startswith(u'caf'))
+
 class SafetyTest(unittest.TestCase):
     def _exccheck(self, fn, exc, data=''):
         fn = os.path.join(_common.RSRC, fn)