diff --git a/pypyodbc.py b/pypyodbc.py
index 7d359cc..1e0ad24 100644
--- a/pypyodbc.py
+++ b/pypyodbc.py
@@ -26,7 +26,7 @@
 paramstyle = 'qmark'
 threadsafety = 1
 version = '1.3.0'
-lowercase=True
+lowercase = False
 
 DEBUG = 0
 # Comment out all "if DEBUG:" statements like below for production
@@ -489,15 +489,26 @@ def UCS_dec(buffer):
 
 # This is the common case on Linux, which uses wide Python build together with
 # the default unixODBC without the "-DSQL_WCHART_CONVERT" CFLAGS.
-if sys.platform not in ('win32','cli'):
-    if UNICODE_SIZE >= SQLWCHAR_SIZE:
+if sys.platform not in ('win32','cli') and UNICODE_SIZE != SQLWCHAR_SIZE:
+    if UNICODE_SIZE > SQLWCHAR_SIZE:
         # We can only use unicode buffer if the size of wchar_t (UNICODE_SIZE) is
         # the same as the size expected by the driver manager (SQLWCHAR_SIZE).
-        create_buffer_u = create_buffer
+        chars_to_bytes = lambda chars: chars * SQLWCHAR_SIZE
+        def create_buffer_u(init_or_size, *size_if_init):
+            if isinstance(init_or_size, basestring):
+                if size_if_init:
+                    return create_buffer(init_or_size, chars_to_bytes(size_if_init[0]))
+                else:
+                    return create_buffer(init_or_size)
+            else:
+                return create_buffer(chars_to_bytes(init_or_size))
         wchar_pointer = ctypes.c_char_p
 
         def UCS_buf(s):
-            return s.encode(odbc_encoding)
+            # c_char_p adds a single NUL-terminating byte because it assumes its argument is being
+            # passed to a function expecting a single NUL byte. But these functions actually take an
+            # array of two-byte integers, and so expect two NUL bytes' termination.
+            return (s + u'\x00').encode(odbc_encoding)
 
         from_buffer_u = UCS_dec
 
@@ -1017,8 +1028,6 @@ def AllocateEnv():
 
 def TupleRow(cursor):
     """Normal tuple with added attribute `cursor_description`, as in pyodbc.
-
-    This is the default.
     """
     class Row(tuple):
         cursor_description = cursor.description
@@ -1040,7 +1049,9 @@ def __getitem__(self, field):
 
 
 def NamedTupleRow(cursor):
-    """Named tuple to allow attribute lookup by name.
+    """Named tuple to allow attribute lookup by name, as in pyodbc.
+
+    This is the default.
 
     Requires py2.6 or above.
     """
@@ -1152,7 +1163,7 @@ def __init__(self, conx, row_type_callable=None):
         self.stmt_h = ctypes.c_void_p()
         self.connection = conx
         self.ansi = conx.ansi
-        self.row_type_callable = row_type_callable or TupleRow
+        self.row_type_callable = row_type_callable or NamedTupleRow
         self.statement = None
         self._last_param_types = None
         self._ParamBufferList = []
@@ -1766,11 +1777,12 @@ def _UpdateDesc(self):
         
         if len(ColDescr) > 0:
             self.description = ColDescr
-            # Create the row type before fetching.
-            self._row_type = self.row_type_callable(self)
         else:
             self.description = None
         self._CreateColBuf()
+
+        # Create the row type before fetching.
+        self._row_type = self.row_type_callable(self)
             
     
     def _NumOfRows(self):
@@ -2522,6 +2534,11 @@ def cursor(self, row_type_callable=None):
         # self._cursors.append(cur)
         return cur
 
+    def execute(self, sql, *args, **kwargs):
+        cur = self.cursor(row_type_callable=kwargs.pop('row_type_callable', None))
+        cur.execute(sql, *args, **kwargs)
+        return cur
+
     def update_db_special_info(self):
         for sql_type in (
             SQL_TYPE_TIMESTAMP,