1 // -*- C++ -*-
2 //===----------------------------------------------------------------------===//
3 //
4 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
5 // See https://llvm.org/LICENSE.txt for license information.
6 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
7 //
8 //===----------------------------------------------------------------------===//
9
10 #ifndef _LIBCPP___FORMAT_FORMAT_STRING_H
11 #define _LIBCPP___FORMAT_FORMAT_STRING_H
12
13 #include <__assert>
14 #include <__config>
15 #include <__format/format_error.h>
16 #include <__iterator/concepts.h>
17 #include <__iterator/readable_traits.h> // iter_value_t
18 #include <cstddef>
19 #include <cstdint>
20
21 #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
22 # pragma GCC system_header
23 #endif
24
25 _LIBCPP_BEGIN_NAMESPACE_STD
26
27 #if _LIBCPP_STD_VER >= 20
28
29 namespace __format {
30
31 template <contiguous_iterator _Iterator>
32 struct _LIBCPP_TEMPLATE_VIS __parse_number_result {
33 _Iterator __last;
34 uint32_t __value;
35 };
36
37 template <contiguous_iterator _Iterator>
38 __parse_number_result(_Iterator, uint32_t) -> __parse_number_result<_Iterator>;
39
40 template <contiguous_iterator _Iterator>
41 _LIBCPP_HIDE_FROM_ABI constexpr __parse_number_result<_Iterator>
42 __parse_number(_Iterator __begin, _Iterator __end);
43
44 /**
45 * The maximum value of a numeric argument.
46 *
47 * This is used for:
48 * * arg-id
49 * * width as value or arg-id.
50 * * precision as value or arg-id.
51 *
52 * The value is compatible with the maximum formatting width and precision
53 * using the `%*` syntax on a 32-bit system.
54 */
55 inline constexpr uint32_t __number_max = INT32_MAX;
56
57 namespace __detail {
58 template <contiguous_iterator _Iterator>
59 _LIBCPP_HIDE_FROM_ABI constexpr __parse_number_result<_Iterator>
__parse_zero(_Iterator __begin,_Iterator,auto & __parse_ctx)60 __parse_zero(_Iterator __begin, _Iterator, auto& __parse_ctx) {
61 __parse_ctx.check_arg_id(0);
62 return {++__begin, 0}; // can never be larger than the maximum.
63 }
64
65 template <contiguous_iterator _Iterator>
66 _LIBCPP_HIDE_FROM_ABI constexpr __parse_number_result<_Iterator>
__parse_automatic(_Iterator __begin,_Iterator,auto & __parse_ctx)67 __parse_automatic(_Iterator __begin, _Iterator, auto& __parse_ctx) {
68 size_t __value = __parse_ctx.next_arg_id();
69 _LIBCPP_ASSERT(__value <= __number_max,
70 "Compilers don't support this number of arguments");
71
72 return {__begin, uint32_t(__value)};
73 }
74
75 template <contiguous_iterator _Iterator>
76 _LIBCPP_HIDE_FROM_ABI constexpr __parse_number_result<_Iterator>
__parse_manual(_Iterator __begin,_Iterator __end,auto & __parse_ctx)77 __parse_manual(_Iterator __begin, _Iterator __end, auto& __parse_ctx) {
78 __parse_number_result<_Iterator> __r = __format::__parse_number(__begin, __end);
79 __parse_ctx.check_arg_id(__r.__value);
80 return __r;
81 }
82
83 } // namespace __detail
84
85 /**
86 * Parses a number.
87 *
88 * The number is used for the 31-bit values @em width and @em precision. This
89 * allows a maximum value of 2147483647.
90 */
91 template <contiguous_iterator _Iterator>
92 _LIBCPP_HIDE_FROM_ABI constexpr __parse_number_result<_Iterator>
__parse_number(_Iterator __begin,_Iterator __end_input)93 __parse_number(_Iterator __begin, _Iterator __end_input) {
94 using _CharT = iter_value_t<_Iterator>;
95 static_assert(__format::__number_max == INT32_MAX,
96 "The algorithm is implemented based on this value.");
97 /*
98 * Limit the input to 9 digits, otherwise we need two checks during every
99 * iteration:
100 * - Are we at the end of the input?
101 * - Does the value exceed width of an uint32_t? (Switching to uint64_t would
102 * have the same issue, but with a higher maximum.)
103 */
104 _Iterator __end = __end_input - __begin > 9 ? __begin + 9 : __end_input;
105 uint32_t __value = *__begin - _CharT('0');
106 while (++__begin != __end) {
107 if (*__begin < _CharT('0') || *__begin > _CharT('9'))
108 return {__begin, __value};
109
110 __value = __value * 10 + *__begin - _CharT('0');
111 }
112
113 if (__begin != __end_input && *__begin >= _CharT('0') &&
114 *__begin <= _CharT('9')) {
115
116 /*
117 * There are more than 9 digits, do additional validations:
118 * - Does the 10th digit exceed the maximum allowed value?
119 * - Are there more than 10 digits?
120 * (More than 10 digits always overflows the maximum.)
121 */
122 uint64_t __v = uint64_t(__value) * 10 + *__begin++ - _CharT('0');
123 if (__v > __number_max ||
124 (__begin != __end_input && *__begin >= _CharT('0') &&
125 *__begin <= _CharT('9')))
126 std::__throw_format_error("The numeric value of the format-spec is too large");
127
128 __value = __v;
129 }
130
131 return {__begin, __value};
132 }
133
134 /**
135 * Multiplexer for all parse functions.
136 *
137 * The parser will return a pointer beyond the last consumed character. This
138 * should be the closing '}' of the arg-id.
139 */
140 template <contiguous_iterator _Iterator>
141 _LIBCPP_HIDE_FROM_ABI constexpr __parse_number_result<_Iterator>
__parse_arg_id(_Iterator __begin,_Iterator __end,auto & __parse_ctx)142 __parse_arg_id(_Iterator __begin, _Iterator __end, auto& __parse_ctx) {
143 using _CharT = iter_value_t<_Iterator>;
144 switch (*__begin) {
145 case _CharT('0'):
146 return __detail::__parse_zero(__begin, __end, __parse_ctx);
147
148 case _CharT(':'):
149 // This case is conditionally valid. It's allowed in an arg-id in the
150 // replacement-field, but not in the std-format-spec. The caller can
151 // provide a better diagnostic, so accept it here unconditionally.
152 case _CharT('}'):
153 return __detail::__parse_automatic(__begin, __end, __parse_ctx);
154 }
155 if (*__begin < _CharT('0') || *__begin > _CharT('9'))
156 std::__throw_format_error("The arg-id of the format-spec starts with an invalid character");
157
158 return __detail::__parse_manual(__begin, __end, __parse_ctx);
159 }
160
161 } // namespace __format
162
163 #endif //_LIBCPP_STD_VER >= 20
164
165 _LIBCPP_END_NAMESPACE_STD
166
167 #endif // _LIBCPP___FORMAT_FORMAT_STRING_H
168