1//===-- String to integer conversion utils ----------------------*- C++ -*-===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8
9// -----------------------------------------------------------------------------
10// **** WARNING ****
11// This file is shared with libc++. You should also be careful when adding
12// dependencies to this file, since it needs to build for all libc++ targets.
13// -----------------------------------------------------------------------------
14
15#ifndef LLVM_LIBC_SRC___SUPPORT_STR_TO_INTEGER_H
16#define LLVM_LIBC_SRC___SUPPORT_STR_TO_INTEGER_H
17
18#include "hdr/errno_macros.h" // For ERANGE
19#include "src/__support/CPP/limits.h"
20#include "src/__support/CPP/type_traits.h"
21#include "src/__support/CPP/type_traits/make_unsigned.h"
22#include "src/__support/big_int.h"
23#include "src/__support/common.h"
24#include "src/__support/ctype_utils.h"
25#include "src/__support/macros/config.h"
26#include "src/__support/str_to_num_result.h"
27#include "src/__support/uint128.h"
28#include "src/__support/wctype_utils.h"
29
30namespace LIBC_NAMESPACE_DECL {
31namespace internal {
32
33// Returns the idx to the first character in src that is not a whitespace
34// character (as determined by isspace())
35template <typename CharType>
36LIBC_INLINE size_t
37first_non_whitespace(const CharType *__restrict src,
38 size_t src_len = cpp::numeric_limits<size_t>::max()) {
39 size_t src_cur = 0;
40 for (; src_cur < src_len && internal::isspace(src[src_cur]); ++src_cur)
41 ;
42 return src_cur;
43}
44
45// Returns +1, -1, or 0 if 'src' starts with (respectively)
46// plus sign, minus sign, or neither.
47template <typename CharType>
48LIBC_INLINE static int get_sign(const CharType *__restrict src) {
49 if (is_char_or_wchar(src[0], '+', L'+'))
50 return 1;
51 if (is_char_or_wchar(src[0], '-', L'-'))
52 return -1;
53 return 0;
54}
55
56// checks if the next 3 characters of the string pointer are the start of a
57// hexadecimal number. Does not advance the string pointer.
58template <typename CharType>
59LIBC_INLINE static bool is_hex_start(const CharType *__restrict src,
60 size_t src_len) {
61 if (src_len < 3)
62 return false;
63 return is_char_or_wchar(src[0], '0', L'0') &&
64 is_char_or_wchar(tolower(src[1]), 'x', L'x') && isalnum(src[2]) &&
65 b36_char_to_int(src[2]) < 16;
66}
67
68// Takes the address of the string pointer and parses the base from the start of
69// it.
70template <typename CharType>
71LIBC_INLINE static int infer_base(const CharType *__restrict src,
72 size_t src_len) {
73 // A hexadecimal number is defined as "the prefix 0x or 0X followed by a
74 // sequence of the decimal digits and the letters a (or A) through f (or F)
75 // with values 10 through 15 respectively." (C standard 6.4.4.1)
76 if (is_hex_start(src, src_len))
77 return 16;
78 // An octal number is defined as "the prefix 0 optionally followed by a
79 // sequence of the digits 0 through 7 only" (C standard 6.4.4.1) and so any
80 // number that starts with 0, including just 0, is an octal number.
81 if (src_len > 0 && is_char_or_wchar(src[0], '0', L'0')) {
82 return 8;
83 }
84 // A decimal number is defined as beginning "with a nonzero digit and
85 // consist[ing] of a sequence of decimal digits." (C standard 6.4.4.1)
86 return 10;
87}
88
89// -----------------------------------------------------------------------------
90// **** WARNING ****
91// This interface is shared with libc++, if you change this interface you need
92// to update it in both libc and libc++.
93// -----------------------------------------------------------------------------
94// Takes a pointer to a string and the base to convert to. This function is used
95// as the backend for all of the string to int functions.
96template <typename T, typename CharType>
97LIBC_INLINE StrToNumResult<T>
98strtointeger(const CharType *__restrict src, int base,
99 const size_t src_len = cpp::numeric_limits<size_t>::max()) {
100 using ResultType = make_integral_or_big_int_unsigned_t<T>;
101
102 if (src_len == 0)
103 return {0, 0, 0};
104
105 if (base < 0 || base == 1 || base > 36)
106 return {0, 0, EINVAL};
107
108 size_t src_cur = first_non_whitespace(src, src_len);
109 if (src_cur == src_len) {
110 return {0, 0, 0};
111 }
112
113 int sign = get_sign(src + src_cur);
114 bool is_positive = (sign >= 0);
115 src_cur += (sign != 0);
116
117 if (base == 0)
118 base = infer_base(src + src_cur, src_len - src_cur);
119
120 if (base == 16 && is_hex_start(src + src_cur, src_len - src_cur))
121 src_cur = src_cur + 2;
122
123 constexpr bool IS_UNSIGNED = cpp::is_unsigned_v<T>;
124 ResultType constexpr NEGATIVE_MAX =
125 !IS_UNSIGNED ? static_cast<ResultType>(cpp::numeric_limits<T>::max()) + 1
126 : cpp::numeric_limits<T>::max();
127 ResultType const abs_max =
128 (is_positive ? cpp::numeric_limits<T>::max() : NEGATIVE_MAX);
129 ResultType const abs_max_div_by_base =
130 abs_max / static_cast<ResultType>(base);
131
132 bool is_number = false;
133 int error_val = 0;
134 ResultType result = 0;
135 while (src_cur < src_len && isalnum(src[src_cur])) {
136 int cur_digit = b36_char_to_int(src[src_cur]);
137 if (cur_digit >= base)
138 break;
139
140 is_number = true;
141 ++src_cur;
142
143 // If the number has already hit the maximum value for the current type then
144 // the result cannot change, but we still need to advance src to the end of
145 // the number.
146 if (result == abs_max) {
147 error_val = ERANGE;
148 continue;
149 }
150
151 if (result > abs_max_div_by_base) {
152 result = abs_max;
153 error_val = ERANGE;
154 } else {
155 result = result * static_cast<ResultType>(base);
156 }
157 if (result > abs_max - static_cast<ResultType>(cur_digit)) {
158 result = abs_max;
159 error_val = ERANGE;
160 } else {
161 result = result + static_cast<ResultType>(cur_digit);
162 }
163 }
164
165 ptrdiff_t str_len = is_number ? static_cast<ptrdiff_t>(src_cur) : 0;
166
167 if (error_val == ERANGE) {
168 if (is_positive || IS_UNSIGNED)
169 return {cpp::numeric_limits<T>::max(), str_len, error_val};
170 else // T is signed and there is a negative overflow
171 return {cpp::numeric_limits<T>::min(), str_len, error_val};
172 }
173
174 return {static_cast<T>(is_positive ? result : -result), str_len, error_val};
175}
176
177} // namespace internal
178} // namespace LIBC_NAMESPACE_DECL
179
180#endif // LLVM_LIBC_SRC___SUPPORT_STR_TO_INTEGER_H
181