xref: /netbsd-src/lib/libc/locale/c8rtomb.c (revision 49c2256c44868b72d61c5781b2aef98ae9ecd7d5)
1 /*	$NetBSD: c8rtomb.c,v 1.9 2024/10/12 16:44:44 rillig Exp $	*/
2 
3 /*-
4  * Copyright (c) 2024 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
17  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
18  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
19  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
20  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
21  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
22  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
23  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
24  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
25  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
26  * POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 /*
30  * c8rtomb(s, c8, ps)
31  *
32  *	Encode the Unicode UTF-8 code unit c8 into the multibyte buffer
33  *	s under the current locale, using multibyte encoding state ps.
34  *
35  *	If c8 is not the last byte of a UTF-8 scalar value sequence, no
36  *	output will be produced, but c8 will be remembered; this must
37  *	be followed by another call passing the following bytes.
38  *
39  *	Return the number of bytes stored on success, or (size_t)-1 on
40  *	error with errno set to EILSEQ.
41  *
42  *	At most MB_CUR_MAX bytes will be stored.
43  *
44  * References:
45  *
46  *	The Unicode Standard, Version 15.0 -- Core Specification, The
47  *	Unicode Consortium, Sec. 3.9 `Unicode Encoding Forms': UTF-8,
48  *	p. 124.
49  *	https://www.unicode.org/versions/Unicode15.0.0/UnicodeStandard-15.0.pdf#page=150
50  *	https://web.archive.org/web/20240718101254/https://www.unicode.org/versions/Unicode15.0.0/UnicodeStandard-15.0.pdf#page=150
51  *
52  *	F. Yergeau, `UTF-8, a transformation format of ISO 10646',
53  *	RFC 3629, Internet Engineering Task Force, November 2003.
54  *	https://datatracker.ietf.org/doc/html/rfc3629
55  */
56 
57 #include <sys/cdefs.h>
58 __RCSID("$NetBSD: c8rtomb.c,v 1.9 2024/10/12 16:44:44 rillig Exp $");
59 
60 #include "namespace.h"
61 
62 #include <assert.h>
63 #include <errno.h>
64 #include <limits.h>
65 #include <locale.h>
66 #include <stdalign.h>
67 #include <stddef.h>
68 #include <stdint.h>
69 #include <uchar.h>
70 
71 #include "c32rtomb.h"
72 #include "setlocale_local.h"
73 
74 struct c8rtombstate {
75 	char32_t	state_c32; /* 8-bit state and 24-bit buffer */
76 	mbstate_t	mbs;
77 };
78 __CTASSERT(offsetof(struct c8rtombstate, mbs) <= sizeof(mbstate_t));
79 __CTASSERT(sizeof(struct c32rtombstate) <= sizeof(mbstate_t) -
80     offsetof(struct c8rtombstate, mbs));
81 __CTASSERT(alignof(struct c8rtombstate) <= alignof(mbstate_t));
82 
83 /*
84  * UTF-8 validation, inspired by Bjoern Hoermann's UTF-8 decoder at
85  * <http://bjoern.hoehrmann.de/utf-8/decoder/dfa/>, but reimplemented
86  * from scratch.
87  */
88 
89 #define	UTF8_ACCEPT	0
90 #define	UTF8_REJECT	96
91 
92 typedef uint8_t utf8_class_t;
93 typedef uint8_t utf8_state_t;
94 
95 static const uint8_t utf8_classtab[] = {
96     0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
97     0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
98     0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
99     0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
100     8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
101     1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
102     8,8,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,
103    11,3,3,3,3,3,3,3,3,3,3,3,3,4,3,3, 7,6,6,6,5,8,8,8,8,8,8,8,8,8,8,8,
104 };
105 
106 static const uint8_t utf8_statetab[] = {
107      0,96,12,36,48,84,72,60,96,96,96,24, 96, 0,96,96,96,96,96,96, 0, 0,96,96,
108     96,12,96,96,96,96,96,96,96,96,96,96, 96,12,96,96,96,96,96,96,12,12,96,96,
109     96,96,96,96,96,96,96,96,12,12,96,96, 96,36,96,96,96,96,96,96,96,36,96,96,
110     96,36,96,96,96,96,96,96,36,36,96,96, 96,96,96,96,96,96,96,96,36,96,96,96,
111     96,96,96,96,96,96,96,96,96,96,96,96,
112 };
113 
114 static utf8_state_t
115 utf8_decode_step(utf8_state_t state, char8_t c8, char32_t *pc32)
116 {
117 	const utf8_class_t class = utf8_classtab[c8];
118 
119 	*pc32 = (state == UTF8_ACCEPT
120 	    ? (c8 & (0xff >> class))
121 	    : ((c8 & 0x3f) | (*pc32 << 6)));
122 
123 	return utf8_statetab[state + class];
124 }
125 
126 #ifdef __weak_alias
127 __weak_alias(c8rtomb_l,_c8rtomb_l)
128 #endif
129 
130 size_t
131 c8rtomb(char *restrict s, char8_t c8, mbstate_t *restrict ps)
132 {
133 
134 	return c8rtomb_l(s, c8, ps, _current_locale());
135 }
136 
137 size_t
138 c8rtomb_l(char *restrict s, char8_t c8, mbstate_t *restrict ps, locale_t loc)
139 {
140 	static mbstate_t psbuf;
141 	char buf[MB_LEN_MAX];
142 	struct c8rtombstate *S;
143 	utf8_state_t state;
144 	char32_t c32;
145 
146 	/*
147 	 * `If ps is a null pointer, each function uses its own
148 	 *  internal mbstate_t object instead, which is initialized at
149 	 *  program startup to the initial conversion state; the
150 	 *  functions are not required to avoid data races with other
151 	 *  calls to the same function in this case.  The
152 	 *  implementation behaves as if no library function calls
153 	 *  these functions with a null pointer for ps.'
154 	 */
155 	if (ps == NULL)
156 		ps = &psbuf;
157 
158 	/*
159 	 * `If s is a null pointer, the c8rtomb function is equivalent
160 	 *  to the call
161 	 *
162 	 *	c8rtomb(buf, u8'\0', ps)
163 	 *
164 	 *  where buf is an internal buffer.
165 	 */
166 	if (s == NULL) {
167 		s = buf;
168 		c8 = 0;		/* XXX u8'\0' */
169 	}
170 
171 	/*
172 	 * Open the private UTF-8 decoding state.
173 	 */
174 	S = (struct c8rtombstate *)(void *)ps;
175 
176 	/*
177 	 * `If c8 is a null character, a null byte is stored, preceded
178 	 *  by any shift sequence needed to restore the initial shift
179 	 *  state; the resulting state described is the initial
180 	 *  conversion state.'
181 	 *
182 	 * So if c8 is null, discard any buffered input -- there's
183 	 * nothing we can legitimately do with it -- and convert a null
184 	 * scalar value, which by definition of c32rtomb writes out any
185 	 * shift sequence reset followed by a null byte.
186 	 */
187 	if (c8 == '\0') {
188 		c32 = 0;
189 		goto accept;
190 	}
191 
192 	/*
193 	 * Get the current state and buffer.
194 	 */
195 	__CTASSERT(UTF8_ACCEPT == 0); /* initial conversion state */
196 	state = __SHIFTOUT(S->state_c32, __BITS(31,24));
197 	c32 = __SHIFTOUT(S->state_c32, __BITS(23,0));
198 
199 	/*
200 	 * Feed the byte into the state machine to update the state.
201 	 */
202 	state = utf8_decode_step(state, c8, &c32);
203 	switch (state) {
204 	case UTF8_REJECT:
205 		/*
206 		 * Invalid UTF-8.  Fail with EILSEQ.
207 		 */
208 		errno = EILSEQ;
209 		return (size_t)-1;
210 	default:
211 		/*
212 		 * Valid UTF-8 so far but incomplete.  Update state and
213 		 * output nothing.
214 		 */
215 		S->state_c32 =
216 		    __SHIFTIN(state, __BITS(31,24)) |
217 		    __SHIFTIN(c32, __BITS(23,0));
218 		return 0;
219 	case UTF8_ACCEPT:
220 	accept:
221 		/*
222 		 * We have a scalar value.  Clear the state and output
223 		 * the scalar value.
224 		 */
225 		S->state_c32 = 0;
226 		return c32rtomb_l(s, c32, &S->mbs, loc);
227 	}
228 }
229