master
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
29namespace LIBC_NAMESPACE_DECL {
30namespace internal {
31
32// Returns the idx to the first character in src that is not a whitespace
33// character (as determined by isspace())
34LIBC_INLINE size_t
35first_non_whitespace(const char *__restrict src,
36 size_t src_len = cpp::numeric_limits<size_t>::max()) {
37 size_t src_cur = 0;
38 while (src_cur < src_len && internal::isspace(src[src_cur])) {
39 ++src_cur;
40 }
41 return src_cur;
42}
43
44// checks if the next 3 characters of the string pointer are the start of a
45// hexadecimal number. Does not advance the string pointer.
46LIBC_INLINE bool
47is_hex_start(const char *__restrict src,
48 size_t src_len = cpp::numeric_limits<size_t>::max()) {
49 if (src_len < 3)
50 return false;
51 return *src == '0' && tolower(*(src + 1)) == 'x' && isalnum(*(src + 2)) &&
52 b36_char_to_int(*(src + 2)) < 16;
53}
54
55// Takes the address of the string pointer and parses the base from the start of
56// it.
57LIBC_INLINE int infer_base(const char *__restrict src, size_t src_len) {
58 // A hexadecimal number is defined as "the prefix 0x or 0X followed by a
59 // sequence of the decimal digits and the letters a (or A) through f (or F)
60 // with values 10 through 15 respectively." (C standard 6.4.4.1)
61 if (is_hex_start(src, src_len))
62 return 16;
63 // An octal number is defined as "the prefix 0 optionally followed by a
64 // sequence of the digits 0 through 7 only" (C standard 6.4.4.1) and so any
65 // number that starts with 0, including just 0, is an octal number.
66 if (src_len > 0 && src[0] == '0')
67 return 8;
68 // A decimal number is defined as beginning "with a nonzero digit and
69 // consist[ing] of a sequence of decimal digits." (C standard 6.4.4.1)
70 return 10;
71}
72
73// -----------------------------------------------------------------------------
74// **** WARNING ****
75// This interface is shared with libc++, if you change this interface you need
76// to update it in both libc and libc++.
77// -----------------------------------------------------------------------------
78// Takes a pointer to a string and the base to convert to. This function is used
79// as the backend for all of the string to int functions.
80template <class T>
81LIBC_INLINE StrToNumResult<T>
82strtointeger(const char *__restrict src, int base,
83 const size_t src_len = cpp::numeric_limits<size_t>::max()) {
84 using ResultType = make_integral_or_big_int_unsigned_t<T>;
85
86 ResultType result = 0;
87
88 bool is_number = false;
89 size_t src_cur = 0;
90 int error_val = 0;
91
92 if (src_len == 0)
93 return {0, 0, 0};
94
95 if (base < 0 || base == 1 || base > 36)
96 return {0, 0, EINVAL};
97
98 src_cur = first_non_whitespace(src, src_len);
99
100 char result_sign = '+';
101 if (src[src_cur] == '+' || src[src_cur] == '-') {
102 result_sign = src[src_cur];
103 ++src_cur;
104 }
105
106 if (base == 0)
107 base = infer_base(src + src_cur, src_len - src_cur);
108
109 if (base == 16 && is_hex_start(src + src_cur, src_len - src_cur))
110 src_cur = src_cur + 2;
111
112 constexpr bool IS_UNSIGNED = cpp::is_unsigned_v<T>;
113 const bool is_positive = (result_sign == '+');
114
115 ResultType constexpr NEGATIVE_MAX =
116 !IS_UNSIGNED ? static_cast<ResultType>(cpp::numeric_limits<T>::max()) + 1
117 : cpp::numeric_limits<T>::max();
118 ResultType const abs_max =
119 (is_positive ? cpp::numeric_limits<T>::max() : NEGATIVE_MAX);
120 ResultType const abs_max_div_by_base =
121 abs_max / static_cast<ResultType>(base);
122
123 while (src_cur < src_len && isalnum(src[src_cur])) {
124 int cur_digit = b36_char_to_int(src[src_cur]);
125 if (cur_digit >= base)
126 break;
127
128 is_number = true;
129 ++src_cur;
130
131 // If the number has already hit the maximum value for the current type then
132 // the result cannot change, but we still need to advance src to the end of
133 // the number.
134 if (result == abs_max) {
135 error_val = ERANGE;
136 continue;
137 }
138
139 if (result > abs_max_div_by_base) {
140 result = abs_max;
141 error_val = ERANGE;
142 } else {
143 result = result * static_cast<ResultType>(base);
144 }
145 if (result > abs_max - static_cast<ResultType>(cur_digit)) {
146 result = abs_max;
147 error_val = ERANGE;
148 } else {
149 result = result + static_cast<ResultType>(cur_digit);
150 }
151 }
152
153 ptrdiff_t str_len = is_number ? static_cast<ptrdiff_t>(src_cur) : 0;
154
155 if (error_val == ERANGE) {
156 if (is_positive || IS_UNSIGNED)
157 return {cpp::numeric_limits<T>::max(), str_len, error_val};
158 else // T is signed and there is a negative overflow
159 return {cpp::numeric_limits<T>::min(), str_len, error_val};
160 }
161
162 return {static_cast<T>(is_positive ? result : -result), str_len, error_val};
163}
164
165} // namespace internal
166} // namespace LIBC_NAMESPACE_DECL
167
168#endif // LLVM_LIBC_SRC___SUPPORT_STR_TO_INTEGER_H