-
Notifications
You must be signed in to change notification settings - Fork 1
/
p3234r1.html
213 lines (213 loc) · 7.61 KB
/
p3234r1.html
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
<!DOCTYPE html>
<html lang="en">
<head>
<title>Utility to check if a pointer is in a given range</title>
<meta charset="utf-8">
<meta name="author" content="Glen Joseph Fernandes">
<meta name="viewport" content="width=device-width, initial-scale=1">
<style>
ins { background-color: #CCFFCC; }
.insert { background-color: #CCFFCC; }
</style>
</head>
<body>
<p><strong>Document Number:</strong> P3234R1<br>
<strong>Date:</strong> 2024-04-16<br>
<strong>Project:</strong> Programming Language C++<br>
<strong>Audience:</strong> LEWG, EWG<br>
<strong>Author:</strong> Glen Joseph Fernandes
(<a href="mailto:[email protected]">[email protected]</a>)</p>
<h1>Utility to check if a pointer is in a given range</h1>
<p>This paper proposes adding a new function template
<code>pointer_in_range</code>, a utility that can check if a pointer is in a
given range, and can be used in a constant expression.</p>
<h2>Changes in Revision 1</h2>
<p>Added rationale for the choice of two pointer parameters instead of a single
range parameter.</p>
<h2>Motivation</h2>
<p>Library authors often need this functionality and implement it themselves.
A solution in the standard library would be more convenient, portable, optimal,
and correct.</p>
<h3>Common usage</h3>
<p>A common use is determining if string ranges overlap, to be able to use a
fast copy operation:</p>
<blockquote>
<code>if (!pointer_in_range(ptr, data_, data_ + size_)) {<br>
std::copy(ptr, ptr + size, data_);<br>
}</code>
</blockquote>
<p>Another use is allocators that first use automatic storage, falling back to
dynamic allocation:</p>
<blockquote>
<code>if (!pointer_in_range(ptr, store_, store_ + size_)) {<br>
::operator delete(ptr);<br>
}</code>
</blockquote>
<h3>Existing practice</h3>
<p>This function appears in some form in many projects, including large library
collections such as:</p>
<ul>
<li>libc++: <code>__is_pointer_in_range</code></li>
<li>Qt: <code>q_points_into_range</code></li>
<li>Boost: <code>pointer_in_range</code>, <code>ptr_in_range</code></li>
</ul>
<p>Other libraries contain the same check inline even if they do not define a
function for it, such as:</p>
<ul>
<li>libstdc++: <code>basic_string::_M_disjunct</code></li>
<li>BSL: <code>basic_string::privateReplaceRaw</code></li>
<li>EASTL: <code>basic_string::replace</code></li>
</ul>
<h3>Imperfect solutions</h3>
<p>Users can not implement this perfectly. For example, the following solution
is not correct when the built-in operators for pointers do not yield a strict
total order:</p>
<blockquote>
<code>template<class T><br>
bool pointer_in_range(T* ptr, T* begin, T* end)<br>
{<br>
return begin <= ptr && ptr < end;<br>
}</code>
</blockquote>
<p>The following solution uses comparisons consistent with the
implementation-defined total order but is still not correct when two arrays
of <code>T</code> may be interleaved:</p>
<blockquote>
<code>template<class T><br>
bool pointer_in_range(T* ptr, T* begin, T* end)<br>
{<br>
return std::less_equal<>()(begin, ptr) &&
std::less<>()(ptr, end);<br>
}</code>
</blockquote>
<p>Neither function above can be used in constant expressions. The following
solution is correct and usable in a constant expression but is not optimal at
runtime:</p>
<blockquote>
<code>template<class T><br>
constexpr bool pointer_in_range(T* ptr, T* begin, T* end)<br>
{<br>
for (; begin != end; ++begin) {<br>
if (begin == ptr) {<br>
return true;<br>
}<br>
}<br>
return false;<br>
}</code>
</blockquote>
<h3>Argument order</h3>
<p>The argument order (<code>ptr</code>, <code>begin</code>, <code>end</code>)
is consistent with English and mathematical notation:</p>
<ul>
<li>the pointer <code>ptr</code> is in the range [<code>begin</code>,
<code>end</code>)</li>
<li><code>ptr</code> ∈ [<code>begin</code>, <code>end</code>)</li>
</ul>
<p>This order is also consistent with other standard library names that are
read left to right:</p>
<ul>
<li><code>is_base_of_v<Base, Derived></code></li>
<li><code>is_convertible_v<From, To></code></li>
<li><code>less<T>(lhs, rhs)</code></li>
</ul>
<h3>Why not span?</h3>
<p>Span's convenience is also a double edged sword. Now it can even be
implicitly constructed from an initializer list of two elements which
allows:</p>
<blockquote>
<code>if (!pointer_in_range(p, {x, y})) {<br>
// always true<br>
}</code>
</blockquote>
<p>This function will typically use an intrinsic that operates on raw pointers
and thus its interface should least inhibit the analyzer or the optimizer:</p>
<blockquote>
<code>template<class T><br>
constexpr bool pointer_in_range(const T* ptr, const T* begin, const T* end)<br>
{<br>
return __builtin_pointer_in_range(ptr, begin, end);<br>
}</code>
</blockquote>
<h3>Header choice</h3>
<p>The header <code><memory></code> is the home of other functionality
for dealing with pointers such as <code>align</code> and
<code>to_address</code>.</p>
<h2>Implementation</h2>
<p>The <a href="http://www.boost.org/">Boost</a> C++ library collection now
also has the following implementation in the Core library, releasing in version
1.86, for supported platforms.</p>
<blockquote>
<code>template<class T><br>
constexpr bool pointer_in_range(const T* ptr, const T* begin, const T* end)<br>
{<br>
if (std::is_constant_evaluated()) {<br>
for (; begin != end; ++begin) {<br>
if (begin == ptr) {<br>
return true;<br>
}<br>
}<br>
return false;<br>
}<br>
return std::less_equal<>()(begin, ptr) &&
std::less<>()(ptr, end);<br>
}</code>
</blockquote>
<h3>Limitations</h3>
<p>At runtime, this targets only the platforms that Boost supports, which does
not include implementations where two arrays of <code>T</code> may be
interleaved.</p>
<h2>Proposed Wording</h2>
<p>All changes are relative to <a href="#N4971">N4971</a>.</p>
<p>1. Insert into 20.2.2 [memory.syn] as follows:</p>
<blockquote>
<code>
// [pointer.conversion], pointer conversion<br>
template <class T>
constexpr T* to_address(T* p) noexcept;<br>
template <class Ptr>
constexpr auto to_address(const Ptr& p) noexcept;<br>
<br>
<ins>// [pointer.range.check], pointer range check<br>
template <class T>
constexpr bool
pointer_in_range(const T* ptr, const T* begin, const T* end);</ins>
</code>
</blockquote>
<p>2. Insert after 20.2.4 [pointer.conversion] as follows:</p>
<blockquote class="insert">
<strong>20.2.5 Pointer range check [pointer.range.check]</strong>
<dl>
<dt><code>template <class T>
constexpr bool
pointer_in_range(const T* ptr, const T* begin, const T* end);</code></dt>
<dd>
<p><em>Mandates:</em> <code>T</code> is not a function type.</p>
<p><em>Preconditions:</em> <code>end</code> is reachable from
<code>begin</code>.</p>
<p><em>Returns:</em> As if:</p>
<blockquote>
<code> for (; begin != end; ++begin) {<br>
if (begin == ptr) {<br>
return true;<br>
}<br>
}<br>
return false;</code>
</blockquote>
<p><em>Recommended practice:</em> Implementations should be O(1) on platforms
when possible.</p>
</dd>
</dl>
</blockquote>
<h2>Acknowledgments</h2>
<p>Peter Dimov and Jens Maurer provided feedback that improved the first
revision of this paper. Peter Dimov also reviewed the Boost implementation.</p>
<h2>References</h2>
<ul>
<li id="#N4971">N4971, Working Draft, Standard for Programming Language C++,
<br>
<a href="https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2023/n4971.pdf">
https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2023/n4971.pdf</a>
</li>
</ul>
</body>
</html>