xref: /dragonfly/contrib/mdocml/term.c (revision de78d61c)
1 /*	$Id: term.c,v 1.280 2019/01/15 12:16:18 schwarze Exp $ */
2 /*
3  * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4  * Copyright (c) 2010-2019 Ingo Schwarze <schwarze@openbsd.org>
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 #include "config.h"
19 
20 #include <sys/types.h>
21 
22 #include <assert.h>
23 #include <ctype.h>
24 #include <stdint.h>
25 #include <stdio.h>
26 #include <stdlib.h>
27 #include <string.h>
28 
29 #include "mandoc.h"
30 #include "mandoc_aux.h"
31 #include "out.h"
32 #include "term.h"
33 #include "main.h"
34 
35 static	size_t		 cond_width(const struct termp *, int, int *);
36 static	void		 adjbuf(struct termp_col *, size_t);
37 static	void		 bufferc(struct termp *, char);
38 static	void		 encode(struct termp *, const char *, size_t);
39 static	void		 encode1(struct termp *, int);
40 static	void		 endline(struct termp *);
41 static	void		 term_field(struct termp *, size_t, size_t,
42 				size_t, size_t);
43 static	void		 term_fill(struct termp *, size_t *, size_t *,
44 				size_t);
45 
46 
47 void
48 term_setcol(struct termp *p, size_t maxtcol)
49 {
50 	if (maxtcol > p->maxtcol) {
51 		p->tcols = mandoc_recallocarray(p->tcols,
52 		    p->maxtcol, maxtcol, sizeof(*p->tcols));
53 		p->maxtcol = maxtcol;
54 	}
55 	p->lasttcol = maxtcol - 1;
56 	p->tcol = p->tcols;
57 }
58 
59 void
60 term_free(struct termp *p)
61 {
62 	for (p->tcol = p->tcols; p->tcol < p->tcols + p->maxtcol; p->tcol++)
63 		free(p->tcol->buf);
64 	free(p->tcols);
65 	free(p->fontq);
66 	free(p);
67 }
68 
69 void
70 term_begin(struct termp *p, term_margin head,
71 		term_margin foot, const struct roff_meta *arg)
72 {
73 
74 	p->headf = head;
75 	p->footf = foot;
76 	p->argf = arg;
77 	(*p->begin)(p);
78 }
79 
80 void
81 term_end(struct termp *p)
82 {
83 
84 	(*p->end)(p);
85 }
86 
87 /*
88  * Flush a chunk of text.  By default, break the output line each time
89  * the right margin is reached, and continue output on the next line
90  * at the same offset as the chunk itself.  By default, also break the
91  * output line at the end of the chunk.  There are many flags modifying
92  * this behaviour, see the comments in the body of the function.
93  */
94 void
95 term_flushln(struct termp *p)
96 {
97 	size_t	 vbl;      /* Number of blanks to prepend to the output. */
98 	size_t	 vbr;      /* Actual visual position of the end of field. */
99 	size_t	 vfield;   /* Desired visual field width. */
100 	size_t	 vtarget;  /* Desired visual position of the right margin. */
101 	size_t	 ic;       /* Character position in the input buffer. */
102 	size_t	 nbr;      /* Number of characters to print in this field. */
103 
104 	/*
105 	 * Normally, start writing at the left margin, but with the
106 	 * NOPAD flag, start writing at the current position instead.
107 	 */
108 
109 	vbl = (p->flags & TERMP_NOPAD) || p->tcol->offset < p->viscol ?
110 	    0 : p->tcol->offset - p->viscol;
111 	if (p->minbl && vbl < p->minbl)
112 		vbl = p->minbl;
113 
114 	if ((p->flags & TERMP_MULTICOL) == 0)
115 		p->tcol->col = 0;
116 
117 	/* Loop over output lines. */
118 
119 	for (;;) {
120 		vfield = p->tcol->rmargin > p->viscol + vbl ?
121 		    p->tcol->rmargin - p->viscol - vbl : 0;
122 
123 		/*
124 		 * Normally, break the line at the the right margin
125 		 * of the field, but with the NOBREAK flag, only
126 		 * break it at the max right margin of the screen,
127 		 * and with the BRNEVER flag, never break it at all.
128 		 */
129 
130 		vtarget = p->flags & TERMP_BRNEVER ? SIZE_MAX :
131 		    (p->flags & TERMP_NOBREAK) == 0 ? vfield :
132 		    p->maxrmargin > p->viscol + vbl ?
133 		    p->maxrmargin - p->viscol - vbl : 0;
134 
135 		/*
136 		 * Figure out how much text will fit in the field.
137 		 * If there is whitespace only, print nothing.
138 		 */
139 
140 		term_fill(p, &nbr, &vbr, vtarget);
141 		if (nbr == 0)
142 			break;
143 
144 		/*
145 		 * With the CENTER or RIGHT flag, increase the indentation
146 		 * to center the text between the left and right margins
147 		 * or to adjust it to the right margin, respectively.
148 		 */
149 
150 		if (vbr < vtarget) {
151 			if (p->flags & TERMP_CENTER)
152 				vbl += (vtarget - vbr) / 2;
153 			else if (p->flags & TERMP_RIGHT)
154 				vbl += vtarget - vbr;
155 		}
156 
157 		/* Finally, print the field content. */
158 
159 		term_field(p, vbl, nbr, vbr, vtarget);
160 
161 		/*
162 		 * If there is no text left in the field, exit the loop.
163 		 * If the BRTRSP flag is set, consider trailing
164 		 * whitespace significant when deciding whether
165 		 * the field fits or not.
166 		 */
167 
168 		for (ic = p->tcol->col; ic < p->tcol->lastcol; ic++) {
169 			switch (p->tcol->buf[ic]) {
170 			case '\t':
171 				if (p->flags & TERMP_BRTRSP)
172 					vbr = term_tab_next(vbr);
173 				continue;
174 			case ' ':
175 				if (p->flags & TERMP_BRTRSP)
176 					vbr += (*p->width)(p, ' ');
177 				continue;
178 			case '\n':
179 			case ASCII_BREAK:
180 				continue;
181 			default:
182 				break;
183 			}
184 			break;
185 		}
186 		if (ic == p->tcol->lastcol)
187 			break;
188 
189 		/*
190 		 * At the location of an automtic line break, input
191 		 * space characters are consumed by the line break.
192 		 */
193 
194 		while (p->tcol->col < p->tcol->lastcol &&
195 		    p->tcol->buf[p->tcol->col] == ' ')
196 			p->tcol->col++;
197 
198 		/*
199 		 * In multi-column mode, leave the rest of the text
200 		 * in the buffer to be handled by a subsequent
201 		 * invocation, such that the other columns of the
202 		 * table can be handled first.
203 		 * In single-column mode, simply break the line.
204 		 */
205 
206 		if (p->flags & TERMP_MULTICOL)
207 			return;
208 
209 		endline(p);
210 		p->viscol = 0;
211 
212 		/*
213 		 * Normally, start the next line at the same indentation
214 		 * as this one, but with the BRIND flag, start it at the
215 		 * right margin instead.  This is used together with
216 		 * NOBREAK for the tags in various kinds of tagged lists.
217 		 */
218 
219 		vbl = p->flags & TERMP_BRIND ?
220 		    p->tcol->rmargin : p->tcol->offset;
221 	}
222 
223 	/* Reset output state in preparation for the next field. */
224 
225 	p->col = p->tcol->col = p->tcol->lastcol = 0;
226 	p->minbl = p->trailspace;
227 	p->flags &= ~(TERMP_BACKAFTER | TERMP_BACKBEFORE | TERMP_NOPAD);
228 
229 	if (p->flags & TERMP_MULTICOL)
230 		return;
231 
232 	/*
233 	 * The HANG flag means that the next field
234 	 * always follows on the same line.
235 	 * The NOBREAK flag means that the next field
236 	 * follows on the same line unless the field was overrun.
237 	 * Normally, break the line at the end of each field.
238 	 */
239 
240 	if ((p->flags & TERMP_HANG) == 0 &&
241 	    ((p->flags & TERMP_NOBREAK) == 0 ||
242 	     vbr + term_len(p, p->trailspace) > vfield))
243 		endline(p);
244 }
245 
246 /*
247  * Store the number of input characters to print in this field in *nbr
248  * and their total visual width to print in *vbr.
249  * If there is only whitespace in the field, both remain zero.
250  * The desired visual width of the field is provided by vtarget.
251  * If the first word is longer, the field will be overrun.
252  */
253 static void
254 term_fill(struct termp *p, size_t *nbr, size_t *vbr, size_t vtarget)
255 {
256 	size_t	 ic;        /* Character position in the input buffer. */
257 	size_t	 vis;       /* Visual position of the current character. */
258 	size_t	 vn;        /* Visual position of the next character. */
259 	int	 breakline; /* Break at the end of this word. */
260 	int	 graph;     /* Last character was non-blank. */
261 
262 	*nbr = *vbr = vis = 0;
263 	breakline = graph = 0;
264 	for (ic = p->tcol->col; ic < p->tcol->lastcol; ic++) {
265 		switch (p->tcol->buf[ic]) {
266 		case '\b':  /* Escape \o (overstrike) or backspace markup. */
267 			assert(ic > 0);
268 			vis -= (*p->width)(p, p->tcol->buf[ic - 1]);
269 			continue;
270 
271 		case '\t':  /* Normal ASCII whitespace. */
272 		case ' ':
273 		case ASCII_BREAK:  /* Escape \: (breakpoint). */
274 			switch (p->tcol->buf[ic]) {
275 			case '\t':
276 				vn = term_tab_next(vis);
277 				break;
278 			case ' ':
279 				vn = vis + (*p->width)(p, ' ');
280 				break;
281 			case ASCII_BREAK:
282 				vn = vis;
283 				break;
284 			}
285 			/* Can break at the end of a word. */
286 			if (breakline || vn > vtarget)
287 				break;
288 			if (graph) {
289 				*nbr = ic;
290 				*vbr = vis;
291 				graph = 0;
292 			}
293 			vis = vn;
294 			continue;
295 
296 		case '\n':  /* Escape \p (break at the end of the word). */
297 			breakline = 1;
298 			continue;
299 
300 		case ASCII_HYPH:  /* Breakable hyphen. */
301 			graph = 1;
302 			/*
303 			 * We are about to decide whether to break the
304 			 * line or not, so we no longer need this hyphen
305 			 * to be marked as breakable.  Put back a real
306 			 * hyphen such that we get the correct width.
307 			 */
308 			p->tcol->buf[ic] = '-';
309 			vis += (*p->width)(p, '-');
310 			if (vis > vtarget) {
311 				ic++;
312 				break;
313 			}
314 			*nbr = ic + 1;
315 			*vbr = vis;
316 			continue;
317 
318 		case ASCII_NBRSP:  /* Non-breakable space. */
319 			p->tcol->buf[ic] = ' ';
320 			/* FALLTHROUGH */
321 		default:  /* Printable character. */
322 			graph = 1;
323 			vis += (*p->width)(p, p->tcol->buf[ic]);
324 			if (vis > vtarget && *nbr > 0)
325 				return;
326 			continue;
327 		}
328 		break;
329 	}
330 
331 	/*
332 	 * If the last word extends to the end of the field without any
333 	 * trailing whitespace, the loop could not check yet whether it
334 	 * can remain on this line.  So do the check now.
335 	 */
336 
337 	if (graph && (vis <= vtarget || *nbr == 0)) {
338 		*nbr = ic;
339 		*vbr = vis;
340 	}
341 }
342 
343 /*
344  * Print the contents of one field
345  * with an indentation of	 vbl	  visual columns,
346  * an input string length of	 nbr	  characters,
347  * an output width of		 vbr	  visual columns,
348  * and a desired field width of	 vtarget  visual columns.
349  */
350 static void
351 term_field(struct termp *p, size_t vbl, size_t nbr, size_t vbr, size_t vtarget)
352 {
353 	size_t	 ic;	/* Character position in the input buffer. */
354 	size_t	 vis;	/* Visual position of the current character. */
355 	size_t	 dv;	/* Visual width of the current character. */
356 	size_t	 vn;	/* Visual position of the next character. */
357 
358 	vis = 0;
359 	for (ic = p->tcol->col; ic < nbr; ic++) {
360 
361 		/*
362 		 * To avoid the printing of trailing whitespace,
363 		 * do not print whitespace right away, only count it.
364 		 */
365 
366 		switch (p->tcol->buf[ic]) {
367 		case '\n':
368 		case ASCII_BREAK:
369 			continue;
370 		case '\t':
371 			vn = term_tab_next(vis);
372 			vbl += vn - vis;
373 			vis = vn;
374 			continue;
375 		case ' ':
376 		case ASCII_NBRSP:
377 			dv = (*p->width)(p, ' ');
378 			vbl += dv;
379 			vis += dv;
380 			continue;
381 		default:
382 			break;
383 		}
384 
385 		/*
386 		 * We found a non-blank character to print,
387 		 * so write preceding white space now.
388 		 */
389 
390 		if (vbl > 0) {
391 			(*p->advance)(p, vbl);
392 			p->viscol += vbl;
393 			vbl = 0;
394 		}
395 
396 		/* Print the character and adjust the visual position. */
397 
398 		(*p->letter)(p, p->tcol->buf[ic]);
399 		if (p->tcol->buf[ic] == '\b') {
400 			dv = (*p->width)(p, p->tcol->buf[ic - 1]);
401 			p->viscol -= dv;
402 			vis -= dv;
403 		} else {
404 			dv = (*p->width)(p, p->tcol->buf[ic]);
405 			p->viscol += dv;
406 			vis += dv;
407 		}
408 	}
409 	p->tcol->col = nbr;
410 }
411 
412 static void
413 endline(struct termp *p)
414 {
415 	if ((p->flags & (TERMP_NEWMC | TERMP_ENDMC)) == TERMP_ENDMC) {
416 		p->mc = NULL;
417 		p->flags &= ~TERMP_ENDMC;
418 	}
419 	if (p->mc != NULL) {
420 		if (p->viscol && p->maxrmargin >= p->viscol)
421 			(*p->advance)(p, p->maxrmargin - p->viscol + 1);
422 		p->flags |= TERMP_NOBUF | TERMP_NOSPACE;
423 		term_word(p, p->mc);
424 		p->flags &= ~(TERMP_NOBUF | TERMP_NEWMC);
425 	}
426 	p->viscol = 0;
427 	p->minbl = 0;
428 	(*p->endline)(p);
429 }
430 
431 /*
432  * A newline only breaks an existing line; it won't assert vertical
433  * space.  All data in the output buffer is flushed prior to the newline
434  * assertion.
435  */
436 void
437 term_newln(struct termp *p)
438 {
439 
440 	p->flags |= TERMP_NOSPACE;
441 	if (p->tcol->lastcol || p->viscol)
442 		term_flushln(p);
443 }
444 
445 /*
446  * Asserts a vertical space (a full, empty line-break between lines).
447  * Note that if used twice, this will cause two blank spaces and so on.
448  * All data in the output buffer is flushed prior to the newline
449  * assertion.
450  */
451 void
452 term_vspace(struct termp *p)
453 {
454 
455 	term_newln(p);
456 	p->viscol = 0;
457 	p->minbl = 0;
458 	if (0 < p->skipvsp)
459 		p->skipvsp--;
460 	else
461 		(*p->endline)(p);
462 }
463 
464 /* Swap current and previous font; for \fP and .ft P */
465 void
466 term_fontlast(struct termp *p)
467 {
468 	enum termfont	 f;
469 
470 	f = p->fontl;
471 	p->fontl = p->fontq[p->fonti];
472 	p->fontq[p->fonti] = f;
473 }
474 
475 /* Set font, save current, discard previous; for \f, .ft, .B etc. */
476 void
477 term_fontrepl(struct termp *p, enum termfont f)
478 {
479 
480 	p->fontl = p->fontq[p->fonti];
481 	p->fontq[p->fonti] = f;
482 }
483 
484 /* Set font, save previous. */
485 void
486 term_fontpush(struct termp *p, enum termfont f)
487 {
488 
489 	p->fontl = p->fontq[p->fonti];
490 	if (++p->fonti == p->fontsz) {
491 		p->fontsz += 8;
492 		p->fontq = mandoc_reallocarray(p->fontq,
493 		    p->fontsz, sizeof(*p->fontq));
494 	}
495 	p->fontq[p->fonti] = f;
496 }
497 
498 /* Flush to make the saved pointer current again. */
499 void
500 term_fontpopq(struct termp *p, int i)
501 {
502 
503 	assert(i >= 0);
504 	if (p->fonti > i)
505 		p->fonti = i;
506 }
507 
508 /* Pop one font off the stack. */
509 void
510 term_fontpop(struct termp *p)
511 {
512 
513 	assert(p->fonti);
514 	p->fonti--;
515 }
516 
517 /*
518  * Handle pwords, partial words, which may be either a single word or a
519  * phrase that cannot be broken down (such as a literal string).  This
520  * handles word styling.
521  */
522 void
523 term_word(struct termp *p, const char *word)
524 {
525 	struct roffsu	 su;
526 	const char	 nbrsp[2] = { ASCII_NBRSP, 0 };
527 	const char	*seq, *cp;
528 	int		 sz, uc;
529 	size_t		 csz, lsz, ssz;
530 	enum mandoc_esc	 esc;
531 
532 	if ((p->flags & TERMP_NOBUF) == 0) {
533 		if ((p->flags & TERMP_NOSPACE) == 0) {
534 			if ((p->flags & TERMP_KEEP) == 0) {
535 				bufferc(p, ' ');
536 				if (p->flags & TERMP_SENTENCE)
537 					bufferc(p, ' ');
538 			} else
539 				bufferc(p, ASCII_NBRSP);
540 		}
541 		if (p->flags & TERMP_PREKEEP)
542 			p->flags |= TERMP_KEEP;
543 		if (p->flags & TERMP_NONOSPACE)
544 			p->flags |= TERMP_NOSPACE;
545 		else
546 			p->flags &= ~TERMP_NOSPACE;
547 		p->flags &= ~(TERMP_SENTENCE | TERMP_NONEWLINE);
548 		p->skipvsp = 0;
549 	}
550 
551 	while ('\0' != *word) {
552 		if ('\\' != *word) {
553 			if (TERMP_NBRWORD & p->flags) {
554 				if (' ' == *word) {
555 					encode(p, nbrsp, 1);
556 					word++;
557 					continue;
558 				}
559 				ssz = strcspn(word, "\\ ");
560 			} else
561 				ssz = strcspn(word, "\\");
562 			encode(p, word, ssz);
563 			word += (int)ssz;
564 			continue;
565 		}
566 
567 		word++;
568 		esc = mandoc_escape(&word, &seq, &sz);
569 		switch (esc) {
570 		case ESCAPE_UNICODE:
571 			uc = mchars_num2uc(seq + 1, sz - 1);
572 			break;
573 		case ESCAPE_NUMBERED:
574 			uc = mchars_num2char(seq, sz);
575 			if (uc < 0)
576 				continue;
577 			break;
578 		case ESCAPE_SPECIAL:
579 			if (p->enc == TERMENC_ASCII) {
580 				cp = mchars_spec2str(seq, sz, &ssz);
581 				if (cp != NULL)
582 					encode(p, cp, ssz);
583 			} else {
584 				uc = mchars_spec2cp(seq, sz);
585 				if (uc > 0)
586 					encode1(p, uc);
587 			}
588 			continue;
589 		case ESCAPE_UNDEF:
590 			uc = *seq;
591 			break;
592 		case ESCAPE_FONTBOLD:
593 			term_fontrepl(p, TERMFONT_BOLD);
594 			continue;
595 		case ESCAPE_FONTITALIC:
596 			term_fontrepl(p, TERMFONT_UNDER);
597 			continue;
598 		case ESCAPE_FONTBI:
599 			term_fontrepl(p, TERMFONT_BI);
600 			continue;
601 		case ESCAPE_FONT:
602 		case ESCAPE_FONTCW:
603 		case ESCAPE_FONTROMAN:
604 			term_fontrepl(p, TERMFONT_NONE);
605 			continue;
606 		case ESCAPE_FONTPREV:
607 			term_fontlast(p);
608 			continue;
609 		case ESCAPE_BREAK:
610 			bufferc(p, '\n');
611 			continue;
612 		case ESCAPE_NOSPACE:
613 			if (p->flags & TERMP_BACKAFTER)
614 				p->flags &= ~TERMP_BACKAFTER;
615 			else if (*word == '\0')
616 				p->flags |= (TERMP_NOSPACE | TERMP_NONEWLINE);
617 			continue;
618 		case ESCAPE_DEVICE:
619 			if (p->type == TERMTYPE_PDF)
620 				encode(p, "pdf", 3);
621 			else if (p->type == TERMTYPE_PS)
622 				encode(p, "ps", 2);
623 			else if (p->enc == TERMENC_ASCII)
624 				encode(p, "ascii", 5);
625 			else
626 				encode(p, "utf8", 4);
627 			continue;
628 		case ESCAPE_HORIZ:
629 			if (*seq == '|') {
630 				seq++;
631 				uc = -p->col;
632 			} else
633 				uc = 0;
634 			if (a2roffsu(seq, &su, SCALE_EM) == NULL)
635 				continue;
636 			uc += term_hen(p, &su);
637 			if (uc > 0)
638 				while (uc-- > 0)
639 					bufferc(p, ASCII_NBRSP);
640 			else if (p->col > (size_t)(-uc))
641 				p->col += uc;
642 			else {
643 				uc += p->col;
644 				p->col = 0;
645 				if (p->tcol->offset > (size_t)(-uc)) {
646 					p->ti += uc;
647 					p->tcol->offset += uc;
648 				} else {
649 					p->ti -= p->tcol->offset;
650 					p->tcol->offset = 0;
651 				}
652 			}
653 			continue;
654 		case ESCAPE_HLINE:
655 			if ((cp = a2roffsu(seq, &su, SCALE_EM)) == NULL)
656 				continue;
657 			uc = term_hen(p, &su);
658 			if (uc <= 0) {
659 				if (p->tcol->rmargin <= p->tcol->offset)
660 					continue;
661 				lsz = p->tcol->rmargin - p->tcol->offset;
662 			} else
663 				lsz = uc;
664 			if (*cp == seq[-1])
665 				uc = -1;
666 			else if (*cp == '\\') {
667 				seq = cp + 1;
668 				esc = mandoc_escape(&seq, &cp, &sz);
669 				switch (esc) {
670 				case ESCAPE_UNICODE:
671 					uc = mchars_num2uc(cp + 1, sz - 1);
672 					break;
673 				case ESCAPE_NUMBERED:
674 					uc = mchars_num2char(cp, sz);
675 					break;
676 				case ESCAPE_SPECIAL:
677 					uc = mchars_spec2cp(cp, sz);
678 					break;
679 				case ESCAPE_UNDEF:
680 					uc = *seq;
681 					break;
682 				default:
683 					uc = -1;
684 					break;
685 				}
686 			} else
687 				uc = *cp;
688 			if (uc < 0x20 || (uc > 0x7E && uc < 0xA0))
689 				uc = '_';
690 			if (p->enc == TERMENC_ASCII) {
691 				cp = ascii_uc2str(uc);
692 				csz = term_strlen(p, cp);
693 				ssz = strlen(cp);
694 			} else
695 				csz = (*p->width)(p, uc);
696 			while (lsz >= csz) {
697 				if (p->enc == TERMENC_ASCII)
698 					encode(p, cp, ssz);
699 				else
700 					encode1(p, uc);
701 				lsz -= csz;
702 			}
703 			continue;
704 		case ESCAPE_SKIPCHAR:
705 			p->flags |= TERMP_BACKAFTER;
706 			continue;
707 		case ESCAPE_OVERSTRIKE:
708 			cp = seq + sz;
709 			while (seq < cp) {
710 				if (*seq == '\\') {
711 					mandoc_escape(&seq, NULL, NULL);
712 					continue;
713 				}
714 				encode1(p, *seq++);
715 				if (seq < cp) {
716 					if (p->flags & TERMP_BACKBEFORE)
717 						p->flags |= TERMP_BACKAFTER;
718 					else
719 						p->flags |= TERMP_BACKBEFORE;
720 				}
721 			}
722 			/* Trim trailing backspace/blank pair. */
723 			if (p->tcol->lastcol > 2 &&
724 			    (p->tcol->buf[p->tcol->lastcol - 1] == ' ' ||
725 			     p->tcol->buf[p->tcol->lastcol - 1] == '\t'))
726 				p->tcol->lastcol -= 2;
727 			if (p->col > p->tcol->lastcol)
728 				p->col = p->tcol->lastcol;
729 			continue;
730 		default:
731 			continue;
732 		}
733 
734 		/*
735 		 * Common handling for Unicode and numbered
736 		 * character escape sequences.
737 		 */
738 
739 		if (p->enc == TERMENC_ASCII) {
740 			cp = ascii_uc2str(uc);
741 			encode(p, cp, strlen(cp));
742 		} else {
743 			if ((uc < 0x20 && uc != 0x09) ||
744 			    (uc > 0x7E && uc < 0xA0))
745 				uc = 0xFFFD;
746 			encode1(p, uc);
747 		}
748 	}
749 	p->flags &= ~TERMP_NBRWORD;
750 }
751 
752 static void
753 adjbuf(struct termp_col *c, size_t sz)
754 {
755 	if (c->maxcols == 0)
756 		c->maxcols = 1024;
757 	while (c->maxcols <= sz)
758 		c->maxcols <<= 2;
759 	c->buf = mandoc_reallocarray(c->buf, c->maxcols, sizeof(*c->buf));
760 }
761 
762 static void
763 bufferc(struct termp *p, char c)
764 {
765 	if (p->flags & TERMP_NOBUF) {
766 		(*p->letter)(p, c);
767 		return;
768 	}
769 	if (p->col + 1 >= p->tcol->maxcols)
770 		adjbuf(p->tcol, p->col + 1);
771 	if (p->tcol->lastcol <= p->col || (c != ' ' && c != ASCII_NBRSP))
772 		p->tcol->buf[p->col] = c;
773 	if (p->tcol->lastcol < ++p->col)
774 		p->tcol->lastcol = p->col;
775 }
776 
777 /*
778  * See encode().
779  * Do this for a single (probably unicode) value.
780  * Does not check for non-decorated glyphs.
781  */
782 static void
783 encode1(struct termp *p, int c)
784 {
785 	enum termfont	  f;
786 
787 	if (p->flags & TERMP_NOBUF) {
788 		(*p->letter)(p, c);
789 		return;
790 	}
791 
792 	if (p->col + 7 >= p->tcol->maxcols)
793 		adjbuf(p->tcol, p->col + 7);
794 
795 	f = (c == ASCII_HYPH || c > 127 || isgraph(c)) ?
796 	    p->fontq[p->fonti] : TERMFONT_NONE;
797 
798 	if (p->flags & TERMP_BACKBEFORE) {
799 		if (p->tcol->buf[p->col - 1] == ' ' ||
800 		    p->tcol->buf[p->col - 1] == '\t')
801 			p->col--;
802 		else
803 			p->tcol->buf[p->col++] = '\b';
804 		p->flags &= ~TERMP_BACKBEFORE;
805 	}
806 	if (f == TERMFONT_UNDER || f == TERMFONT_BI) {
807 		p->tcol->buf[p->col++] = '_';
808 		p->tcol->buf[p->col++] = '\b';
809 	}
810 	if (f == TERMFONT_BOLD || f == TERMFONT_BI) {
811 		if (c == ASCII_HYPH)
812 			p->tcol->buf[p->col++] = '-';
813 		else
814 			p->tcol->buf[p->col++] = c;
815 		p->tcol->buf[p->col++] = '\b';
816 	}
817 	if (p->tcol->lastcol <= p->col || (c != ' ' && c != ASCII_NBRSP))
818 		p->tcol->buf[p->col] = c;
819 	if (p->tcol->lastcol < ++p->col)
820 		p->tcol->lastcol = p->col;
821 	if (p->flags & TERMP_BACKAFTER) {
822 		p->flags |= TERMP_BACKBEFORE;
823 		p->flags &= ~TERMP_BACKAFTER;
824 	}
825 }
826 
827 static void
828 encode(struct termp *p, const char *word, size_t sz)
829 {
830 	size_t		  i;
831 
832 	if (p->flags & TERMP_NOBUF) {
833 		for (i = 0; i < sz; i++)
834 			(*p->letter)(p, word[i]);
835 		return;
836 	}
837 
838 	if (p->col + 2 + (sz * 5) >= p->tcol->maxcols)
839 		adjbuf(p->tcol, p->col + 2 + (sz * 5));
840 
841 	for (i = 0; i < sz; i++) {
842 		if (ASCII_HYPH == word[i] ||
843 		    isgraph((unsigned char)word[i]))
844 			encode1(p, word[i]);
845 		else {
846 			if (p->tcol->lastcol <= p->col ||
847 			    (word[i] != ' ' && word[i] != ASCII_NBRSP))
848 				p->tcol->buf[p->col] = word[i];
849 			p->col++;
850 
851 			/*
852 			 * Postpone the effect of \z while handling
853 			 * an overstrike sequence from ascii_uc2str().
854 			 */
855 
856 			if (word[i] == '\b' &&
857 			    (p->flags & TERMP_BACKBEFORE)) {
858 				p->flags &= ~TERMP_BACKBEFORE;
859 				p->flags |= TERMP_BACKAFTER;
860 			}
861 		}
862 	}
863 	if (p->tcol->lastcol < p->col)
864 		p->tcol->lastcol = p->col;
865 }
866 
867 void
868 term_setwidth(struct termp *p, const char *wstr)
869 {
870 	struct roffsu	 su;
871 	int		 iop, width;
872 
873 	iop = 0;
874 	width = 0;
875 	if (NULL != wstr) {
876 		switch (*wstr) {
877 		case '+':
878 			iop = 1;
879 			wstr++;
880 			break;
881 		case '-':
882 			iop = -1;
883 			wstr++;
884 			break;
885 		default:
886 			break;
887 		}
888 		if (a2roffsu(wstr, &su, SCALE_MAX) != NULL)
889 			width = term_hspan(p, &su);
890 		else
891 			iop = 0;
892 	}
893 	(*p->setwidth)(p, iop, width);
894 }
895 
896 size_t
897 term_len(const struct termp *p, size_t sz)
898 {
899 
900 	return (*p->width)(p, ' ') * sz;
901 }
902 
903 static size_t
904 cond_width(const struct termp *p, int c, int *skip)
905 {
906 
907 	if (*skip) {
908 		(*skip) = 0;
909 		return 0;
910 	} else
911 		return (*p->width)(p, c);
912 }
913 
914 size_t
915 term_strlen(const struct termp *p, const char *cp)
916 {
917 	size_t		 sz, rsz, i;
918 	int		 ssz, skip, uc;
919 	const char	*seq, *rhs;
920 	enum mandoc_esc	 esc;
921 	static const char rej[] = { '\\', ASCII_NBRSP, ASCII_HYPH,
922 			ASCII_BREAK, '\0' };
923 
924 	/*
925 	 * Account for escaped sequences within string length
926 	 * calculations.  This follows the logic in term_word() as we
927 	 * must calculate the width of produced strings.
928 	 */
929 
930 	sz = 0;
931 	skip = 0;
932 	while ('\0' != *cp) {
933 		rsz = strcspn(cp, rej);
934 		for (i = 0; i < rsz; i++)
935 			sz += cond_width(p, *cp++, &skip);
936 
937 		switch (*cp) {
938 		case '\\':
939 			cp++;
940 			rhs = NULL;
941 			esc = mandoc_escape(&cp, &seq, &ssz);
942 			switch (esc) {
943 			case ESCAPE_UNICODE:
944 				uc = mchars_num2uc(seq + 1, ssz - 1);
945 				break;
946 			case ESCAPE_NUMBERED:
947 				uc = mchars_num2char(seq, ssz);
948 				if (uc < 0)
949 					continue;
950 				break;
951 			case ESCAPE_SPECIAL:
952 				if (p->enc == TERMENC_ASCII) {
953 					rhs = mchars_spec2str(seq, ssz, &rsz);
954 					if (rhs != NULL)
955 						break;
956 				} else {
957 					uc = mchars_spec2cp(seq, ssz);
958 					if (uc > 0)
959 						sz += cond_width(p, uc, &skip);
960 				}
961 				continue;
962 			case ESCAPE_UNDEF:
963 				uc = *seq;
964 				break;
965 			case ESCAPE_DEVICE:
966 				if (p->type == TERMTYPE_PDF) {
967 					rhs = "pdf";
968 					rsz = 3;
969 				} else if (p->type == TERMTYPE_PS) {
970 					rhs = "ps";
971 					rsz = 2;
972 				} else if (p->enc == TERMENC_ASCII) {
973 					rhs = "ascii";
974 					rsz = 5;
975 				} else {
976 					rhs = "utf8";
977 					rsz = 4;
978 				}
979 				break;
980 			case ESCAPE_SKIPCHAR:
981 				skip = 1;
982 				continue;
983 			case ESCAPE_OVERSTRIKE:
984 				rsz = 0;
985 				rhs = seq + ssz;
986 				while (seq < rhs) {
987 					if (*seq == '\\') {
988 						mandoc_escape(&seq, NULL, NULL);
989 						continue;
990 					}
991 					i = (*p->width)(p, *seq++);
992 					if (rsz < i)
993 						rsz = i;
994 				}
995 				sz += rsz;
996 				continue;
997 			default:
998 				continue;
999 			}
1000 
1001 			/*
1002 			 * Common handling for Unicode and numbered
1003 			 * character escape sequences.
1004 			 */
1005 
1006 			if (rhs == NULL) {
1007 				if (p->enc == TERMENC_ASCII) {
1008 					rhs = ascii_uc2str(uc);
1009 					rsz = strlen(rhs);
1010 				} else {
1011 					if ((uc < 0x20 && uc != 0x09) ||
1012 					    (uc > 0x7E && uc < 0xA0))
1013 						uc = 0xFFFD;
1014 					sz += cond_width(p, uc, &skip);
1015 					continue;
1016 				}
1017 			}
1018 
1019 			if (skip) {
1020 				skip = 0;
1021 				break;
1022 			}
1023 
1024 			/*
1025 			 * Common handling for all escape sequences
1026 			 * printing more than one character.
1027 			 */
1028 
1029 			for (i = 0; i < rsz; i++)
1030 				sz += (*p->width)(p, *rhs++);
1031 			break;
1032 		case ASCII_NBRSP:
1033 			sz += cond_width(p, ' ', &skip);
1034 			cp++;
1035 			break;
1036 		case ASCII_HYPH:
1037 			sz += cond_width(p, '-', &skip);
1038 			cp++;
1039 			break;
1040 		default:
1041 			break;
1042 		}
1043 	}
1044 
1045 	return sz;
1046 }
1047 
1048 int
1049 term_vspan(const struct termp *p, const struct roffsu *su)
1050 {
1051 	double		 r;
1052 	int		 ri;
1053 
1054 	switch (su->unit) {
1055 	case SCALE_BU:
1056 		r = su->scale / 40.0;
1057 		break;
1058 	case SCALE_CM:
1059 		r = su->scale * 6.0 / 2.54;
1060 		break;
1061 	case SCALE_FS:
1062 		r = su->scale * 65536.0 / 40.0;
1063 		break;
1064 	case SCALE_IN:
1065 		r = su->scale * 6.0;
1066 		break;
1067 	case SCALE_MM:
1068 		r = su->scale * 0.006;
1069 		break;
1070 	case SCALE_PC:
1071 		r = su->scale;
1072 		break;
1073 	case SCALE_PT:
1074 		r = su->scale / 12.0;
1075 		break;
1076 	case SCALE_EN:
1077 	case SCALE_EM:
1078 		r = su->scale * 0.6;
1079 		break;
1080 	case SCALE_VS:
1081 		r = su->scale;
1082 		break;
1083 	default:
1084 		abort();
1085 	}
1086 	ri = r > 0.0 ? r + 0.4995 : r - 0.4995;
1087 	return ri < 66 ? ri : 1;
1088 }
1089 
1090 /*
1091  * Convert a scaling width to basic units, rounding towards 0.
1092  */
1093 int
1094 term_hspan(const struct termp *p, const struct roffsu *su)
1095 {
1096 
1097 	return (*p->hspan)(p, su);
1098 }
1099 
1100 /*
1101  * Convert a scaling width to basic units, rounding to closest.
1102  */
1103 int
1104 term_hen(const struct termp *p, const struct roffsu *su)
1105 {
1106 	int bu;
1107 
1108 	if ((bu = (*p->hspan)(p, su)) >= 0)
1109 		return (bu + 11) / 24;
1110 	else
1111 		return -((-bu + 11) / 24);
1112 }
1113