1 #ifndef MB_WC_INCLUDED
2 #define MB_WC_INCLUDED
3
4 /* Copyright (c) 2016, Oracle and/or its affiliates. All rights reserved.
5
6 This library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Library General Public
8 License, version 2.0, as published by the Free Software Foundation.
9
10 This library is also distributed with certain software (including
11 but not limited to OpenSSL) that is licensed under separate terms,
12 as designated in a particular file or component or in included license
13 documentation. The authors of MySQL hereby grant you an additional
14 permission to link the library and your derivative works with the
15 separately licensed software that they have included with MySQL.
16
17 This library is distributed in the hope that it will be useful,
18 but WITHOUT ANY WARRANTY; without even the implied warranty of
19 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 Library General Public License, version 2.0, for more details.
21
22 You should have received a copy of the GNU Library General Public
23 License along with this library; if not, write to the Free
24 Software Foundation, Inc., 51 Franklin St, Fifth Floor, Boston,
25 MA 02110-1301 USA */
26
27 /**
28 @file mb_wc.h
29
30 Definitions of mb_wc (multibyte to wide character, ie., effectively
31 “parse a UTF-8 character”) functions for UTF-8 (both three- and four-byte).
32 These are available both as inline functions, as C-style thunks so that they
33 can fit into MY_CHARSET_HANDLER, and as functors.
34
35 The functors exist so that you can specialize a class on them and get them
36 inlined instead of having to call them through the function pointer in
37 MY_CHARSET_HANDLER; mb_wc is in itself so cheap (the most common case is
38 just a single byte load and a predictable compare) that the call overhead
39 in a tight loop is significant, and these routines tend to take up a lot
40 of CPU time when sorting. Typically, at the outermost level, you'd simply
41 compare cs->cset->mb_wc with my_mb_wc_{utf8,utf8mb4}_thunk, and if so,
42 instantiate your function with the given class. If it doesn't match,
43 you can use Mb_wc_through_function_pointer, which calls through the
44 function pointer as usual. (It will cache the function pointer for you,
45 which is typically faster than looking it up all the time -- the compiler
46 cannot always figure out on its own that it doesn't change.)
47
48 The Mb_wc_* classes should be sent by _value_, not by reference, since
49 they are never larger than two pointers (and usually simply zero).
50 */
51
52 #include "m_ctype.h"
53 #include "my_compiler.h"
54 #include "my_config.h"
55
56 template <bool RANGE_CHECK, bool SUPPORT_MB4>
57 static int my_mb_wc_utf8_prototype(my_wc_t *pwc, const uchar *s,
58 const uchar *e);
59
60 static int my_mb_wc_utf8(my_wc_t *pwc, const uchar *s, const uchar *e);
61 static int my_mb_wc_utf8mb4(my_wc_t *pwc, const uchar *s, const uchar *e);
62
63 /**
64 Functor that converts a UTF-8 multibyte sequence (up to three bytes)
65 to a wide character.
66 */
67 struct Mb_wc_utf8 {
Mb_wc_utf8Mb_wc_utf868 Mb_wc_utf8() {}
69
70 ALWAYS_INLINE
operatorMb_wc_utf871 int operator()(my_wc_t *pwc, const uchar *s, const uchar *e) const {
72 return my_mb_wc_utf8(pwc, s, e);
73 }
74 };
75
76 /**
77 Functor that converts a UTF-8 multibyte sequence (up to four bytes)
78 to a wide character.
79 */
80 struct Mb_wc_utf8mb4 {
Mb_wc_utf8mb4Mb_wc_utf8mb481 Mb_wc_utf8mb4() {}
82
83 ALWAYS_INLINE
operatorMb_wc_utf8mb484 int operator()(my_wc_t *pwc, const uchar *s, const uchar *e) const {
85 return my_mb_wc_utf8mb4(pwc, s, e);
86 }
87 };
88
89 /**
90 Functor that uses a function pointer to convert a multibyte sequence
91 to a wide character.
92 */
93 class Mb_wc_through_function_pointer {
94 public:
Mb_wc_through_function_pointer(const CHARSET_INFO * cs)95 explicit Mb_wc_through_function_pointer(const CHARSET_INFO *cs)
96 : m_funcptr(cs->cset->mb_wc), m_cs(cs) {}
97
operator()98 int operator()(my_wc_t *pwc, const uchar *s, const uchar *e) const {
99 return m_funcptr(m_cs, pwc, s, e);
100 }
101
102 private:
103 typedef int (*mbwc_func_t)(const CHARSET_INFO *, my_wc_t *, const uchar *,
104 const uchar *);
105
106 const mbwc_func_t m_funcptr;
107 const CHARSET_INFO *const m_cs;
108 };
109
110 template <bool RANGE_CHECK, bool SUPPORT_MB4>
my_mb_wc_utf8_prototype(my_wc_t * pwc,const uchar * s,const uchar * e)111 static ALWAYS_INLINE int my_mb_wc_utf8_prototype(my_wc_t *pwc, const uchar *s,
112 const uchar *e) {
113 if (RANGE_CHECK && s >= e) return MY_CS_TOOSMALL;
114
115 uchar c = s[0];
116 if (c < 0x80) {
117 *pwc = c;
118 return 1;
119 }
120
121 if (c < 0xe0) {
122 if (c < 0xc2) // Resulting code point would be less than 0x80.
123 return MY_CS_ILSEQ;
124
125 if (RANGE_CHECK && s + 2 > e) return MY_CS_TOOSMALL2;
126
127 if ((s[1] & 0xc0) != 0x80) // Next byte must be a continuation byte.
128 return MY_CS_ILSEQ;
129
130 *pwc = ((my_wc_t)(c & 0x1f) << 6) + (my_wc_t)(s[1] & 0x3f);
131 return 2;
132 }
133
134 if (c < 0xf0) {
135 if (RANGE_CHECK && s + 3 > e) return MY_CS_TOOSMALL3;
136
137 // Next two bytes must be continuation bytes.
138 uint16 two_bytes;
139 memcpy(&two_bytes, s + 1, sizeof(two_bytes));
140 if ((two_bytes & 0xc0c0) != 0x8080) // Endianness does not matter.
141 return MY_CS_ILSEQ;
142
143 *pwc = ((my_wc_t)(c & 0x0f) << 12) + ((my_wc_t)(s[1] & 0x3f) << 6) +
144 (my_wc_t)(s[2] & 0x3f);
145 if (*pwc < 0x800) return MY_CS_ILSEQ;
146 /*
147 According to RFC 3629, UTF-8 should prohibit characters between
148 U+D800 and U+DFFF, which are reserved for surrogate pairs and do
149 not directly represent characters.
150 */
151 if (*pwc >= 0xd800 && *pwc <= 0xdfff) return MY_CS_ILSEQ;
152 return 3;
153 }
154
155 if (SUPPORT_MB4) {
156 if (RANGE_CHECK && s + 4 > e) /* We need 4 characters */
157 return MY_CS_TOOSMALL4;
158
159 /*
160 This byte must be of the form 11110xxx, and the next three bytes
161 must be continuation bytes.
162 */
163 uint32 four_bytes;
164 memcpy(&four_bytes, s, sizeof(four_bytes));
165 #ifdef WORDS_BIGENDIAN
166 if ((four_bytes & 0xf8c0c0c0) != 0xf0808080)
167 #else
168 if ((four_bytes & 0xc0c0c0f8) != 0x808080f0)
169 #endif
170 return MY_CS_ILSEQ;
171
172 *pwc = ((my_wc_t)(c & 0x07) << 18) + ((my_wc_t)(s[1] & 0x3f) << 12) +
173 ((my_wc_t)(s[2] & 0x3f) << 6) + (my_wc_t)(s[3] & 0x3f);
174 if (*pwc < 0x10000 || *pwc > 0x10ffff) return MY_CS_ILSEQ;
175 return 4;
176 }
177
178 return MY_CS_ILSEQ;
179 }
180
181 /**
182 Parses a single UTF-8 character from a byte string.
183
184 @param[out] pwc the parsed character, if any
185 @param s the string to read from
186 @param e the end of the string; will not read past this
187
188 @return the number of bytes read from s, or a value <= 0 for failure
189 (see m_ctype.h)
190 */
my_mb_wc_utf8(my_wc_t * pwc,const uchar * s,const uchar * e)191 static inline int my_mb_wc_utf8(my_wc_t *pwc, const uchar *s, const uchar *e) {
192 return my_mb_wc_utf8_prototype</*RANGE_CHECK=*/true, /*SUPPORT_MB4=*/false>(
193 pwc, s, e);
194 }
195
196 /**
197 Parses a single UTF-8 character from a byte string. The difference
198 between this and my_mb_wc_utf8 is that this function also can handle
199 four-byte UTF-8 characters.
200
201 @param[out] pwc the parsed character, if any
202 @param s the string to read from
203 @param e the end of the string; will not read past this
204
205 @return the number of bytes read from s, or a value <= 0 for failure
206 (see m_ctype.h)
207 */
my_mb_wc_utf8mb4(my_wc_t * pwc,const uchar * s,const uchar * e)208 static ALWAYS_INLINE int my_mb_wc_utf8mb4(my_wc_t *pwc, const uchar *s,
209 const uchar *e) {
210 return my_mb_wc_utf8_prototype</*RANGE_CHECK=*/true, /*SUPPORT_MB4=*/true>(
211 pwc, s, e);
212 }
213
214 // Non-inlined versions of the above. These are used as function pointers
215 // in MY_CHARSET_HANDLER structs, and you can compare againt them to see
216 // if using the Mb_wc_utf8* functors would be appropriate.
217
218 extern "C" int my_mb_wc_utf8_thunk(const CHARSET_INFO *cs, my_wc_t *pwc,
219 const uchar *s, const uchar *e);
220
221 extern "C" int my_mb_wc_utf8mb4_thunk(const CHARSET_INFO *cs, my_wc_t *pwc,
222 const uchar *s, const uchar *e);
223
224 #endif // MB_WC_INCLUDED
225