1 /* $Id: preconv.c,v 1.5 2011/07/24 18:15:14 kristaps Exp $ */
3 * Copyright (c) 2011 Kristaps Dzonsons <kristaps@bsd.lv>
5 * Permission to use, copy, modify, and distribute this software for any
6 * purpose with or without fee is hereby granted, provided that the above
7 * copyright notice and this permission notice appear in all copies.
9 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
10 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
11 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
12 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
13 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
14 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
15 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
34 * The read_whole_file() and resize_buf() functions are copied from
35 * read.c, including all dependency code (MAP_FILE, etc.).
43 ENC_UTF_8, /* UTF-8 */
44 ENC_US_ASCII, /* US-ASCII */
45 ENC_LATIN_1, /* Latin-1 */
50 char *buf; /* binary input buffer */
51 size_t sz; /* size of binary buffer */
52 size_t offs; /* starting buffer offset */
57 int (*conv)(const struct buf *);
60 static int cue_enc(const struct buf *, size_t *, enum enc *);
61 static int conv_latin_1(const struct buf *);
62 static int conv_us_ascii(const struct buf *);
63 static int conv_utf_8(const struct buf *);
64 static int read_whole_file(const char *, int,
66 static void resize_buf(struct buf *, size_t);
67 static void usage(void);
69 static const struct encode encs[ENC__MAX] = {
70 { "utf-8", conv_utf_8 }, /* ENC_UTF_8 */
71 { "us-ascii", conv_us_ascii }, /* ENC_US_ASCII */
72 { "latin-1", conv_latin_1 }, /* ENC_LATIN_1 */
75 static const char *progname;
81 fprintf(stderr, "usage: %s "
84 "[file]\n", progname);
88 conv_latin_1(const struct buf *b)
94 cp = b->buf + (int)b->offs;
97 * Latin-1 falls into the first 256 code-points of Unicode, so
98 * there's no need for any sort of translation. Just make the
99 * 8-bit characters use the Unicode escape.
100 * Note that binary values 128 < v < 160 are passed through
101 * unmodified to mandoc.
104 for (i = b->offs; i < b->sz; i++) {
105 cu = (unsigned char)*cp++;
106 cu < 160U ? putchar(cu) : printf("\\[u%.4X]", cu);
113 conv_us_ascii(const struct buf *b)
117 * US-ASCII has no conversion since it falls into the first 128
121 fwrite(b->buf, 1, b->sz, stdout);
126 conv_utf_8(const struct buf *b)
135 cp = b->buf + (int)b->offs;
140 /* Quick test for big-endian value. */
142 if ( ! (*((const char *)(&one))))
145 for (i = b->offs; i < b->sz; i++) {
146 cu = (unsigned char)*cp++;
148 if ( ! (cu & 128) || (cu & 64)) {
149 /* Bad sequence header. */
153 /* Accept only legitimate bit patterns. */
155 if (cu > 191 || cu < 128) {
156 /* Bad in-sequence bits. */
160 accum |= (cu & 63) << --state * 6;
163 * Accum is held in little-endian order as
164 * stipulated by the UTF-8 sequence coding. We
165 * need to convert to a native big-endian if our
166 * architecture requires it.
169 if (0 == state && be)
170 accum = (accum >> 24) |
171 ((accum << 8) & 0x00FF0000) |
172 ((accum >> 8) & 0x0000FF00) |
176 accum < 128U ? putchar(accum) :
177 printf("\\[u%.4X]", accum);
180 } else if (cu & (1 << 7)) {
182 * Entering a UTF-8 state: if we encounter a
183 * UTF-8 bitmask, calculate the expected UTF-8
186 for (state = 0; state < 7; state++)
187 if ( ! (cu & (1 << (7 - state))))
190 /* Accept only legitimate bit patterns. */
194 if (cu <= 244 && cu >= 240) {
195 accum = (cu & 7) << 18;
198 /* Bad 4-sequence start bits. */
201 if (cu <= 239 && cu >= 224) {
202 accum = (cu & 15) << 12;
205 /* Bad 3-sequence start bits. */
208 if (cu <= 223 && cu >= 194) {
209 accum = (cu & 31) << 6;
212 /* Bad 2-sequence start bits. */
215 /* Bad sequence bit mask. */
224 /* Bad trailing bits. */
232 resize_buf(struct buf *buf, size_t initial)
235 buf->sz = buf->sz > initial / 2 ?
236 2 * buf->sz : initial;
238 buf->buf = realloc(buf->buf, buf->sz);
239 if (NULL == buf->buf) {
246 read_whole_file(const char *f, int fd,
247 struct buf *fb, int *with_mmap)
254 if (-1 == fstat(fd, &st)) {
260 * If we're a regular file, try just reading in the whole entry
261 * via mmap(). This is faster than reading it into blocks, and
262 * since each file is only a few bytes to begin with, I'm not
263 * concerned that this is going to tank any machines.
266 if (S_ISREG(st.st_mode) && st.st_size >= (1U << 31)) {
267 fprintf(stderr, "%s: input too large\n", f);
271 if (S_ISREG(st.st_mode)) {
273 fb->sz = (size_t)st.st_size;
274 fb->buf = mmap(NULL, fb->sz, PROT_READ,
275 MAP_FILE|MAP_SHARED, fd, 0);
276 if (fb->buf != MAP_FAILED)
282 * If this isn't a regular file (like, say, stdin), then we must
283 * go the old way and just read things in bit by bit.
291 if (off == fb->sz && fb->sz == (1U << 31)) {
292 fprintf(stderr, "%s: input too large\n", f);
297 resize_buf(fb, 65536);
299 ssz = read(fd, fb->buf + (int)off, fb->sz - off);
317 cue_enc(const struct buf *b, size_t *offs, enum enc *enc)
319 const char *ln, *eoln, *eoph;
320 size_t sz, phsz, nsz;
323 ln = b->buf + (int)*offs;
326 /* Look for the end-of-line. */
328 if (NULL == (eoln = memchr(ln, '\n', sz)))
331 /* Set next-line marker. */
333 *offs = (size_t)((eoln + 1) - b->buf);
335 /* Check if we have the correct header/trailer. */
337 if ((sz = (size_t)(eoln - ln)) < 10 ||
338 memcmp(ln, ".\\\" -*-", 7) ||
339 memcmp(eoln - 3, "-*-", 3))
342 /* Move after the header and adjust for the trailer. */
348 while (sz > 0 && ' ' == *ln) {
355 /* Find the end-of-phrase marker (or eoln). */
357 if (NULL == (eoph = memchr(ln, ';', sz)))
362 /* Only account for the "coding" phrase. */
364 if ((phsz = (size_t)(eoph - ln)) < 7 ||
365 strncasecmp(ln, "coding:", 7)) {
374 while (sz > 0 && ' ' == *ln) {
381 /* Check us against known encodings. */
383 for (i = 0; i < (int)ENC__MAX; i++) {
384 nsz = strlen(encs[i].name);
387 if (strncasecmp(ln, encs[i].name, nsz))
394 /* Unknown encoding. */
404 main(int argc, char *argv[])
406 int i, ch, map, fd, rc;
410 unsigned char bom[3] = { 0xEF, 0xBB, 0xBF };
415 progname = strrchr(argv[0], '/');
416 if (progname == NULL)
424 enc = def = ENC__MAX;
427 memset(&b, 0, sizeof(struct buf));
429 while (-1 != (ch = getopt(argc, argv, "D:e:rdvh")))
434 for (i = 0; i < (int)ENC__MAX; i++) {
435 if (strcasecmp(optarg, encs[i].name))
439 if (i < (int)ENC__MAX) {
447 fprintf(stderr, "%s: Bad encoding\n", optarg);
448 return(EXIT_FAILURE);
454 /* Compatibility with GNU preconv. */
457 /* Compatibility with GNU preconv. */
461 return(EXIT_FAILURE);
468 * Open and read the first argument on the command-line.
469 * If we don't have one, we default to stdin.
474 fd = open(fn, O_RDONLY, 0);
477 return(EXIT_FAILURE);
481 if ( ! read_whole_file(fn, fd, &b, &map))
484 /* Try to read the UTF-8 BOM. */
487 if (b.sz > 3 && 0 == memcmp(b.buf, bom, 3)) {
492 /* Try reading from the "-*-" cue. */
494 if (ENC__MAX == enc) {
496 ch = cue_enc(&b, &offs, &enc);
498 ch = cue_enc(&b, &offs, &enc);
502 * No encoding has been detected.
503 * Thus, we either fall into our default encoder, if specified,
504 * or use Latin-1 if all else fails.
508 enc = ENC__MAX == def ? ENC_LATIN_1 : def;
510 if ( ! (*encs[(int)enc].conv)(&b)) {
511 fprintf(stderr, "%s: Bad encoding\n", fn);
524 if (fd > STDIN_FILENO)