Use _binary prefix for bytes/bytearray parameters

- Based on PyMySQL#106 but now disabled by default
- Can be enabled via 'binary_prefix' connection parameter
- Added unit tests to verify behaviour

fix falsely prefixing strings with _binary type identifier (PyMySQL#123)

MySQLdb/__init__.py

@@ -27,6 +27,7 @@
 paramstyle = "format"
 
 from _mysql import *
+from MySQLdb.compat import PY2
 from MySQLdb.constants import FIELD_TYPE
 from MySQLdb.times import Date, Time, Timestamp, \
     DateFromTicks, TimeFromTicks, TimestampFromTicks
@@ -72,8 +73,12 @@ def test_DBAPISet_set_equality_membership():
 def test_DBAPISet_set_inequality_membership():
     assert FIELD_TYPE.DATE != STRING
 
-def Binary(x):
-    return bytes(x)
+if PY2:
+    def Binary(x):
+        return bytearray(x)
+else:
+    def Binary(x):
+        return bytes(x)
 
 def Connect(*args, **kwargs):
     """Factory function for connections.Connection."""

MySQLdb/connections.py

@@ -137,6 +137,10 @@ class object, used to create cursors (keyword only)
             If True, autocommit is enabled.
             If None, autocommit isn't set and server default is used.
 
+        :param bool binary_prefix:
+            If set, the '_binary' prefix will be used for raw byte query
+            arguments (e.g. Binary). This is disabled by default.
+
         There are a number of undocumented, non-standard methods. See the
         documentation for the MySQL C API for some hints on what they do.
         """
@@ -174,6 +178,7 @@ class object, used to create cursors (keyword only)
 
         use_unicode = kwargs2.pop('use_unicode', use_unicode)
         sql_mode = kwargs2.pop('sql_mode', '')
+        binary_prefix = kwargs2.pop('binary_prefix', False)
 
         client_flag = kwargs.get('client_flag', 0)
         client_version = tuple([ numeric_part(n) for n in _mysql.get_client_info().split('.')[:2] ])
@@ -197,7 +202,7 @@ class object, used to create cursors (keyword only)
 
         db = proxy(self)
         def _get_string_literal():
-            # Note: string_literal() is called for bytes object on Python 3.
+            # Note: string_literal() is called for bytes object on Python 3 (via bytes_literal)
             def string_literal(obj, dummy=None):
                 return db.string_literal(obj)
             return string_literal
@@ -206,20 +211,26 @@ def _get_unicode_literal():
             if PY2:
                 # unicode_literal is called for only unicode object.
                 def unicode_literal(u, dummy=None):
-                    return db.literal(u.encode(unicode_literal.charset))
+                    return db.string_literal(u.encode(unicode_literal.charset))
             else:
                 # unicode_literal() is called for arbitrary object.
                 def unicode_literal(u, dummy=None):
-                    return db.literal(str(u).encode(unicode_literal.charset))
+                    return db.string_literal(str(u).encode(unicode_literal.charset))
             return unicode_literal
 
+        def _get_bytes_literal():
+            def bytes_literal(obj, dummy=None):
+                return b'_binary' + db.string_literal(obj)
+            return bytes_literal
+
         def _get_string_decoder():
             def string_decoder(s):
                 return s.decode(string_decoder.charset)
             return string_decoder
 
         string_literal = _get_string_literal()
         self.unicode_literal = unicode_literal = _get_unicode_literal()
+        bytes_literal = _get_bytes_literal()
         self.string_decoder = string_decoder = _get_string_decoder()
         if not charset:
             charset = self.character_set_name()
@@ -234,7 +245,12 @@ def string_decoder(s):
             self.converter[FIELD_TYPE.VARCHAR].append((None, string_decoder))
             self.converter[FIELD_TYPE.BLOB].append((None, string_decoder))
 
-        self.encoders[bytes] = string_literal
+        if binary_prefix:
+            self.encoders[bytes] = string_literal if PY2 else bytes_literal
+            self.encoders[bytearray] = bytes_literal
+        else:
+            self.encoders[bytes] = string_literal
+
         self.encoders[unicode] = unicode_literal
         self._transactional = self.server_capabilities & CLIENT.TRANSACTIONS
         if self._transactional:

tests/test_MySQLdb_capabilities.py

@@ -2,10 +2,12 @@
 # -*- coding: utf-8 -*-
 import capabilities
 from datetime import timedelta
+from contextlib import closing
 import unittest
 import MySQLdb
 from MySQLdb.compat import unicode
 from MySQLdb import cursors
+from configdb import connection_factory
 import warnings
 
 
@@ -180,6 +182,23 @@ def test_warning_propagation(self):
                 finally:
                     c.close()
 
+    def test_binary_prefix(self):
+        # verify prefix behaviour when enabled, disabled and for default (disabled)
+        for binary_prefix in (True, False, None):
+            kwargs = self.connect_kwargs.copy()
+            # needs to be set to can guarantee CHARSET response for normal strings
+            kwargs['charset'] = 'utf8'
+            if binary_prefix != None:
+                kwargs['binary_prefix'] = binary_prefix
+
+            with closing(connection_factory(**kwargs)) as conn:
+                with closing(conn.cursor()) as c:
+                    c.execute('SELECT CHARSET(%s)', (MySQLdb.Binary(b'raw bytes'),))
+                    self.assertEqual(c.fetchall()[0][0], 'binary' if binary_prefix else 'utf8')
+                    # normal strings should not get prefix
+                    c.execute('SELECT CHARSET(%s)', ('str',))
+                    self.assertEqual(c.fetchall()[0][0], 'utf8')
+
 
 if __name__ == '__main__':
     if test_MySQLdb.leak_test: