Merge tag 'refs/tags/sync-piper' into sync-stage

# Conflicts:
#	src/google/protobuf/compiler/objectivec/helpers.cc
#	src/google/protobuf/compiler/objectivec/text_format_decode_data.cc
#	src/google/protobuf/util/message_differencer.h

java/core/src/main/java/com/google/protobuf/CanIgnoreReturnValue.java

@@ -31,7 +31,6 @@
 package com.google.protobuf;
 
 import static java.lang.annotation.ElementType.METHOD;
-import static java.lang.annotation.ElementType.TYPE;
 import static java.lang.annotation.RetentionPolicy.CLASS;
 
 import java.lang.annotation.Documented;
@@ -45,6 +44,6 @@ import java.lang.annotation.Target;
  * annotated with {@code @CheckReturnValue} to exempt specific methods from the default.
  */
 @Documented
-@Target({METHOD, TYPE})
+@Target(METHOD) // TODO(kak): consider adding CONSTRUCTOR later if necessary
 @Retention(CLASS)
 @interface CanIgnoreReturnValue {}

java/core/src/test/proto/com/google/protobuf/proto2_message.proto

@@ -36,7 +36,6 @@ package protobuf.experimental;
 option java_package = "com.google.protobuf.testing";
 option java_outer_classname = "Proto2Testing";
 
-
 message Proto2SpecialFieldName {
   optional double regular_name = 1;
   optional int32 cached_size = 2;

java/core/src/test/proto/com/google/protobuf/proto2_message_lite.proto

@@ -36,7 +36,6 @@ package protobuf.experimental.lite;
 option java_package = "com.google.protobuf.testing";
 option java_outer_classname = "Proto2TestingLite";
 
-
 message Proto2MessageLite {
 
   enum TestEnum {

protobuf.bzl

@@ -40,13 +40,14 @@ def _SourceDir(ctx):
 def _ObjcBase(srcs):
     return [
         "".join([token.capitalize() for token in src[:-len(".proto")].split("_")])
-    for src in srcs]
+        for src in srcs
+    ]
 
 def _ObjcHdrs(srcs):
-    return[src + ".pbobjc.h" for src in _ObjcBase(srcs)]
+    return [src + ".pbobjc.h" for src in _ObjcBase(srcs)]
 
 def _ObjcSrcs(srcs):
-    return[src + ".pbobjc.m" for src in _ObjcBase(srcs)]
+    return [src + ".pbobjc.m" for src in _ObjcBase(srcs)]
 
 def _ObjcOuts(srcs, out_type):
     if out_type == "hdrs":
@@ -67,7 +68,8 @@ def _RubyOuts(srcs):
 def _CsharpOuts(srcs):
     return [
         "".join([token.capitalize() for token in src[:-len(".proto")].split("_")]) + ".cs"
-    for src in srcs]
+        for src in srcs
+    ]
 
 ProtoGenInfo = provider(
     fields = ["srcs", "import_flags", "deps"],
@@ -96,7 +98,7 @@ def _proto_gen_impl(ctx):
 
     if ctx.attr.includes:
         for include in ctx.attr.includes:
-            import_flags += ["-I"+_GetPath(ctx,include)]
+            import_flags += ["-I" + _GetPath(ctx, include)]
 
     import_flags = depset(direct = import_flags)
 
@@ -149,6 +151,7 @@ def _proto_gen_impl(ctx):
                 outs.extend(_PyOuts([src.basename], use_grpc_plugin = use_grpc_plugin))
             elif lang == "ruby":
                 outs.extend(_RubyOuts([src.basename]))
+
             # Otherwise, rely on user-supplied outs.
             args += [("--%s_out=" + path_tpl) % (lang, gen_dir)]
 
@@ -261,7 +264,7 @@ _proto_gen = rule(
         "langs": attr.string_list(),
         "outs": attr.string_list(),
         "out_type": attr.string(
-            default = "all"
+            default = "all",
         ),
     },
     output_to_genfiles = True,
@@ -641,7 +644,7 @@ def _source_proto_library(
 
     native.filegroup(
         name = name,
-        srcs = [":%s_genproto"%name],
+        srcs = [":%s_genproto" % name],
         testonly = testonly,
         visibility = visibility,
         **kwargs

python/BUILD.bazel

@@ -275,6 +275,13 @@ py_test(
     deps = [":python_test_lib"],
 )
 
+py_test(
+    name = "field_mask_test",
+    srcs = ["google/protobuf/internal/field_mask_test.py"],
+    imports = ["."],
+    deps = [":python_test_lib"],
+)
+
 py_test(
     name = "generator_test",
     srcs = ["google/protobuf/internal/generator_test.py"],

python/google/protobuf/internal/field_mask.py

@@ -0,0 +1,333 @@
+# Protocol Buffers - Google's data interchange format
+# Copyright 2008 Google Inc.  All rights reserved.
+# https://developers.google.com/protocol-buffers/
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are
+# met:
+#
+#     * Redistributions of source code must retain the above copyright
+# notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above
+# copyright notice, this list of conditions and the following disclaimer
+# in the documentation and/or other materials provided with the
+# distribution.
+#     * Neither the name of Google Inc. nor the names of its
+# contributors may be used to endorse or promote products derived from
+# this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+"""Contains FieldMask class."""
+
+from google.protobuf.descriptor import FieldDescriptor
+
+
+class FieldMask(object):
+  """Class for FieldMask message type."""
+
+  __slots__ = ()
+
+  def ToJsonString(self):
+    """Converts FieldMask to string according to proto3 JSON spec."""
+    camelcase_paths = []
+    for path in self.paths:
+      camelcase_paths.append(_SnakeCaseToCamelCase(path))
+    return ','.join(camelcase_paths)
+
+  def FromJsonString(self, value):
+    """Converts string to FieldMask according to proto3 JSON spec."""
+    if not isinstance(value, str):
+      raise ValueError('FieldMask JSON value not a string: {!r}'.format(value))
+    self.Clear()
+    if value:
+      for path in value.split(','):
+        self.paths.append(_CamelCaseToSnakeCase(path))
+
+  def IsValidForDescriptor(self, message_descriptor):
+    """Checks whether the FieldMask is valid for Message Descriptor."""
+    for path in self.paths:
+      if not _IsValidPath(message_descriptor, path):
+        return False
+    return True
+
+  def AllFieldsFromDescriptor(self, message_descriptor):
+    """Gets all direct fields of Message Descriptor to FieldMask."""
+    self.Clear()
+    for field in message_descriptor.fields:
+      self.paths.append(field.name)
+
+  def CanonicalFormFromMask(self, mask):
+    """Converts a FieldMask to the canonical form.
+
+    Removes paths that are covered by another path. For example,
+    "foo.bar" is covered by "foo" and will be removed if "foo"
+    is also in the FieldMask. Then sorts all paths in alphabetical order.
+
+    Args:
+      mask: The original FieldMask to be converted.
+    """
+    tree = _FieldMaskTree(mask)
+    tree.ToFieldMask(self)
+
+  def Union(self, mask1, mask2):
+    """Merges mask1 and mask2 into this FieldMask."""
+    _CheckFieldMaskMessage(mask1)
+    _CheckFieldMaskMessage(mask2)
+    tree = _FieldMaskTree(mask1)
+    tree.MergeFromFieldMask(mask2)
+    tree.ToFieldMask(self)
+
+  def Intersect(self, mask1, mask2):
+    """Intersects mask1 and mask2 into this FieldMask."""
+    _CheckFieldMaskMessage(mask1)
+    _CheckFieldMaskMessage(mask2)
+    tree = _FieldMaskTree(mask1)
+    intersection = _FieldMaskTree()
+    for path in mask2.paths:
+      tree.IntersectPath(path, intersection)
+    intersection.ToFieldMask(self)
+
+  def MergeMessage(
+      self, source, destination,
+      replace_message_field=False, replace_repeated_field=False):
+    """Merges fields specified in FieldMask from source to destination.
+
+    Args:
+      source: Source message.
+      destination: The destination message to be merged into.
+      replace_message_field: Replace message field if True. Merge message
+          field if False.
+      replace_repeated_field: Replace repeated field if True. Append
+          elements of repeated field if False.
+    """
+    tree = _FieldMaskTree(self)
+    tree.MergeMessage(
+        source, destination, replace_message_field, replace_repeated_field)
+
+
+def _IsValidPath(message_descriptor, path):
+  """Checks whether the path is valid for Message Descriptor."""
+  parts = path.split('.')
+  last = parts.pop()
+  for name in parts:
+    field = message_descriptor.fields_by_name.get(name)
+    if (field is None or
+        field.label == FieldDescriptor.LABEL_REPEATED or
+        field.type != FieldDescriptor.TYPE_MESSAGE):
+      return False
+    message_descriptor = field.message_type
+  return last in message_descriptor.fields_by_name
+
+
+def _CheckFieldMaskMessage(message):
+  """Raises ValueError if message is not a FieldMask."""
+  message_descriptor = message.DESCRIPTOR
+  if (message_descriptor.name != 'FieldMask' or
+      message_descriptor.file.name != 'google/protobuf/field_mask.proto'):
+    raise ValueError('Message {0} is not a FieldMask.'.format(
+        message_descriptor.full_name))
+
+
+def _SnakeCaseToCamelCase(path_name):
+  """Converts a path name from snake_case to camelCase."""
+  result = []
+  after_underscore = False
+  for c in path_name:
+    if c.isupper():
+      raise ValueError(
+          'Fail to print FieldMask to Json string: Path name '
+          '{0} must not contain uppercase letters.'.format(path_name))
+    if after_underscore:
+      if c.islower():
+        result.append(c.upper())
+        after_underscore = False
+      else:
+        raise ValueError(
+            'Fail to print FieldMask to Json string: The '
+            'character after a "_" must be a lowercase letter '
+            'in path name {0}.'.format(path_name))
+    elif c == '_':
+      after_underscore = True
+    else:
+      result += c
+
+  if after_underscore:
+    raise ValueError('Fail to print FieldMask to Json string: Trailing "_" '
+                     'in path name {0}.'.format(path_name))
+  return ''.join(result)
+
+
+def _CamelCaseToSnakeCase(path_name):
+  """Converts a field name from camelCase to snake_case."""
+  result = []
+  for c in path_name:
+    if c == '_':
+      raise ValueError('Fail to parse FieldMask: Path name '
+                       '{0} must not contain "_"s.'.format(path_name))
+    if c.isupper():
+      result += '_'
+      result += c.lower()
+    else:
+      result += c
+  return ''.join(result)
+
+
+class _FieldMaskTree(object):
+  """Represents a FieldMask in a tree structure.
+
+  For example, given a FieldMask "foo.bar,foo.baz,bar.baz",
+  the FieldMaskTree will be:
+      [_root] -+- foo -+- bar
+            |       |
+            |       +- baz
+            |
+            +- bar --- baz
+  In the tree, each leaf node represents a field path.
+  """
+
+  __slots__ = ('_root',)
+
+  def __init__(self, field_mask=None):
+    """Initializes the tree by FieldMask."""
+    self._root = {}
+    if field_mask:
+      self.MergeFromFieldMask(field_mask)
+
+  def MergeFromFieldMask(self, field_mask):
+    """Merges a FieldMask to the tree."""
+    for path in field_mask.paths:
+      self.AddPath(path)
+
+  def AddPath(self, path):
+    """Adds a field path into the tree.
+
+    If the field path to add is a sub-path of an existing field path
+    in the tree (i.e., a leaf node), it means the tree already matches
+    the given path so nothing will be added to the tree. If the path
+    matches an existing non-leaf node in the tree, that non-leaf node
+    will be turned into a leaf node with all its children removed because
+    the path matches all the node's children. Otherwise, a new path will
+    be added.
+
+    Args:
+      path: The field path to add.
+    """
+    node = self._root
+    for name in path.split('.'):
+      if name not in node:
+        node[name] = {}
+      elif not node[name]:
+        # Pre-existing empty node implies we already have this entire tree.
+        return
+      node = node[name]
+    # Remove any sub-trees we might have had.
+    node.clear()
+
+  def ToFieldMask(self, field_mask):
+    """Converts the tree to a FieldMask."""
+    field_mask.Clear()
+    _AddFieldPaths(self._root, '', field_mask)
+
+  def IntersectPath(self, path, intersection):
+    """Calculates the intersection part of a field path with this tree.
+
+    Args:
+      path: The field path to calculates.
+      intersection: The out tree to record the intersection part.
+    """
+    node = self._root
+    for name in path.split('.'):
+      if name not in node:
+        return
+      elif not node[name]:
+        intersection.AddPath(path)
+        return
+      node = node[name]
+    intersection.AddLeafNodes(path, node)
+
+  def AddLeafNodes(self, prefix, node):
+    """Adds leaf nodes begin with prefix to this tree."""
+    if not node:
+      self.AddPath(prefix)
+    for name in node:
+      child_path = prefix + '.' + name
+      self.AddLeafNodes(child_path, node[name])
+
+  def MergeMessage(
+      self, source, destination,
+      replace_message, replace_repeated):
+    """Merge all fields specified by this tree from source to destination."""
+    _MergeMessage(
+        self._root, source, destination, replace_message, replace_repeated)
+
+
+def _StrConvert(value):
+  """Converts value to str if it is not."""
+  # This file is imported by c extension and some methods like ClearField
+  # requires string for the field name. py2/py3 has different text
+  # type and may use unicode.
+  if not isinstance(value, str):
+    return value.encode('utf-8')
+  return value
+
+
+def _MergeMessage(
+    node, source, destination, replace_message, replace_repeated):
+  """Merge all fields specified by a sub-tree from source to destination."""
+  source_descriptor = source.DESCRIPTOR
+  for name in node:
+    child = node[name]
+    field = source_descriptor.fields_by_name[name]
+    if field is None:
+      raise ValueError('Error: Can\'t find field {0} in message {1}.'.format(
+          name, source_descriptor.full_name))
+    if child:
+      # Sub-paths are only allowed for singular message fields.
+      if (field.label == FieldDescriptor.LABEL_REPEATED or
+          field.cpp_type != FieldDescriptor.CPPTYPE_MESSAGE):
+        raise ValueError('Error: Field {0} in message {1} is not a singular '
+                         'message field and cannot have sub-fields.'.format(
+                             name, source_descriptor.full_name))
+      if source.HasField(name):
+        _MergeMessage(
+            child, getattr(source, name), getattr(destination, name),
+            replace_message, replace_repeated)
+      continue
+    if field.label == FieldDescriptor.LABEL_REPEATED:
+      if replace_repeated:
+        destination.ClearField(_StrConvert(name))
+      repeated_source = getattr(source, name)
+      repeated_destination = getattr(destination, name)
+      repeated_destination.MergeFrom(repeated_source)
+    else:
+      if field.cpp_type == FieldDescriptor.CPPTYPE_MESSAGE:
+        if replace_message:
+          destination.ClearField(_StrConvert(name))
+        if source.HasField(name):
+          getattr(destination, name).MergeFrom(getattr(source, name))
+      else:
+        setattr(destination, name, getattr(source, name))
+
+
+def _AddFieldPaths(node, prefix, field_mask):
+  """Adds the field paths descended from node to field_mask."""
+  if not node and prefix:
+    field_mask.paths.append(prefix)
+    return
+  for name in sorted(node):
+    if prefix:
+      child_path = prefix + '.' + name
+    else:
+      child_path = name
+    _AddFieldPaths(node[name], child_path, field_mask)

python/google/protobuf/internal/field_mask_test.py

@@ -0,0 +1,400 @@
+# Protocol Buffers - Google's data interchange format
+# Copyright 2008 Google Inc.  All rights reserved.
+# https://developers.google.com/protocol-buffers/
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are
+# met:
+#
+#     * Redistributions of source code must retain the above copyright
+# notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above
+# copyright notice, this list of conditions and the following disclaimer
+# in the documentation and/or other materials provided with the
+# distribution.
+#     * Neither the name of Google Inc. nor the names of its
+# contributors may be used to endorse or promote products derived from
+# this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+"""Test for google.protobuf.internal.well_known_types."""
+
+import unittest
+
+from google.protobuf import field_mask_pb2
+from google.protobuf import map_unittest_pb2
+from google.protobuf import unittest_pb2
+from google.protobuf.internal import field_mask
+from google.protobuf.internal import test_util
+from google.protobuf import descriptor
+
+
+class FieldMaskTest(unittest.TestCase):
+
+  def testStringFormat(self):
+    mask = field_mask_pb2.FieldMask()
+    self.assertEqual('', mask.ToJsonString())
+    mask.paths.append('foo')
+    self.assertEqual('foo', mask.ToJsonString())
+    mask.paths.append('bar')
+    self.assertEqual('foo,bar', mask.ToJsonString())
+
+    mask.FromJsonString('')
+    self.assertEqual('', mask.ToJsonString())
+    mask.FromJsonString('foo')
+    self.assertEqual(['foo'], mask.paths)
+    mask.FromJsonString('foo,bar')
+    self.assertEqual(['foo', 'bar'], mask.paths)
+
+    # Test camel case
+    mask.Clear()
+    mask.paths.append('foo_bar')
+    self.assertEqual('fooBar', mask.ToJsonString())
+    mask.paths.append('bar_quz')
+    self.assertEqual('fooBar,barQuz', mask.ToJsonString())
+
+    mask.FromJsonString('')
+    self.assertEqual('', mask.ToJsonString())
+    self.assertEqual([], mask.paths)
+    mask.FromJsonString('fooBar')
+    self.assertEqual(['foo_bar'], mask.paths)
+    mask.FromJsonString('fooBar,barQuz')
+    self.assertEqual(['foo_bar', 'bar_quz'], mask.paths)
+
+  def testDescriptorToFieldMask(self):
+    mask = field_mask_pb2.FieldMask()
+    msg_descriptor = unittest_pb2.TestAllTypes.DESCRIPTOR
+    mask.AllFieldsFromDescriptor(msg_descriptor)
+    self.assertEqual(76, len(mask.paths))
+    self.assertTrue(mask.IsValidForDescriptor(msg_descriptor))
+    for field in msg_descriptor.fields:
+      self.assertTrue(field.name in mask.paths)
+
+  def testIsValidForDescriptor(self):
+    msg_descriptor = unittest_pb2.TestAllTypes.DESCRIPTOR
+    # Empty mask
+    mask = field_mask_pb2.FieldMask()
+    self.assertTrue(mask.IsValidForDescriptor(msg_descriptor))
+    # All fields from descriptor
+    mask.AllFieldsFromDescriptor(msg_descriptor)
+    self.assertTrue(mask.IsValidForDescriptor(msg_descriptor))
+    # Child under optional message
+    mask.paths.append('optional_nested_message.bb')
+    self.assertTrue(mask.IsValidForDescriptor(msg_descriptor))
+    # Repeated field is only allowed in the last position of path
+    mask.paths.append('repeated_nested_message.bb')
+    self.assertFalse(mask.IsValidForDescriptor(msg_descriptor))
+    # Invalid top level field
+    mask = field_mask_pb2.FieldMask()
+    mask.paths.append('xxx')
+    self.assertFalse(mask.IsValidForDescriptor(msg_descriptor))
+    # Invalid field in root
+    mask = field_mask_pb2.FieldMask()
+    mask.paths.append('xxx.zzz')
+    self.assertFalse(mask.IsValidForDescriptor(msg_descriptor))
+    # Invalid field in internal node
+    mask = field_mask_pb2.FieldMask()
+    mask.paths.append('optional_nested_message.xxx.zzz')
+    self.assertFalse(mask.IsValidForDescriptor(msg_descriptor))
+    # Invalid field in leaf
+    mask = field_mask_pb2.FieldMask()
+    mask.paths.append('optional_nested_message.xxx')
+    self.assertFalse(mask.IsValidForDescriptor(msg_descriptor))
+
+  def testCanonicalFrom(self):
+    mask = field_mask_pb2.FieldMask()
+    out_mask = field_mask_pb2.FieldMask()
+    # Paths will be sorted.
+    mask.FromJsonString('baz.quz,bar,foo')
+    out_mask.CanonicalFormFromMask(mask)
+    self.assertEqual('bar,baz.quz,foo', out_mask.ToJsonString())
+    # Duplicated paths will be removed.
+    mask.FromJsonString('foo,bar,foo')
+    out_mask.CanonicalFormFromMask(mask)
+    self.assertEqual('bar,foo', out_mask.ToJsonString())
+    # Sub-paths of other paths will be removed.
+    mask.FromJsonString('foo.b1,bar.b1,foo.b2,bar')
+    out_mask.CanonicalFormFromMask(mask)
+    self.assertEqual('bar,foo.b1,foo.b2', out_mask.ToJsonString())
+
+    # Test more deeply nested cases.
+    mask.FromJsonString(
+        'foo.bar.baz1,foo.bar.baz2.quz,foo.bar.baz2')
+    out_mask.CanonicalFormFromMask(mask)
+    self.assertEqual('foo.bar.baz1,foo.bar.baz2',
+                     out_mask.ToJsonString())
+    mask.FromJsonString(
+        'foo.bar.baz1,foo.bar.baz2,foo.bar.baz2.quz')
+    out_mask.CanonicalFormFromMask(mask)
+    self.assertEqual('foo.bar.baz1,foo.bar.baz2',
+                     out_mask.ToJsonString())
+    mask.FromJsonString(
+        'foo.bar.baz1,foo.bar.baz2,foo.bar.baz2.quz,foo.bar')
+    out_mask.CanonicalFormFromMask(mask)
+    self.assertEqual('foo.bar', out_mask.ToJsonString())
+    mask.FromJsonString(
+        'foo.bar.baz1,foo.bar.baz2,foo.bar.baz2.quz,foo')
+    out_mask.CanonicalFormFromMask(mask)
+    self.assertEqual('foo', out_mask.ToJsonString())
+
+  def testUnion(self):
+    mask1 = field_mask_pb2.FieldMask()
+    mask2 = field_mask_pb2.FieldMask()
+    out_mask = field_mask_pb2.FieldMask()
+    mask1.FromJsonString('foo,baz')
+    mask2.FromJsonString('bar,quz')
+    out_mask.Union(mask1, mask2)
+    self.assertEqual('bar,baz,foo,quz', out_mask.ToJsonString())
+    # Overlap with duplicated paths.
+    mask1.FromJsonString('foo,baz.bb')
+    mask2.FromJsonString('baz.bb,quz')
+    out_mask.Union(mask1, mask2)
+    self.assertEqual('baz.bb,foo,quz', out_mask.ToJsonString())
+    # Overlap with paths covering some other paths.
+    mask1.FromJsonString('foo.bar.baz,quz')
+    mask2.FromJsonString('foo.bar,bar')
+    out_mask.Union(mask1, mask2)
+    self.assertEqual('bar,foo.bar,quz', out_mask.ToJsonString())
+    src = unittest_pb2.TestAllTypes()
+    with self.assertRaises(ValueError):
+      out_mask.Union(src, mask2)
+
+  def testIntersect(self):
+    mask1 = field_mask_pb2.FieldMask()
+    mask2 = field_mask_pb2.FieldMask()
+    out_mask = field_mask_pb2.FieldMask()
+    # Test cases without overlapping.
+    mask1.FromJsonString('foo,baz')
+    mask2.FromJsonString('bar,quz')
+    out_mask.Intersect(mask1, mask2)
+    self.assertEqual('', out_mask.ToJsonString())
+    self.assertEqual(len(out_mask.paths), 0)
+    self.assertEqual(out_mask.paths, [])
+    # Overlap with duplicated paths.
+    mask1.FromJsonString('foo,baz.bb')
+    mask2.FromJsonString('baz.bb,quz')
+    out_mask.Intersect(mask1, mask2)
+    self.assertEqual('baz.bb', out_mask.ToJsonString())
+    # Overlap with paths covering some other paths.
+    mask1.FromJsonString('foo.bar.baz,quz')
+    mask2.FromJsonString('foo.bar,bar')
+    out_mask.Intersect(mask1, mask2)
+    self.assertEqual('foo.bar.baz', out_mask.ToJsonString())
+    mask1.FromJsonString('foo.bar,bar')
+    mask2.FromJsonString('foo.bar.baz,quz')
+    out_mask.Intersect(mask1, mask2)
+    self.assertEqual('foo.bar.baz', out_mask.ToJsonString())
+    # Intersect '' with ''
+    mask1.Clear()
+    mask2.Clear()
+    mask1.paths.append('')
+    mask2.paths.append('')
+    self.assertEqual(mask1.paths, [''])
+    self.assertEqual('', mask1.ToJsonString())
+    out_mask.Intersect(mask1, mask2)
+    self.assertEqual(out_mask.paths, [])
+
+  def testMergeMessageWithoutMapFields(self):
+    # Test merge one field.
+    src = unittest_pb2.TestAllTypes()
+    test_util.SetAllFields(src)
+    for field in src.DESCRIPTOR.fields:
+      if field.containing_oneof:
+        continue
+      field_name = field.name
+      dst = unittest_pb2.TestAllTypes()
+      # Only set one path to mask.
+      mask = field_mask_pb2.FieldMask()
+      mask.paths.append(field_name)
+      mask.MergeMessage(src, dst)
+      # The expected result message.
+      msg = unittest_pb2.TestAllTypes()
+      if field.label == descriptor.FieldDescriptor.LABEL_REPEATED:
+        repeated_src = getattr(src, field_name)
+        repeated_msg = getattr(msg, field_name)
+        if field.cpp_type == descriptor.FieldDescriptor.CPPTYPE_MESSAGE:
+          for item in repeated_src:
+            repeated_msg.add().CopyFrom(item)
+        else:
+          repeated_msg.extend(repeated_src)
+      elif field.cpp_type == descriptor.FieldDescriptor.CPPTYPE_MESSAGE:
+        getattr(msg, field_name).CopyFrom(getattr(src, field_name))
+      else:
+        setattr(msg, field_name, getattr(src, field_name))
+      # Only field specified in mask is merged.
+      self.assertEqual(msg, dst)
+
+    # Test merge nested fields.
+    nested_src = unittest_pb2.NestedTestAllTypes()
+    nested_dst = unittest_pb2.NestedTestAllTypes()
+    nested_src.child.payload.optional_int32 = 1234
+    nested_src.child.child.payload.optional_int32 = 5678
+    mask = field_mask_pb2.FieldMask()
+    mask.FromJsonString('child.payload')
+    mask.MergeMessage(nested_src, nested_dst)
+    self.assertEqual(1234, nested_dst.child.payload.optional_int32)
+    self.assertEqual(0, nested_dst.child.child.payload.optional_int32)
+
+    mask.FromJsonString('child.child.payload')
+    mask.MergeMessage(nested_src, nested_dst)
+    self.assertEqual(1234, nested_dst.child.payload.optional_int32)
+    self.assertEqual(5678, nested_dst.child.child.payload.optional_int32)
+
+    nested_dst.Clear()
+    mask.FromJsonString('child.child.payload')
+    mask.MergeMessage(nested_src, nested_dst)
+    self.assertEqual(0, nested_dst.child.payload.optional_int32)
+    self.assertEqual(5678, nested_dst.child.child.payload.optional_int32)
+
+    nested_dst.Clear()
+    mask.FromJsonString('child')
+    mask.MergeMessage(nested_src, nested_dst)
+    self.assertEqual(1234, nested_dst.child.payload.optional_int32)
+    self.assertEqual(5678, nested_dst.child.child.payload.optional_int32)
+
+    # Test MergeOptions.
+    nested_dst.Clear()
+    nested_dst.child.payload.optional_int64 = 4321
+    # Message fields will be merged by default.
+    mask.FromJsonString('child.payload')
+    mask.MergeMessage(nested_src, nested_dst)
+    self.assertEqual(1234, nested_dst.child.payload.optional_int32)
+    self.assertEqual(4321, nested_dst.child.payload.optional_int64)
+    # Change the behavior to replace message fields.
+    mask.FromJsonString('child.payload')
+    mask.MergeMessage(nested_src, nested_dst, True, False)
+    self.assertEqual(1234, nested_dst.child.payload.optional_int32)
+    self.assertEqual(0, nested_dst.child.payload.optional_int64)
+
+    # By default, fields missing in source are not cleared in destination.
+    nested_dst.payload.optional_int32 = 1234
+    self.assertTrue(nested_dst.HasField('payload'))
+    mask.FromJsonString('payload')
+    mask.MergeMessage(nested_src, nested_dst)
+    self.assertTrue(nested_dst.HasField('payload'))
+    # But they are cleared when replacing message fields.
+    nested_dst.Clear()
+    nested_dst.payload.optional_int32 = 1234
+    mask.FromJsonString('payload')
+    mask.MergeMessage(nested_src, nested_dst, True, False)
+    self.assertFalse(nested_dst.HasField('payload'))
+
+    nested_src.payload.repeated_int32.append(1234)
+    nested_dst.payload.repeated_int32.append(5678)
+    # Repeated fields will be appended by default.
+    mask.FromJsonString('payload.repeatedInt32')
+    mask.MergeMessage(nested_src, nested_dst)
+    self.assertEqual(2, len(nested_dst.payload.repeated_int32))
+    self.assertEqual(5678, nested_dst.payload.repeated_int32[0])
+    self.assertEqual(1234, nested_dst.payload.repeated_int32[1])
+    # Change the behavior to replace repeated fields.
+    mask.FromJsonString('payload.repeatedInt32')
+    mask.MergeMessage(nested_src, nested_dst, False, True)
+    self.assertEqual(1, len(nested_dst.payload.repeated_int32))
+    self.assertEqual(1234, nested_dst.payload.repeated_int32[0])
+
+    # Test Merge oneof field.
+    new_msg = unittest_pb2.TestOneof2()
+    dst = unittest_pb2.TestOneof2()
+    dst.foo_message.moo_int = 1
+    mask = field_mask_pb2.FieldMask()
+    mask.FromJsonString('fooMessage,fooLazyMessage.mooInt')
+    mask.MergeMessage(new_msg, dst)
+    self.assertTrue(dst.HasField('foo_message'))
+    self.assertFalse(dst.HasField('foo_lazy_message'))
+
+  def testMergeMessageWithMapField(self):
+    empty_map = map_unittest_pb2.TestRecursiveMapMessage()
+    src_level_2 = map_unittest_pb2.TestRecursiveMapMessage()
+    src_level_2.a['src level 2'].CopyFrom(empty_map)
+    src = map_unittest_pb2.TestRecursiveMapMessage()
+    src.a['common key'].CopyFrom(src_level_2)
+    src.a['src level 1'].CopyFrom(src_level_2)
+
+    dst_level_2 = map_unittest_pb2.TestRecursiveMapMessage()
+    dst_level_2.a['dst level 2'].CopyFrom(empty_map)
+    dst = map_unittest_pb2.TestRecursiveMapMessage()
+    dst.a['common key'].CopyFrom(dst_level_2)
+    dst.a['dst level 1'].CopyFrom(empty_map)
+
+    mask = field_mask_pb2.FieldMask()
+    mask.FromJsonString('a')
+    mask.MergeMessage(src, dst)
+
+    # map from dst is replaced with map from src.
+    self.assertEqual(dst.a['common key'], src_level_2)
+    self.assertEqual(dst.a['src level 1'], src_level_2)
+    self.assertEqual(dst.a['dst level 1'], empty_map)
+
+  def testMergeErrors(self):
+    src = unittest_pb2.TestAllTypes()
+    dst = unittest_pb2.TestAllTypes()
+    mask = field_mask_pb2.FieldMask()
+    test_util.SetAllFields(src)
+    mask.FromJsonString('optionalInt32.field')
+    with self.assertRaises(ValueError) as e:
+      mask.MergeMessage(src, dst)
+    self.assertEqual('Error: Field optional_int32 in message '
+                     'protobuf_unittest.TestAllTypes is not a singular '
+                     'message field and cannot have sub-fields.',
+                     str(e.exception))
+
+  def testSnakeCaseToCamelCase(self):
+    self.assertEqual('fooBar',
+                     field_mask._SnakeCaseToCamelCase('foo_bar'))
+    self.assertEqual('FooBar',
+                     field_mask._SnakeCaseToCamelCase('_foo_bar'))
+    self.assertEqual('foo3Bar',
+                     field_mask._SnakeCaseToCamelCase('foo3_bar'))
+
+    # No uppercase letter is allowed.
+    self.assertRaisesRegex(
+        ValueError,
+        'Fail to print FieldMask to Json string: Path name Foo must '
+        'not contain uppercase letters.',
+        field_mask._SnakeCaseToCamelCase, 'Foo')
+    # Any character after a "_" must be a lowercase letter.
+    #   1. "_" cannot be followed by another "_".
+    #   2. "_" cannot be followed by a digit.
+    #   3. "_" cannot appear as the last character.
+    self.assertRaisesRegex(
+        ValueError,
+        'Fail to print FieldMask to Json string: The character after a '
+        '"_" must be a lowercase letter in path name foo__bar.',
+        field_mask._SnakeCaseToCamelCase, 'foo__bar')
+    self.assertRaisesRegex(
+        ValueError,
+        'Fail to print FieldMask to Json string: The character after a '
+        '"_" must be a lowercase letter in path name foo_3bar.',
+        field_mask._SnakeCaseToCamelCase, 'foo_3bar')
+    self.assertRaisesRegex(
+        ValueError,
+        'Fail to print FieldMask to Json string: Trailing "_" in path '
+        'name foo_bar_.', field_mask._SnakeCaseToCamelCase, 'foo_bar_')
+
+  def testCamelCaseToSnakeCase(self):
+    self.assertEqual('foo_bar',
+                     field_mask._CamelCaseToSnakeCase('fooBar'))
+    self.assertEqual('_foo_bar',
+                     field_mask._CamelCaseToSnakeCase('FooBar'))
+    self.assertEqual('foo3_bar',
+                     field_mask._CamelCaseToSnakeCase('foo3Bar'))
+    self.assertRaisesRegex(
+        ValueError,
+        'Fail to parse FieldMask: Path name foo_bar must not contain "_"s.',
+        field_mask._CamelCaseToSnakeCase, 'foo_bar')
+
+
+if __name__ == '__main__':
+  unittest.main()

python/google/protobuf/internal/missing_enum_values.proto

@@ -30,7 +30,6 @@
 
 syntax = "proto2";
 
-
 package google.protobuf.python.internal;
 
 message TestEnumValues {
@@ -53,4 +52,3 @@ message TestMissingEnumValues {
 message JustString {
   required string dummy = 1;
 }
-

python/google/protobuf/internal/well_known_types.py

@@ -44,7 +44,9 @@ import calendar
 import collections.abc
 import datetime
 
-from google.protobuf.descriptor import FieldDescriptor
+from google.protobuf.internal import field_mask
+
+FieldMask = field_mask.FieldMask
 
 _TIMESTAMPFOMAT = '%Y-%m-%dT%H:%M:%S'
 _NANOS_PER_SECOND = 1000000000
@@ -430,306 +432,6 @@ def _RoundTowardZero(value, divider):
     return result
 
 
-class FieldMask(object):
-  """Class for FieldMask message type."""
-
-  __slots__ = ()
-
-  def ToJsonString(self):
-    """Converts FieldMask to string according to proto3 JSON spec."""
-    camelcase_paths = []
-    for path in self.paths:
-      camelcase_paths.append(_SnakeCaseToCamelCase(path))
-    return ','.join(camelcase_paths)
-
-  def FromJsonString(self, value):
-    """Converts string to FieldMask according to proto3 JSON spec."""
-    if not isinstance(value, str):
-      raise ValueError('FieldMask JSON value not a string: {!r}'.format(value))
-    self.Clear()
-    if value:
-      for path in value.split(','):
-        self.paths.append(_CamelCaseToSnakeCase(path))
-
-  def IsValidForDescriptor(self, message_descriptor):
-    """Checks whether the FieldMask is valid for Message Descriptor."""
-    for path in self.paths:
-      if not _IsValidPath(message_descriptor, path):
-        return False
-    return True
-
-  def AllFieldsFromDescriptor(self, message_descriptor):
-    """Gets all direct fields of Message Descriptor to FieldMask."""
-    self.Clear()
-    for field in message_descriptor.fields:
-      self.paths.append(field.name)
-
-  def CanonicalFormFromMask(self, mask):
-    """Converts a FieldMask to the canonical form.
-
-    Removes paths that are covered by another path. For example,
-    "foo.bar" is covered by "foo" and will be removed if "foo"
-    is also in the FieldMask. Then sorts all paths in alphabetical order.
-
-    Args:
-      mask: The original FieldMask to be converted.
-    """
-    tree = _FieldMaskTree(mask)
-    tree.ToFieldMask(self)
-
-  def Union(self, mask1, mask2):
-    """Merges mask1 and mask2 into this FieldMask."""
-    _CheckFieldMaskMessage(mask1)
-    _CheckFieldMaskMessage(mask2)
-    tree = _FieldMaskTree(mask1)
-    tree.MergeFromFieldMask(mask2)
-    tree.ToFieldMask(self)
-
-  def Intersect(self, mask1, mask2):
-    """Intersects mask1 and mask2 into this FieldMask."""
-    _CheckFieldMaskMessage(mask1)
-    _CheckFieldMaskMessage(mask2)
-    tree = _FieldMaskTree(mask1)
-    intersection = _FieldMaskTree()
-    for path in mask2.paths:
-      tree.IntersectPath(path, intersection)
-    intersection.ToFieldMask(self)
-
-  def MergeMessage(
-      self, source, destination,
-      replace_message_field=False, replace_repeated_field=False):
-    """Merges fields specified in FieldMask from source to destination.
-
-    Args:
-      source: Source message.
-      destination: The destination message to be merged into.
-      replace_message_field: Replace message field if True. Merge message
-          field if False.
-      replace_repeated_field: Replace repeated field if True. Append
-          elements of repeated field if False.
-    """
-    tree = _FieldMaskTree(self)
-    tree.MergeMessage(
-        source, destination, replace_message_field, replace_repeated_field)
-
-
-def _IsValidPath(message_descriptor, path):
-  """Checks whether the path is valid for Message Descriptor."""
-  parts = path.split('.')
-  last = parts.pop()
-  for name in parts:
-    field = message_descriptor.fields_by_name.get(name)
-    if (field is None or
-        field.label == FieldDescriptor.LABEL_REPEATED or
-        field.type != FieldDescriptor.TYPE_MESSAGE):
-      return False
-    message_descriptor = field.message_type
-  return last in message_descriptor.fields_by_name
-
-
-def _CheckFieldMaskMessage(message):
-  """Raises ValueError if message is not a FieldMask."""
-  message_descriptor = message.DESCRIPTOR
-  if (message_descriptor.name != 'FieldMask' or
-      message_descriptor.file.name != 'google/protobuf/field_mask.proto'):
-    raise ValueError('Message {0} is not a FieldMask.'.format(
-        message_descriptor.full_name))
-
-
-def _SnakeCaseToCamelCase(path_name):
-  """Converts a path name from snake_case to camelCase."""
-  result = []
-  after_underscore = False
-  for c in path_name:
-    if c.isupper():
-      raise ValueError(
-          'Fail to print FieldMask to Json string: Path name '
-          '{0} must not contain uppercase letters.'.format(path_name))
-    if after_underscore:
-      if c.islower():
-        result.append(c.upper())
-        after_underscore = False
-      else:
-        raise ValueError(
-            'Fail to print FieldMask to Json string: The '
-            'character after a "_" must be a lowercase letter '
-            'in path name {0}.'.format(path_name))
-    elif c == '_':
-      after_underscore = True
-    else:
-      result += c
-
-  if after_underscore:
-    raise ValueError('Fail to print FieldMask to Json string: Trailing "_" '
-                     'in path name {0}.'.format(path_name))
-  return ''.join(result)
-
-
-def _CamelCaseToSnakeCase(path_name):
-  """Converts a field name from camelCase to snake_case."""
-  result = []
-  for c in path_name:
-    if c == '_':
-      raise ValueError('Fail to parse FieldMask: Path name '
-                       '{0} must not contain "_"s.'.format(path_name))
-    if c.isupper():
-      result += '_'
-      result += c.lower()
-    else:
-      result += c
-  return ''.join(result)
-
-
-class _FieldMaskTree(object):
-  """Represents a FieldMask in a tree structure.
-
-  For example, given a FieldMask "foo.bar,foo.baz,bar.baz",
-  the FieldMaskTree will be:
-      [_root] -+- foo -+- bar
-            |       |
-            |       +- baz
-            |
-            +- bar --- baz
-  In the tree, each leaf node represents a field path.
-  """
-
-  __slots__ = ('_root',)
-
-  def __init__(self, field_mask=None):
-    """Initializes the tree by FieldMask."""
-    self._root = {}
-    if field_mask:
-      self.MergeFromFieldMask(field_mask)
-
-  def MergeFromFieldMask(self, field_mask):
-    """Merges a FieldMask to the tree."""
-    for path in field_mask.paths:
-      self.AddPath(path)
-
-  def AddPath(self, path):
-    """Adds a field path into the tree.
-
-    If the field path to add is a sub-path of an existing field path
-    in the tree (i.e., a leaf node), it means the tree already matches
-    the given path so nothing will be added to the tree. If the path
-    matches an existing non-leaf node in the tree, that non-leaf node
-    will be turned into a leaf node with all its children removed because
-    the path matches all the node's children. Otherwise, a new path will
-    be added.
-
-    Args:
-      path: The field path to add.
-    """
-    node = self._root
-    for name in path.split('.'):
-      if name not in node:
-        node[name] = {}
-      elif not node[name]:
-        # Pre-existing empty node implies we already have this entire tree.
-        return
-      node = node[name]
-    # Remove any sub-trees we might have had.
-    node.clear()
-
-  def ToFieldMask(self, field_mask):
-    """Converts the tree to a FieldMask."""
-    field_mask.Clear()
-    _AddFieldPaths(self._root, '', field_mask)
-
-  def IntersectPath(self, path, intersection):
-    """Calculates the intersection part of a field path with this tree.
-
-    Args:
-      path: The field path to calculates.
-      intersection: The out tree to record the intersection part.
-    """
-    node = self._root
-    for name in path.split('.'):
-      if name not in node:
-        return
-      elif not node[name]:
-        intersection.AddPath(path)
-        return
-      node = node[name]
-    intersection.AddLeafNodes(path, node)
-
-  def AddLeafNodes(self, prefix, node):
-    """Adds leaf nodes begin with prefix to this tree."""
-    if not node:
-      self.AddPath(prefix)
-    for name in node:
-      child_path = prefix + '.' + name
-      self.AddLeafNodes(child_path, node[name])
-
-  def MergeMessage(
-      self, source, destination,
-      replace_message, replace_repeated):
-    """Merge all fields specified by this tree from source to destination."""
-    _MergeMessage(
-        self._root, source, destination, replace_message, replace_repeated)
-
-
-def _StrConvert(value):
-  """Converts value to str if it is not."""
-  # This file is imported by c extension and some methods like ClearField
-  # requires string for the field name. py2/py3 has different text
-  # type and may use unicode.
-  if not isinstance(value, str):
-    return value.encode('utf-8')
-  return value
-
-
-def _MergeMessage(
-    node, source, destination, replace_message, replace_repeated):
-  """Merge all fields specified by a sub-tree from source to destination."""
-  source_descriptor = source.DESCRIPTOR
-  for name in node:
-    child = node[name]
-    field = source_descriptor.fields_by_name[name]
-    if field is None:
-      raise ValueError('Error: Can\'t find field {0} in message {1}.'.format(
-          name, source_descriptor.full_name))
-    if child:
-      # Sub-paths are only allowed for singular message fields.
-      if (field.label == FieldDescriptor.LABEL_REPEATED or
-          field.cpp_type != FieldDescriptor.CPPTYPE_MESSAGE):
-        raise ValueError('Error: Field {0} in message {1} is not a singular '
-                         'message field and cannot have sub-fields.'.format(
-                             name, source_descriptor.full_name))
-      if source.HasField(name):
-        _MergeMessage(
-            child, getattr(source, name), getattr(destination, name),
-            replace_message, replace_repeated)
-      continue
-    if field.label == FieldDescriptor.LABEL_REPEATED:
-      if replace_repeated:
-        destination.ClearField(_StrConvert(name))
-      repeated_source = getattr(source, name)
-      repeated_destination = getattr(destination, name)
-      repeated_destination.MergeFrom(repeated_source)
-    else:
-      if field.cpp_type == FieldDescriptor.CPPTYPE_MESSAGE:
-        if replace_message:
-          destination.ClearField(_StrConvert(name))
-        if source.HasField(name):
-          getattr(destination, name).MergeFrom(getattr(source, name))
-      else:
-        setattr(destination, name, getattr(source, name))
-
-
-def _AddFieldPaths(node, prefix, field_mask):
-  """Adds the field paths descended from node to field_mask."""
-  if not node and prefix:
-    field_mask.paths.append(prefix)
-    return
-  for name in sorted(node):
-    if prefix:
-      child_path = prefix + '.' + name
-    else:
-      child_path = name
-    _AddFieldPaths(node[name], child_path, field_mask)
-
-
 def _SetStructValue(struct_value, value):
   if value is None:
     struct_value.null_value = 0

python/google/protobuf/internal/well_known_types_test.py

@@ -38,15 +38,11 @@ import unittest
 
 from google.protobuf import any_pb2
 from google.protobuf import duration_pb2
-from google.protobuf import field_mask_pb2
 from google.protobuf import struct_pb2
 from google.protobuf import timestamp_pb2
-from google.protobuf import map_unittest_pb2
 from google.protobuf import unittest_pb2
 from google.protobuf.internal import any_test_pb2
-from google.protobuf.internal import test_util
 from google.protobuf.internal import well_known_types
-from google.protobuf import descriptor
 from google.protobuf import text_format
 from google.protobuf.internal import _parameterized
 
@@ -390,362 +386,6 @@ class TimeUtilTest(TimeUtilTestBase):
                            message.ToJsonString)
 
 
-class FieldMaskTest(unittest.TestCase):
-
-  def testStringFormat(self):
-    mask = field_mask_pb2.FieldMask()
-    self.assertEqual('', mask.ToJsonString())
-    mask.paths.append('foo')
-    self.assertEqual('foo', mask.ToJsonString())
-    mask.paths.append('bar')
-    self.assertEqual('foo,bar', mask.ToJsonString())
-
-    mask.FromJsonString('')
-    self.assertEqual('', mask.ToJsonString())
-    mask.FromJsonString('foo')
-    self.assertEqual(['foo'], mask.paths)
-    mask.FromJsonString('foo,bar')
-    self.assertEqual(['foo', 'bar'], mask.paths)
-
-    # Test camel case
-    mask.Clear()
-    mask.paths.append('foo_bar')
-    self.assertEqual('fooBar', mask.ToJsonString())
-    mask.paths.append('bar_quz')
-    self.assertEqual('fooBar,barQuz', mask.ToJsonString())
-
-    mask.FromJsonString('')
-    self.assertEqual('', mask.ToJsonString())
-    self.assertEqual([], mask.paths)
-    mask.FromJsonString('fooBar')
-    self.assertEqual(['foo_bar'], mask.paths)
-    mask.FromJsonString('fooBar,barQuz')
-    self.assertEqual(['foo_bar', 'bar_quz'], mask.paths)
-
-  def testDescriptorToFieldMask(self):
-    mask = field_mask_pb2.FieldMask()
-    msg_descriptor = unittest_pb2.TestAllTypes.DESCRIPTOR
-    mask.AllFieldsFromDescriptor(msg_descriptor)
-    self.assertEqual(76, len(mask.paths))
-    self.assertTrue(mask.IsValidForDescriptor(msg_descriptor))
-    for field in msg_descriptor.fields:
-      self.assertTrue(field.name in mask.paths)
-
-  def testIsValidForDescriptor(self):
-    msg_descriptor = unittest_pb2.TestAllTypes.DESCRIPTOR
-    # Empty mask
-    mask = field_mask_pb2.FieldMask()
-    self.assertTrue(mask.IsValidForDescriptor(msg_descriptor))
-    # All fields from descriptor
-    mask.AllFieldsFromDescriptor(msg_descriptor)
-    self.assertTrue(mask.IsValidForDescriptor(msg_descriptor))
-    # Child under optional message
-    mask.paths.append('optional_nested_message.bb')
-    self.assertTrue(mask.IsValidForDescriptor(msg_descriptor))
-    # Repeated field is only allowed in the last position of path
-    mask.paths.append('repeated_nested_message.bb')
-    self.assertFalse(mask.IsValidForDescriptor(msg_descriptor))
-    # Invalid top level field
-    mask = field_mask_pb2.FieldMask()
-    mask.paths.append('xxx')
-    self.assertFalse(mask.IsValidForDescriptor(msg_descriptor))
-    # Invalid field in root
-    mask = field_mask_pb2.FieldMask()
-    mask.paths.append('xxx.zzz')
-    self.assertFalse(mask.IsValidForDescriptor(msg_descriptor))
-    # Invalid field in internal node
-    mask = field_mask_pb2.FieldMask()
-    mask.paths.append('optional_nested_message.xxx.zzz')
-    self.assertFalse(mask.IsValidForDescriptor(msg_descriptor))
-    # Invalid field in leaf
-    mask = field_mask_pb2.FieldMask()
-    mask.paths.append('optional_nested_message.xxx')
-    self.assertFalse(mask.IsValidForDescriptor(msg_descriptor))
-
-  def testCanonicalFrom(self):
-    mask = field_mask_pb2.FieldMask()
-    out_mask = field_mask_pb2.FieldMask()
-    # Paths will be sorted.
-    mask.FromJsonString('baz.quz,bar,foo')
-    out_mask.CanonicalFormFromMask(mask)
-    self.assertEqual('bar,baz.quz,foo', out_mask.ToJsonString())
-    # Duplicated paths will be removed.
-    mask.FromJsonString('foo,bar,foo')
-    out_mask.CanonicalFormFromMask(mask)
-    self.assertEqual('bar,foo', out_mask.ToJsonString())
-    # Sub-paths of other paths will be removed.
-    mask.FromJsonString('foo.b1,bar.b1,foo.b2,bar')
-    out_mask.CanonicalFormFromMask(mask)
-    self.assertEqual('bar,foo.b1,foo.b2', out_mask.ToJsonString())
-
-    # Test more deeply nested cases.
-    mask.FromJsonString(
-        'foo.bar.baz1,foo.bar.baz2.quz,foo.bar.baz2')
-    out_mask.CanonicalFormFromMask(mask)
-    self.assertEqual('foo.bar.baz1,foo.bar.baz2',
-                     out_mask.ToJsonString())
-    mask.FromJsonString(
-        'foo.bar.baz1,foo.bar.baz2,foo.bar.baz2.quz')
-    out_mask.CanonicalFormFromMask(mask)
-    self.assertEqual('foo.bar.baz1,foo.bar.baz2',
-                     out_mask.ToJsonString())
-    mask.FromJsonString(
-        'foo.bar.baz1,foo.bar.baz2,foo.bar.baz2.quz,foo.bar')
-    out_mask.CanonicalFormFromMask(mask)
-    self.assertEqual('foo.bar', out_mask.ToJsonString())
-    mask.FromJsonString(
-        'foo.bar.baz1,foo.bar.baz2,foo.bar.baz2.quz,foo')
-    out_mask.CanonicalFormFromMask(mask)
-    self.assertEqual('foo', out_mask.ToJsonString())
-
-  def testUnion(self):
-    mask1 = field_mask_pb2.FieldMask()
-    mask2 = field_mask_pb2.FieldMask()
-    out_mask = field_mask_pb2.FieldMask()
-    mask1.FromJsonString('foo,baz')
-    mask2.FromJsonString('bar,quz')
-    out_mask.Union(mask1, mask2)
-    self.assertEqual('bar,baz,foo,quz', out_mask.ToJsonString())
-    # Overlap with duplicated paths.
-    mask1.FromJsonString('foo,baz.bb')
-    mask2.FromJsonString('baz.bb,quz')
-    out_mask.Union(mask1, mask2)
-    self.assertEqual('baz.bb,foo,quz', out_mask.ToJsonString())
-    # Overlap with paths covering some other paths.
-    mask1.FromJsonString('foo.bar.baz,quz')
-    mask2.FromJsonString('foo.bar,bar')
-    out_mask.Union(mask1, mask2)
-    self.assertEqual('bar,foo.bar,quz', out_mask.ToJsonString())
-    src = unittest_pb2.TestAllTypes()
-    with self.assertRaises(ValueError):
-      out_mask.Union(src, mask2)
-
-  def testIntersect(self):
-    mask1 = field_mask_pb2.FieldMask()
-    mask2 = field_mask_pb2.FieldMask()
-    out_mask = field_mask_pb2.FieldMask()
-    # Test cases without overlapping.
-    mask1.FromJsonString('foo,baz')
-    mask2.FromJsonString('bar,quz')
-    out_mask.Intersect(mask1, mask2)
-    self.assertEqual('', out_mask.ToJsonString())
-    self.assertEqual(len(out_mask.paths), 0)
-    self.assertEqual(out_mask.paths, [])
-    # Overlap with duplicated paths.
-    mask1.FromJsonString('foo,baz.bb')
-    mask2.FromJsonString('baz.bb,quz')
-    out_mask.Intersect(mask1, mask2)
-    self.assertEqual('baz.bb', out_mask.ToJsonString())
-    # Overlap with paths covering some other paths.
-    mask1.FromJsonString('foo.bar.baz,quz')
-    mask2.FromJsonString('foo.bar,bar')
-    out_mask.Intersect(mask1, mask2)
-    self.assertEqual('foo.bar.baz', out_mask.ToJsonString())
-    mask1.FromJsonString('foo.bar,bar')
-    mask2.FromJsonString('foo.bar.baz,quz')
-    out_mask.Intersect(mask1, mask2)
-    self.assertEqual('foo.bar.baz', out_mask.ToJsonString())
-    # Intersect '' with ''
-    mask1.Clear()
-    mask2.Clear()
-    mask1.paths.append('')
-    mask2.paths.append('')
-    self.assertEqual(mask1.paths, [''])
-    self.assertEqual('', mask1.ToJsonString())
-    out_mask.Intersect(mask1, mask2)
-    self.assertEqual(out_mask.paths, [])
-
-  def testMergeMessageWithoutMapFields(self):
-    # Test merge one field.
-    src = unittest_pb2.TestAllTypes()
-    test_util.SetAllFields(src)
-    for field in src.DESCRIPTOR.fields:
-      if field.containing_oneof:
-        continue
-      field_name = field.name
-      dst = unittest_pb2.TestAllTypes()
-      # Only set one path to mask.
-      mask = field_mask_pb2.FieldMask()
-      mask.paths.append(field_name)
-      mask.MergeMessage(src, dst)
-      # The expected result message.
-      msg = unittest_pb2.TestAllTypes()
-      if field.label == descriptor.FieldDescriptor.LABEL_REPEATED:
-        repeated_src = getattr(src, field_name)
-        repeated_msg = getattr(msg, field_name)
-        if field.cpp_type == descriptor.FieldDescriptor.CPPTYPE_MESSAGE:
-          for item in repeated_src:
-            repeated_msg.add().CopyFrom(item)
-        else:
-          repeated_msg.extend(repeated_src)
-      elif field.cpp_type == descriptor.FieldDescriptor.CPPTYPE_MESSAGE:
-        getattr(msg, field_name).CopyFrom(getattr(src, field_name))
-      else:
-        setattr(msg, field_name, getattr(src, field_name))
-      # Only field specified in mask is merged.
-      self.assertEqual(msg, dst)
-
-    # Test merge nested fields.
-    nested_src = unittest_pb2.NestedTestAllTypes()
-    nested_dst = unittest_pb2.NestedTestAllTypes()
-    nested_src.child.payload.optional_int32 = 1234
-    nested_src.child.child.payload.optional_int32 = 5678
-    mask = field_mask_pb2.FieldMask()
-    mask.FromJsonString('child.payload')
-    mask.MergeMessage(nested_src, nested_dst)
-    self.assertEqual(1234, nested_dst.child.payload.optional_int32)
-    self.assertEqual(0, nested_dst.child.child.payload.optional_int32)
-
-    mask.FromJsonString('child.child.payload')
-    mask.MergeMessage(nested_src, nested_dst)
-    self.assertEqual(1234, nested_dst.child.payload.optional_int32)
-    self.assertEqual(5678, nested_dst.child.child.payload.optional_int32)
-
-    nested_dst.Clear()
-    mask.FromJsonString('child.child.payload')
-    mask.MergeMessage(nested_src, nested_dst)
-    self.assertEqual(0, nested_dst.child.payload.optional_int32)
-    self.assertEqual(5678, nested_dst.child.child.payload.optional_int32)
-
-    nested_dst.Clear()
-    mask.FromJsonString('child')
-    mask.MergeMessage(nested_src, nested_dst)
-    self.assertEqual(1234, nested_dst.child.payload.optional_int32)
-    self.assertEqual(5678, nested_dst.child.child.payload.optional_int32)
-
-    # Test MergeOptions.
-    nested_dst.Clear()
-    nested_dst.child.payload.optional_int64 = 4321
-    # Message fields will be merged by default.
-    mask.FromJsonString('child.payload')
-    mask.MergeMessage(nested_src, nested_dst)
-    self.assertEqual(1234, nested_dst.child.payload.optional_int32)
-    self.assertEqual(4321, nested_dst.child.payload.optional_int64)
-    # Change the behavior to replace message fields.
-    mask.FromJsonString('child.payload')
-    mask.MergeMessage(nested_src, nested_dst, True, False)
-    self.assertEqual(1234, nested_dst.child.payload.optional_int32)
-    self.assertEqual(0, nested_dst.child.payload.optional_int64)
-
-    # By default, fields missing in source are not cleared in destination.
-    nested_dst.payload.optional_int32 = 1234
-    self.assertTrue(nested_dst.HasField('payload'))
-    mask.FromJsonString('payload')
-    mask.MergeMessage(nested_src, nested_dst)
-    self.assertTrue(nested_dst.HasField('payload'))
-    # But they are cleared when replacing message fields.
-    nested_dst.Clear()
-    nested_dst.payload.optional_int32 = 1234
-    mask.FromJsonString('payload')
-    mask.MergeMessage(nested_src, nested_dst, True, False)
-    self.assertFalse(nested_dst.HasField('payload'))
-
-    nested_src.payload.repeated_int32.append(1234)
-    nested_dst.payload.repeated_int32.append(5678)
-    # Repeated fields will be appended by default.
-    mask.FromJsonString('payload.repeatedInt32')
-    mask.MergeMessage(nested_src, nested_dst)
-    self.assertEqual(2, len(nested_dst.payload.repeated_int32))
-    self.assertEqual(5678, nested_dst.payload.repeated_int32[0])
-    self.assertEqual(1234, nested_dst.payload.repeated_int32[1])
-    # Change the behavior to replace repeated fields.
-    mask.FromJsonString('payload.repeatedInt32')
-    mask.MergeMessage(nested_src, nested_dst, False, True)
-    self.assertEqual(1, len(nested_dst.payload.repeated_int32))
-    self.assertEqual(1234, nested_dst.payload.repeated_int32[0])
-
-    # Test Merge oneof field.
-    new_msg = unittest_pb2.TestOneof2()
-    dst = unittest_pb2.TestOneof2()
-    dst.foo_message.moo_int = 1
-    mask = field_mask_pb2.FieldMask()
-    mask.FromJsonString('fooMessage,fooLazyMessage.mooInt')
-    mask.MergeMessage(new_msg, dst)
-    self.assertTrue(dst.HasField('foo_message'))
-    self.assertFalse(dst.HasField('foo_lazy_message'))
-
-  def testMergeMessageWithMapField(self):
-    empty_map = map_unittest_pb2.TestRecursiveMapMessage()
-    src_level_2 = map_unittest_pb2.TestRecursiveMapMessage()
-    src_level_2.a['src level 2'].CopyFrom(empty_map)
-    src = map_unittest_pb2.TestRecursiveMapMessage()
-    src.a['common key'].CopyFrom(src_level_2)
-    src.a['src level 1'].CopyFrom(src_level_2)
-
-    dst_level_2 = map_unittest_pb2.TestRecursiveMapMessage()
-    dst_level_2.a['dst level 2'].CopyFrom(empty_map)
-    dst = map_unittest_pb2.TestRecursiveMapMessage()
-    dst.a['common key'].CopyFrom(dst_level_2)
-    dst.a['dst level 1'].CopyFrom(empty_map)
-
-    mask = field_mask_pb2.FieldMask()
-    mask.FromJsonString('a')
-    mask.MergeMessage(src, dst)
-
-    # map from dst is replaced with map from src.
-    self.assertEqual(dst.a['common key'], src_level_2)
-    self.assertEqual(dst.a['src level 1'], src_level_2)
-    self.assertEqual(dst.a['dst level 1'], empty_map)
-
-  def testMergeErrors(self):
-    src = unittest_pb2.TestAllTypes()
-    dst = unittest_pb2.TestAllTypes()
-    mask = field_mask_pb2.FieldMask()
-    test_util.SetAllFields(src)
-    mask.FromJsonString('optionalInt32.field')
-    with self.assertRaises(ValueError) as e:
-      mask.MergeMessage(src, dst)
-    self.assertEqual('Error: Field optional_int32 in message '
-                     'protobuf_unittest.TestAllTypes is not a singular '
-                     'message field and cannot have sub-fields.',
-                     str(e.exception))
-
-  def testSnakeCaseToCamelCase(self):
-    self.assertEqual('fooBar',
-                     well_known_types._SnakeCaseToCamelCase('foo_bar'))
-    self.assertEqual('FooBar',
-                     well_known_types._SnakeCaseToCamelCase('_foo_bar'))
-    self.assertEqual('foo3Bar',
-                     well_known_types._SnakeCaseToCamelCase('foo3_bar'))
-
-    # No uppercase letter is allowed.
-    self.assertRaisesRegex(
-        ValueError,
-        'Fail to print FieldMask to Json string: Path name Foo must '
-        'not contain uppercase letters.',
-        well_known_types._SnakeCaseToCamelCase, 'Foo')
-    # Any character after a "_" must be a lowercase letter.
-    #   1. "_" cannot be followed by another "_".
-    #   2. "_" cannot be followed by a digit.
-    #   3. "_" cannot appear as the last character.
-    self.assertRaisesRegex(
-        ValueError,
-        'Fail to print FieldMask to Json string: The character after a '
-        '"_" must be a lowercase letter in path name foo__bar.',
-        well_known_types._SnakeCaseToCamelCase, 'foo__bar')
-    self.assertRaisesRegex(
-        ValueError,
-        'Fail to print FieldMask to Json string: The character after a '
-        '"_" must be a lowercase letter in path name foo_3bar.',
-        well_known_types._SnakeCaseToCamelCase, 'foo_3bar')
-    self.assertRaisesRegex(
-        ValueError,
-        'Fail to print FieldMask to Json string: Trailing "_" in path '
-        'name foo_bar_.', well_known_types._SnakeCaseToCamelCase, 'foo_bar_')
-
-  def testCamelCaseToSnakeCase(self):
-    self.assertEqual('foo_bar',
-                     well_known_types._CamelCaseToSnakeCase('fooBar'))
-    self.assertEqual('_foo_bar',
-                     well_known_types._CamelCaseToSnakeCase('FooBar'))
-    self.assertEqual('foo3_bar',
-                     well_known_types._CamelCaseToSnakeCase('foo3Bar'))
-    self.assertRaisesRegex(
-        ValueError,
-        'Fail to parse FieldMask: Path name foo_bar must not contain "_"s.',
-        well_known_types._CamelCaseToSnakeCase, 'foo_bar')
-
-
 class StructTest(unittest.TestCase):
 
   def testStruct(self):

src/file_lists.cmake

@@ -181,7 +181,6 @@ set(libprotobuf_hdrs
   ${protobuf_SOURCE_DIR}/src/google/protobuf/stubs/platform_macros.h
   ${protobuf_SOURCE_DIR}/src/google/protobuf/stubs/port.h
   ${protobuf_SOURCE_DIR}/src/google/protobuf/stubs/status_macros.h
-  ${protobuf_SOURCE_DIR}/src/google/protobuf/stubs/stl_util.h
   ${protobuf_SOURCE_DIR}/src/google/protobuf/stubs/strutil.h
   ${protobuf_SOURCE_DIR}/src/google/protobuf/text_format.h
   ${protobuf_SOURCE_DIR}/src/google/protobuf/unknown_field_set.h
@@ -273,7 +272,6 @@ set(libprotobuf_lite_hdrs
   ${protobuf_SOURCE_DIR}/src/google/protobuf/stubs/platform_macros.h
   ${protobuf_SOURCE_DIR}/src/google/protobuf/stubs/port.h
   ${protobuf_SOURCE_DIR}/src/google/protobuf/stubs/status_macros.h
-  ${protobuf_SOURCE_DIR}/src/google/protobuf/stubs/stl_util.h
   ${protobuf_SOURCE_DIR}/src/google/protobuf/stubs/strutil.h
   ${protobuf_SOURCE_DIR}/src/google/protobuf/wire_format_lite.h
 )

src/google/protobuf/any.proto

@@ -93,7 +93,6 @@ option objc_class_prefix = "GPB";
 // in the type URL, for example "foo.bar.com/x/y.z" will yield type
 // name "y.z".
 //
-//
 // JSON
 //
 // The JSON representation of an `Any` value uses the regular

src/google/protobuf/api.proto

@@ -82,7 +82,6 @@ message Api {
   // be omitted. Zero major versions must only be used for
   // experimental, non-GA interfaces.
   //
-  //
   string version = 4;
 
   // Source context for the protocol buffer service represented by this

src/google/protobuf/arenastring.cc

@@ -39,7 +39,6 @@
 #include "absl/synchronization/mutex.h"
 #include "google/protobuf/message_lite.h"
 #include "google/protobuf/parse_context.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 // clang-format off
 #include "google/protobuf/port_def.inc"

src/google/protobuf/compiler/command_line_interface.cc

@@ -89,7 +89,6 @@
 #include "google/protobuf/io/printer.h"
 #include "google/protobuf/io/zero_copy_stream_impl.h"
 #include "google/protobuf/text_format.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 
 // Must be included last.
@@ -872,7 +871,7 @@ CommandLineInterface::MemoryOutputStream::~MemoryOutputStream() {
 
       // Now copy in the data.
       std::string::size_type data_pos = 0;
-      char* target_ptr = ::google::protobuf::string_as_array(target) + pos;
+      char* target_ptr = &(*target)[pos];
       while (data_pos < data_.size()) {
         // Copy indent.
         memcpy(target_ptr, indent_.data(), indent_.size());
@@ -889,8 +888,7 @@ CommandLineInterface::MemoryOutputStream::~MemoryOutputStream() {
       }
       UpdateMetadata(data_, pos, data_.size() + indent_size, indent_.size());
 
-      GOOGLE_CHECK_EQ(target_ptr,
+      GOOGLE_CHECK_EQ(target_ptr, &(*target)[pos] + data_.size() + indent_size);
-               ::google::protobuf::string_as_array(target) + pos + data_.size() + indent_size);
     }
   }
 }

src/google/protobuf/compiler/cpp/bootstrap_unittest.cc

@@ -56,7 +56,6 @@
 #include "absl/container/flat_hash_map.h"
 #include "absl/strings/substitute.h"
 #include "google/protobuf/compiler/cpp/helpers.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 namespace google {
 namespace protobuf {

src/google/protobuf/compiler/cpp/cpp_generator.h

@@ -1,3 +1,33 @@
+// Protocol Buffers - Google's data interchange format
+// Copyright 2008 Google Inc.  All rights reserved.
+// https://developers.google.com/protocol-buffers/
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+//     * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+//     * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
 #ifndef GOOGLE_PROTOBUF_COMPILER_CPP_CPP_GENERATOR_H_
 #define GOOGLE_PROTOBUF_COMPILER_CPP_CPP_GENERATOR_H_
 

src/google/protobuf/compiler/cpp/parse_function_generator.cc

@@ -271,6 +271,12 @@ struct NumToEntryTable {
 static NumToEntryTable MakeNumToEntryTable(
     const std::vector<const FieldDescriptor*>& field_descriptors);
 
+static int FieldNameDataSize(const std::vector<uint8_t>& data) {
+  // We add a +1 here to allow for a NUL termination character. It makes the
+  // codegen nicer.
+  return data.empty() ? 0 : data.size() + 1;
+}
+
 void ParseFunctionGenerator::GenerateDataDecls(io::Printer* printer) {
   if (!should_generate_tctable()) {
     return;
@@ -288,9 +294,7 @@ void ParseFunctionGenerator::GenerateDataDecls(io::Printer* printer) {
       "TcParseTable<$1$, $2$, $3$, $4$, $5$> _table_;\n",
       tc_table_info_->table_size_log2, ordered_fields_.size(),
       tc_table_info_->aux_entries.size(),
-      // We add a +1 here to allow for a NUL termination character. It makes the
+      FieldNameDataSize(tc_table_info_->field_name_data),
-      // codegen nicer.
-      tc_table_info_->field_name_data.size() + 1,
       field_num_to_entry_table.size16());
   if (should_generate_guarded_tctable()) {
     format.Outdent();
@@ -450,7 +454,7 @@ void ParseFunctionGenerator::GenerateTailCallTable(Formatter& format) {
       "{\n",
       tc_table_info_->table_size_log2, ordered_fields_.size(),
       tc_table_info_->aux_entries.size(),
-      tc_table_info_->field_name_data.size() + 1,  // See above for why +1
+      FieldNameDataSize(tc_table_info_->field_name_data),
       field_num_to_entry_table.size16());
   {
     auto table_scope = format.ScopedIndent();
@@ -608,12 +612,12 @@ void ParseFunctionGenerator::GenerateTailCallTable(Formatter& format) {
         format("}}, {{\n");
       }
     }  // ordered_fields_.empty()
-    {
+      {
-      // field_names[]
+        // field_names[]
-      auto field_name_scope = format.ScopedIndent();
+        auto field_name_scope = format.ScopedIndent();
-      GenerateFieldNames(format);
+        GenerateFieldNames(format);
-    }
+      }
-    format("}},\n");
+      format("}},\n");
   }
   format("};\n\n");  // _table_
 }
@@ -832,6 +836,11 @@ void ParseFunctionGenerator::GenerateFieldEntries(Formatter& format) {
 }
 
 void ParseFunctionGenerator::GenerateFieldNames(Formatter& format) {
+  if (tc_table_info_->field_name_data.empty()) {
+    // No names to output.
+    return;
+  }
+
   // We could just output the bytes directly, but we want it to look better than
   // that in the source code. Also, it is more efficient for compilation time to
   // have a literal string than an initializer list of chars.
@@ -854,8 +863,8 @@ void ParseFunctionGenerator::GenerateFieldNames(Formatter& format) {
   format("\"\n");
 
   // Then print each name in a line of its own
-  for (; sizes < sizes_end && sizes[0] != 0; p += *sizes++) {
+  for (; sizes < sizes_end; p += *sizes++) {
-    format("\"$1$\"\n", std::string(p, p + *sizes));
+    if (*sizes != 0) format("\"$1$\"\n", std::string(p, p + *sizes));
   }
 }
 

src/google/protobuf/compiler/cpp/unittest.inc

@@ -76,7 +76,6 @@
 #include "google/protobuf/io/coded_stream.h"
 #include "google/protobuf/io/zero_copy_stream_impl.h"
 #include "google/protobuf/test_util2.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 // Must be included last.
 #include "google/protobuf/port_def.inc"
@@ -664,7 +663,7 @@ TEST(GENERATED_MESSAGE_TEST_NAME, SerializationToArray) {
   TestUtil::SetAllFields(&message1);
   int size = message1.ByteSizeLong();
   data.resize(size);
-  uint8_t* start = reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&data));
+  uint8_t* start = reinterpret_cast<uint8_t*>(&data[0]);
   uint8_t* end = message1.SerializeWithCachedSizesToArray(start);
   EXPECT_EQ(size, end - start);
   EXPECT_TRUE(message2.ParseFromString(data));
@@ -678,8 +677,7 @@ TEST(GENERATED_MESSAGE_TEST_NAME, PackedFieldsSerializationToArray) {
   TestUtil::SetPackedFields(&packed_message1);
   int packed_size = packed_message1.ByteSizeLong();
   packed_data.resize(packed_size);
-  uint8_t* start =
+  uint8_t* start = reinterpret_cast<uint8_t*>(&packed_data[0]);
-      reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&packed_data));
   uint8_t* end = packed_message1.SerializeWithCachedSizesToArray(start);
   EXPECT_EQ(packed_size, end - start);
   EXPECT_TRUE(packed_message2.ParseFromString(packed_data));
@@ -696,7 +694,7 @@ TEST(GENERATED_MESSAGE_TEST_NAME, SerializationToStream) {
   data.resize(size);
   {
     // Allow the output stream to buffer only one byte at a time.
-    io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+    io::ArrayOutputStream array_stream(&data[0], size, 1);
     io::CodedOutputStream output_stream(&array_stream);
     message1.SerializeWithCachedSizes(&output_stream);
     EXPECT_FALSE(output_stream.HadError());
@@ -715,7 +713,7 @@ TEST(GENERATED_MESSAGE_TEST_NAME, PackedFieldsSerializationToStream) {
   data.resize(size);
   {
     // Allow the output stream to buffer only one byte at a time.
-    io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+    io::ArrayOutputStream array_stream(&data[0], size, 1);
     io::CodedOutputStream output_stream(&array_stream);
     message1.SerializeWithCachedSizes(&output_stream);
     EXPECT_FALSE(output_stream.HadError());
@@ -1841,7 +1839,7 @@ std::string data;
 message1.set_foo_int(123);
 int size = message1.ByteSizeLong();
 data.resize(size);
-uint8_t* start = reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&data));
+uint8_t* start = reinterpret_cast<uint8_t*>(&data[0]);
 uint8_t* end = message1.SerializeWithCachedSizesToArray(start);
 EXPECT_EQ(size, end - start);
 EXPECT_TRUE(message2.ParseFromString(data));
@@ -1855,7 +1853,7 @@ EXPECT_EQ(message2.foo_int(), 123);
     message1.set_foo_string("foo");
     int size = message1.ByteSizeLong();
     data.resize(size);
-    uint8_t* start = reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&data));
+    uint8_t* start = reinterpret_cast<uint8_t*>(&data[0]);
     uint8_t* end = message1.SerializeWithCachedSizesToArray(start);
     EXPECT_EQ(size, end - start);
     EXPECT_TRUE(message2.ParseFromString(data));
@@ -1870,7 +1868,7 @@ EXPECT_EQ(message2.foo_int(), 123);
     message1.set_foo_bytes("moo");
     int size = message1.ByteSizeLong();
     data.resize(size);
-    uint8_t* start = reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&data));
+    uint8_t* start = reinterpret_cast<uint8_t*>(&data[0]);
     uint8_t* end = message1.SerializeWithCachedSizesToArray(start);
     EXPECT_EQ(size, end - start);
     EXPECT_TRUE(message2.ParseFromString(data));
@@ -1884,7 +1882,7 @@ EXPECT_EQ(message2.foo_int(), 123);
     message1.set_foo_enum(UNITTEST::TestOneof2::FOO);
     int size = message1.ByteSizeLong();
     data.resize(size);
-    uint8_t* start = reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&data));
+    uint8_t* start = reinterpret_cast<uint8_t*>(&data[0]);
     uint8_t* end = message1.SerializeWithCachedSizesToArray(start);
     EXPECT_EQ(size, end - start);
     EXPECT_TRUE(message2.ParseFromString(data));
@@ -1898,7 +1896,7 @@ EXPECT_EQ(message2.foo_int(), 123);
     message1.mutable_foo_message()->set_moo_int(234);
     int size = message1.ByteSizeLong();
     data.resize(size);
-    uint8_t* start = reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&data));
+    uint8_t* start = reinterpret_cast<uint8_t*>(&data[0]);
     uint8_t* end = message1.SerializeWithCachedSizesToArray(start);
     EXPECT_EQ(size, end - start);
     EXPECT_TRUE(message2.ParseFromString(data));
@@ -1912,7 +1910,7 @@ EXPECT_EQ(message2.foo_int(), 123);
     message1.mutable_foogroup()->set_a(345);
     int size = message1.ByteSizeLong();
     data.resize(size);
-    uint8_t* start = reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&data));
+    uint8_t* start = reinterpret_cast<uint8_t*>(&data[0]);
     uint8_t* end = message1.SerializeWithCachedSizesToArray(start);
     EXPECT_EQ(size, end - start);
     EXPECT_TRUE(message2.ParseFromString(data));
@@ -1937,11 +1935,11 @@ data.resize(size);
 
 {
   // Allow the output stream to buffer only one byte at a time.
-  io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+    io::ArrayOutputStream array_stream(&data[0], size, 1);
-  io::CodedOutputStream output_stream(&array_stream);
+    io::CodedOutputStream output_stream(&array_stream);
-  message1.SerializeWithCachedSizes(&output_stream);
+    message1.SerializeWithCachedSizes(&output_stream);
-  EXPECT_FALSE(output_stream.HadError());
+    EXPECT_FALSE(output_stream.HadError());
-  EXPECT_EQ(size, output_stream.ByteCount());
+    EXPECT_EQ(size, output_stream.ByteCount());
 }
 
     EXPECT_TRUE(message2.ParseFromString(data));
@@ -1958,11 +1956,11 @@ data.resize(size);
 
     {
       // Allow the output stream to buffer only one byte at a time.
-      io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+    io::ArrayOutputStream array_stream(&data[0], size, 1);
-      io::CodedOutputStream output_stream(&array_stream);
+    io::CodedOutputStream output_stream(&array_stream);
-      message1.SerializeWithCachedSizes(&output_stream);
+    message1.SerializeWithCachedSizes(&output_stream);
-      EXPECT_FALSE(output_stream.HadError());
+    EXPECT_FALSE(output_stream.HadError());
-      EXPECT_EQ(size, output_stream.ByteCount());
+    EXPECT_EQ(size, output_stream.ByteCount());
     }
 
     EXPECT_TRUE(message2.ParseFromString(data));
@@ -1980,11 +1978,11 @@ data.resize(size);
 
     {
       // Allow the output stream to buffer only one byte at a time.
-      io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+    io::ArrayOutputStream array_stream(&data[0], size, 1);
-      io::CodedOutputStream output_stream(&array_stream);
+    io::CodedOutputStream output_stream(&array_stream);
-      message1.SerializeWithCachedSizes(&output_stream);
+    message1.SerializeWithCachedSizes(&output_stream);
-      EXPECT_FALSE(output_stream.HadError());
+    EXPECT_FALSE(output_stream.HadError());
-      EXPECT_EQ(size, output_stream.ByteCount());
+    EXPECT_EQ(size, output_stream.ByteCount());
     }
 
     EXPECT_TRUE(message2.ParseFromString(data));
@@ -2001,11 +1999,11 @@ data.resize(size);
 
     {
       // Allow the output stream to buffer only one byte at a time.
-      io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+    io::ArrayOutputStream array_stream(&data[0], size, 1);
-      io::CodedOutputStream output_stream(&array_stream);
+    io::CodedOutputStream output_stream(&array_stream);
-      message1.SerializeWithCachedSizes(&output_stream);
+    message1.SerializeWithCachedSizes(&output_stream);
-      EXPECT_FALSE(output_stream.HadError());
+    EXPECT_FALSE(output_stream.HadError());
-      EXPECT_EQ(size, output_stream.ByteCount());
+    EXPECT_EQ(size, output_stream.ByteCount());
     }
 
     EXPECT_TRUE(message2.ParseFromString(data));
@@ -2022,11 +2020,11 @@ data.resize(size);
 
     {
       // Allow the output stream to buffer only one byte at a time.
-      io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+    io::ArrayOutputStream array_stream(&data[0], size, 1);
-      io::CodedOutputStream output_stream(&array_stream);
+    io::CodedOutputStream output_stream(&array_stream);
-      message1.SerializeWithCachedSizes(&output_stream);
+    message1.SerializeWithCachedSizes(&output_stream);
-      EXPECT_FALSE(output_stream.HadError());
+    EXPECT_FALSE(output_stream.HadError());
-      EXPECT_EQ(size, output_stream.ByteCount());
+    EXPECT_EQ(size, output_stream.ByteCount());
     }
 
     EXPECT_TRUE(message2.ParseFromString(data));
@@ -2043,11 +2041,11 @@ data.resize(size);
 
     {
       // Allow the output stream to buffer only one byte at a time.
-      io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+    io::ArrayOutputStream array_stream(&data[0], size, 1);
-      io::CodedOutputStream output_stream(&array_stream);
+    io::CodedOutputStream output_stream(&array_stream);
-      message1.SerializeWithCachedSizes(&output_stream);
+    message1.SerializeWithCachedSizes(&output_stream);
-      EXPECT_FALSE(output_stream.HadError());
+    EXPECT_FALSE(output_stream.HadError());
-      EXPECT_EQ(size, output_stream.ByteCount());
+    EXPECT_EQ(size, output_stream.ByteCount());
     }
 
     EXPECT_TRUE(message2.ParseFromString(data));

src/google/protobuf/compiler/csharp/csharp_bootstrap_unittest.cc

@@ -51,7 +51,6 @@
 #include "google/protobuf/compiler/importer.h"
 #include "google/protobuf/descriptor.h"
 #include "google/protobuf/io/zero_copy_stream_impl.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 namespace google {
 namespace protobuf {

src/google/protobuf/compiler/java/java_generator.h

@@ -1,3 +1,33 @@
+// Protocol Buffers - Google's data interchange format
+// Copyright 2008 Google Inc.  All rights reserved.
+// https://developers.google.com/protocol-buffers/
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+//     * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+//     * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
 #ifndef GOOGLE_PROTOBUF_COMPILER_JAVA_JAVA_GENERATOR_H_
 #define GOOGLE_PROTOBUF_COMPILER_JAVA_JAVA_GENERATOR_H_
 

src/google/protobuf/compiler/objectivec/text_format_decode_data.cc

@@ -32,14 +32,13 @@
 
 #include "absl/strings/ascii.h"
 #include "absl/strings/escaping.h"
-#include "absl/strings/str_split.h"
 #include "absl/strings/str_replace.h"
+#include "absl/strings/str_split.h"
 #include "google/protobuf/compiler/code_generator.h"
 #include "google/protobuf/compiler/objectivec/names.h"
 #include "google/protobuf/io/coded_stream.h"
 #include "google/protobuf/io/printer.h"
 #include "google/protobuf/io/zero_copy_stream_impl.h"
-#include "google/protobuf/stubs/strutil.h"
 
 // NOTE: src/google/protobuf/compiler/plugin.cc makes use of cerr for some
 // error cases, so it seems to be ok to use as a back door for errors.

src/google/protobuf/compiler/plugin.proto

@@ -89,7 +89,6 @@ message CodeGeneratorRequest {
 
   // The version number of protocol compiler.
   optional Version compiler_version = 3;
-
 }
 
 // The plugin writes an encoded CodeGeneratorResponse to stdout.

src/google/protobuf/compiler/python/python_generator.h

@@ -1,3 +1,33 @@
+// Protocol Buffers - Google's data interchange format
+// Copyright 2008 Google Inc.  All rights reserved.
+// https://developers.google.com/protocol-buffers/
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+//     * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+//     * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
 #ifndef GOOGLE_PROTOBUF_COMPILER_PYTHON_PYTHON_GENERATOR_H_
 #define GOOGLE_PROTOBUF_COMPILER_PYTHON_PYTHON_GENERATOR_H_
 

src/google/protobuf/descriptor.proto

@@ -36,7 +36,6 @@
 // A valid .proto file can be translated directly to a FileDescriptorProto
 // without any other information (e.g. without reading its imports).
 
-
 syntax = "proto2";
 
 package google.protobuf;
@@ -134,7 +133,6 @@ message ExtensionRangeOptions {
   // The parser stores options it doesn't recognize here. See above.
   repeated UninterpretedOption uninterpreted_option = 999;
 
-
   // Clients can define custom options in extensions of this message. See above.
   extensions 1000 to max;
 }
@@ -310,7 +308,6 @@ message MethodDescriptorProto {
   optional bool server_streaming = 6 [default = false];
 }
 
-
 // ===================================================================
 // Options
 
@@ -351,7 +348,6 @@ message FileOptions {
   // domain names.
   optional string java_package = 1;
 
-
   // Controls the name of the wrapper Java class generated for the .proto file.
   // That class will always contain the .proto file's getDescriptor() method as
   // well as any top-level extensions defined in the .proto file.
@@ -378,7 +374,6 @@ message FileOptions {
   // This option has no effect on when used with the lite runtime.
   optional bool java_string_check_utf8 = 27 [default = false];
 
-
   // Generated classes can be optimized for speed or code size.
   enum OptimizeMode {
     SPEED = 1;         // Generate complete code for parsing, serialization,
@@ -395,9 +390,6 @@ message FileOptions {
   //   - Otherwise, the basename of the .proto file, without extension.
   optional string go_package = 11;
 
-
-
-
   // Should generic services be generated in each language?  "Generic" services
   // are not specific to any particular RPC system.  They are generated by the
   // main code generators in each language (without additional plugins).
@@ -423,7 +415,6 @@ message FileOptions {
   // only to generated classes for C++.
   optional bool cc_enable_arenas = 31 [default = true];
 
-
   // Sets the objective c class prefix which is prepended to all objective c
   // generated classes from this .proto. There is no default.
   optional string objc_class_prefix = 36;
@@ -456,7 +447,6 @@ message FileOptions {
   // determining the ruby package.
   optional string ruby_package = 45;
 
-
   // The parser stores options it doesn't recognize here.
   // See the documentation for the "Options" section above.
   repeated UninterpretedOption uninterpreted_option = 999;
@@ -528,7 +518,6 @@ message MessageOptions {
   reserved 8;  // javalite_serializable
   reserved 9;  // javanano_as_lite
 
-
   // The parser stores options it doesn't recognize here. See above.
   repeated UninterpretedOption uninterpreted_option = 999;
 
@@ -597,7 +586,6 @@ message FieldOptions {
   // call from multiple threads concurrently, while non-const methods continue
   // to require exclusive access.
   //
-  //
   // Note that implementations may choose not to check required fields within
   // a lazy sub-message.  That is, calling IsInitialized() on the outer message
   // may return true even if the inner message has missing required fields.
@@ -627,7 +615,6 @@ message FieldOptions {
   // For Google-internal migration only. Do not use.
   optional bool weak = 10 [default = false];
 
-
   // The parser stores options it doesn't recognize here. See above.
   repeated UninterpretedOption uninterpreted_option = 999;
 
@@ -731,7 +718,6 @@ message MethodOptions {
   extensions 1000 to max;
 }
 
-
 // A message representing a option the parser does not recognize. This only
 // appears in options protos created by the compiler::Parser class.
 // DescriptorPool resolves these when building Descriptor objects. Therefore,

src/google/protobuf/descriptor_database.cc

@@ -42,7 +42,6 @@
 #include "absl/strings/match.h"
 #include "absl/strings/str_replace.h"
 #include "google/protobuf/descriptor.pb.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 
 namespace google {

src/google/protobuf/duration.proto

@@ -99,7 +99,6 @@ option objc_class_prefix = "GPB";
 // be expressed in JSON format as "3.000000001s", and 3 seconds and 1
 // microsecond should be expressed in JSON format as "3.000001s".
 //
-//
 message Duration {
   // Signed seconds of the span of time. Must be from -315,576,000,000
   // to +315,576,000,000 inclusive. Note: these bounds are computed from:

src/google/protobuf/extension_set_unittest.cc

@@ -53,7 +53,6 @@
 #include "absl/strings/match.h"
 #include "google/protobuf/test_util.h"
 #include "google/protobuf/test_util2.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 #include "google/protobuf/stubs/strutil.h"
 
@@ -552,7 +551,7 @@ TEST(ExtensionSetTest, SerializationToArray) {
   size_t size = source.ByteSizeLong();
   std::string data;
   data.resize(size);
-  uint8_t* target = reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&data));
+  uint8_t* target = reinterpret_cast<uint8_t*>(&data[0]);
   uint8_t* end = source.SerializeWithCachedSizesToArray(target);
   EXPECT_EQ(size, end - target);
   EXPECT_TRUE(destination.ParseFromString(data));
@@ -574,7 +573,7 @@ TEST(ExtensionSetTest, SerializationToStream) {
   std::string data;
   data.resize(size);
   {
-    io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+    io::ArrayOutputStream array_stream(&data[0], size, 1);
     io::CodedOutputStream output_stream(&array_stream);
     source.SerializeWithCachedSizes(&output_stream);
     ASSERT_FALSE(output_stream.HadError());
@@ -596,7 +595,7 @@ TEST(ExtensionSetTest, PackedSerializationToArray) {
   size_t size = source.ByteSizeLong();
   std::string data;
   data.resize(size);
-  uint8_t* target = reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&data));
+  uint8_t* target = reinterpret_cast<uint8_t*>(&data[0]);
   uint8_t* end = source.SerializeWithCachedSizesToArray(target);
   EXPECT_EQ(size, end - target);
   EXPECT_TRUE(destination.ParseFromString(data));
@@ -618,7 +617,7 @@ TEST(ExtensionSetTest, PackedSerializationToStream) {
   std::string data;
   data.resize(size);
   {
-    io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+    io::ArrayOutputStream array_stream(&data[0], size, 1);
     io::CodedOutputStream output_stream(&array_stream);
     source.SerializeWithCachedSizes(&output_stream);
     ASSERT_FALSE(output_stream.HadError());

src/google/protobuf/generated_message_reflection.cc

@@ -3095,7 +3095,7 @@ const internal::TcParseTableBase* Reflection::CreateTcParseTableForMessageSet()
   // Create a dummy table that only exists to make TcParser::ParseLoop jump
   // into the reflective parse loop.
 
-  using Table = internal::TcParseTable<0, 0, 0, 1, 1>;
+  using Table = internal::TcParseTable<0, 0, 0, 0, 1>;
   // We use `operator new` here because the destruction will be done with
   // `operator delete` unconditionally.
   void* p = ::operator new(sizeof(Table));

src/google/protobuf/generated_message_tctable_decl.h

@@ -347,7 +347,7 @@ struct TcParseTable {
   // Entries for all fields:
   std::array<TcParseTableBase::FieldEntry, kNumFieldEntries> field_entries;
   std::array<TcParseTableBase::FieldAux, kNumFieldAux> aux_entries;
-  std::array<char, kNameTableSize> field_names;
+  std::array<char, kNameTableSize == 0 ? 1 : kNameTableSize> field_names;
 };
 
 // Partial specialization: if there are no aux entries, there will be no array.
@@ -363,7 +363,7 @@ struct TcParseTable<kFastTableSizeLog2, kNumFieldEntries, 0, kNameTableSize,
       fast_entries;
   std::array<uint16_t, kFieldLookupSize> field_lookup_table;
   std::array<TcParseTableBase::FieldEntry, kNumFieldEntries> field_entries;
-  std::array<char, kNameTableSize> field_names;
+  std::array<char, kNameTableSize == 0 ? 1 : kNameTableSize> field_names;
 };
 
 // Partial specialization: if there are no fields at all, then we can save space
@@ -375,7 +375,7 @@ struct TcParseTable<0, 0, 0, kNameTableSize, kFieldLookupSize> {
   // The fast parsing loop will always use this entry, so it must be present.
   std::array<TcParseTableBase::FastFieldEntry, 1> fast_entries;
   std::array<uint16_t, kFieldLookupSize> field_lookup_table;
-  std::array<char, kNameTableSize> field_names;
+  std::array<char, kNameTableSize == 0 ? 1 : kNameTableSize> field_names;
 };
 
 static_assert(std::is_standard_layout<TcParseTable<1>>::value,

src/google/protobuf/generated_message_tctable_gen.cc

@@ -420,10 +420,9 @@ absl::string_view FieldNameForTable(
       case field_layout::kTvUtf8:
       case field_layout::kTvUtf8Debug:
         return field->name();
-        break;
     }
   }
-  return "?";
+  return "";
 }
 
 std::vector<uint8_t> GenerateFieldNames(
@@ -431,6 +430,20 @@ std::vector<uint8_t> GenerateFieldNames(
     const std::vector<TailCallTableInfo::FieldEntryInfo>& entries) {
   static constexpr int kMaxNameLength = 255;
   std::vector<uint8_t> out;
+
+  bool found_needed_name = false;
+  for (const auto& entry : entries) {
+    if (!FieldNameForTable(entry).empty()) {
+      found_needed_name = true;
+      break;
+    }
+  }
+
+  // No names needed. Omit the whole table.
+  if (!found_needed_name) {
+    return out;
+  }
+
   // First, we output the size of each string, as an unsigned byte. The first
   // string is the message name.
   int count = 1;

src/google/protobuf/generated_message_tctable_impl.h

@@ -610,6 +610,7 @@ class PROTOBUF_EXPORT TcParser final {
 
   // For FindFieldEntry tests:
   friend class FindFieldEntryTest;
+  friend struct ParseFunctionGeneratorTestPeer;
   static constexpr const uint32_t kMtSmallScanSize = 4;
 
   // Mini parsing:

src/google/protobuf/io/tokenizer.cc

@@ -96,7 +96,6 @@
 #include "absl/strings/str_format.h"
 #include "google/protobuf/io/strtod.h"
 #include "google/protobuf/io/zero_copy_stream.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 // Must be included last.
 #include "google/protobuf/port_def.inc"

src/google/protobuf/io/zero_copy_stream_impl.cc

@@ -47,7 +47,6 @@
 #include "google/protobuf/stubs/logging.h"
 #include "google/protobuf/io/io_win32.h"
 #include "google/protobuf/io/zero_copy_stream_impl.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 
 namespace google {

src/google/protobuf/io/zero_copy_stream_impl_lite.h

@@ -53,7 +53,6 @@
 #include "google/protobuf/stubs/common.h"
 #include "google/protobuf/io/zero_copy_stream.h"
 #include "google/protobuf/port.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 
 // Must be included last.
@@ -387,24 +386,10 @@ class PROTOBUF_EXPORT LimitingInputStream PROTOBUF_FUTURE_FINAL
 
 // ===================================================================
 
-// mutable_string_data() and as_string_data() are workarounds to improve
-// the performance of writing new data to an existing string.  Unfortunately
-// the methods provided by the string class are suboptimal, and using memcpy()
-// is mildly annoying because it requires its pointer args to be non-NULL even
-// if we ask it to copy 0 bytes.  Furthermore, string_as_array() has the
-// property that it always returns NULL if its arg is the empty string, exactly
-// what we want to avoid if we're using it in conjunction with memcpy()!
-// With C++11, the desired memcpy() boils down to memcpy(..., &(*s)[0], size),
-// where s is a string*.  Without C++11, &(*s)[0] is not guaranteed to be safe,
-// so we use string_as_array(), and live with the extra logic that tests whether
-// *s is empty.
-
 // Return a pointer to mutable characters underlying the given string.  The
 // return value is valid until the next time the string is resized.  We
 // trust the caller to treat the return value as an array of length s->size().
 inline char* mutable_string_data(std::string* s) {
-  // This should be simpler & faster than string_as_array() because the latter
-  // is guaranteed to return NULL when *s is empty, so it has to check for that.
   return &(*s)[0];
 }
 

src/google/protobuf/lite_unittest.cc

@@ -675,8 +675,7 @@ TEST(Lite, AllLite28) {
     MapLiteTestUtil::SetMapFields(&message1);
     size_t size = message1.ByteSizeLong();
     data.resize(size);
-    ::uint8_t* start =
+    ::uint8_t* start = reinterpret_cast<::uint8_t*>(&data[0]);
-        reinterpret_cast<::uint8_t*>(::google::protobuf::string_as_array(&data));
     ::uint8_t* end = message1.SerializeWithCachedSizesToArray(start);
     EXPECT_EQ(size, end - start);
     EXPECT_TRUE(message2.ParseFromString(data));
@@ -696,7 +695,7 @@ TEST(Lite, AllLite29) {
     data.resize(size);
     {
       // Allow the output stream to buffer only one byte at a time.
-      io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+      io::ArrayOutputStream array_stream(&data[0], size, 1);
       io::CodedOutputStream output_stream(&array_stream);
       message1.SerializeWithCachedSizes(&output_stream);
       EXPECT_FALSE(output_stream.HadError());

src/google/protobuf/map_test.inc

@@ -2692,7 +2692,7 @@ TEST(GeneratedMapFieldTest, SerializationToArray) {
   MapTestUtil::SetMapFields(&message1);
   size_t size = message1.ByteSizeLong();
   data.resize(size);
-  uint8_t* start = reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&data));
+  uint8_t* start = reinterpret_cast<uint8_t*>(&data[0]);
   uint8_t* end = message1.SerializeWithCachedSizesToArray(start);
   EXPECT_EQ(size, end - start);
   EXPECT_TRUE(message2.ParseFromString(data));
@@ -2708,7 +2708,7 @@ TEST(GeneratedMapFieldTest, SerializationToStream) {
   data.resize(size);
   {
     // Allow the output stream to buffer only one byte at a time.
-    io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&data), size, 1);
+    io::ArrayOutputStream array_stream(&data[0], size, 1);
     io::CodedOutputStream output_stream(&array_stream);
     message1.SerializeWithCachedSizes(&output_stream);
     EXPECT_FALSE(output_stream.HadError());
@@ -3739,7 +3739,7 @@ static std::string DeterministicSerializationWithSerializePartialToCodedStream(
     const T& t) {
   const size_t size = t.ByteSizeLong();
   std::string result(size, '\0');
-  io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&result), size);
+  io::ArrayOutputStream array_stream(&result[0], size);
   io::CodedOutputStream output_stream(&array_stream);
   output_stream.SetSerializationDeterministic(true);
   t.SerializePartialToCodedStream(&output_stream);
@@ -3753,7 +3753,7 @@ static std::string DeterministicSerializationWithSerializeToCodedStream(
     const T& t) {
   const size_t size = t.ByteSizeLong();
   std::string result(size, '\0');
-  io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&result), size);
+  io::ArrayOutputStream array_stream(&result[0], size);
   io::CodedOutputStream output_stream(&array_stream);
   output_stream.SetSerializationDeterministic(true);
   t.SerializeToCodedStream(&output_stream);
@@ -3766,7 +3766,7 @@ template <typename T>
 static std::string DeterministicSerialization(const T& t) {
   const size_t size = t.ByteSizeLong();
   std::string result(size, '\0');
-  io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&result), size);
+  io::ArrayOutputStream array_stream(&result[0], size);
   {
     io::CodedOutputStream output_stream(&array_stream);
     output_stream.SetSerializationDeterministic(true);

src/google/protobuf/map_unittest.proto

@@ -76,7 +76,6 @@ message TestSameTypeMap {
   map<int32, int32> map2 = 2;
 }
 
-
 enum MapEnum {
   MAP_ENUM_FOO = 0;
   MAP_ENUM_BAR = 1;

src/google/protobuf/message.cc

@@ -61,7 +61,6 @@
 #include "google/protobuf/unknown_field_set.h"
 #include "google/protobuf/wire_format.h"
 #include "google/protobuf/wire_format_lite.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 
 // Must be included last.

src/google/protobuf/repeated_field_unittest.cc

@@ -56,7 +56,6 @@
 #include "absl/strings/cord.h"
 #include "absl/strings/str_cat.h"
 #include "google/protobuf/stubs/strutil.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 // Must be included last.
 #include "google/protobuf/port_def.inc"

src/google/protobuf/stubs/BUILD.bazel

@@ -26,7 +26,6 @@ cc_library(
         "platform_macros.h",
         "port.h",
         "status_macros.h",
-        "stl_util.h",
         "strutil.h",
     ],
     copts = COPTS,
@@ -56,7 +55,6 @@ cc_library(
         "platform_macros.h",
         "port.h",
         "status_macros.h",
-        "stl_util.h",
         "strutil.h",
     ],
     deps = [

src/google/protobuf/stubs/callback.h

@@ -1,3 +1,33 @@
+// Protocol Buffers - Google's data interchange format
+// Copyright 2008 Google Inc.  All rights reserved.
+// https://developers.google.com/protocol-buffers/
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+//     * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+//     * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
 #ifndef GOOGLE_PROTOBUF_STUBS_CALLBACK_H_
 #define GOOGLE_PROTOBUF_STUBS_CALLBACK_H_
 

src/google/protobuf/stubs/stl_util.h

@@ -1,68 +0,0 @@
-// Protocol Buffers - Google's data interchange format
-// Copyright 2008 Google Inc.  All rights reserved.
-// https://developers.google.com/protocol-buffers/
-//
-// Redistribution and use in source and binary forms, with or without
-// modification, are permitted provided that the following conditions are
-// met:
-//
-//     * Redistributions of source code must retain the above copyright
-// notice, this list of conditions and the following disclaimer.
-//     * Redistributions in binary form must reproduce the above
-// copyright notice, this list of conditions and the following disclaimer
-// in the documentation and/or other materials provided with the
-// distribution.
-//     * Neither the name of Google Inc. nor the names of its
-// contributors may be used to endorse or promote products derived from
-// this software without specific prior written permission.
-//
-// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-// from google3/util/gtl/stl_util.h
-
-#ifndef GOOGLE_PROTOBUF_STUBS_STL_UTIL_H__
-#define GOOGLE_PROTOBUF_STUBS_STL_UTIL_H__
-
-#include <algorithm>
-
-#include "google/protobuf/stubs/common.h"
-
-// Must be last.
-#include "google/protobuf/port_def.inc"  // NOLINT
-
-namespace google {
-namespace protobuf {
-
-// Return a mutable char* pointing to a string's internal buffer,
-// which may not be null-terminated. Writing through this pointer will
-// modify the string.
-//
-// string_as_array(&str)[i] is valid for 0 <= i < str.size() until the
-// next call to a string method that invalidates iterators.
-//
-// As of 2006-04, there is no standard-blessed way of getting a
-// mutable reference to a string's internal buffer. However, issue 530
-// (http://www.open-std.org/JTC1/SC22/WG21/docs/lwg-active.html#530)
-// proposes this as the method. According to Matt Austern, this should
-// already work on all current implementations.
-inline char* string_as_array(std::string* str) {
-  // DO NOT USE const_cast<char*>(str->data())! See the unittest for why.
-  return str->empty() ? nullptr : &*str->begin();
-}
-
-}  // namespace protobuf
-}  // namespace google
-
-#include "google/protobuf/port_undef.inc"  // NOLINT
-
-#endif  // GOOGLE_PROTOBUF_STUBS_STL_UTIL_H__

src/google/protobuf/stubs/strutil.cc

@@ -44,7 +44,6 @@
 #include "absl/strings/ascii.h"
 #include "absl/strings/string_view.h"
 #include "google/protobuf/stubs/logging.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 #ifdef _WIN32
 // MSVC has only _snprintf, not snprintf.
@@ -671,9 +670,8 @@ void Base64EscapeInternal(const unsigned char *src, int szsrc,
                           const absl::string_view base64_chars) {
   const int calc_escaped_size = CalculateBase64EscapedLen(szsrc, do_padding);
   dest->resize(calc_escaped_size);
-  const int escaped_len =
+  const int escaped_len = Base64EscapeInternal(
-      Base64EscapeInternal(src, szsrc, string_as_array(dest), dest->size(),
+      src, szsrc, &(*dest)[0], dest->size(), base64_chars, do_padding);
-                           base64_chars, do_padding);
   GOOGLE_DCHECK_EQ(calc_escaped_size, escaped_len);
   dest->erase(escaped_len);
 }

src/google/protobuf/stubs/strutil_unittest.cc

@@ -35,7 +35,6 @@
 #include <gtest/gtest.h>
 #include <locale.h>
 
-#include "google/protobuf/stubs/stl_util.h"
 #include "google/protobuf/testing/googletest.h"
 
 #ifdef _WIN32

src/google/protobuf/text_format.cc

@@ -67,7 +67,6 @@
 #include "google/protobuf/repeated_field.h"
 #include "google/protobuf/unknown_field_set.h"
 #include "google/protobuf/wire_format_lite.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 // Must be included last.
 #include "google/protobuf/port_def.inc"

src/google/protobuf/timestamp.proto

@@ -90,7 +90,6 @@ option objc_class_prefix = "GPB";
 //     Timestamp timestamp = Timestamp.newBuilder().setSeconds(millis / 1000)
 //         .setNanos((int) ((millis % 1000) * 1000000)).build();
 //
-//
 // Example 5: Compute Timestamp from Java `Instant.now()`.
 //
 //     Instant now = Instant.now();
@@ -99,7 +98,6 @@ option objc_class_prefix = "GPB";
 //         Timestamp.newBuilder().setSeconds(now.getEpochSecond())
 //             .setNanos(now.getNano()).build();
 //
-//
 // Example 6: Compute Timestamp from current time in Python.
 //
 //     timestamp = Timestamp()
@@ -132,7 +130,6 @@ option objc_class_prefix = "GPB";
 // http://www.joda.org/joda-time/apidocs/org/joda/time/format/ISODateTimeFormat.html#dateTime%2D%2D
 // ) to obtain a formatter capable of generating timestamps in this format.
 //
-//
 message Timestamp {
   // Represents seconds of UTC time since Unix epoch
   // 1970-01-01T00:00:00Z. Must be from 0001-01-01T00:00:00Z to

src/google/protobuf/unittest.proto

@@ -889,7 +889,6 @@ message TestRequiredOneof {
   }
 }
 
-
 // Test messages for packed fields
 
 message TestPackedTypes {
@@ -1070,7 +1069,6 @@ message TestMessageSize {
   optional int64 m6 = 6;
 }
 
-
 // Test that RPC services work.
 message FooRequest  {}
 message FooResponse {}
@@ -1083,7 +1081,6 @@ service TestService {
   rpc Bar(BarRequest) returns (BarResponse);
 }
 
-
 message BarRequest  {}
 message BarResponse {}
 
@@ -1437,7 +1434,6 @@ message TestVerifyBigFieldNumberUint32 {
   optional Nested optional_nested = 1;
 }
 
-
 // This message contains different kind of enums to exercise the different
 // parsers in table-driven.
 message EnumParseTester {
@@ -1565,4 +1561,3 @@ message StringParseTester {
   repeated string repeated_string_midfield = 1002;
   repeated string repeated_string_hifield = 1000002;
 };
-

src/google/protobuf/unittest_import.proto

@@ -64,7 +64,6 @@ enum ImportEnum {
   IMPORT_BAZ = 9;
 }
 
-
 // To use an enum in a map, it must has the first value as 0.
 enum ImportEnumForMap {
   UNKNOWN = 0;

src/google/protobuf/unknown_field_set_unittest.cc

@@ -59,7 +59,6 @@
 #include "absl/time/clock.h"
 #include "absl/time/time.h"
 #include "google/protobuf/test_util.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 
 namespace google {
@@ -207,15 +206,13 @@ TEST_F(UnknownFieldSetTest, SerializeFastAndSlowAreEquivalent) {
   slow_buffer.resize(size);
   fast_buffer.resize(size);
 
-  uint8_t* target =
+  uint8_t* target = reinterpret_cast<uint8_t*>(&fast_buffer[0]);
-      reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&fast_buffer));
   uint8_t* result = WireFormat::SerializeUnknownFieldsToArray(
       empty_message_.unknown_fields(), target);
   EXPECT_EQ(size, result - target);
 
   {
-    io::ArrayOutputStream raw_stream(::google::protobuf::string_as_array(&slow_buffer), size,
+    io::ArrayOutputStream raw_stream(&slow_buffer[0], size, 1);
-                                     1);
     io::CodedOutputStream output_stream(&raw_stream);
     WireFormat::SerializeUnknownFields(empty_message_.unknown_fields(),
                                        &output_stream);

src/google/protobuf/util/json_format.proto

@@ -41,7 +41,6 @@ package protobuf_unittest;
 import "google/protobuf/any.proto";
 import "google/protobuf/struct.proto";
 
-
 message TestFlagsAndStrings {
   required int32 A = 1;
   repeated group RepeatedGroup = 2 {
@@ -96,7 +95,6 @@ message TestNumbers {
   optional uint32 f = 6;
 }
 
-
 message TestCamelCase {
   optional string normal_field = 1;
   optional int32 CAPITAL_FIELD = 2;

src/google/protobuf/well_known_types_unittest.cc

@@ -30,7 +30,6 @@
 #include <gtest/gtest.h>
 
 #include "google/protobuf/stubs/common.h"
-#include "google/protobuf/stubs/stl_util.h"
 #include "google/protobuf/testing/googletest.h"
 #include "google/protobuf/unittest_well_known_types.pb.h"
 

src/google/protobuf/wire_format_unittest.inc

@@ -49,7 +49,6 @@
 #include "absl/strings/match.h"
 #include "google/protobuf/dynamic_message.h"
 #include "google/protobuf/test_util2.h"
-#include "google/protobuf/stubs/stl_util.h"
 
 // clang-format off
 #include "google/protobuf/port_def.inc"
@@ -494,16 +493,14 @@ TEST(WireFormatTest, SerializeMessageSetVariousWaysAreEqual) {
 
   // Serialize to flat array
   {
-    uint8_t* target =
+    uint8_t* target = reinterpret_cast<uint8_t*>(&flat_data[0]);
-        reinterpret_cast<uint8_t*>(::google::protobuf::string_as_array(&flat_data));
     uint8_t* end = message_set.SerializeWithCachedSizesToArray(target);
     EXPECT_EQ(size, end - target);
   }
 
   // Serialize to buffer
   {
-    io::ArrayOutputStream array_stream(::google::protobuf::string_as_array(&stream_data), size,
+    io::ArrayOutputStream array_stream(&stream_data[0], size, 1);
-                                       1);
     io::CodedOutputStream output_stream(&array_stream);
     message_set.SerializeWithCachedSizes(&output_stream);
     ASSERT_FALSE(output_stream.HadError());