- rework Oracle to no longer do its own unicode conversion; this has been observed

to be very slow.  this now has the effect of producing "conditional" unicode
conversion for the Oracle backend, as it still returns NVARCHAR etc. as unicode
[ticket:2911]
- add new "conditional" functionality to unicode processors; the C-level
function now uses PyUnicode_Check() as a fast alternative to the isinstance()
check in Python

1 files changed, 22 insertions, 1 deletions

diff --git a/lib/sqlalchemy/processors.py b/lib/sqlalchemy/processors.py
index 0abf063b3..d0f52e42b 100644
--- a/lib/sqlalchemy/processors.py
+++ b/lib/sqlalchemy/processors.py
@@ -15,6 +15,7 @@ They all share one common characteristic: None is passed through unchanged.
 import codecs
 import re
 import datetime
+from . import util
 
 
 def str_to_datetime_processor_factory(regexp, type_):
@@ -66,6 +67,21 @@ def py_fallback():
                 return decoder(value, errors)[0]
         return process
 
+    def to_conditional_unicode_processor_factory(encoding, errors=None):
+        decoder = codecs.getdecoder(encoding)
+
+        def process(value):
+            if value is None:
+                return None
+            elif isinstance(value, util.text_type):
+                return value
+            else:
+                # decoder returns a tuple: (value, len). Simply dropping the
+                # len part is safe: it is done that way in the normal
+                # 'xx'.decode(encoding) code path.
+                return decoder(value, errors)[0]
+        return process
+
     def to_decimal_processor_factory(target_class, scale):
         fstring = "%%.%df" % scale
 
@@ -113,12 +129,17 @@ try:
                                        str_to_date
 
     def to_unicode_processor_factory(encoding, errors=None):
-        # this is cumbersome but it would be even more so on the C side
         if errors is not None:
             return UnicodeResultProcessor(encoding, errors).process
         else:
             return UnicodeResultProcessor(encoding).process
 
+    def to_conditional_unicode_processor_factory(encoding, errors=None):
+        if errors is not None:
+            return UnicodeResultProcessor(encoding, errors).conditional_process
+        else:
+            return UnicodeResultProcessor(encoding).conditional_process
+
     def to_decimal_processor_factory(target_class, scale):
         # Note that the scale argument is not taken into account for integer
         # values in the C implementation while it is in the Python one.