busybox/libbb/unicode.c

/* vi: set sw=4 ts=4: */
/*
 * Unicode support routines.
 *
 * Copyright (C) 2009 Denys Vlasenko
 *
 * Licensed under GPL version 2, see file LICENSE in this tarball for details.
 */
#include "libbb.h"
#include "unicode.h"

/* If it's not #defined as a constant in unicode.h... */
#ifndef unicode_status
uint8_t unicode_status;
#endif

/* This file is compiled only if FEATURE_ASSUME_UNICODE is on.
 * We check other options and decide whether to use libc support
 * via locale, or use our own logic:
 */

#if ENABLE_LOCALE_SUPPORT

/* Unicode support using libc locale support. */

void FAST_FUNC init_unicode(void)
{
	/* In unicode, this is a one character string */
	static const char unicode_0x394[] = { 0xce, 0x94, 0 };

	if (unicode_status != UNICODE_UNKNOWN)
		return;

	unicode_status = unicode_strlen(unicode_0x394) == 1 ? UNICODE_ON : UNICODE_OFF;
}

#else

/* Homegrown Unicode support. It knows only C and Unicode locales. */

# if ENABLE_FEATURE_CHECK_UNICODE_IN_ENV
void FAST_FUNC init_unicode(void)
{
	char *lang;

	if (unicode_status != UNICODE_UNKNOWN)
		return;

	unicode_status = UNICODE_OFF;
	lang = getenv("LANG");
	if (!lang || !(strstr(lang, ".utf") || strstr(lang, ".UTF")))
		return;
	unicode_status = UNICODE_ON;
}
# endif

static size_t wcrtomb_internal(char *s, wchar_t wc)
{
	int n, i;
	uint32_t v = wc;

	if (v <= 0x7f) {
		*s = v;
		return 1;
	}

	/* RFC 3629 says that Unicode ends at 10FFFF,
	 * but we cover entire 32 bits */

	/* 4000000-FFFFFFFF -> 111111tt 10tttttt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
	/* 200000-3FFFFFF -> 111110tt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
	/* 10000-1FFFFF -> 11110zzz 10zzyyyy 10yyyyxx 10xxxxxx */
	/* 800-FFFF -> 1110yyyy 10yyyyxx 10xxxxxx */
	/* 80-7FF -> 110yyyxx 10xxxxxx */

	/* How many bytes do we need? */
	n = 2;
	/* (0x80000000+ would result in n = 7, limiting n to 6) */
	while (v >= 0x800 && n < 6) {
		v >>= 5;
		n++;
	}
	/* Fill bytes n-1..1 */
	i = n;
	while (--i) {
		s[i] = (wc & 0x3f) | 0x80;
		wc >>= 6;
	}
	/* Fill byte 0 */
	s[0] = wc | (uint8_t)(0x3f00 >> n);
	return n;
}
size_t FAST_FUNC wcrtomb(char *s, wchar_t wc, mbstate_t *ps UNUSED_PARAM)
{
	if (unicode_status != UNICODE_ON) {
		*s = wc;
		return 1;
	}

	return wcrtomb_internal(s, wc);
}
size_t FAST_FUNC wcstombs(char *dest, const wchar_t *src, size_t n)
{
	size_t org_n = n;

	if (unicode_status != UNICODE_ON) {
		while (n) {
			wchar_t c = *src++;
			*dest++ = c;
			if (c == 0)
				break;
			n--;
		}
		return org_n - n;
	}

	while (n >= MB_CUR_MAX) {
		wchar_t wc = *src++;
		size_t len = wcrtomb_internal(dest, wc);

		if (wc == L'\0')
			return org_n - n;
		dest += len;
		n -= len;
	}
	while (n) {
		char tbuf[MB_CUR_MAX];
		wchar_t wc = *src++;
		size_t len = wcrtomb_internal(tbuf, wc);

		if (len > n)
			len = n;
		memcpy(dest, tbuf, len);
		if (wc == L'\0')
			return org_n - n;
		dest += len;
		n -= len;
	}
	return org_n - n;
}

static const char *mbstowc_internal(wchar_t *res, const char *src)
{
	int bytes;
	unsigned c = (unsigned char) *src++;

	if (c <= 0x7f) {
		*res = c;
		return src;
	}

	/* 80-7FF -> 110yyyxx 10xxxxxx */
	/* 800-FFFF -> 1110yyyy 10yyyyxx 10xxxxxx */
	/* 10000-1FFFFF -> 11110zzz 10zzyyyy 10yyyyxx 10xxxxxx */
	/* 200000-3FFFFFF -> 111110tt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
	/* 4000000-FFFFFFFF -> 111111tt 10tttttt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
	bytes = 0;
	do {
		c <<= 1;
		bytes++;
	} while ((c & 0x80) && bytes < 6);
	if (bytes == 1)
		return NULL;
	c = (uint8_t)(c) >> bytes;

	while (--bytes) {
		unsigned ch = (unsigned char) *src++;
		if ((ch & 0xc0) != 0x80) {
			return NULL;
		}
		c = (c << 6) + (ch & 0x3f);
	}

	/* TODO */
	/* Need to check that c isn't produced by overlong encoding */
	/* Example: 11000000 10000000 converts to NUL */
	/* 11110000 10000000 10000100 10000000 converts to 0x100 */
	/* correct encoding: 11000100 10000000 */
	if (c <= 0x7f) { /* crude check */
		return NULL;
		//or maybe 0xfffd; /* replacement character */
	}

	*res = c;
	return src;
}
size_t FAST_FUNC mbstowcs(wchar_t *dest, const char *src, size_t n)
{
	size_t org_n = n;

	if (unicode_status != UNICODE_ON) {
		while (n) {
			unsigned char c = *src++;

			if (dest)
				*dest++ = c;
			if (c == 0)
				break;
			n--;
		}
		return org_n - n;
	}

	while (n) {
		wchar_t wc;
		src = mbstowc_internal(&wc, src);
		if (src == NULL) /* error */
			return (size_t) -1L;
		if (dest)
			*dest++ = wc;
		if (wc == 0) /* end-of-string */
			break;
		n--;
	}

	return org_n - n;
}

#include "unicode_wcwidth.c"

int FAST_FUNC iswspace(wint_t wc)
{
	return (unsigned)wc <= 0x7f && isspace(wc);
}

int FAST_FUNC iswalnum(wint_t wc)
{
	return (unsigned)wc <= 0x7f && isalnum(wc);
}

int FAST_FUNC iswpunct(wint_t wc)
{
	return (unsigned)wc <= 0x7f && ispunct(wc);
}

#endif /* Homegrown Unicode support */


/* The rest is mostly same for libc and for "homegrown" support */

size_t FAST_FUNC unicode_strlen(const char *string)
{
	size_t width = mbstowcs(NULL, string, INT_MAX);
	if (width == (size_t)-1L)
		return strlen(string);
	return width;
}

char* FAST_FUNC unicode_cut_nchars(unsigned width, const char *src)
{
	char *dst;
	unsigned dst_len;

	if (unicode_status != UNICODE_ON)
		return xasprintf("%-*.*s", width, width, src);

	dst = NULL;
	dst_len = 0;
	while (1) {
		int w;
		wchar_t wc;

		dst = xrealloc(dst, dst_len + 2 * MB_CUR_MAX);
#if ENABLE_LOCALE_SUPPORT
		{
			mbstate_t mbst = { 0 };
			ssize_t rc = mbsrtowcs(&wc, &src, 1, &mbst);
			if (rc <= 0) /* error, or end-of-string */
				break;
		}
#else
		src = mbstowc_internal(&wc, src);
		if (!src || wc == 0) /* error, or end-of-string */
			break;
#endif
		w = wcwidth(wc);
		if (w < 0) /* non-printable wchar */
			break;
		width -= w;
		if ((int)width < 0) { /* string is longer than width */
			width += w;
			while (width) {
				dst[dst_len++] = ' ';
				width--;
			}
			break;
		}
#if ENABLE_LOCALE_SUPPORT
		{
			mbstate_t mbst = { 0 };
			dst_len += wcrtomb(&dst[dst_len], wc, &mbst);
		}
#else
		dst_len += wcrtomb_internal(&dst[dst_len], wc);
#endif
	}
	dst[dst_len] = '\0';
	return dst;
}

unsigned FAST_FUNC unicode_padding_to_width(unsigned width, const char *src)
{
	if (unicode_status != UNICODE_ON) {
		return width - strnlen(src, width);
	}

	while (1) {
		int w;
		wchar_t wc;

#if ENABLE_LOCALE_SUPPORT
		{
			mbstate_t mbst = { 0 };
			ssize_t rc = mbsrtowcs(&wc, &src, 1, &mbst);
			if (rc <= 0) /* error, or end-of-string */
				return width;
		}
#else
		src = mbstowc_internal(&wc, src);
		if (!src || wc == 0) /* error, or end-of-string */
			return width;
#endif
		w = wcwidth(wc);
		if (w < 0) /* non-printable wchar */
			return width;
		width -= w;
		if ((int)width <= 0) /* string is longer than width */
			return 0;
	}
}
1	/* vi: set sw=4 ts=4: */
2	/*
3	* Unicode support routines.
4	*
5	* Copyright (C) 2009 Denys Vlasenko
6	*
7	* Licensed under GPL version 2, see file LICENSE in this tarball for details.
8	*/
9	#include "libbb.h"
10	#include "unicode.h"
11
12	/* If it's not #defined as a constant in unicode.h... */
13	#ifndef unicode_status
14	uint8_t unicode_status;
15	#endif
16
17	/* This file is compiled only if FEATURE_ASSUME_UNICODE is on.
18	* We check other options and decide whether to use libc support
19	* via locale, or use our own logic:
20	*/
21
22	#if ENABLE_LOCALE_SUPPORT
23
24	/* Unicode support using libc locale support. */
25
26	void FAST_FUNC init_unicode(void)
27	{
28	/* In unicode, this is a one character string */
29	static const char unicode_0x394[] = { 0xce, 0x94, 0 };
30
31	if (unicode_status != UNICODE_UNKNOWN)
32	return;
33
34	unicode_status = unicode_strlen(unicode_0x394) == 1 ? UNICODE_ON : UNICODE_OFF;
35	}
36
37	#else
38
39	/* Homegrown Unicode support. It knows only C and Unicode locales. */
40
41	# if ENABLE_FEATURE_CHECK_UNICODE_IN_ENV
42	void FAST_FUNC init_unicode(void)
43	{
44	char *lang;
45
46	if (unicode_status != UNICODE_UNKNOWN)
47	return;
48
49	unicode_status = UNICODE_OFF;
50	lang = getenv("LANG");
51	if (!lang \|\| !(strstr(lang, ".utf") \|\| strstr(lang, ".UTF")))
52	return;
53	unicode_status = UNICODE_ON;
54	}
55	# endif
56
57	static size_t wcrtomb_internal(char *s, wchar_t wc)
58	{
59	int n, i;
60	uint32_t v = wc;
61
62	if (v <= 0x7f) {
63	*s = v;
64	return 1;
65	}
66
67	/* RFC 3629 says that Unicode ends at 10FFFF,
68	* but we cover entire 32 bits */
69
70	/* 4000000-FFFFFFFF -> 111111tt 10tttttt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
71	/* 200000-3FFFFFF -> 111110tt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
72	/* 10000-1FFFFF -> 11110zzz 10zzyyyy 10yyyyxx 10xxxxxx */
73	/* 800-FFFF -> 1110yyyy 10yyyyxx 10xxxxxx */
74	/* 80-7FF -> 110yyyxx 10xxxxxx */
75
76	/* How many bytes do we need? */
77	n = 2;
78	/* (0x80000000+ would result in n = 7, limiting n to 6) */
79	while (v >= 0x800 && n < 6) {
80	v >>= 5;
81	n++;
82	}
83	/* Fill bytes n-1..1 */
84	i = n;
85	while (--i) {
86	s[i] = (wc & 0x3f) \| 0x80;
87	wc >>= 6;
88	}
89	/* Fill byte 0 */
90	s[0] = wc \| (uint8_t)(0x3f00 >> n);
91	return n;
92	}
93	size_t FAST_FUNC wcrtomb(char s, wchar_t wc, mbstate_t ps UNUSED_PARAM)
94	{
95	if (unicode_status != UNICODE_ON) {
96	*s = wc;
97	return 1;
98	}
99
100	return wcrtomb_internal(s, wc);
101	}
102	size_t FAST_FUNC wcstombs(char dest, const wchar_t src, size_t n)
103	{
104	size_t org_n = n;
105
106	if (unicode_status != UNICODE_ON) {
107	while (n) {
108	wchar_t c = *src++;
109	*dest++ = c;
110	if (c == 0)
111	break;
112	n--;
113	}
114	return org_n - n;
115	}
116
117	while (n >= MB_CUR_MAX) {
118	wchar_t wc = *src++;
119	size_t len = wcrtomb_internal(dest, wc);
120
121	if (wc == L'\0')
122	return org_n - n;
123	dest += len;
124	n -= len;
125	}
126	while (n) {
127	char tbuf[MB_CUR_MAX];
128	wchar_t wc = *src++;
129	size_t len = wcrtomb_internal(tbuf, wc);
130
131	if (len > n)
132	len = n;
133	memcpy(dest, tbuf, len);
134	if (wc == L'\0')
135	return org_n - n;
136	dest += len;
137	n -= len;
138	}
139	return org_n - n;
140	}
141
142	static const char mbstowc_internal(wchar_t res, const char *src)
143	{
144	int bytes;
145	unsigned c = (unsigned char) *src++;
146
147	if (c <= 0x7f) {
148	*res = c;
149	return src;
150	}
151
152	/* 80-7FF -> 110yyyxx 10xxxxxx */
153	/* 800-FFFF -> 1110yyyy 10yyyyxx 10xxxxxx */
154	/* 10000-1FFFFF -> 11110zzz 10zzyyyy 10yyyyxx 10xxxxxx */
155	/* 200000-3FFFFFF -> 111110tt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
156	/* 4000000-FFFFFFFF -> 111111tt 10tttttt 10zzzzzz 10zzyyyy 10yyyyxx 10xxxxxx */
157	bytes = 0;
158	do {
159	c <<= 1;
160	bytes++;
161	} while ((c & 0x80) && bytes < 6);
162	if (bytes == 1)
163	return NULL;
164	c = (uint8_t)(c) >> bytes;
165
166	while (--bytes) {
167	unsigned ch = (unsigned char) *src++;
168	if ((ch & 0xc0) != 0x80) {
169	return NULL;
170	}
171	c = (c << 6) + (ch & 0x3f);
172	}
173
174	/* TODO */
175	/* Need to check that c isn't produced by overlong encoding */
176	/* Example: 11000000 10000000 converts to NUL */
177	/* 11110000 10000000 10000100 10000000 converts to 0x100 */
178	/* correct encoding: 11000100 10000000 */
179	if (c <= 0x7f) { /* crude check */
180	return NULL;
181	//or maybe 0xfffd; /* replacement character */
182	}
183
184	*res = c;
185	return src;
186	}
187	size_t FAST_FUNC mbstowcs(wchar_t dest, const char src, size_t n)
188	{
189	size_t org_n = n;
190
191	if (unicode_status != UNICODE_ON) {
192	while (n) {
193	unsigned char c = *src++;
194
195	if (dest)
196	*dest++ = c;
197	if (c == 0)
198	break;
199	n--;
200	}
201	return org_n - n;
202	}
203
204	while (n) {
205	wchar_t wc;
206	src = mbstowc_internal(&wc, src);
207	if (src == NULL) /* error */
208	return (size_t) -1L;
209	if (dest)
210	*dest++ = wc;
211	if (wc == 0) /* end-of-string */
212	break;
213	n--;
214	}
215
216	return org_n - n;
217	}
218
219	#include "unicode_wcwidth.c"
220
221	int FAST_FUNC iswspace(wint_t wc)
222	{
223	return (unsigned)wc <= 0x7f && isspace(wc);
224	}
225
226	int FAST_FUNC iswalnum(wint_t wc)
227	{
228	return (unsigned)wc <= 0x7f && isalnum(wc);
229	}
230
231	int FAST_FUNC iswpunct(wint_t wc)
232	{
233	return (unsigned)wc <= 0x7f && ispunct(wc);
234	}
235
236	#endif /* Homegrown Unicode support */
237
238
239	/* The rest is mostly same for libc and for "homegrown" support */
240
241	size_t FAST_FUNC unicode_strlen(const char *string)
242	{
243	size_t width = mbstowcs(NULL, string, INT_MAX);
244	if (width == (size_t)-1L)
245	return strlen(string);
246	return width;
247	}
248
249	char* FAST_FUNC unicode_cut_nchars(unsigned width, const char *src)
250	{
251	char *dst;
252	unsigned dst_len;
253
254	if (unicode_status != UNICODE_ON)
255	return xasprintf("%-.s", width, width, src);
256
257	dst = NULL;
258	dst_len = 0;
259	while (1) {
260	int w;
261	wchar_t wc;
262
263	dst = xrealloc(dst, dst_len + 2 * MB_CUR_MAX);
264	#if ENABLE_LOCALE_SUPPORT
265	{
266	mbstate_t mbst = { 0 };
267	ssize_t rc = mbsrtowcs(&wc, &src, 1, &mbst);
268	if (rc <= 0) /* error, or end-of-string */
269	break;
270	}
271	#else
272	src = mbstowc_internal(&wc, src);
273	if (!src \|\| wc == 0) /* error, or end-of-string */
274	break;
275	#endif
276	w = wcwidth(wc);
277	if (w < 0) /* non-printable wchar */
278	break;
279	width -= w;
280	if ((int)width < 0) { /* string is longer than width */
281	width += w;
282	while (width) {
283	dst[dst_len++] = ' ';
284	width--;
285	}
286	break;
287	}
288	#if ENABLE_LOCALE_SUPPORT
289	{
290	mbstate_t mbst = { 0 };
291	dst_len += wcrtomb(&dst[dst_len], wc, &mbst);
292	}
293	#else
294	dst_len += wcrtomb_internal(&dst[dst_len], wc);
295	#endif
296	}
297	dst[dst_len] = '\0';
298	return dst;
299	}
300
301	unsigned FAST_FUNC unicode_padding_to_width(unsigned width, const char *src)
302	{
303	if (unicode_status != UNICODE_ON) {
304	return width - strnlen(src, width);
305	}
306
307	while (1) {
308	int w;
309	wchar_t wc;
310
311	#if ENABLE_LOCALE_SUPPORT
312	{
313	mbstate_t mbst = { 0 };
314	ssize_t rc = mbsrtowcs(&wc, &src, 1, &mbst);
315	if (rc <= 0) /* error, or end-of-string */
316	return width;
317	}
318	#else
319	src = mbstowc_internal(&wc, src);
320	if (!src \|\| wc == 0) /* error, or end-of-string */
321	return width;
322	#endif
323	w = wcwidth(wc);
324	if (w < 0) /* non-printable wchar */
325	return width;
326	width -= w;
327	if ((int)width <= 0) /* string is longer than width */
328	return 0;
329	}
330	}