1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
|
// Copyright (c) 2012 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#ifndef BASE_MOVE_H_
#define BASE_MOVE_H_
#include "base/compiler_specific.h"
// Macro with the boilerplate that makes a type move-only in C++03.
//
// USAGE
//
// This macro should be used instead of DISALLOW_COPY_AND_ASSIGN to create
// a "move-only" type. Unlike DISALLOW_COPY_AND_ASSIGN, this macro should be
// the first line in a class declaration.
//
// A class using this macro must call .Pass() (or somehow be an r-value already)
// before it can be:
//
// * Passed as a function argument
// * Used as the right-hand side of an assignment
// * Returned from a function
//
// Each class will still need to define their own "move constructor" and "move
// operator=" to make this useful. Here's an example of the macro, the move
// constructor, and the move operator= from the scoped_ptr class:
//
// template <typename T>
// class scoped_ptr {
// MOVE_ONLY_TYPE_FOR_CPP_03(scoped_ptr, RValue)
// public:
// scoped_ptr(RValue& other) : ptr_(other.release()) { }
// scoped_ptr& operator=(RValue& other) {
// swap(other);
// return *this;
// }
// };
//
// Note that the constructor must NOT be marked explicit.
//
// For consistency, the second parameter to the macro should always be RValue
// unless you have a strong reason to do otherwise. It is only exposed as a
// macro parameter so that the move constructor and move operator= don't look
// like they're using a phantom type.
//
//
// HOW THIS WORKS
//
// For a thorough explanation of this technique, see:
//
// http://en.wikibooks.org/wiki/More_C%2B%2B_Idioms/Move_Constructor
//
// The summary is that we take advantage of 2 properties:
//
// 1) non-const references will not bind to r-values.
// 2) C++ can apply one user-defined conversion when initializing a
// variable.
//
// The first lets us disable the copy constructor and assignment operator
// by declaring private version of them with a non-const reference parameter.
//
// For l-values, direct initialization still fails like in
// DISALLOW_COPY_AND_ASSIGN because the copy constructor and assignment
// operators are private.
//
// For r-values, the situation is different. The copy constructor and
// assignment operator are not viable due to (1), so we are trying to call
// a non-existent constructor and non-existing operator= rather than a private
// one. Since we have not committed an error quite yet, we can provide an
// alternate conversion sequence and a constructor. We add
//
// * a private struct named "RValue"
// * a user-defined conversion "operator RValue()"
// * a "move constructor" and "move operator=" that take the RValue& as
// their sole parameter.
//
// Only r-values will trigger this sequence and execute our "move constructor"
// or "move operator=." L-values will match the private copy constructor and
// operator= first giving a "private in this context" error. This combination
// gives us a move-only type.
//
// For signaling a destructive transfer of data from an l-value, we provide a
// method named Pass() which creates an r-value for the current instance
// triggering the move constructor or move operator=.
//
// Other ways to get r-values is to use the result of an expression like a
// function call.
//
// Here's an example with comments explaining what gets triggered where:
//
// class Foo {
// MOVE_ONLY_TYPE_FOR_CPP_03(Foo, RValue);
//
// public:
// ... API ...
// Foo(RValue other); // Move constructor.
// Foo& operator=(RValue rhs); // Move operator=
// };
//
// Foo MakeFoo(); // Function that returns a Foo.
//
// Foo f;
// Foo f_copy(f); // ERROR: Foo(Foo&) is private in this context.
// Foo f_assign;
// f_assign = f; // ERROR: operator=(Foo&) is private in this context.
//
//
// Foo f(MakeFoo()); // R-value so alternate conversion executed.
// Foo f_copy(f.Pass()); // R-value so alternate conversion executed.
// f = f_copy.Pass(); // R-value so alternate conversion executed.
//
//
// IMPLEMENTATION SUBTLETIES WITH RValue
//
// The RValue struct is just a container for a pointer back to the original
// object. It should only ever be created as a temporary, and no external
// class should ever declare it or use it in a parameter.
//
// It is tempting to want to use the RValue type in function parameters, but
// excluding the limited usage here for the move constructor and move
// operator=, doing so would mean that the function could take both r-values
// and l-values equially which is unexpected. See COMPARED To Boost.Move for
// more details.
//
// An alternate, and incorrect, implementation of the RValue class used by
// Boost.Move makes RValue a fieldless child of the move-only type. RValue&
// is then used in place of RValue in the various operators. The RValue& is
// "created" by doing *reinterpret_cast<RValue*>(this). This has the appeal
// of never creating a temporary RValue struct even with optimizations
// disabled. Also, by virtue of inheritance you can treat the RValue
// reference as if it were the move-only type itself. Unfortunately,
// using the result of this reinterpret_cast<> is actually undefined behavior
// due to C++98 5.2.10.7. In certain compilers (e.g., NaCl) the optimizer
// will generate non-working code.
//
// In optimized builds, both implementations generate the same assembly so we
// choose the one that adheres to the standard.
//
//
// WHY HAVE typedef void MoveOnlyTypeForCPP03
//
// Callback<>/Bind() needs to understand movable-but-not-copyable semantics
// to call .Pass() appropriately when it is expected to transfer the value.
// The cryptic typedef MoveOnlyTypeForCPP03 is added to make this check
// easy and automatic in helper templates for Callback<>/Bind().
// See IsMoveOnlyType template and its usage in base/callback_internal.h
// for more details.
//
//
// COMPARED TO C++11
//
// In C++11, you would implement this functionality using an r-value reference
// and our .Pass() method would be replaced with a call to std::move().
//
// This emulation also has a deficiency where it uses up the single
// user-defined conversion allowed by C++ during initialization. This can
// cause problems in some API edge cases. For instance, in scoped_ptr, it is
// impossible to make a function "void Foo(scoped_ptr<Parent> p)" accept a
// value of type scoped_ptr<Child> even if you add a constructor to
// scoped_ptr<> that would make it look like it should work. C++11 does not
// have this deficiency.
//
//
// COMPARED TO Boost.Move
//
// Our implementation similar to Boost.Move, but we keep the RValue struct
// private to the move-only type, and we don't use the reinterpret_cast<> hack.
//
// In Boost.Move, RValue is the boost::rv<> template. This type can be used
// when writing APIs like:
//
// void MyFunc(boost::rv<Foo>& f)
//
// that can take advantage of rv<> to avoid extra copies of a type. However you
// would still be able to call this version of MyFunc with an l-value:
//
// Foo f;
// MyFunc(f); // Uh oh, we probably just destroyed |f| w/o calling Pass().
//
// unless someone is very careful to also declare a parallel override like:
//
// void MyFunc(const Foo& f)
//
// that would catch the l-values first. This was declared unsafe in C++11 and
// a C++11 compiler will explicitly fail MyFunc(f). Unfortunately, we cannot
// ensure this in C++03.
//
// Since we have no need for writing such APIs yet, our implementation keeps
// RValue private and uses a .Pass() method to do the conversion instead of
// trying to write a version of "std::move()." Writing an API like std::move()
// would require the RValue struct to be public.
//
//
// CAVEATS
//
// If you include a move-only type as a field inside a class that does not
// explicitly declare a copy constructor, the containing class's implicit
// copy constructor will change from Containing(const Containing&) to
// Containing(Containing&). This can cause some unexpected errors.
//
// http://llvm.org/bugs/show_bug.cgi?id=11528
//
// The workaround is to explicitly declare your copy constructor.
//
#define MOVE_ONLY_TYPE_FOR_CPP_03(type, rvalue_type) \
private: \
struct rvalue_type { \
explicit rvalue_type(type* object) : object(object) {} \
type* object; \
}; \
type(type&); \
void operator=(type&); \
public: \
operator rvalue_type() { return rvalue_type(this); } \
type Pass() WARN_UNUSED_RESULT { return type(rvalue_type(this)); } \
typedef void MoveOnlyTypeForCPP03; \
private:
#define MOVE_ONLY_TYPE_WITH_MOVE_CONSTRUCTOR_FOR_CPP_03(type) \
private: \
type(const type&); \
void operator=(const type&); \
public: \
type&& Pass() WARN_UNUSED_RESULT { return static_cast<type&&>(*this); } \
typedef void MoveOnlyTypeForCPP03; \
private:
#define TYPE_WITH_MOVE_CONSTRUCTOR_FOR_CPP_03(type) \
public: \
type&& Pass() WARN_UNUSED_RESULT { return static_cast<type&&>(*this); } \
private:
#endif // BASE_MOVE_H_
|