- A new pickle protocol (protocol 3) is added with explicit support

  for bytes.  This is the default protocol.  It intentionally cannot
  be unpickled by Python 2.x.

- When a pickle	written	by Python 2.x contains an (8-bit) str
  instance, this is now decoded to a (Unicode) str instance.  The
  encoding used to do this defaults to ASCII, but can be overridden
  via two new keyword arguments to the Unpickler class.  Previously
  this would create bytes instances, which is usually wrong: str
  instances are often used to pickle attribute names etc., and text is
  more common than binary data anyway.

6 files changed, 165 insertions, 55 deletions

diff --git a/Lib/pickle.py b/Lib/pickle.py
index e3c112f..ea8d8b5 100644
--- a/Lib/pickle.py
+++ b/Lib/pickle.py
@@ -42,19 +42,22 @@ __all__ = ["PickleError", "PicklingError", "UnpicklingError", "Pickler",
 bytes_types = (bytes, bytearray, memoryview)
 
 # These are purely informational; no code uses these.
-format_version = "2.0"                  # File format version we write
+format_version = "3.0"                  # File format version we write
 compatible_formats = ["1.0",            # Original protocol 0
                       "1.1",            # Protocol 0 with INST added
                       "1.2",            # Original protocol 1
                       "1.3",            # Protocol 1 with BINFLOAT added
                       "2.0",            # Protocol 2
+                      "3.0",            # Protocol 3
                       ]                 # Old format versions we can read
 
 # This is the highest protocol number we know how to read.
-HIGHEST_PROTOCOL = 2
+HIGHEST_PROTOCOL = 3
 
 # The protocol we write by default.  May be less than HIGHEST_PROTOCOL.
-DEFAULT_PROTOCOL = 2
+# We intentionally write a protocol that Python 2.x cannot read;
+# there are too many issues with that.
+DEFAULT_PROTOCOL = 3
 
 # Why use struct.pack() for pickling but marshal.loads() for
 # unpickling?  struct.pack() is 40% faster than marshal.dumps(), but
@@ -161,6 +164,10 @@ LONG4          = b'\x8b'  # push really big long
 
 _tuplesize2code = [EMPTY_TUPLE, TUPLE1, TUPLE2, TUPLE3]
 
+# Protocol 3 (Python 3.x)
+
+BINBYTES       = b'B'   # push bytes; counted binary string argument
+SHORT_BINBYTES = b'C'   #  "     "   ;    "      "       "      " < 256 bytes
 
 __all__.extend([x for x in dir() if re.match("[A-Z][A-Z0-9_]+$",x)])
 
@@ -494,20 +501,19 @@ class Pickler:
             self.write(FLOAT + repr(obj).encode("ascii") + b'\n')
     dispatch[float] = save_float
 
-    def save_string(self, obj, pack=struct.pack):
-        if self.bin:
-            n = len(obj)
-            if n < 256:
-                self.write(SHORT_BINSTRING + bytes([n]) + bytes(obj))
-            else:
-                self.write(BINSTRING + pack("<i", n) + bytes(obj))
+    def save_bytes(self, obj, pack=struct.pack):
+        if self.proto < 3:
+            self.save_reduce(bytes, (list(obj),))
+            return
+        n = len(obj)
+        if n < 256:
+            self.write(SHORT_BINBYTES + bytes([n]) + bytes(obj))
         else:
-            # Strip leading 'b' due to repr() of bytes() returning b'...'
-            self.write(STRING + repr(obj).lstrip("b").encode("ascii") + b'\n')
+            self.write(BINBYTES + pack("<i", n) + bytes(obj))
         self.memoize(obj)
-    dispatch[bytes] = save_string
+    dispatch[bytes] = save_bytes
 
-    def save_unicode(self, obj, pack=struct.pack):
+    def save_str(self, obj, pack=struct.pack):
         if self.bin:
             encoded = obj.encode('utf-8')
             n = len(encoded)
@@ -518,7 +524,7 @@ class Pickler:
             self.write(UNICODE + bytes(obj.encode('raw-unicode-escape')) +
                        b'\n')
         self.memoize(obj)
-    dispatch[str] = save_unicode
+    dispatch[str] = save_str
 
     def save_tuple(self, obj):
         write = self.write
@@ -775,7 +781,7 @@ def whichmodule(func, funcname):
 
 class Unpickler:
 
-    def __init__(self, file):
+    def __init__(self, file, *, encoding="ASCII", errors="strict"):
         """This takes a binary file for reading a pickle data stream.
 
         The protocol version of the pickle is detected automatically, so no
@@ -787,10 +793,16 @@ class Unpickler:
         Thus file-like object can be a binary file object opened for
         reading, a BytesIO object, or any other custom object that
         meets this interface.
+
+        Optional keyword arguments are encoding and errors, which are
+        used to decode 8-bit string instances pickled by Python 2.x.
+        These default to 'ASCII' and 'strict', respectively.
         """
         self.readline = file.readline
         self.read = file.read
         self.memo = {}
+        self.encoding = encoding
+        self.errors = errors
 
     def load(self):
         """Read a pickled object representation from the open file.
@@ -831,7 +843,7 @@ class Unpickler:
 
     def load_proto(self):
         proto = ord(self.read(1))
-        if not 0 <= proto <= 2:
+        if not 0 <= proto <= HIGHEST_PROTOCOL:
             raise ValueError("unsupported pickle protocol: %d" % proto)
     dispatch[PROTO[0]] = load_proto
 
@@ -924,9 +936,16 @@ class Unpickler:
 
     def load_binstring(self):
         len = mloads(b'i' + self.read(4))
-        self.append(self.read(len))
+        data = self.read(len)
+        value = str(data, self.encoding, self.errors)
+        self.append(value)
     dispatch[BINSTRING[0]] = load_binstring
 
+    def load_binbytes(self):
+        len = mloads(b'i' + self.read(4))
+        self.append(self.read(len))
+    dispatch[BINBYTES[0]] = load_binbytes
+
     def load_unicode(self):
         self.append(str(self.readline()[:-1], 'raw-unicode-escape'))
     dispatch[UNICODE[0]] = load_unicode
@@ -938,9 +957,16 @@ class Unpickler:
 
     def load_short_binstring(self):
         len = ord(self.read(1))
-        self.append(bytes(self.read(len)))
+        data = bytes(self.read(len))
+        value = str(data, self.encoding, self.errors)
+        self.append(value)
     dispatch[SHORT_BINSTRING[0]] = load_short_binstring
 
+    def load_short_binbytes(self):
+        len = ord(self.read(1))
+        self.append(bytes(self.read(len)))
+    dispatch[SHORT_BINBYTES[0]] = load_short_binbytes
+
     def load_tuple(self):
         k = self.marker()
         self.stack[k:] = [tuple(self.stack[k+1:])]
diff --git a/Lib/pickletools.py b/Lib/pickletools.py
index ca09c03..37dad9b 100644
--- a/Lib/pickletools.py
+++ b/Lib/pickletools.py
@@ -746,6 +746,11 @@ pyfloat = StackObject(
               doc="A Python float object.")
 
 pystring = StackObject(
+               name='string',
+               obtype=bytes,
+               doc="A Python (8-bit) string object.")
+
+pybytes = StackObject(
                name='bytes',
                obtype=bytes,
                doc="A Python bytes object.")
@@ -753,7 +758,7 @@ pystring = StackObject(
 pyunicode = StackObject(
                 name='str',
                 obtype=str,
-                doc="A Python string object.")
+                doc="A Python (Unicode) string object.")
 
 pynone = StackObject(
              name="None",
@@ -868,7 +873,7 @@ class OpcodeInfo(object):
             assert isinstance(x, StackObject)
         self.stack_after = stack_after
 
-        assert isinstance(proto, int) and 0 <= proto <= 2
+        assert isinstance(proto, int) and 0 <= proto <= 3
         self.proto = proto
 
         assert isinstance(doc, str)
@@ -995,7 +1000,9 @@ opcodes = [
 
       The argument is a repr-style string, with bracketing quote characters,
       and perhaps embedded escapes.  The argument extends until the next
-      newline character.
+      newline character.  (Actually, they are decoded into a str instance
+      using the encoding given to the Unpickler constructor. or the default,
+      'ASCII'.)
       """),
 
     I(name='BINSTRING',
@@ -1008,7 +1015,9 @@ opcodes = [
 
       There are two arguments:  the first is a 4-byte little-endian signed int
       giving the number of bytes in the string, and the second is that many
-      bytes, which are taken literally as the string content.
+      bytes, which are taken literally as the string content.  (Actually,
+      they are decoded into a str instance using the encoding given to the
+      Unpickler constructor. or the default, 'ASCII'.)
       """),
 
     I(name='SHORT_BINSTRING',
@@ -1021,6 +1030,36 @@ opcodes = [
 
       There are two arguments:  the first is a 1-byte unsigned int giving
       the number of bytes in the string, and the second is that many bytes,
+      which are taken literally as the string content.  (Actually, they
+      are decoded into a str instance using the encoding given to the
+      Unpickler constructor. or the default, 'ASCII'.)
+      """),
+
+    # Bytes (protocol 3 only; older protocols don't support bytes at all)
+
+    I(name='BINBYTES',
+      code='B',
+      arg=string4,
+      stack_before=[],
+      stack_after=[pybytes],
+      proto=3,
+      doc="""Push a Python bytes object.
+
+      There are two arguments:  the first is a 4-byte little-endian signed int
+      giving the number of bytes in the string, and the second is that many
+      bytes, which are taken literally as the bytes content.
+      """),
+
+    I(name='SHORT_BINBYTES',
+      code='C',
+      arg=string1,
+      stack_before=[],
+      stack_after=[pybytes],
+      proto=1,
+      doc="""Push a Python string object.
+
+      There are two arguments:  the first is a 1-byte unsigned int giving
+      the number of bytes in the string, and the second is that many bytes,
       which are taken literally as the string content.
       """),
 
@@ -2006,9 +2045,9 @@ class _Example:
 
 _dis_test = r"""
 >>> import pickle
->>> x = [1, 2, (3, 4), {bytes(b'abc'): "def"}]
->>> pkl = pickle.dumps(x, 0)
->>> dis(pkl)
+>>> x = [1, 2, (3, 4), {b'abc': "def"}]
+>>> pkl0 = pickle.dumps(x, 0)
+>>> dis(pkl0)
     0: (    MARK
     1: l        LIST       (MARK at 0)
     2: p    PUT        0
@@ -2025,19 +2064,32 @@ _dis_test = r"""
    25: (    MARK
    26: d        DICT       (MARK at 25)
    27: p    PUT        2
-   30: S    STRING     'abc'
-   37: p    PUT        3
-   40: V    UNICODE    'def'
-   45: p    PUT        4
-   48: s    SETITEM
-   49: a    APPEND
-   50: .    STOP
+   30: c    GLOBAL     'builtins bytes'
+   46: p    PUT        3
+   49: (    MARK
+   50: (        MARK
+   51: l            LIST       (MARK at 50)
+   52: p        PUT        4
+   55: L        LONG       97
+   59: a        APPEND
+   60: L        LONG       98
+   64: a        APPEND
+   65: L        LONG       99
+   69: a        APPEND
+   70: t        TUPLE      (MARK at 49)
+   71: p    PUT        5
+   74: R    REDUCE
+   75: V    UNICODE    'def'
+   80: p    PUT        6
+   83: s    SETITEM
+   84: a    APPEND
+   85: .    STOP
 highest protocol among opcodes = 0
 
 Try again with a "binary" pickle.
 
->>> pkl = pickle.dumps(x, 1)
->>> dis(pkl)
+>>> pkl1 = pickle.dumps(x, 1)
+>>> dis(pkl1)
     0: ]    EMPTY_LIST
     1: q    BINPUT     0
     3: (    MARK
@@ -2050,13 +2102,24 @@ Try again with a "binary" pickle.
    14: q        BINPUT     1
    16: }        EMPTY_DICT
    17: q        BINPUT     2
-   19: U        SHORT_BINSTRING 'abc'
-   24: q        BINPUT     3
-   26: X        BINUNICODE 'def'
-   34: q        BINPUT     4
-   36: s        SETITEM
-   37: e        APPENDS    (MARK at 3)
-   38: .    STOP
+   19: c        GLOBAL     'builtins bytes'
+   35: q        BINPUT     3
+   37: (        MARK
+   38: ]            EMPTY_LIST
+   39: q            BINPUT     4
+   41: (            MARK
+   42: K                BININT1    97
+   44: K                BININT1    98
+   46: K                BININT1    99
+   48: e                APPENDS    (MARK at 41)
+   49: t            TUPLE      (MARK at 37)
+   50: q        BINPUT     5
+   52: R        REDUCE
+   53: X        BINUNICODE 'def'
+   61: q        BINPUT     6
+   63: s        SETITEM
+   64: e        APPENDS    (MARK at 3)
+   65: .    STOP
 highest protocol among opcodes = 1
 
 Exercise the INST/OBJ/BUILD family.
diff --git a/Lib/test/pickletester.py b/Lib/test/pickletester.py
index 5ac2bdc..0230f3c 100644
--- a/Lib/test/pickletester.py
+++ b/Lib/test/pickletester.py
@@ -490,6 +490,12 @@ class AbstractPickleTests(unittest.TestCase):
                 u2 = self.loads(p)
                 self.assertEqual(u2, u)
 
+    def test_bytes(self):
+        for proto in protocols:
+            for u in b'', b'xyz', b'xyz'*100:
+                p = self.dumps(u)
+                self.assertEqual(self.loads(p), u)
+
     def test_ints(self):
         import sys
         for proto in protocols:
@@ -532,8 +538,8 @@ class AbstractPickleTests(unittest.TestCase):
 
     @run_with_locale('LC_ALL', 'de_DE', 'fr_FR')
     def test_float_format(self):
-        # make sure that floats are formatted locale independent
-        self.assertEqual(self.dumps(1.2)[0:3], b'F1.')
+        # make sure that floats are formatted locale independent with proto 0
+        self.assertEqual(self.dumps(1.2, 0)[0:3], b'F1.')
 
     def test_reduce(self):
         pass
@@ -624,6 +630,12 @@ class AbstractPickleTests(unittest.TestCase):
                            (2, 2): pickle.TUPLE2,
                            (2, 3): pickle.TUPLE3,
                            (2, 4): pickle.TUPLE,
+
+                           (3, 0): pickle.EMPTY_TUPLE,
+                           (3, 1): pickle.TUPLE1,
+                           (3, 2): pickle.TUPLE2,
+                           (3, 3): pickle.TUPLE3,
+                           (3, 4): pickle.TUPLE,
                           }
         a = ()
         b = (1,)
@@ -643,14 +655,17 @@ class AbstractPickleTests(unittest.TestCase):
         expected_opcode = {(0, None): pickle.NONE,
                            (1, None): pickle.NONE,
                            (2, None): pickle.NONE,
+                           (3, None): pickle.NONE,
 
                            (0, True): pickle.INT,
                            (1, True): pickle.INT,
                            (2, True): pickle.NEWTRUE,
+                           (3, True): pickle.NEWTRUE,
 
                            (0, False): pickle.INT,
                            (1, False): pickle.INT,
                            (2, False): pickle.NEWFALSE,
+                           (3, False): pickle.NEWFALSE,
                           }
         for proto in protocols:
             for x in None, False, True:
@@ -955,7 +970,7 @@ class AbstractPickleModuleTests(unittest.TestCase):
 
     def test_highest_protocol(self):
         # Of course this needs to be changed when HIGHEST_PROTOCOL changes.
-        self.assertEqual(self.module.HIGHEST_PROTOCOL, 2)
+        self.assertEqual(self.module.HIGHEST_PROTOCOL, 3)
 
     def test_callapi(self):
         from io import BytesIO
diff --git a/Lib/test/test_pickle.py b/Lib/test/test_pickle.py
index aa09a6a..67d83c7c 100644
--- a/Lib/test/test_pickle.py
+++ b/Lib/test/test_pickle.py
@@ -12,23 +12,19 @@ class PickleTests(AbstractPickleTests, AbstractPickleModuleTests):
     module = pickle
     error = KeyError
 
-    def dumps(self, arg, proto=0, fast=0):
-        # Ignore fast
+    def dumps(self, arg, proto=None):
         return pickle.dumps(arg, proto)
 
     def loads(self, buf):
-        # Ignore fast
         return pickle.loads(buf)
 
 class PicklerTests(AbstractPickleTests):
 
     error = KeyError
 
-    def dumps(self, arg, proto=0, fast=0):
+    def dumps(self, arg, proto=None):
         f = io.BytesIO()
         p = pickle.Pickler(f, proto)
-        if fast:
-            p.fast = fast
         p.dump(arg)
         f.seek(0)
         return bytes(f.read())
@@ -40,14 +36,12 @@ class PicklerTests(AbstractPickleTests):
 
 class PersPicklerTests(AbstractPersistentPicklerTests):
 
-    def dumps(self, arg, proto=0, fast=0):
+    def dumps(self, arg, proto=None):
         class PersPickler(pickle.Pickler):
             def persistent_id(subself, obj):
                 return self.persistent_id(obj)
         f = io.BytesIO()
         p = PersPickler(f, proto)
-        if fast:
-            p.fast = fast
         p.dump(arg)
         f.seek(0)
         return f.read()
diff --git a/Lib/test/test_pickletools.py b/Lib/test/test_pickletools.py
index 3e5b35a..932dcd1 100644
--- a/Lib/test/test_pickletools.py
+++ b/Lib/test/test_pickletools.py
@@ -6,7 +6,7 @@ from test.pickletester import AbstractPickleModuleTests
 
 class OptimizedPickleTests(AbstractPickleTests, AbstractPickleModuleTests):
 
-    def dumps(self, arg, proto=0, fast=0):
+    def dumps(self, arg, proto=None):
         return pickletools.optimize(pickle.dumps(arg, proto))
 
     def loads(self, buf):
diff --git a/Misc/NEWS b/Misc/NEWS
index 65643bf..db686cd 100644
--- a/Misc/NEWS
+++ b/Misc/NEWS
@@ -23,6 +23,18 @@ Extension Modules
 Library
 -------
 
+- A new pickle protocol (protocol 3) is added with explicit support
+  for bytes.  This is the default protocol.  It intentionally cannot
+  be unpickled by Python 2.x.
+
+- When a pickle written by Python 2.x contains an (8-bit) str
+  instance, this is now decoded to a (Unicode) str instance.  The
+  encoding used to do this defaults to ASCII, but can be overridden
+  via two new keyword arguments to the Unpickler class.  Previously
+  this would create bytes instances, which is usually wrong: str
+  instances are often used to pickle attribute names etc., and text is
+  more common than binary data anyway.
+
 - Default to ASCII as the locale.getpreferredencoding, if the POSIX
   system doesn't support CODESET and LANG isn't set or doesn't
   allow deduction of an encoding.