line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
1
|
|
|
|
|
|
|
// Numeric functions implementation -*- C++ -*- |
2
|
|
|
|
|
|
|
|
3
|
|
|
|
|
|
|
// Copyright (C) 2001-2015 Free Software Foundation, Inc. |
4
|
|
|
|
|
|
|
// |
5
|
|
|
|
|
|
|
// This file is part of the GNU ISO C++ Library. This library is free |
6
|
|
|
|
|
|
|
// software; you can redistribute it and/or modify it under the |
7
|
|
|
|
|
|
|
// terms of the GNU General Public License as published by the |
8
|
|
|
|
|
|
|
// Free Software Foundation; either version 3, or (at your option) |
9
|
|
|
|
|
|
|
// any later version. |
10
|
|
|
|
|
|
|
|
11
|
|
|
|
|
|
|
// This library is distributed in the hope that it will be useful, |
12
|
|
|
|
|
|
|
// but WITHOUT ANY WARRANTY; without even the implied warranty of |
13
|
|
|
|
|
|
|
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
14
|
|
|
|
|
|
|
// GNU General Public License for more details. |
15
|
|
|
|
|
|
|
|
16
|
|
|
|
|
|
|
// Under Section 7 of GPL version 3, you are granted additional |
17
|
|
|
|
|
|
|
// permissions described in the GCC Runtime Library Exception, version |
18
|
|
|
|
|
|
|
// 3.1, as published by the Free Software Foundation. |
19
|
|
|
|
|
|
|
|
20
|
|
|
|
|
|
|
// You should have received a copy of the GNU General Public License and |
21
|
|
|
|
|
|
|
// a copy of the GCC Runtime Library Exception along with this program; |
22
|
|
|
|
|
|
|
// see the files COPYING3 and COPYING.RUNTIME respectively. If not, see |
23
|
|
|
|
|
|
|
// . |
24
|
|
|
|
|
|
|
|
25
|
|
|
|
|
|
|
/* |
26
|
|
|
|
|
|
|
* |
27
|
|
|
|
|
|
|
* Copyright (c) 1994 |
28
|
|
|
|
|
|
|
* Hewlett-Packard Company |
29
|
|
|
|
|
|
|
* |
30
|
|
|
|
|
|
|
* Permission to use, copy, modify, distribute and sell this software |
31
|
|
|
|
|
|
|
* and its documentation for any purpose is hereby granted without fee, |
32
|
|
|
|
|
|
|
* provided that the above copyright notice appear in all copies and |
33
|
|
|
|
|
|
|
* that both that copyright notice and this permission notice appear |
34
|
|
|
|
|
|
|
* in supporting documentation. Hewlett-Packard Company makes no |
35
|
|
|
|
|
|
|
* representations about the suitability of this software for any |
36
|
|
|
|
|
|
|
* purpose. It is provided "as is" without express or implied warranty. |
37
|
|
|
|
|
|
|
* |
38
|
|
|
|
|
|
|
* |
39
|
|
|
|
|
|
|
* Copyright (c) 1996,1997 |
40
|
|
|
|
|
|
|
* Silicon Graphics Computer Systems, Inc. |
41
|
|
|
|
|
|
|
* |
42
|
|
|
|
|
|
|
* Permission to use, copy, modify, distribute and sell this software |
43
|
|
|
|
|
|
|
* and its documentation for any purpose is hereby granted without fee, |
44
|
|
|
|
|
|
|
* provided that the above copyright notice appear in all copies and |
45
|
|
|
|
|
|
|
* that both that copyright notice and this permission notice appear |
46
|
|
|
|
|
|
|
* in supporting documentation. Silicon Graphics makes no |
47
|
|
|
|
|
|
|
* representations about the suitability of this software for any |
48
|
|
|
|
|
|
|
* purpose. It is provided "as is" without express or implied warranty. |
49
|
|
|
|
|
|
|
*/ |
50
|
|
|
|
|
|
|
|
51
|
|
|
|
|
|
|
/** @file bits/stl_numeric.h |
52
|
|
|
|
|
|
|
* This is an internal header file, included by other library headers. |
53
|
|
|
|
|
|
|
* Do not attempt to use it directly. @headername{numeric} |
54
|
|
|
|
|
|
|
*/ |
55
|
|
|
|
|
|
|
|
56
|
|
|
|
|
|
|
#ifndef _STL_NUMERIC_H |
57
|
|
|
|
|
|
|
#define _STL_NUMERIC_H 1 |
58
|
|
|
|
|
|
|
|
59
|
|
|
|
|
|
|
#include |
60
|
|
|
|
|
|
|
#include |
61
|
|
|
|
|
|
|
#include // For _GLIBCXX_MOVE |
62
|
|
|
|
|
|
|
|
63
|
|
|
|
|
|
|
#if __cplusplus >= 201103L |
64
|
|
|
|
|
|
|
|
65
|
|
|
|
|
|
|
namespace std _GLIBCXX_VISIBILITY(default) |
66
|
|
|
|
|
|
|
{ |
67
|
|
|
|
|
|
|
_GLIBCXX_BEGIN_NAMESPACE_VERSION |
68
|
|
|
|
|
|
|
|
69
|
|
|
|
|
|
|
/** |
70
|
|
|
|
|
|
|
* @brief Create a range of sequentially increasing values. |
71
|
|
|
|
|
|
|
* |
72
|
|
|
|
|
|
|
* For each element in the range @p [first,last) assigns @p value and |
73
|
|
|
|
|
|
|
* increments @p value as if by @p ++value. |
74
|
|
|
|
|
|
|
* |
75
|
|
|
|
|
|
|
* @param __first Start of range. |
76
|
|
|
|
|
|
|
* @param __last End of range. |
77
|
|
|
|
|
|
|
* @param __value Starting value. |
78
|
|
|
|
|
|
|
* @return Nothing. |
79
|
|
|
|
|
|
|
*/ |
80
|
|
|
|
|
|
|
template |
81
|
|
|
|
|
|
|
void |
82
|
|
|
|
|
|
|
iota(_ForwardIterator __first, _ForwardIterator __last, _Tp __value) |
83
|
|
|
|
|
|
|
{ |
84
|
|
|
|
|
|
|
// concept requirements |
85
|
|
|
|
|
|
|
__glibcxx_function_requires(_Mutable_ForwardIteratorConcept< |
86
|
|
|
|
|
|
|
_ForwardIterator>) |
87
|
|
|
|
|
|
|
__glibcxx_function_requires(_ConvertibleConcept<_Tp, |
88
|
|
|
|
|
|
|
typename iterator_traits<_ForwardIterator>::value_type>) |
89
|
|
|
|
|
|
|
__glibcxx_requires_valid_range(__first, __last); |
90
|
|
|
|
|
|
|
|
91
|
|
|
|
|
|
|
for (; __first != __last; ++__first) |
92
|
|
|
|
|
|
|
{ |
93
|
|
|
|
|
|
|
*__first = __value; |
94
|
|
|
|
|
|
|
++__value; |
95
|
|
|
|
|
|
|
} |
96
|
|
|
|
|
|
|
} |
97
|
|
|
|
|
|
|
|
98
|
|
|
|
|
|
|
_GLIBCXX_END_NAMESPACE_VERSION |
99
|
|
|
|
|
|
|
} // namespace std |
100
|
|
|
|
|
|
|
|
101
|
|
|
|
|
|
|
#endif |
102
|
|
|
|
|
|
|
|
103
|
|
|
|
|
|
|
namespace std _GLIBCXX_VISIBILITY(default) |
104
|
|
|
|
|
|
|
{ |
105
|
|
|
|
|
|
|
_GLIBCXX_BEGIN_NAMESPACE_ALGO |
106
|
|
|
|
|
|
|
|
107
|
|
|
|
|
|
|
/** |
108
|
|
|
|
|
|
|
* @brief Accumulate values in a range. |
109
|
|
|
|
|
|
|
* |
110
|
|
|
|
|
|
|
* Accumulates the values in the range [first,last) using operator+(). The |
111
|
|
|
|
|
|
|
* initial value is @a init. The values are processed in order. |
112
|
|
|
|
|
|
|
* |
113
|
|
|
|
|
|
|
* @param __first Start of range. |
114
|
|
|
|
|
|
|
* @param __last End of range. |
115
|
|
|
|
|
|
|
* @param __init Starting value to add other values to. |
116
|
|
|
|
|
|
|
* @return The final sum. |
117
|
|
|
|
|
|
|
*/ |
118
|
|
|
|
|
|
|
template |
119
|
|
|
|
|
|
|
inline _Tp |
120
|
0
|
|
|
|
|
|
accumulate(_InputIterator __first, _InputIterator __last, _Tp __init) |
121
|
|
|
|
|
|
|
{ |
122
|
|
|
|
|
|
|
// concept requirements |
123
|
|
|
|
|
|
|
__glibcxx_function_requires(_InputIteratorConcept<_InputIterator>) |
124
|
|
|
|
|
|
|
__glibcxx_requires_valid_range(__first, __last); |
125
|
|
|
|
|
|
|
|
126
|
0
|
0
|
|
|
|
|
for (; __first != __last; ++__first) |
127
|
0
|
|
|
|
|
|
__init = __init + *__first; |
128
|
0
|
|
|
|
|
|
return __init; |
129
|
|
|
|
|
|
|
} |
130
|
|
|
|
|
|
|
|
131
|
|
|
|
|
|
|
/** |
132
|
|
|
|
|
|
|
* @brief Accumulate values in a range with operation. |
133
|
|
|
|
|
|
|
* |
134
|
|
|
|
|
|
|
* Accumulates the values in the range [first,last) using the function |
135
|
|
|
|
|
|
|
* object @p __binary_op. The initial value is @p __init. The values are |
136
|
|
|
|
|
|
|
* processed in order. |
137
|
|
|
|
|
|
|
* |
138
|
|
|
|
|
|
|
* @param __first Start of range. |
139
|
|
|
|
|
|
|
* @param __last End of range. |
140
|
|
|
|
|
|
|
* @param __init Starting value to add other values to. |
141
|
|
|
|
|
|
|
* @param __binary_op Function object to accumulate with. |
142
|
|
|
|
|
|
|
* @return The final sum. |
143
|
|
|
|
|
|
|
*/ |
144
|
|
|
|
|
|
|
template |
145
|
|
|
|
|
|
|
inline _Tp |
146
|
|
|
|
|
|
|
accumulate(_InputIterator __first, _InputIterator __last, _Tp __init, |
147
|
|
|
|
|
|
|
_BinaryOperation __binary_op) |
148
|
|
|
|
|
|
|
{ |
149
|
|
|
|
|
|
|
// concept requirements |
150
|
|
|
|
|
|
|
__glibcxx_function_requires(_InputIteratorConcept<_InputIterator>) |
151
|
|
|
|
|
|
|
__glibcxx_requires_valid_range(__first, __last); |
152
|
|
|
|
|
|
|
|
153
|
|
|
|
|
|
|
for (; __first != __last; ++__first) |
154
|
|
|
|
|
|
|
__init = __binary_op(__init, *__first); |
155
|
|
|
|
|
|
|
return __init; |
156
|
|
|
|
|
|
|
} |
157
|
|
|
|
|
|
|
|
158
|
|
|
|
|
|
|
/** |
159
|
|
|
|
|
|
|
* @brief Compute inner product of two ranges. |
160
|
|
|
|
|
|
|
* |
161
|
|
|
|
|
|
|
* Starting with an initial value of @p __init, multiplies successive |
162
|
|
|
|
|
|
|
* elements from the two ranges and adds each product into the accumulated |
163
|
|
|
|
|
|
|
* value using operator+(). The values in the ranges are processed in |
164
|
|
|
|
|
|
|
* order. |
165
|
|
|
|
|
|
|
* |
166
|
|
|
|
|
|
|
* @param __first1 Start of range 1. |
167
|
|
|
|
|
|
|
* @param __last1 End of range 1. |
168
|
|
|
|
|
|
|
* @param __first2 Start of range 2. |
169
|
|
|
|
|
|
|
* @param __init Starting value to add other values to. |
170
|
|
|
|
|
|
|
* @return The final inner product. |
171
|
|
|
|
|
|
|
*/ |
172
|
|
|
|
|
|
|
template |
173
|
|
|
|
|
|
|
inline _Tp |
174
|
|
|
|
|
|
|
inner_product(_InputIterator1 __first1, _InputIterator1 __last1, |
175
|
|
|
|
|
|
|
_InputIterator2 __first2, _Tp __init) |
176
|
|
|
|
|
|
|
{ |
177
|
|
|
|
|
|
|
// concept requirements |
178
|
|
|
|
|
|
|
__glibcxx_function_requires(_InputIteratorConcept<_InputIterator1>) |
179
|
|
|
|
|
|
|
__glibcxx_function_requires(_InputIteratorConcept<_InputIterator2>) |
180
|
|
|
|
|
|
|
__glibcxx_requires_valid_range(__first1, __last1); |
181
|
|
|
|
|
|
|
|
182
|
|
|
|
|
|
|
for (; __first1 != __last1; ++__first1, ++__first2) |
183
|
|
|
|
|
|
|
__init = __init + (*__first1 * *__first2); |
184
|
|
|
|
|
|
|
return __init; |
185
|
|
|
|
|
|
|
} |
186
|
|
|
|
|
|
|
|
187
|
|
|
|
|
|
|
/** |
188
|
|
|
|
|
|
|
* @brief Compute inner product of two ranges. |
189
|
|
|
|
|
|
|
* |
190
|
|
|
|
|
|
|
* Starting with an initial value of @p __init, applies @p __binary_op2 to |
191
|
|
|
|
|
|
|
* successive elements from the two ranges and accumulates each result into |
192
|
|
|
|
|
|
|
* the accumulated value using @p __binary_op1. The values in the ranges are |
193
|
|
|
|
|
|
|
* processed in order. |
194
|
|
|
|
|
|
|
* |
195
|
|
|
|
|
|
|
* @param __first1 Start of range 1. |
196
|
|
|
|
|
|
|
* @param __last1 End of range 1. |
197
|
|
|
|
|
|
|
* @param __first2 Start of range 2. |
198
|
|
|
|
|
|
|
* @param __init Starting value to add other values to. |
199
|
|
|
|
|
|
|
* @param __binary_op1 Function object to accumulate with. |
200
|
|
|
|
|
|
|
* @param __binary_op2 Function object to apply to pairs of input values. |
201
|
|
|
|
|
|
|
* @return The final inner product. |
202
|
|
|
|
|
|
|
*/ |
203
|
|
|
|
|
|
|
template
|
204
|
|
|
|
|
|
|
typename _BinaryOperation1, typename _BinaryOperation2> |
205
|
|
|
|
|
|
|
inline _Tp |
206
|
|
|
|
|
|
|
inner_product(_InputIterator1 __first1, _InputIterator1 __last1, |
207
|
|
|
|
|
|
|
_InputIterator2 __first2, _Tp __init, |
208
|
|
|
|
|
|
|
_BinaryOperation1 __binary_op1, |
209
|
|
|
|
|
|
|
_BinaryOperation2 __binary_op2) |
210
|
|
|
|
|
|
|
{ |
211
|
|
|
|
|
|
|
// concept requirements |
212
|
|
|
|
|
|
|
__glibcxx_function_requires(_InputIteratorConcept<_InputIterator1>) |
213
|
|
|
|
|
|
|
__glibcxx_function_requires(_InputIteratorConcept<_InputIterator2>) |
214
|
|
|
|
|
|
|
__glibcxx_requires_valid_range(__first1, __last1); |
215
|
|
|
|
|
|
|
|
216
|
|
|
|
|
|
|
for (; __first1 != __last1; ++__first1, ++__first2) |
217
|
|
|
|
|
|
|
__init = __binary_op1(__init, __binary_op2(*__first1, *__first2)); |
218
|
|
|
|
|
|
|
return __init; |
219
|
|
|
|
|
|
|
} |
220
|
|
|
|
|
|
|
|
221
|
|
|
|
|
|
|
/** |
222
|
|
|
|
|
|
|
* @brief Return list of partial sums |
223
|
|
|
|
|
|
|
* |
224
|
|
|
|
|
|
|
* Accumulates the values in the range [first,last) using the @c + operator. |
225
|
|
|
|
|
|
|
* As each successive input value is added into the total, that partial sum |
226
|
|
|
|
|
|
|
* is written to @p __result. Therefore, the first value in @p __result is |
227
|
|
|
|
|
|
|
* the first value of the input, the second value in @p __result is the sum |
228
|
|
|
|
|
|
|
* of the first and second input values, and so on. |
229
|
|
|
|
|
|
|
* |
230
|
|
|
|
|
|
|
* @param __first Start of input range. |
231
|
|
|
|
|
|
|
* @param __last End of input range. |
232
|
|
|
|
|
|
|
* @param __result Output sum. |
233
|
|
|
|
|
|
|
* @return Iterator pointing just beyond the values written to __result. |
234
|
|
|
|
|
|
|
*/ |
235
|
|
|
|
|
|
|
template |
236
|
|
|
|
|
|
|
_OutputIterator |
237
|
|
|
|
|
|
|
partial_sum(_InputIterator __first, _InputIterator __last, |
238
|
|
|
|
|
|
|
_OutputIterator __result) |
239
|
|
|
|
|
|
|
{ |
240
|
|
|
|
|
|
|
typedef typename iterator_traits<_InputIterator>::value_type _ValueType; |
241
|
|
|
|
|
|
|
|
242
|
|
|
|
|
|
|
// concept requirements |
243
|
|
|
|
|
|
|
__glibcxx_function_requires(_InputIteratorConcept<_InputIterator>) |
244
|
|
|
|
|
|
|
__glibcxx_function_requires(_OutputIteratorConcept<_OutputIterator, |
245
|
|
|
|
|
|
|
_ValueType>) |
246
|
|
|
|
|
|
|
__glibcxx_requires_valid_range(__first, __last); |
247
|
|
|
|
|
|
|
|
248
|
|
|
|
|
|
|
if (__first == __last) |
249
|
|
|
|
|
|
|
return __result; |
250
|
|
|
|
|
|
|
_ValueType __value = *__first; |
251
|
|
|
|
|
|
|
*__result = __value; |
252
|
|
|
|
|
|
|
while (++__first != __last) |
253
|
|
|
|
|
|
|
{ |
254
|
|
|
|
|
|
|
__value = __value + *__first; |
255
|
|
|
|
|
|
|
*++__result = __value; |
256
|
|
|
|
|
|
|
} |
257
|
|
|
|
|
|
|
return ++__result; |
258
|
|
|
|
|
|
|
} |
259
|
|
|
|
|
|
|
|
260
|
|
|
|
|
|
|
/** |
261
|
|
|
|
|
|
|
* @brief Return list of partial sums |
262
|
|
|
|
|
|
|
* |
263
|
|
|
|
|
|
|
* Accumulates the values in the range [first,last) using @p __binary_op. |
264
|
|
|
|
|
|
|
* As each successive input value is added into the total, that partial sum |
265
|
|
|
|
|
|
|
* is written to @p __result. Therefore, the first value in @p __result is |
266
|
|
|
|
|
|
|
* the first value of the input, the second value in @p __result is the sum |
267
|
|
|
|
|
|
|
* of the first and second input values, and so on. |
268
|
|
|
|
|
|
|
* |
269
|
|
|
|
|
|
|
* @param __first Start of input range. |
270
|
|
|
|
|
|
|
* @param __last End of input range. |
271
|
|
|
|
|
|
|
* @param __result Output sum. |
272
|
|
|
|
|
|
|
* @param __binary_op Function object. |
273
|
|
|
|
|
|
|
* @return Iterator pointing just beyond the values written to __result. |
274
|
|
|
|
|
|
|
*/ |
275
|
|
|
|
|
|
|
template
|
276
|
|
|
|
|
|
|
typename _BinaryOperation> |
277
|
|
|
|
|
|
|
_OutputIterator |
278
|
|
|
|
|
|
|
partial_sum(_InputIterator __first, _InputIterator __last, |
279
|
|
|
|
|
|
|
_OutputIterator __result, _BinaryOperation __binary_op) |
280
|
|
|
|
|
|
|
{ |
281
|
|
|
|
|
|
|
typedef typename iterator_traits<_InputIterator>::value_type _ValueType; |
282
|
|
|
|
|
|
|
|
283
|
|
|
|
|
|
|
// concept requirements |
284
|
|
|
|
|
|
|
__glibcxx_function_requires(_InputIteratorConcept<_InputIterator>) |
285
|
|
|
|
|
|
|
__glibcxx_function_requires(_OutputIteratorConcept<_OutputIterator, |
286
|
|
|
|
|
|
|
_ValueType>) |
287
|
|
|
|
|
|
|
__glibcxx_requires_valid_range(__first, __last); |
288
|
|
|
|
|
|
|
|
289
|
|
|
|
|
|
|
if (__first == __last) |
290
|
|
|
|
|
|
|
return __result; |
291
|
|
|
|
|
|
|
_ValueType __value = *__first; |
292
|
|
|
|
|
|
|
*__result = __value; |
293
|
|
|
|
|
|
|
while (++__first != __last) |
294
|
|
|
|
|
|
|
{ |
295
|
|
|
|
|
|
|
__value = __binary_op(__value, *__first); |
296
|
|
|
|
|
|
|
*++__result = __value; |
297
|
|
|
|
|
|
|
} |
298
|
|
|
|
|
|
|
return ++__result; |
299
|
|
|
|
|
|
|
} |
300
|
|
|
|
|
|
|
|
301
|
|
|
|
|
|
|
/** |
302
|
|
|
|
|
|
|
* @brief Return differences between adjacent values. |
303
|
|
|
|
|
|
|
* |
304
|
|
|
|
|
|
|
* Computes the difference between adjacent values in the range |
305
|
|
|
|
|
|
|
* [first,last) using operator-() and writes the result to @p __result. |
306
|
|
|
|
|
|
|
* |
307
|
|
|
|
|
|
|
* @param __first Start of input range. |
308
|
|
|
|
|
|
|
* @param __last End of input range. |
309
|
|
|
|
|
|
|
* @param __result Output sums. |
310
|
|
|
|
|
|
|
* @return Iterator pointing just beyond the values written to result. |
311
|
|
|
|
|
|
|
* |
312
|
|
|
|
|
|
|
* _GLIBCXX_RESOLVE_LIB_DEFECTS |
313
|
|
|
|
|
|
|
* DR 539. partial_sum and adjacent_difference should mention requirements |
314
|
|
|
|
|
|
|
*/ |
315
|
|
|
|
|
|
|
template |
316
|
|
|
|
|
|
|
_OutputIterator |
317
|
|
|
|
|
|
|
adjacent_difference(_InputIterator __first, |
318
|
|
|
|
|
|
|
_InputIterator __last, _OutputIterator __result) |
319
|
|
|
|
|
|
|
{ |
320
|
|
|
|
|
|
|
typedef typename iterator_traits<_InputIterator>::value_type _ValueType; |
321
|
|
|
|
|
|
|
|
322
|
|
|
|
|
|
|
// concept requirements |
323
|
|
|
|
|
|
|
__glibcxx_function_requires(_InputIteratorConcept<_InputIterator>) |
324
|
|
|
|
|
|
|
__glibcxx_function_requires(_OutputIteratorConcept<_OutputIterator, |
325
|
|
|
|
|
|
|
_ValueType>) |
326
|
|
|
|
|
|
|
__glibcxx_requires_valid_range(__first, __last); |
327
|
|
|
|
|
|
|
|
328
|
|
|
|
|
|
|
if (__first == __last) |
329
|
|
|
|
|
|
|
return __result; |
330
|
|
|
|
|
|
|
_ValueType __value = *__first; |
331
|
|
|
|
|
|
|
*__result = __value; |
332
|
|
|
|
|
|
|
while (++__first != __last) |
333
|
|
|
|
|
|
|
{ |
334
|
|
|
|
|
|
|
_ValueType __tmp = *__first; |
335
|
|
|
|
|
|
|
*++__result = __tmp - __value; |
336
|
|
|
|
|
|
|
__value = _GLIBCXX_MOVE(__tmp); |
337
|
|
|
|
|
|
|
} |
338
|
|
|
|
|
|
|
return ++__result; |
339
|
|
|
|
|
|
|
} |
340
|
|
|
|
|
|
|
|
341
|
|
|
|
|
|
|
/** |
342
|
|
|
|
|
|
|
* @brief Return differences between adjacent values. |
343
|
|
|
|
|
|
|
* |
344
|
|
|
|
|
|
|
* Computes the difference between adjacent values in the range |
345
|
|
|
|
|
|
|
* [__first,__last) using the function object @p __binary_op and writes the |
346
|
|
|
|
|
|
|
* result to @p __result. |
347
|
|
|
|
|
|
|
* |
348
|
|
|
|
|
|
|
* @param __first Start of input range. |
349
|
|
|
|
|
|
|
* @param __last End of input range. |
350
|
|
|
|
|
|
|
* @param __result Output sum. |
351
|
|
|
|
|
|
|
* @param __binary_op Function object. |
352
|
|
|
|
|
|
|
* @return Iterator pointing just beyond the values written to result. |
353
|
|
|
|
|
|
|
* |
354
|
|
|
|
|
|
|
* _GLIBCXX_RESOLVE_LIB_DEFECTS |
355
|
|
|
|
|
|
|
* DR 539. partial_sum and adjacent_difference should mention requirements |
356
|
|
|
|
|
|
|
*/ |
357
|
|
|
|
|
|
|
template
|
358
|
|
|
|
|
|
|
typename _BinaryOperation> |
359
|
|
|
|
|
|
|
_OutputIterator |
360
|
|
|
|
|
|
|
adjacent_difference(_InputIterator __first, _InputIterator __last, |
361
|
|
|
|
|
|
|
_OutputIterator __result, _BinaryOperation __binary_op) |
362
|
|
|
|
|
|
|
{ |
363
|
|
|
|
|
|
|
typedef typename iterator_traits<_InputIterator>::value_type _ValueType; |
364
|
|
|
|
|
|
|
|
365
|
|
|
|
|
|
|
// concept requirements |
366
|
|
|
|
|
|
|
__glibcxx_function_requires(_InputIteratorConcept<_InputIterator>) |
367
|
|
|
|
|
|
|
__glibcxx_function_requires(_OutputIteratorConcept<_OutputIterator, |
368
|
|
|
|
|
|
|
_ValueType>) |
369
|
|
|
|
|
|
|
__glibcxx_requires_valid_range(__first, __last); |
370
|
|
|
|
|
|
|
|
371
|
|
|
|
|
|
|
if (__first == __last) |
372
|
|
|
|
|
|
|
return __result; |
373
|
|
|
|
|
|
|
_ValueType __value = *__first; |
374
|
|
|
|
|
|
|
*__result = __value; |
375
|
|
|
|
|
|
|
while (++__first != __last) |
376
|
|
|
|
|
|
|
{ |
377
|
|
|
|
|
|
|
_ValueType __tmp = *__first; |
378
|
|
|
|
|
|
|
*++__result = __binary_op(__tmp, __value); |
379
|
|
|
|
|
|
|
__value = _GLIBCXX_MOVE(__tmp); |
380
|
|
|
|
|
|
|
} |
381
|
|
|
|
|
|
|
return ++__result; |
382
|
|
|
|
|
|
|
} |
383
|
|
|
|
|
|
|
|
384
|
|
|
|
|
|
|
_GLIBCXX_END_NAMESPACE_ALGO |
385
|
|
|
|
|
|
|
} // namespace std |
386
|
|
|
|
|
|
|
|
387
|
|
|
|
|
|
|
#endif /* _STL_NUMERIC_H */ |