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 "src/__support/CPP/limits.h" |
19 | #include "src/__support/CPP/type_traits.h" |
20 | #include "src/__support/CPP/type_traits/make_unsigned.h" |
21 | #include "src/__support/big_int.h" |
22 | #include "src/__support/common.h" |
23 | #include "src/__support/ctype_utils.h" |
24 | #include "src/__support/macros/config.h" |
25 | #include "src/__support/str_to_num_result.h" |
26 | #include "src/__support/uint128.h" |
27 | #include "src/errno/libc_errno.h" // For ERANGE |
28 | |
29 | namespace LIBC_NAMESPACE_DECL { |
30 | namespace internal { |
31 | |
32 | // Returns a pointer to the first character in src that is not a whitespace |
33 | // character (as determined by isspace()) |
34 | // TODO: Change from returning a pointer to returning a length. |
35 | LIBC_INLINE const char * |
36 | first_non_whitespace(const char *__restrict src, |
37 | size_t src_len = cpp::numeric_limits<size_t>::max()) { |
38 | size_t src_cur = 0; |
39 | while (src_cur < src_len && internal::isspace(ch: src[src_cur])) { |
40 | ++src_cur; |
41 | } |
42 | return src + src_cur; |
43 | } |
44 | |
45 | // checks if the next 3 characters of the string pointer are the start of a |
46 | // hexadecimal number. Does not advance the string pointer. |
47 | LIBC_INLINE bool |
48 | is_hex_start(const char *__restrict src, |
49 | size_t src_len = cpp::numeric_limits<size_t>::max()) { |
50 | if (src_len < 3) |
51 | return false; |
52 | return *src == '0' && tolower(ch: *(src + 1)) == 'x' && isalnum(ch: *(src + 2)) && |
53 | b36_char_to_int(ch: *(src + 2)) < 16; |
54 | } |
55 | |
56 | // Takes the address of the string pointer and parses the base from the start of |
57 | // it. |
58 | LIBC_INLINE int infer_base(const char *__restrict src, size_t src_len) { |
59 | // A hexadecimal number is defined as "the prefix 0x or 0X followed by a |
60 | // sequence of the decimal digits and the letters a (or A) through f (or F) |
61 | // with values 10 through 15 respectively." (C standard 6.4.4.1) |
62 | if (is_hex_start(src, src_len)) |
63 | return 16; |
64 | // An octal number is defined as "the prefix 0 optionally followed by a |
65 | // sequence of the digits 0 through 7 only" (C standard 6.4.4.1) and so any |
66 | // number that starts with 0, including just 0, is an octal number. |
67 | if (src_len > 0 && src[0] == '0') |
68 | return 8; |
69 | // A decimal number is defined as beginning "with a nonzero digit and |
70 | // consist[ing] of a sequence of decimal digits." (C standard 6.4.4.1) |
71 | return 10; |
72 | } |
73 | |
74 | // ----------------------------------------------------------------------------- |
75 | // **** WARNING **** |
76 | // This interface is shared with libc++, if you change this interface you need |
77 | // to update it in both libc and libc++. |
78 | // ----------------------------------------------------------------------------- |
79 | // Takes a pointer to a string and the base to convert to. This function is used |
80 | // as the backend for all of the string to int functions. |
81 | template <class T> |
82 | LIBC_INLINE StrToNumResult<T> |
83 | strtointeger(const char *__restrict src, int base, |
84 | const size_t src_len = cpp::numeric_limits<size_t>::max()) { |
85 | using ResultType = make_integral_or_big_int_unsigned_t<T>; |
86 | |
87 | ResultType result = 0; |
88 | |
89 | bool is_number = false; |
90 | size_t src_cur = 0; |
91 | int error_val = 0; |
92 | |
93 | if (src_len == 0) |
94 | return {0, 0, 0}; |
95 | |
96 | if (base < 0 || base == 1 || base > 36) |
97 | return {0, 0, EINVAL}; |
98 | |
99 | src_cur = first_non_whitespace(src, src_len) - src; |
100 | |
101 | char result_sign = '+'; |
102 | if (src[src_cur] == '+' || src[src_cur] == '-') { |
103 | result_sign = src[src_cur]; |
104 | ++src_cur; |
105 | } |
106 | |
107 | if (base == 0) |
108 | base = infer_base(src: src + src_cur, src_len: src_len - src_cur); |
109 | |
110 | if (base == 16 && is_hex_start(src: src + src_cur, src_len: src_len - src_cur)) |
111 | src_cur = src_cur + 2; |
112 | |
113 | constexpr bool IS_UNSIGNED = cpp::is_unsigned_v<T>; |
114 | const bool is_positive = (result_sign == '+'); |
115 | |
116 | ResultType constexpr NEGATIVE_MAX = |
117 | !IS_UNSIGNED ? static_cast<ResultType>(cpp::numeric_limits<T>::max()) + 1 |
118 | : cpp::numeric_limits<T>::max(); |
119 | ResultType const abs_max = |
120 | (is_positive ? cpp::numeric_limits<T>::max() : NEGATIVE_MAX); |
121 | ResultType const abs_max_div_by_base = |
122 | static_cast<ResultType>(abs_max / base); |
123 | |
124 | while (src_cur < src_len && isalnum(ch: src[src_cur])) { |
125 | int cur_digit = b36_char_to_int(ch: src[src_cur]); |
126 | if (cur_digit >= base) |
127 | break; |
128 | |
129 | is_number = true; |
130 | ++src_cur; |
131 | |
132 | // If the number has already hit the maximum value for the current type then |
133 | // the result cannot change, but we still need to advance src to the end of |
134 | // the number. |
135 | if (result == abs_max) { |
136 | error_val = ERANGE; |
137 | continue; |
138 | } |
139 | |
140 | if (result > abs_max_div_by_base) { |
141 | result = abs_max; |
142 | error_val = ERANGE; |
143 | } else { |
144 | result = static_cast<ResultType>(result * base); |
145 | } |
146 | if (result > abs_max - cur_digit) { |
147 | result = abs_max; |
148 | error_val = ERANGE; |
149 | } else { |
150 | result = static_cast<ResultType>(result + cur_digit); |
151 | } |
152 | } |
153 | |
154 | ptrdiff_t str_len = is_number ? (src_cur) : 0; |
155 | |
156 | if (error_val == ERANGE) { |
157 | if (is_positive || IS_UNSIGNED) |
158 | return {cpp::numeric_limits<T>::max(), str_len, error_val}; |
159 | else // T is signed and there is a negative overflow |
160 | return {cpp::numeric_limits<T>::min(), str_len, error_val}; |
161 | } |
162 | |
163 | return {static_cast<T>(is_positive ? result : -result), str_len, error_val}; |
164 | } |
165 | |
166 | } // namespace internal |
167 | } // namespace LIBC_NAMESPACE_DECL |
168 | |
169 | #endif // LLVM_LIBC_SRC___SUPPORT_STR_TO_INTEGER_H |
170 | |