Fix pandas-dev#777: Handle zero columns when converting to unicode

Author: shashank88

arctic/serialization/numpy_records.py

@@ -301,8 +301,8 @@ def deserialize(self, item, force_bytes_to_unicode=False):
                 if type(df.index[0]) == bytes:
                     df.index = df.index.astype('unicode')
 
-            if type(df.columns[0]) == bytes:
-                df.columns = df.index.astype('unicode')
+            if not df.columns.empty and type(df.columns[0]) == bytes:
+                df.columns = df.columns.astype('unicode')
 
         return df