xref: /titanic_44/usr/src/cmd/ast/msgcc/msgcvt.c (revision a38ddfee9c8c6b6c5a2947ff52fd2338362a4444) 
1 /***********************************************************************
2 *                                                                      *
3 *               This software is part of the ast package               *
4 *           Copyright (c) 2000-2007 AT&T Knowledge Ventures            *
5 *                      and is licensed under the                       *
6 *                  Common Public License, Version 1.0                  *
7 *                      by AT&T Knowledge Ventures                      *
8 *                                                                      *
9 *                A copy of the License is available at                 *
10 *            http://www.opensource.org/licenses/cpl1.0.txt             *
11 *         (with md5 checksum 059e8cd6165cb4c31e351f2b69388fd9)         *
12 *                                                                      *
13 *              Information and Software Systems Research               *
14 *                            AT&T Research                             *
15 *                           Florham Park NJ                            *
16 *                                                                      *
17 *                 Glenn Fowler <gsf@research.att.com>                  *
18 *                                                                      *
19 ***********************************************************************/
20 #pragma prototyped
21 /*
22  * Glenn Fowler
23  * AT&T Research
24  */
25 
26 static const char usage[] =
27 "[-?\n@(#)$Id: msgcvt (AT&T Research) 2000-05-01 $\n]"
28 USAGE_LICENSE
29 "[+NAME?msgcvt - convert message file to/from html]"
30 "[+DESCRIPTION?\bmsgcvt\b reads a \bgencat\b(1) format file on the standard"
31 "	input and converts it to \bhtml\b on the standard output. The input"
32 "	file must contain the control statement \b$quote \"\b and use the \""
33 "	character to quote message text. The output is in a form suitable for"
34 "	automatic translation by web sites like"
35 "	\bhttp://babelfish.altavista.com/\b or filters like"
36 "	\btranslate\b(1).]"
37 "[h:html?Generate \bhtml\b from \bgencat\b(1) input. This is the default.]"
38 "[m:msg?Generate a \bgencat\b(1) message file from (presumably translated)"
39 "	\bhtml\b. Wide characters are UTF-8 encoded.]"
40 "[r:raw?The message file is raw message text, one message per line, with no"
41 "	quoting or line numbering.]"
42 "[+SEE ALSO?\bgencat\b(1), \bmsgcc\b(1), \bmsggen\b(1), \btranslate\b(1)]"
43 ;
44 
45 #include <ast.h>
46 #include <ctype.h>
47 #include <error.h>
48 
49 #define MSG_RAW		(1<<0)
50 #define MSG_SPLICE	(1<<1)
51 
52 #define SPACE(s)	(isspace(*s)&&(s+=1)||*s=='\\'&&(*(s+1)=='n'||*(s+1)=='t')&&(s+=2))
53 
54 typedef void (*Convert_f)(Sfio_t*, Sfio_t*, int);
55 
56 typedef struct
57 {
58 	const char*	name;
59 	int		code;
60 } Code_t;
61 
62 static const Code_t	codes[] =
63 {
64 	"aacute",	225,
65 	"Aacute",	193,
66 	"acirc",	226,
67 	"Acirc",	194,
68 	"aelig",	230,
69 	"AElig",	198,
70 	"agrave",	224,
71 	"Agrave",	192,
72 	"amp",		'&',
73 	"aring",	229,
74 	"Aring",	197,
75 	"atilde",	227,
76 	"Atilde",	195,
77 	"auml",		228,
78 	"Auml",		196,
79 	"ccedil",	231,
80 	"Ccedil",	199,
81 	"copy",		169,
82 	"eacute",	233,
83 	"Eacute",	201,
84 	"ecirc",	234,
85 	"Ecirc",	202,
86 	"egrave",	232,
87 	"Egrave",	200,
88 	"euml",		235,
89 	"Euml",		203,
90 	"gt",		'>',
91 	"iacute",	237,
92 	"Iacute",	205,
93 	"icirc",	238,
94 	"Icirc",	206,
95 	"igrave",	236,
96 	"Igrave",	204,
97 	"iuml",		239,
98 	"Iuml",		207,
99 	"lt",		'<',
100 	"nbsp",		' ',
101 	"ntilde",	241,
102 	"Ntilde",	209,
103 	"oacute",	243,
104 	"Oacute",	211,
105 	"ocirc",	244,
106 	"Ocirc",	212,
107 	"ograve",	242,
108 	"Ograve",	210,
109 	"oslash",	248,
110 	"Oslash",	216,
111 	"otilde",	245,
112 	"Otilde",	213,
113 	"ouml",		246,
114 	"Ouml",		214,
115 	"quot",		'"',
116 	"reg",		174,
117 	"szlig",	223,
118 	"uacute",	250,
119 	"Uacute",	218,
120 	"ucirc",	251,
121 	"Ucirc",	219,
122 	"ugrave",	249,
123 	"Ugrave",	217,
124 	"uuml",		252,
125 	"Uuml",		220,
126 	"yuml",		255,
127 };
128 
129 static int
130 decode(Sfio_t* ip)
131 {
132 	register int	c;
133 	register int	i;
134 	char		name[32];
135 
136 	if ((c = sfgetc(ip)) == EOF)
137 		return '&';
138 	name[0] = c;
139 	i = 1;
140 	if (c != '#' && !isalpha(c))
141 		goto bad;
142 	while ((c = sfgetc(ip)) != EOF && c != ';')
143 	{
144 		if (c == '&')
145 			i = 0;
146 		else
147 		{
148 			name[i++] = c;
149 			if (!isalnum(c) && (i > 1 || c != '#') || i >= (elementsof(name) - 1))
150 				goto bad;
151 		}
152 	}
153 	name[i] = 0;
154 	if (name[0] == '#')
155 	{
156 		switch (c = strtol(name + 1, NiL, 10))
157 		{
158 		case 91:
159 			c = '[';
160 			break;
161 		case 93:
162 			c = ']';
163 			break;
164 		}
165 	}
166 	else
167 	{
168 		for (i = 0; i < elementsof(codes); i++)
169 			if (streq(codes[i].name, name))
170 			{
171 				c = codes[i].code;
172 				break;
173 			}
174 		if (i >= elementsof(codes))
175 			goto bad;
176 	}
177 	return c;
178  bad:
179 	name[i] = 0;
180 	if (c == ';')
181 		error(1, "&%s: unknown HTML special character -- & assumed", name);
182 	else
183 		error(1, "&%s: invalid HTML special character -- & assumed", name);
184 	while (i--)
185 		sfungetc(ip, name[i]);
186 	return '&';
187 }
188 
189 static int
190 sfpututf(Sfio_t* op, register int w)
191 {
192 	if (!(w & ~0x7F))
193 		return sfputc(op, w);
194 	else if (!(w & ~0x7FF))
195 		sfputc(op, 0xC0 + (w >> 6));
196 	else if (!(w & ~0xFFFF))
197 	{
198 		sfputc(op, 0xE0 + (w >> 12));
199 		sfputc(op, 0x80 + (w >> 6 ) & 0x3F);
200 	}
201 	else
202 		return sfputc(op, '?');
203 	return sfputc(op, 0x80 + (w & 0x3F));
204 }
205 
206 static int
207 sfnext(Sfio_t* ip)
208 {
209 	register int	c;
210 
211 	while (isspace(c = sfgetc(ip)));
212 	return c;
213 }
214 
215 static void
216 html2msg(register Sfio_t* ip, register Sfio_t* op, int flags)
217 {
218 	register int	c;
219 	register int	q;
220 
221  again:
222 	while ((c = sfgetc(ip)) != EOF)
223 		if (c == '<')
224 		{
225 			if ((c = sfnext(ip)) == 'O' &&
226 			    (c = sfnext(ip)) == 'L' &&
227 			    isspace(c = sfgetc(ip)) &&
228 			    (c = sfnext(ip)) == 'S' &&
229 			    (c = sfnext(ip)) == 'T' &&
230 			    (c = sfnext(ip)) == 'A' &&
231 			    (c = sfnext(ip)) == 'R' &&
232 			    (c = sfnext(ip)) == 'T' &&
233 			    (c = sfnext(ip)) == '=' &&
234 			    (c = sfnext(ip)) == '"' &&
235 			    (c = sfnext(ip)) == '5' &&
236 			    (c = sfnext(ip)) == '5' &&
237 			    (c = sfnext(ip)) == '0' &&
238 			    (c = sfnext(ip)) == '7' &&
239 			    (c = sfnext(ip)) == '1' &&
240 			    (c = sfnext(ip)) == '7' &&
241 			    (c = sfnext(ip)) == '"' &&
242 			    (c = sfnext(ip)) == '>')
243 				break;
244 			while (c != EOF && c != '>')
245 				c = sfgetc(ip);
246 		}
247 	if ((c = sfnext(ip)) != EOF)
248 		sfungetc(ip, c);
249 	q = 0;
250 	for (;;)
251 	{
252 		switch (c = sfgetc(ip))
253 		{
254 		case EOF:
255 			break;
256 		case '&':
257 			c = decode(ip);
258 			sfpututf(op, c);
259 			if (isspace(c))
260 			{
261 				while (isspace(c = sfgetc(ip)));
262 				if (c == EOF)
263 					break;
264 				sfungetc(ip, c);
265 			}
266 			continue;
267 		case '<':
268 			switch (c = sfnext(ip))
269 			{
270 			case '/':
271 				if ((c = sfnext(ip)) == 'O' &&
272 				    (c = sfgetc(ip)) == 'L' &&
273 				    (c = sfnext(ip)) == '>')
274 				{
275 					if (q)
276 					{
277 						sfputc(op, q);
278 						q = '"';
279 					}
280 					goto again;
281 				}
282 				break;
283 			case 'B':
284 				if ((c = sfgetc(ip)) == 'R' &&
285 				    (c = sfnext(ip)) == '>')
286 					sfputc(op, ' ');
287 				break;
288 			case 'L':
289 				if ((c = sfgetc(ip)) == 'I' &&
290 				    (c = sfnext(ip)) == '>' &&
291 				     isdigit(c = sfnext(ip)))
292 				{
293 					if (q)
294 						sfputc(op, q);
295 					else
296 						q = '"';
297 					sfputc(op, '\n');
298 					do
299 					{
300 						sfputc(op, c);
301 					} while (isdigit(c = sfgetc(ip)));
302 					if (c == EOF)
303 						break;
304 					sfputc(op, ' ');
305 					sfputc(op, '"');
306 					if (isspace(c))
307 						c = sfnext(ip);
308 					if (c == '<' &&
309 					    (c = sfnext(ip)) == 'L' &&
310 					    (c = sfgetc(ip)) == 'I' &&
311 					    (c = sfnext(ip)) == '>')
312 						/* great */;
313 					continue;
314 				}
315 				break;
316 			case 'P':
317 				if ((c = sfnext(ip)) == '>')
318 					sfputc(op, '\n');
319 				else if (c == 'C' &&
320 					 (c = sfgetc(ip)) == 'L' &&
321 					 (c = sfgetc(ip)) == 'A' &&
322 					 (c = sfgetc(ip)) == 'S' &&
323 					 (c = sfgetc(ip)) == 'S' &&
324 					 (c = sfnext(ip)) == '=' &&
325 					 (c = sfnext(ip)) == '"')
326 					for (;;)
327 					{
328 						switch (c = sfgetc(ip))
329 						{
330 						case EOF:
331 						case '"':
332 							break;
333 						case '&':
334 							c = decode(ip);
335 							sfpututf(op, c);
336 							continue;
337 						default:
338 							sfpututf(op, c);
339 							continue;
340 						}
341 						break;
342 					}
343 				break;
344 			}
345 			while (c != EOF && c != '>')
346 				c = sfgetc(ip);
347 			if (c == EOF || (c = sfgetc(ip)) == EOF)
348 				break;
349 			sfungetc(ip, c);
350 			continue;
351 		case '"':
352 			if (!flags)
353 				sfputc(op, '\\');
354 			sfputc(op, c);
355 			continue;
356 		case '\n':
357 			if (flags)
358 			{
359 				sfputc(op, c);
360 				continue;
361 			}
362 			/*FALLTHROUGH*/
363 		case ' ':
364 		case '\t':
365 			while ((c = sfgetc(ip)) != EOF)
366 				if (c == '&')
367 				{
368 					c = decode(ip);
369 					if (!isspace(c))
370 						sfputc(op, ' ');
371 					sfpututf(op, c);
372 					break;
373 				}
374 				else if (!isspace(c))
375 				{
376 					if (c == '<')
377 					{
378 						c = sfgetc(ip);
379 						if (c == EOF)
380 							break;
381 						sfungetc(ip, c);
382 						sfungetc(ip, '<');
383 						if (c != 'L' && c != '/')
384 							sfputc(op, ' ');
385 					}
386 					else
387 					{
388 						if (c != EOF)
389 							sfungetc(ip, c);
390 						sfputc(op, ' ');
391 					}
392 					break;
393 				}
394 			continue;
395 		case '\r':
396 		case '[':
397 		case ']':
398 			continue;
399 		default:
400 			sfpututf(op, c);
401 			continue;
402 		}
403 		break;
404 	}
405 	if (q)
406 		sfputc(op, q);
407 	sfputc(op, '\n');
408 }
409 
410 static void
411 encode(Sfio_t* op, register int c)
412 {
413 	if (c == '<')
414 		sfprintf(op, "&lt;");
415 	else if (c == '>')
416 		sfprintf(op, "&gt;");
417 	else if (c == '"')
418 		sfprintf(op, "&quot;");
419 	else if (c == '&')
420 		sfprintf(op, "&amp;");
421 	else if (c == '[')
422 		sfprintf(op, "&#091;");
423 	else if (c == ']')
424 		sfprintf(op, "&#093;");
425 	else
426 		sfputc(op, c);
427 }
428 
429 static void
430 msg2html(register Sfio_t* ip, register Sfio_t* op, register int flags)
431 {
432 	register char*	s;
433 	register int	c;
434 	register int	q;
435 	register int	p;
436 
437 	sfprintf(op, "<!DOCTYPE HTML PUBLIC \"-//W3C//DTD HTML 4.0//EN\"><HTML><HEAD><!-- text massaged for external translation --></HEAD><BODY>\n");
438 	sfprintf(op, "<OL START=\"550717\">\n");
439 	p = q = 0;
440 	while (s = sfgetr(ip, '\n', 1))
441 	{
442 		error_info.line++;
443 		if (flags)
444 			sfprintf(op, "<P>");
445 		else
446 		{
447 			if (*s == '$')
448 			{
449 				if (p)
450 					sfprintf(op, "<P>");
451 				else
452 					p = 1;
453 				sfprintf(op, "<P CLASS=\"", s);
454 				while (c = *s++)
455 					encode(op, c);
456 				sfprintf(op, "\">\n");
457 				continue;
458 			}
459 			p = 0;
460 			if (!isdigit(*s))
461 				continue;
462 			sfprintf(op, "<LI>");
463 			while (isdigit(c = *s++))
464 				sfputc(op, c);
465 			sfprintf(op, "<LI>");
466 			while (c && c != '"')
467 				c = *s++;
468 			if (!c)
469 				s--;
470 			else if (isspace(*s))
471 			{
472 				s++;
473 				sfprintf(op, "<BR>");
474 			}
475 		}
476 		for (;;)
477 		{
478 			switch (c = *s++)
479 			{
480 			case 0:
481 				flags &= ~MSG_SPLICE;
482 				if (q)
483 				{
484 					q = 0;
485 					sfprintf(op, "\">");
486 				}
487 				sfputc(op, '\n');
488 				break;
489 			case '<':
490 				sfprintf(op, "&lt;");
491 				continue;
492 			case '>':
493 				sfprintf(op, "&gt;");
494 				continue;
495 			case '&':
496 				sfprintf(op, "&amp;");
497 				continue;
498 			case '[':
499 				sfprintf(op, "&#091;");
500 				continue;
501 			case ']':
502 				sfprintf(op, "&#093;");
503 				continue;
504 			case '$':
505 				if (!q)
506 				{
507 					q = 1;
508 					sfprintf(op, "<P CLASS=\"");
509 				}
510 				sfputc(op, c);
511 				while (isalnum(c = *s++))
512 					sfputc(op, c);
513 				s--;
514 				continue;
515 			case '%':
516 				if (!q)
517 				{
518 					q = 1;
519 					sfprintf(op, "<P CLASS=\"");
520 				}
521 				sfputc(op, c);
522 				if (*s == '%')
523 					sfputc(op, *s++);
524 				else
525 					do
526 					{
527 						if (!(c = *s++) || c == '"')
528 						{
529 							s--;
530 							break;
531 						}
532 						encode(op, c);
533 					} while (!isalpha(c) || (!islower(c) || c == 'h' || c == 'l') && isalpha(*s));
534 				if (SPACE(s))
535 					sfprintf(op, "&nbsp;");
536 				continue;
537 			case '"':
538 				if (!(flags & MSG_RAW))
539 				{
540 					s = "";
541 					continue;
542 				}
543 				/*FALLTHROUGH*/
544 			case '\'':
545 			case ':':
546 			case '/':
547 			case '+':
548 			case '@':
549 				if (!q)
550 				{
551 					q = 1;
552 					sfprintf(op, "<P CLASS=\"");
553 				}
554 				/*FALLTHROUGH*/
555 			case '.':
556 			case ',':
557 				sfputc(op, c);
558 				if (SPACE(s))
559 					sfprintf(op, "&nbsp;");
560 				continue;
561 			case '\\':
562 				if (!(c = *s++))
563 				{
564 					flags |= MSG_SPLICE;
565 					break;
566 				}
567 				if (c != 'n' && c != 't')
568 				{
569 					if (!q)
570 					{
571 						q = 1;
572 						sfprintf(op, "<P CLASS=\"");
573 					}
574 					sfputc(op, '\\');
575 					encode(op, c);
576 					if (c == 'b')
577 					{
578 						for (;;)
579 						{
580 							if (!(c = *s++) || c == '"')
581 							{
582 								s--;
583 								break;
584 							}
585 							if (c == '?')
586 							{
587 								if (*s != '?')
588 								{
589 									s--;
590 									break;
591 								}
592 								sfputc(op, c);
593 								sfputc(op, *s++);
594 								continue;
595 							}
596 							if (c == '\\')
597 							{
598 								if (!*s)
599 									break;
600 								sfputc(op, c);
601 								if (*s == 'a' || *s == 'b' || *s == '0')
602 								{
603 									sfputc(op, *s++);
604 									break;
605 								}
606 								c = *s++;
607 							}
608 							encode(op, c);
609 						}
610 					}
611 					else if (isdigit(c) && isdigit(*s))
612 					{
613 						sfputc(op, *s++);
614 						if (isdigit(*s))
615 							sfputc(op, *s++);
616 					}
617 					if (SPACE(s))
618 						sfprintf(op, "&nbsp;");
619 					continue;
620 				}
621 				/*FALLTHROUGH*/
622 			case ' ':
623 			case '\t':
624 				while (isspace(*s) || *s == '\\' && (*(s + 1) == 'n' || *(s + 1) == 't') && s++)
625 					s++;
626 				if (*s == '"')
627 				{
628 					if (q)
629 					{
630 						q = 0;
631 						sfprintf(op, " \">");
632 					}
633 					else
634 						sfprintf(op, "<BR>");
635 					continue;
636 				}
637 				c = ' ';
638 				/*FALLTHROUGH*/
639 			default:
640 				if (q)
641 				{
642 					q = 0;
643 					sfprintf(op, "\">");
644 				}
645 				sfputc(op, c);
646 				continue;
647 			}
648 			break;
649 		}
650 	}
651 	sfprintf(op, "</OL>\n");
652 	sfprintf(op, "</BODY></HTML>\n");
653 	error_info.line = 0;
654 }
655 
656 int
657 main(int argc, char** argv)
658 {
659 	int		flags = 0;
660 	Convert_f	convert = msg2html;
661 
662 	NoP(argc);
663 	error_info.id = "msgcvt";
664 	for (;;)
665 	{
666 		switch (optget(argv, usage))
667 		{
668 		case 'h':
669 			convert = msg2html;
670 			continue;
671 		case 'm':
672 			convert = html2msg;
673 			continue;
674 		case 'r':
675 			flags |= MSG_RAW;
676 			continue;
677 		case '?':
678 			error(ERROR_USAGE|4, "%s", opt_info.arg);
679 			continue;
680 		case ':':
681 			error(2, "%s", opt_info.arg);
682 			continue;
683 		}
684 		break;
685 	}
686 	argv += opt_info.index;
687 	if (error_info.errors)
688 		error(ERROR_USAGE|4, "%s", optusage(NiL));
689 	(*convert)(sfstdin, sfstdout, flags);
690 	return error_info.errors != 0;
691 }
692