xref: /openbsd-src/usr.bin/mandoc/html.c (revision d0fc3bb68efd6c434b4053cd7adb29023cbec341)
1 /* $OpenBSD: html.c,v 1.144 2021/05/22 05:49:32 anton Exp $ */
2 /*
3  * Copyright (c) 2011-2015, 2017-2020 Ingo Schwarze <schwarze@openbsd.org>
4  * Copyright (c) 2008-2011, 2014 Kristaps Dzonsons <kristaps@bsd.lv>
5  *
6  * Permission to use, copy, modify, and distribute this software for any
7  * purpose with or without fee is hereby granted, provided that the above
8  * copyright notice and this permission notice appear in all copies.
9  *
10  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
13  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17  *
18  * Common functions for mandoc(1) HTML formatters.
19  * For use by individual formatters and by the main program.
20  */
21 #include <sys/types.h>
22 #include <sys/stat.h>
23 
24 #include <assert.h>
25 #include <ctype.h>
26 #include <stdarg.h>
27 #include <stddef.h>
28 #include <stdio.h>
29 #include <stdint.h>
30 #include <stdlib.h>
31 #include <string.h>
32 #include <unistd.h>
33 
34 #include "mandoc_aux.h"
35 #include "mandoc_ohash.h"
36 #include "mandoc.h"
37 #include "roff.h"
38 #include "out.h"
39 #include "html.h"
40 #include "manconf.h"
41 #include "main.h"
42 
43 struct	htmldata {
44 	const char	 *name;
45 	int		  flags;
46 #define	HTML_INPHRASE	 (1 << 0)  /* Can appear in phrasing context. */
47 #define	HTML_TOPHRASE	 (1 << 1)  /* Establishes phrasing context. */
48 #define	HTML_NOSTACK	 (1 << 2)  /* Does not have an end tag. */
49 #define	HTML_NLBEFORE	 (1 << 3)  /* Output line break before opening. */
50 #define	HTML_NLBEGIN	 (1 << 4)  /* Output line break after opening. */
51 #define	HTML_NLEND	 (1 << 5)  /* Output line break before closing. */
52 #define	HTML_NLAFTER	 (1 << 6)  /* Output line break after closing. */
53 #define	HTML_NLAROUND	 (HTML_NLBEFORE | HTML_NLAFTER)
54 #define	HTML_NLINSIDE	 (HTML_NLBEGIN | HTML_NLEND)
55 #define	HTML_NLALL	 (HTML_NLAROUND | HTML_NLINSIDE)
56 #define	HTML_INDENT	 (1 << 7)  /* Indent content by two spaces. */
57 #define	HTML_NOINDENT	 (1 << 8)  /* Exception: never indent content. */
58 };
59 
60 static	const struct htmldata htmltags[TAG_MAX] = {
61 	{"html",	HTML_NLALL},
62 	{"head",	HTML_NLALL | HTML_INDENT},
63 	{"meta",	HTML_NOSTACK | HTML_NLALL},
64 	{"link",	HTML_NOSTACK | HTML_NLALL},
65 	{"style",	HTML_NLALL | HTML_INDENT},
66 	{"title",	HTML_NLAROUND},
67 	{"body",	HTML_NLALL},
68 	{"div",		HTML_NLAROUND},
69 	{"section",	HTML_NLALL},
70 	{"table",	HTML_NLALL | HTML_INDENT},
71 	{"tr",		HTML_NLALL | HTML_INDENT},
72 	{"td",		HTML_NLAROUND},
73 	{"li",		HTML_NLAROUND | HTML_INDENT},
74 	{"ul",		HTML_NLALL | HTML_INDENT},
75 	{"ol",		HTML_NLALL | HTML_INDENT},
76 	{"dl",		HTML_NLALL | HTML_INDENT},
77 	{"dt",		HTML_NLAROUND},
78 	{"dd",		HTML_NLAROUND | HTML_INDENT},
79 	{"h1",		HTML_TOPHRASE | HTML_NLAROUND},
80 	{"h2",		HTML_TOPHRASE | HTML_NLAROUND},
81 	{"p",		HTML_TOPHRASE | HTML_NLAROUND | HTML_INDENT},
82 	{"pre",		HTML_TOPHRASE | HTML_NLAROUND | HTML_NOINDENT},
83 	{"a",		HTML_INPHRASE | HTML_TOPHRASE},
84 	{"b",		HTML_INPHRASE | HTML_TOPHRASE},
85 	{"cite",	HTML_INPHRASE | HTML_TOPHRASE},
86 	{"code",	HTML_INPHRASE | HTML_TOPHRASE},
87 	{"i",		HTML_INPHRASE | HTML_TOPHRASE},
88 	{"small",	HTML_INPHRASE | HTML_TOPHRASE},
89 	{"span",	HTML_INPHRASE | HTML_TOPHRASE},
90 	{"var",		HTML_INPHRASE | HTML_TOPHRASE},
91 	{"br",		HTML_INPHRASE | HTML_NOSTACK | HTML_NLALL},
92 	{"mark",	HTML_INPHRASE },
93 	{"math",	HTML_INPHRASE | HTML_NLALL | HTML_INDENT},
94 	{"mrow",	0},
95 	{"mi",		0},
96 	{"mn",		0},
97 	{"mo",		0},
98 	{"msup",	0},
99 	{"msub",	0},
100 	{"msubsup",	0},
101 	{"mfrac",	0},
102 	{"msqrt",	0},
103 	{"mfenced",	0},
104 	{"mtable",	0},
105 	{"mtr",		0},
106 	{"mtd",		0},
107 	{"munderover",	0},
108 	{"munder",	0},
109 	{"mover",	0},
110 };
111 
112 /* Avoid duplicate HTML id= attributes. */
113 
114 struct	id_entry {
115 	int	 ord;	/* Ordinal number of the latest occurrence. */
116 	char	 id[];	/* The id= attribute without any ordinal suffix. */
117 };
118 static	struct ohash	 id_unique;
119 
120 static	void	 html_reset_internal(struct html *);
121 static	void	 print_byte(struct html *, char);
122 static	void	 print_endword(struct html *);
123 static	void	 print_indent(struct html *);
124 static	void	 print_word(struct html *, const char *);
125 
126 static	void	 print_ctag(struct html *, struct tag *);
127 static	int	 print_escape(struct html *, char);
128 static	int	 print_encode(struct html *, const char *, const char *, int);
129 static	void	 print_href(struct html *, const char *, const char *, int);
130 static	void	 print_metaf(struct html *);
131 
132 
133 void *
134 html_alloc(const struct manoutput *outopts)
135 {
136 	struct html	*h;
137 
138 	h = mandoc_calloc(1, sizeof(struct html));
139 
140 	h->tag = NULL;
141 	h->metac = h->metal = ESCAPE_FONTROMAN;
142 	h->style = outopts->style;
143 	if ((h->base_man1 = outopts->man) == NULL)
144 		h->base_man2 = NULL;
145 	else if ((h->base_man2 = strchr(h->base_man1, ';')) != NULL)
146 		*h->base_man2++ = '\0';
147 	h->base_includes = outopts->includes;
148 	if (outopts->fragment)
149 		h->oflags |= HTML_FRAGMENT;
150 	if (outopts->toc)
151 		h->oflags |= HTML_TOC;
152 
153 	mandoc_ohash_init(&id_unique, 4, offsetof(struct id_entry, id));
154 
155 	return h;
156 }
157 
158 static void
159 html_reset_internal(struct html *h)
160 {
161 	struct tag	*tag;
162 	struct id_entry	*entry;
163 	unsigned int	 slot;
164 
165 	while ((tag = h->tag) != NULL) {
166 		h->tag = tag->next;
167 		free(tag);
168 	}
169 	entry = ohash_first(&id_unique, &slot);
170 	while (entry != NULL) {
171 		free(entry);
172 		entry = ohash_next(&id_unique, &slot);
173 	}
174 	ohash_delete(&id_unique);
175 }
176 
177 void
178 html_reset(void *p)
179 {
180 	html_reset_internal(p);
181 	mandoc_ohash_init(&id_unique, 4, offsetof(struct id_entry, id));
182 }
183 
184 void
185 html_free(void *p)
186 {
187 	html_reset_internal(p);
188 	free(p);
189 }
190 
191 void
192 print_gen_head(struct html *h)
193 {
194 	struct tag	*t;
195 
196 	print_otag(h, TAG_META, "?", "charset", "utf-8");
197 	print_otag(h, TAG_META, "??", "name", "viewport",
198 	    "content", "width=device-width, initial-scale=1.0");
199 	if (h->style != NULL) {
200 		print_otag(h, TAG_LINK, "?h??", "rel", "stylesheet",
201 		    h->style, "type", "text/css", "media", "all");
202 		return;
203 	}
204 
205 	/*
206 	 * Print a minimal embedded style sheet.
207 	 */
208 
209 	t = print_otag(h, TAG_STYLE, "");
210 	print_text(h, "table.head, table.foot { width: 100%; }");
211 	print_endline(h);
212 	print_text(h, "td.head-rtitle, td.foot-os { text-align: right; }");
213 	print_endline(h);
214 	print_text(h, "td.head-vol { text-align: center; }");
215 	print_endline(h);
216 	print_text(h, ".Nd, .Bf, .Op { display: inline; }");
217 	print_endline(h);
218 	print_text(h, ".Pa, .Ad { font-style: italic; }");
219 	print_endline(h);
220 	print_text(h, ".Ms { font-weight: bold; }");
221 	print_endline(h);
222 	print_text(h, ".Bl-diag ");
223 	print_byte(h, '>');
224 	print_text(h, " dt { font-weight: bold; }");
225 	print_endline(h);
226 	print_text(h, "code.Nm, .Fl, .Cm, .Ic, code.In, .Fd, .Fn, .Cd "
227 	    "{ font-weight: bold; font-family: inherit; }");
228 	print_tagq(h, t);
229 }
230 
231 int
232 html_setfont(struct html *h, enum mandoc_esc font)
233 {
234 	switch (font) {
235 	case ESCAPE_FONTPREV:
236 		font = h->metal;
237 		break;
238 	case ESCAPE_FONTITALIC:
239 	case ESCAPE_FONTBOLD:
240 	case ESCAPE_FONTBI:
241 	case ESCAPE_FONTCW:
242 	case ESCAPE_FONTROMAN:
243 		break;
244 	case ESCAPE_FONT:
245 		font = ESCAPE_FONTROMAN;
246 		break;
247 	default:
248 		return 0;
249 	}
250 	h->metal = h->metac;
251 	h->metac = font;
252 	return 1;
253 }
254 
255 static void
256 print_metaf(struct html *h)
257 {
258 	if (h->metaf) {
259 		print_tagq(h, h->metaf);
260 		h->metaf = NULL;
261 	}
262 	switch (h->metac) {
263 	case ESCAPE_FONTITALIC:
264 		h->metaf = print_otag(h, TAG_I, "");
265 		break;
266 	case ESCAPE_FONTBOLD:
267 		h->metaf = print_otag(h, TAG_B, "");
268 		break;
269 	case ESCAPE_FONTBI:
270 		h->metaf = print_otag(h, TAG_B, "");
271 		print_otag(h, TAG_I, "");
272 		break;
273 	case ESCAPE_FONTCW:
274 		h->metaf = print_otag(h, TAG_SPAN, "c", "Li");
275 		break;
276 	default:
277 		break;
278 	}
279 }
280 
281 void
282 html_close_paragraph(struct html *h)
283 {
284 	struct tag	*this, *next;
285 	int		 flags;
286 
287 	this = h->tag;
288 	for (;;) {
289 		next = this->next;
290 		flags = htmltags[this->tag].flags;
291 		if (flags & (HTML_INPHRASE | HTML_TOPHRASE))
292 			print_ctag(h, this);
293 		if ((flags & HTML_INPHRASE) == 0)
294 			break;
295 		this = next;
296 	}
297 }
298 
299 /*
300  * ROFF_nf switches to no-fill mode, ROFF_fi to fill mode.
301  * TOKEN_NONE does not switch.  The old mode is returned.
302  */
303 enum roff_tok
304 html_fillmode(struct html *h, enum roff_tok want)
305 {
306 	struct tag	*t;
307 	enum roff_tok	 had;
308 
309 	for (t = h->tag; t != NULL; t = t->next)
310 		if (t->tag == TAG_PRE)
311 			break;
312 
313 	had = t == NULL ? ROFF_fi : ROFF_nf;
314 
315 	if (want != had) {
316 		switch (want) {
317 		case ROFF_fi:
318 			print_tagq(h, t);
319 			break;
320 		case ROFF_nf:
321 			html_close_paragraph(h);
322 			print_otag(h, TAG_PRE, "");
323 			break;
324 		case TOKEN_NONE:
325 			break;
326 		default:
327 			abort();
328 		}
329 	}
330 	return had;
331 }
332 
333 /*
334  * Allocate a string to be used for the "id=" attribute of an HTML
335  * element and/or as a segment identifier for a URI in an <a> element.
336  * The function may fail and return NULL if the node lacks text data
337  * to create the attribute from.
338  * The caller is responsible for free(3)ing the returned string.
339  *
340  * If the "unique" argument is non-zero, the "id_unique" ohash table
341  * is used for de-duplication.  If the "unique" argument is 1,
342  * it is the first time the function is called for this tag and
343  * location, so if an ordinal suffix is needed, it is incremented.
344  * If the "unique" argument is 2, it is the second time the function
345  * is called for this tag and location, so the ordinal suffix
346  * remains unchanged.
347  */
348 char *
349 html_make_id(const struct roff_node *n, int unique)
350 {
351 	const struct roff_node	*nch;
352 	struct id_entry		*entry;
353 	char			*buf, *cp;
354 	size_t			 len;
355 	unsigned int		 slot;
356 
357 	if (n->tag != NULL)
358 		buf = mandoc_strdup(n->tag);
359 	else {
360 		switch (n->tok) {
361 		case MDOC_Sh:
362 		case MDOC_Ss:
363 		case MDOC_Sx:
364 		case MAN_SH:
365 		case MAN_SS:
366 			for (nch = n->child; nch != NULL; nch = nch->next)
367 				if (nch->type != ROFFT_TEXT)
368 					return NULL;
369 			buf = NULL;
370 			deroff(&buf, n);
371 			if (buf == NULL)
372 				return NULL;
373 			break;
374 		default:
375 			if (n->child == NULL || n->child->type != ROFFT_TEXT)
376 				return NULL;
377 			buf = mandoc_strdup(n->child->string);
378 			break;
379 		}
380 	}
381 
382 	/*
383 	 * In ID attributes, only use ASCII characters that are
384 	 * permitted in URL-fragment strings according to the
385 	 * explicit list at:
386 	 * https://url.spec.whatwg.org/#url-fragment-string
387 	 * In addition, reserve '~' for ordinal suffixes.
388 	 */
389 
390 	for (cp = buf; *cp != '\0'; cp++)
391 		if (isalnum((unsigned char)*cp) == 0 &&
392 		    strchr("!$&'()*+,-./:;=?@_", *cp) == NULL)
393 			*cp = '_';
394 
395 	if (unique == 0)
396 		return buf;
397 
398 	/* Avoid duplicate HTML id= attributes. */
399 
400 	slot = ohash_qlookup(&id_unique, buf);
401 	if ((entry = ohash_find(&id_unique, slot)) == NULL) {
402 		len = strlen(buf) + 1;
403 		entry = mandoc_malloc(sizeof(*entry) + len);
404 		entry->ord = 1;
405 		memcpy(entry->id, buf, len);
406 		ohash_insert(&id_unique, slot, entry);
407 	} else if (unique == 1)
408 		entry->ord++;
409 
410 	if (entry->ord > 1) {
411 		cp = buf;
412 		mandoc_asprintf(&buf, "%s~%d", cp, entry->ord);
413 		free(cp);
414 	}
415 	return buf;
416 }
417 
418 static int
419 print_escape(struct html *h, char c)
420 {
421 
422 	switch (c) {
423 	case '<':
424 		print_word(h, "&lt;");
425 		break;
426 	case '>':
427 		print_word(h, "&gt;");
428 		break;
429 	case '&':
430 		print_word(h, "&amp;");
431 		break;
432 	case '"':
433 		print_word(h, "&quot;");
434 		break;
435 	case ASCII_NBRSP:
436 		print_word(h, "&nbsp;");
437 		break;
438 	case ASCII_HYPH:
439 		print_byte(h, '-');
440 		break;
441 	case ASCII_BREAK:
442 		break;
443 	default:
444 		return 0;
445 	}
446 	return 1;
447 }
448 
449 static int
450 print_encode(struct html *h, const char *p, const char *pend, int norecurse)
451 {
452 	char		 numbuf[16];
453 	const char	*seq;
454 	size_t		 sz;
455 	int		 c, len, breakline, nospace;
456 	enum mandoc_esc	 esc;
457 	static const char rejs[10] = { ' ', '\\', '<', '>', '&', '"',
458 		ASCII_NBRSP, ASCII_HYPH, ASCII_BREAK, '\0' };
459 
460 	if (pend == NULL)
461 		pend = strchr(p, '\0');
462 
463 	breakline = 0;
464 	nospace = 0;
465 
466 	while (p < pend) {
467 		if (HTML_SKIPCHAR & h->flags && '\\' != *p) {
468 			h->flags &= ~HTML_SKIPCHAR;
469 			p++;
470 			continue;
471 		}
472 
473 		for (sz = strcspn(p, rejs); sz-- && p < pend; p++)
474 			print_byte(h, *p);
475 
476 		if (breakline &&
477 		    (p >= pend || *p == ' ' || *p == ASCII_NBRSP)) {
478 			print_otag(h, TAG_BR, "");
479 			breakline = 0;
480 			while (p < pend && (*p == ' ' || *p == ASCII_NBRSP))
481 				p++;
482 			continue;
483 		}
484 
485 		if (p >= pend)
486 			break;
487 
488 		if (*p == ' ') {
489 			print_endword(h);
490 			p++;
491 			continue;
492 		}
493 
494 		if (print_escape(h, *p++))
495 			continue;
496 
497 		esc = mandoc_escape(&p, &seq, &len);
498 		switch (esc) {
499 		case ESCAPE_FONT:
500 		case ESCAPE_FONTPREV:
501 		case ESCAPE_FONTBOLD:
502 		case ESCAPE_FONTITALIC:
503 		case ESCAPE_FONTBI:
504 		case ESCAPE_FONTCW:
505 		case ESCAPE_FONTROMAN:
506 			if (0 == norecurse) {
507 				h->flags |= HTML_NOSPACE;
508 				if (html_setfont(h, esc))
509 					print_metaf(h);
510 				h->flags &= ~HTML_NOSPACE;
511 			}
512 			continue;
513 		case ESCAPE_SKIPCHAR:
514 			h->flags |= HTML_SKIPCHAR;
515 			continue;
516 		case ESCAPE_ERROR:
517 			continue;
518 		default:
519 			break;
520 		}
521 
522 		if (h->flags & HTML_SKIPCHAR) {
523 			h->flags &= ~HTML_SKIPCHAR;
524 			continue;
525 		}
526 
527 		switch (esc) {
528 		case ESCAPE_UNICODE:
529 			/* Skip past "u" header. */
530 			c = mchars_num2uc(seq + 1, len - 1);
531 			break;
532 		case ESCAPE_NUMBERED:
533 			c = mchars_num2char(seq, len);
534 			if (c < 0)
535 				continue;
536 			break;
537 		case ESCAPE_SPECIAL:
538 			c = mchars_spec2cp(seq, len);
539 			if (c <= 0)
540 				continue;
541 			break;
542 		case ESCAPE_UNDEF:
543 			c = *seq;
544 			break;
545 		case ESCAPE_DEVICE:
546 			print_word(h, "html");
547 			continue;
548 		case ESCAPE_BREAK:
549 			breakline = 1;
550 			continue;
551 		case ESCAPE_NOSPACE:
552 			if ('\0' == *p)
553 				nospace = 1;
554 			continue;
555 		case ESCAPE_OVERSTRIKE:
556 			if (len == 0)
557 				continue;
558 			c = seq[len - 1];
559 			break;
560 		default:
561 			continue;
562 		}
563 		if ((c < 0x20 && c != 0x09) ||
564 		    (c > 0x7E && c < 0xA0))
565 			c = 0xFFFD;
566 		if (c > 0x7E) {
567 			(void)snprintf(numbuf, sizeof(numbuf), "&#x%.4X;", c);
568 			print_word(h, numbuf);
569 		} else if (print_escape(h, c) == 0)
570 			print_byte(h, c);
571 	}
572 
573 	return nospace;
574 }
575 
576 static void
577 print_href(struct html *h, const char *name, const char *sec, int man)
578 {
579 	struct stat	 sb;
580 	const char	*p, *pp;
581 	char		*filename;
582 
583 	if (man) {
584 		pp = h->base_man1;
585 		if (h->base_man2 != NULL) {
586 			mandoc_asprintf(&filename, "%s.%s", name, sec);
587 			if (stat(filename, &sb) == -1)
588 				pp = h->base_man2;
589 			free(filename);
590 		}
591 	} else
592 		pp = h->base_includes;
593 
594 	while ((p = strchr(pp, '%')) != NULL) {
595 		print_encode(h, pp, p, 1);
596 		if (man && p[1] == 'S') {
597 			if (sec == NULL)
598 				print_byte(h, '1');
599 			else
600 				print_encode(h, sec, NULL, 1);
601 		} else if ((man && p[1] == 'N') ||
602 		    (man == 0 && p[1] == 'I'))
603 			print_encode(h, name, NULL, 1);
604 		else
605 			print_encode(h, p, p + 2, 1);
606 		pp = p + 2;
607 	}
608 	if (*pp != '\0')
609 		print_encode(h, pp, NULL, 1);
610 }
611 
612 struct tag *
613 print_otag(struct html *h, enum htmltag tag, const char *fmt, ...)
614 {
615 	va_list		 ap;
616 	struct tag	*t;
617 	const char	*attr;
618 	char		*arg1, *arg2;
619 	int		 style_written, tflags;
620 
621 	tflags = htmltags[tag].flags;
622 
623 	/* Flow content is not allowed in phrasing context. */
624 
625 	if ((tflags & HTML_INPHRASE) == 0) {
626 		for (t = h->tag; t != NULL; t = t->next) {
627 			if (t->closed)
628 				continue;
629 			assert((htmltags[t->tag].flags & HTML_TOPHRASE) == 0);
630 			break;
631 		}
632 
633 	/*
634 	 * Always wrap phrasing elements in a paragraph
635 	 * unless already contained in some flow container;
636 	 * never put them directly into a section.
637 	 */
638 
639 	} else if (tflags & HTML_TOPHRASE && h->tag->tag == TAG_SECTION)
640 		print_otag(h, TAG_P, "c", "Pp");
641 
642 	/* Push this tag onto the stack of open scopes. */
643 
644 	if ((tflags & HTML_NOSTACK) == 0) {
645 		t = mandoc_malloc(sizeof(struct tag));
646 		t->tag = tag;
647 		t->next = h->tag;
648 		t->refcnt = 0;
649 		t->closed = 0;
650 		h->tag = t;
651 	} else
652 		t = NULL;
653 
654 	if (tflags & HTML_NLBEFORE)
655 		print_endline(h);
656 	if (h->col == 0)
657 		print_indent(h);
658 	else if ((h->flags & HTML_NOSPACE) == 0) {
659 		if (h->flags & HTML_KEEP)
660 			print_word(h, "&#x00A0;");
661 		else {
662 			if (h->flags & HTML_PREKEEP)
663 				h->flags |= HTML_KEEP;
664 			print_endword(h);
665 		}
666 	}
667 
668 	if ( ! (h->flags & HTML_NONOSPACE))
669 		h->flags &= ~HTML_NOSPACE;
670 	else
671 		h->flags |= HTML_NOSPACE;
672 
673 	/* Print out the tag name and attributes. */
674 
675 	print_byte(h, '<');
676 	print_word(h, htmltags[tag].name);
677 
678 	va_start(ap, fmt);
679 
680 	while (*fmt != '\0' && *fmt != 's') {
681 
682 		/* Parse attributes and arguments. */
683 
684 		arg1 = va_arg(ap, char *);
685 		arg2 = NULL;
686 		switch (*fmt++) {
687 		case 'c':
688 			attr = "class";
689 			break;
690 		case 'h':
691 			attr = "href";
692 			break;
693 		case 'i':
694 			attr = "id";
695 			break;
696 		case '?':
697 			attr = arg1;
698 			arg1 = va_arg(ap, char *);
699 			break;
700 		default:
701 			abort();
702 		}
703 		if (*fmt == 'M')
704 			arg2 = va_arg(ap, char *);
705 		if (arg1 == NULL)
706 			continue;
707 
708 		/* Print the attributes. */
709 
710 		print_byte(h, ' ');
711 		print_word(h, attr);
712 		print_byte(h, '=');
713 		print_byte(h, '"');
714 		switch (*fmt) {
715 		case 'I':
716 			print_href(h, arg1, NULL, 0);
717 			fmt++;
718 			break;
719 		case 'M':
720 			print_href(h, arg1, arg2, 1);
721 			fmt++;
722 			break;
723 		case 'R':
724 			print_byte(h, '#');
725 			print_encode(h, arg1, NULL, 1);
726 			fmt++;
727 			break;
728 		default:
729 			print_encode(h, arg1, NULL, 1);
730 			break;
731 		}
732 		print_byte(h, '"');
733 	}
734 
735 	style_written = 0;
736 	while (*fmt++ == 's') {
737 		arg1 = va_arg(ap, char *);
738 		arg2 = va_arg(ap, char *);
739 		if (arg2 == NULL)
740 			continue;
741 		print_byte(h, ' ');
742 		if (style_written == 0) {
743 			print_word(h, "style=\"");
744 			style_written = 1;
745 		}
746 		print_word(h, arg1);
747 		print_byte(h, ':');
748 		print_byte(h, ' ');
749 		print_word(h, arg2);
750 		print_byte(h, ';');
751 	}
752 	if (style_written)
753 		print_byte(h, '"');
754 
755 	va_end(ap);
756 
757 	/* Accommodate for "well-formed" singleton escaping. */
758 
759 	if (htmltags[tag].flags & HTML_NOSTACK)
760 		print_byte(h, '/');
761 
762 	print_byte(h, '>');
763 
764 	if (tflags & HTML_NLBEGIN)
765 		print_endline(h);
766 	else
767 		h->flags |= HTML_NOSPACE;
768 
769 	if (tflags & HTML_INDENT)
770 		h->indent++;
771 	if (tflags & HTML_NOINDENT)
772 		h->noindent++;
773 
774 	return t;
775 }
776 
777 /*
778  * Print an element with an optional "id=" attribute.
779  * If the element has phrasing content and an "id=" attribute,
780  * also add a permalink: outside if it can be in phrasing context,
781  * inside otherwise.
782  */
783 struct tag *
784 print_otag_id(struct html *h, enum htmltag elemtype, const char *cattr,
785     struct roff_node *n)
786 {
787 	struct roff_node *nch;
788 	struct tag	*ret, *t;
789 	char		*id, *href;
790 
791 	ret = NULL;
792 	id = href = NULL;
793 	if (n->flags & NODE_ID)
794 		id = html_make_id(n, 1);
795 	if (n->flags & NODE_HREF)
796 		href = id == NULL ? html_make_id(n, 2) : id;
797 	if (href != NULL && htmltags[elemtype].flags & HTML_INPHRASE)
798 		ret = print_otag(h, TAG_A, "chR", "permalink", href);
799 	t = print_otag(h, elemtype, "ci", cattr, id);
800 	if (ret == NULL) {
801 		ret = t;
802 		if (href != NULL && (nch = n->child) != NULL) {
803 			/* man(7) is safe, it tags phrasing content only. */
804 			if (n->tok > MDOC_MAX ||
805 			    htmltags[elemtype].flags & HTML_TOPHRASE)
806 				nch = NULL;
807 			else  /* For mdoc(7), beware of nested blocks. */
808 				while (nch != NULL && nch->type == ROFFT_TEXT)
809 					nch = nch->next;
810 			if (nch == NULL)
811 				print_otag(h, TAG_A, "chR", "permalink", href);
812 		}
813 	}
814 	free(id);
815 	if (id == NULL)
816 		free(href);
817 	return ret;
818 }
819 
820 static void
821 print_ctag(struct html *h, struct tag *tag)
822 {
823 	int	 tflags;
824 
825 	if (tag->closed == 0) {
826 		tag->closed = 1;
827 		if (tag == h->metaf)
828 			h->metaf = NULL;
829 		if (tag == h->tblt)
830 			h->tblt = NULL;
831 
832 		tflags = htmltags[tag->tag].flags;
833 		if (tflags & HTML_INDENT)
834 			h->indent--;
835 		if (tflags & HTML_NOINDENT)
836 			h->noindent--;
837 		if (tflags & HTML_NLEND)
838 			print_endline(h);
839 		print_indent(h);
840 		print_byte(h, '<');
841 		print_byte(h, '/');
842 		print_word(h, htmltags[tag->tag].name);
843 		print_byte(h, '>');
844 		if (tflags & HTML_NLAFTER)
845 			print_endline(h);
846 	}
847 	if (tag->refcnt == 0) {
848 		h->tag = tag->next;
849 		free(tag);
850 	}
851 }
852 
853 void
854 print_gen_decls(struct html *h)
855 {
856 	print_word(h, "<!DOCTYPE html>");
857 	print_endline(h);
858 }
859 
860 void
861 print_gen_comment(struct html *h, struct roff_node *n)
862 {
863 	int	 wantblank;
864 
865 	print_word(h, "<!-- This is an automatically generated file."
866 	    "  Do not edit.");
867 	h->indent = 1;
868 	wantblank = 0;
869 	while (n != NULL && n->type == ROFFT_COMMENT) {
870 		if (strstr(n->string, "-->") == NULL &&
871 		    (wantblank || *n->string != '\0')) {
872 			print_endline(h);
873 			print_indent(h);
874 			print_word(h, n->string);
875 			wantblank = *n->string != '\0';
876 		}
877 		n = n->next;
878 	}
879 	if (wantblank)
880 		print_endline(h);
881 	print_word(h, " -->");
882 	print_endline(h);
883 	h->indent = 0;
884 }
885 
886 void
887 print_text(struct html *h, const char *word)
888 {
889 	print_tagged_text(h, word, NULL);
890 }
891 
892 void
893 print_tagged_text(struct html *h, const char *word, struct roff_node *n)
894 {
895 	struct tag	*t;
896 	char		*href;
897 
898 	/*
899 	 * Always wrap text in a paragraph unless already contained in
900 	 * some flow container; never put it directly into a section.
901 	 */
902 
903 	if (h->tag->tag == TAG_SECTION)
904 		print_otag(h, TAG_P, "c", "Pp");
905 
906 	/* Output whitespace before this text? */
907 
908 	if (h->col && (h->flags & HTML_NOSPACE) == 0) {
909 		if ( ! (HTML_KEEP & h->flags)) {
910 			if (HTML_PREKEEP & h->flags)
911 				h->flags |= HTML_KEEP;
912 			print_endword(h);
913 		} else
914 			print_word(h, "&#x00A0;");
915 	}
916 
917 	/*
918 	 * Optionally switch fonts, optionally write a permalink, then
919 	 * print the text, optionally surrounded by HTML whitespace.
920 	 */
921 
922 	assert(h->metaf == NULL);
923 	print_metaf(h);
924 	print_indent(h);
925 
926 	if (n != NULL && (href = html_make_id(n, 2)) != NULL) {
927 		t = print_otag(h, TAG_A, "chR", "permalink", href);
928 		free(href);
929 	} else
930 		t = NULL;
931 
932 	if ( ! print_encode(h, word, NULL, 0)) {
933 		if ( ! (h->flags & HTML_NONOSPACE))
934 			h->flags &= ~HTML_NOSPACE;
935 		h->flags &= ~HTML_NONEWLINE;
936 	} else
937 		h->flags |= HTML_NOSPACE | HTML_NONEWLINE;
938 
939 	if (h->metaf != NULL) {
940 		print_tagq(h, h->metaf);
941 		h->metaf = NULL;
942 	} else if (t != NULL)
943 		print_tagq(h, t);
944 
945 	h->flags &= ~HTML_IGNDELIM;
946 }
947 
948 void
949 print_tagq(struct html *h, const struct tag *until)
950 {
951 	struct tag	*this, *next;
952 
953 	for (this = h->tag; this != NULL; this = next) {
954 		next = this == until ? NULL : this->next;
955 		print_ctag(h, this);
956 	}
957 }
958 
959 /*
960  * Close out all open elements up to but excluding suntil.
961  * Note that a paragraph just inside stays open together with it
962  * because paragraphs include subsequent phrasing content.
963  */
964 void
965 print_stagq(struct html *h, const struct tag *suntil)
966 {
967 	struct tag	*this, *next;
968 
969 	for (this = h->tag; this != NULL; this = next) {
970 		next = this->next;
971 		if (this == suntil || (next == suntil &&
972 		    (this->tag == TAG_P || this->tag == TAG_PRE)))
973 			break;
974 		print_ctag(h, this);
975 	}
976 }
977 
978 
979 /***********************************************************************
980  * Low level output functions.
981  * They implement line breaking using a short static buffer.
982  ***********************************************************************/
983 
984 /*
985  * Buffer one HTML output byte.
986  * If the buffer is full, flush and deactivate it and start a new line.
987  * If the buffer is inactive, print directly.
988  */
989 static void
990 print_byte(struct html *h, char c)
991 {
992 	if ((h->flags & HTML_BUFFER) == 0) {
993 		putchar(c);
994 		h->col++;
995 		return;
996 	}
997 
998 	if (h->col + h->bufcol < sizeof(h->buf)) {
999 		h->buf[h->bufcol++] = c;
1000 		return;
1001 	}
1002 
1003 	putchar('\n');
1004 	h->col = 0;
1005 	print_indent(h);
1006 	putchar(' ');
1007 	putchar(' ');
1008 	fwrite(h->buf, h->bufcol, 1, stdout);
1009 	putchar(c);
1010 	h->col = (h->indent + 1) * 2 + h->bufcol + 1;
1011 	h->bufcol = 0;
1012 	h->flags &= ~HTML_BUFFER;
1013 }
1014 
1015 /*
1016  * If something was printed on the current output line, end it.
1017  * Not to be called right after print_indent().
1018  */
1019 void
1020 print_endline(struct html *h)
1021 {
1022 	if (h->col == 0)
1023 		return;
1024 
1025 	if (h->bufcol) {
1026 		putchar(' ');
1027 		fwrite(h->buf, h->bufcol, 1, stdout);
1028 		h->bufcol = 0;
1029 	}
1030 	putchar('\n');
1031 	h->col = 0;
1032 	h->flags |= HTML_NOSPACE;
1033 	h->flags &= ~HTML_BUFFER;
1034 }
1035 
1036 /*
1037  * Flush the HTML output buffer.
1038  * If it is inactive, activate it.
1039  */
1040 static void
1041 print_endword(struct html *h)
1042 {
1043 	if (h->noindent) {
1044 		print_byte(h, ' ');
1045 		return;
1046 	}
1047 
1048 	if ((h->flags & HTML_BUFFER) == 0) {
1049 		h->col++;
1050 		h->flags |= HTML_BUFFER;
1051 	} else if (h->bufcol) {
1052 		putchar(' ');
1053 		fwrite(h->buf, h->bufcol, 1, stdout);
1054 		h->col += h->bufcol + 1;
1055 	}
1056 	h->bufcol = 0;
1057 }
1058 
1059 /*
1060  * If at the beginning of a new output line,
1061  * perform indentation and mark the line as containing output.
1062  * Make sure to really produce some output right afterwards,
1063  * but do not use print_otag() for producing it.
1064  */
1065 static void
1066 print_indent(struct html *h)
1067 {
1068 	size_t	 i;
1069 
1070 	if (h->col || h->noindent)
1071 		return;
1072 
1073 	h->col = h->indent * 2;
1074 	for (i = 0; i < h->col; i++)
1075 		putchar(' ');
1076 }
1077 
1078 /*
1079  * Print or buffer some characters
1080  * depending on the current HTML output buffer state.
1081  */
1082 static void
1083 print_word(struct html *h, const char *cp)
1084 {
1085 	while (*cp != '\0')
1086 		print_byte(h, *cp++);
1087 }
1088