===================================================================
RCS file: /cvs/mandoc/term.c,v
retrieving revision 1.74
retrieving revision 1.260
diff -u -p -r1.74 -r1.260
--- mandoc/term.c	2009/04/12 19:45:26	1.74
+++ mandoc/term.c	2017/05/07 17:31:45	1.260
@@ -1,327 +1,313 @@
-/*	$Id: term.c,v 1.74 2009/04/12 19:45:26 kristaps Exp $ */
+/*	$Id: term.c,v 1.260 2017/05/07 17:31:45 schwarze Exp $ */
 /*
- * Copyright (c) 2008, 2009 Kristaps Dzonsons <kristaps@openbsd.org>
+ * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
+ * Copyright (c) 2010-2017 Ingo Schwarze <schwarze@openbsd.org>
  *
  * Permission to use, copy, modify, and distribute this software for any
  * purpose with or without fee is hereby granted, provided that the above
  * copyright notice and this permission notice appear in all copies.
  *
- * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
- * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  */
+#include "config.h"
+
+#include <sys/types.h>
+
 #include <assert.h>
-#include <err.h>
+#include <ctype.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
 
+#include "mandoc.h"
+#include "mandoc_aux.h"
+#include "out.h"
 #include "term.h"
-#include "man.h"
-#include "mdoc.h"
+#include "main.h"
 
-extern	int		  man_run(struct termp *, 
-				const struct man *);
-extern	int		  mdoc_run(struct termp *, 
-				const struct mdoc *);
+static	size_t		 cond_width(const struct termp *, int, int *);
+static	void		 adjbuf(struct termp *p, size_t);
+static	void		 bufferc(struct termp *, char);
+static	void		 encode(struct termp *, const char *, size_t);
+static	void		 encode1(struct termp *, int);
 
-static	struct termp	 *term_alloc(enum termenc);
-static	void		  term_free(struct termp *);
-static	void		  term_pword(struct termp *, const char *, int);
-static	void		  term_pescape(struct termp *, 
-				const char *, int *, int);
-static	void		  term_nescape(struct termp *,
-				const char *, size_t);
-static	void		  term_chara(struct termp *, char);
-static	void		  term_stringa(struct termp *, 
-				const char *, size_t);
-static	int		  term_isopendelim(const char *, int);
-static	int		  term_isclosedelim(const char *, int);
 
-
-void *
-ascii_alloc(void)
-{
-
-	return(term_alloc(TERMENC_ASCII));
-}
-
-
-int
-terminal_man(void *arg, const struct man *man)
-{
-	struct termp	*p;
-
-	p = (struct termp *)arg;
-	if (NULL == p->symtab)
-		p->symtab = term_ascii2htab();
-
-	return(man_run(p, man));
-}
-
-
-int
-terminal_mdoc(void *arg, const struct mdoc *mdoc)
-{
-	struct termp	*p;
-
-	p = (struct termp *)arg;
-	if (NULL == p->symtab)
-		p->symtab = term_ascii2htab();
-
-	return(mdoc_run(p, mdoc));
-}
-
-
 void
-terminal_free(void *arg)
-{
-
-	term_free((struct termp *)arg);
-}
-
-
-static void
 term_free(struct termp *p)
 {
 
-	if (p->buf)
-		free(p->buf);
-	if (TERMENC_ASCII == p->enc && p->symtab)
-		term_asciifree(p->symtab);
-
+	free(p->buf);
+	free(p->fontq);
 	free(p);
 }
 
-
-static struct termp *
-term_alloc(enum termenc enc)
+void
+term_begin(struct termp *p, term_margin head,
+		term_margin foot, const struct roff_meta *arg)
 {
-	struct termp *p;
 
-	if (NULL == (p = malloc(sizeof(struct termp))))
-		err(1, "malloc");
-	bzero(p, sizeof(struct termp));
-	p->maxrmargin = 78;
-	p->enc = enc;
-	return(p);
+	p->headf = head;
+	p->footf = foot;
+	p->argf = arg;
+	(*p->begin)(p);
 }
 
-
-static int
-term_isclosedelim(const char *p, int len)
+void
+term_end(struct termp *p)
 {
 
-	if (1 != len)
-		return(0);
-
-	switch (*p) {
-	case('.'):
-		/* FALLTHROUGH */
-	case(','):
-		/* FALLTHROUGH */
-	case(';'):
-		/* FALLTHROUGH */
-	case(':'):
-		/* FALLTHROUGH */
-	case('?'):
-		/* FALLTHROUGH */
-	case('!'):
-		/* FALLTHROUGH */
-	case(')'):
-		/* FALLTHROUGH */
-	case(']'):
-		/* FALLTHROUGH */
-	case('}'):
-		return(1);
-	default:
-		break;
-	}
-
-	return(0);
+	(*p->end)(p);
 }
 
-
-static int
-term_isopendelim(const char *p, int len)
-{
-
-	if (1 != len)
-		return(0);
-
-	switch (*p) {
-	case('('):
-		/* FALLTHROUGH */
-	case('['):
-		/* FALLTHROUGH */
-	case('{'):
-		return(1);
-	default:
-		break;
-	}
-
-	return(0);
-}
-
-
 /*
- * Flush a line of text.  A "line" is loosely defined as being something
- * that should be followed by a newline, regardless of whether it's
- * broken apart by newlines getting there.  A line can also be a
- * fragment of a columnar list.
+ * Flush a chunk of text.  By default, break the output line each time
+ * the right margin is reached, and continue output on the next line
+ * at the same offset as the chunk itself.  By default, also break the
+ * output line at the end of the chunk.
+ * The following flags may be specified:
  *
- * Specifically, a line is whatever's in p->buf of length p->col, which
- * is zeroed after this function returns.
- *
- * The variables TERMP_NOLPAD, TERMP_LITERAL and TERMP_NOBREAK are of
- * critical importance here.  Their behaviour follows:
- *
- *  - TERMP_NOLPAD: when beginning to write the line, don't left-pad the
- *    offset value.  This is useful when doing columnar lists where the
- *    prior column has right-padded.
- *
- *  - TERMP_NOBREAK: this is the most important and is used when making
- *    columns.  In short: don't print a newline and instead pad to the
- *    right margin.  Used in conjunction with TERMP_NOLPAD.
- *
- *  - TERMP_NONOBREAK: don't newline when TERMP_NOBREAK is specified.
- *
- *  In-line line breaking:
- *
- *  If TERMP_NOBREAK is specified and the line overruns the right
- *  margin, it will break and pad-right to the right margin after
- *  writing.  If maxrmargin is violated, it will break and continue
- *  writing from the right-margin, which will lead to the above
- *  scenario upon exit.
- *
- *  Otherwise, the line will break at the right margin.  Extremely long
- *  lines will cause the system to emit a warning (TODO: hyphenate, if
- *  possible).
+ *  - TERMP_NOBREAK: Do not break the output line at the right margin,
+ *    but only at the max right margin.  Also, do not break the output
+ *    line at the end of the chunk, such that the next call can pad to
+ *    the next column.  However, if less than p->trailspace blanks,
+ *    which can be 0, 1, or 2, remain to the right margin, the line
+ *    will be broken.
+ *  - TERMP_BRTRSP: Consider trailing whitespace significant
+ *    when deciding whether the chunk fits or not.
+ *  - TERMP_BRIND: If the chunk does not fit and the output line has
+ *    to be broken, start the next line at the right margin instead
+ *    of at the offset.  Used together with TERMP_NOBREAK for the tags
+ *    in various kinds of tagged lists.
+ *  - TERMP_DANGLE: Do not break the output line at the right margin,
+ *    append the next chunk after it even if this one is too long.
+ *    To be used together with TERMP_NOBREAK.
+ *  - TERMP_HANG: Like TERMP_DANGLE, and also suppress padding before
+ *    the next chunk if this column is not full.
  */
 void
 term_flushln(struct termp *p)
 {
-	int		 i, j;
-	size_t		 vsz, vis, maxvis, mmax, bp;
+	size_t		 i;     /* current input position in p->buf */
+	int		 ntab;	/* number of tabs to prepend */
+	size_t		 vis;   /* current visual position on output */
+	size_t		 vbl;   /* number of blanks to prepend to output */
+	size_t		 vend;	/* end of word visual position on output */
+	size_t		 bp;    /* visual right border position */
+	size_t		 dv;    /* temporary for visual pos calculations */
+	size_t		 j;     /* temporary loop index for p->buf */
+	size_t		 jhy;	/* last hyph before overflow w/r/t j */
+	size_t		 maxvis; /* output position of visible boundary */
 
 	/*
 	 * First, establish the maximum columns of "visible" content.
 	 * This is usually the difference between the right-margin and
 	 * an indentation, but can be, for tagged lists or columns, a
 	 * small set of values.
+	 *
+	 * The following unsigned-signed subtractions look strange,
+	 * but they are actually correct.  If the int p->overstep
+	 * is negative, it gets sign extended.  Subtracting that
+	 * very large size_t effectively adds a small number to dv.
 	 */
+	dv = p->rmargin > p->offset ? p->rmargin - p->offset : 0;
+	maxvis = (int)dv > p->overstep ? dv - (size_t)p->overstep : 0;
 
-	assert(p->offset < p->rmargin);
-	maxvis = p->rmargin - p->offset;
-	mmax = p->maxrmargin - p->offset;
-	bp = TERMP_NOBREAK & p->flags ? mmax : maxvis;
-	vis = 0;
+	if (p->flags & TERMP_NOBREAK) {
+		dv = p->maxrmargin > p->offset ?
+		     p->maxrmargin - p->offset : 0;
+		bp = (int)dv > p->overstep ?
+		     dv - (size_t)p->overstep : 0;
+	} else
+		bp = maxvis;
 
 	/*
-	 * If in the standard case (left-justified), then begin with our
-	 * indentation, otherwise (columns, etc.) just start spitting
-	 * out text.
+	 * Calculate the required amount of padding.
 	 */
+	vbl = p->offset + p->overstep > p->viscol ?
+	      p->offset + p->overstep - p->viscol : 0;
 
-	if ( ! (p->flags & TERMP_NOLPAD))
-		/* LINTED */
-		for (j = 0; j < (int)p->offset; j++)
-			putchar(' ');
+	vis = vend = 0;
+	i = 0;
 
-	for (i = 0; i < (int)p->col; i++) {
+	while (i < p->col) {
 		/*
+		 * Handle literal tab characters: collapse all
+		 * subsequent tabs into a single huge set of spaces.
+		 */
+		ntab = 0;
+		while (i < p->col && p->buf[i] == '\t') {
+			vend = term_tab_next(vis);
+			vbl += vend - vis;
+			vis = vend;
+			ntab++;
+			i++;
+		}
+
+		/*
 		 * Count up visible word characters.  Control sequences
 		 * (starting with the CSI) aren't counted.  A space
 		 * generates a non-printing word, which is valid (the
 		 * space is printed according to regular spacing rules).
 		 */
 
-		/* LINTED */
-		for (j = i, vsz = 0; j < (int)p->col; j++) {
-			if (' ' == p->buf[j])
+		for (j = i, jhy = 0; j < p->col; j++) {
+			if (' ' == p->buf[j] || '\t' == p->buf[j])
 				break;
-			else if (8 == p->buf[j])
-				j += 1;
-			else
-				vsz++;
+
+			/* Back over the last printed character. */
+			if (8 == p->buf[j]) {
+				assert(j);
+				vend -= (*p->width)(p, p->buf[j - 1]);
+				continue;
+			}
+
+			/* Regular word. */
+			/* Break at the hyphen point if we overrun. */
+			if (vend > vis && vend < bp &&
+			    (ASCII_HYPH == p->buf[j] ||
+			     ASCII_BREAK == p->buf[j]))
+				jhy = j;
+
+			/*
+			 * Hyphenation now decided, put back a real
+			 * hyphen such that we get the correct width.
+			 */
+			if (ASCII_HYPH == p->buf[j])
+				p->buf[j] = '-';
+
+			vend += (*p->width)(p, p->buf[j]);
 		}
 
 		/*
-		 * Do line-breaking.  If we're greater than our
-		 * break-point and already in-line, break to the next
-		 * line and start writing.  If we're at the line start,
-		 * then write out the word (TODO: hyphenate) and break
-		 * in a subsequent loop invocation.
+		 * Find out whether we would exceed the right margin.
+		 * If so, break to the next line.
 		 */
+		if (vend > bp && 0 == jhy && vis > 0) {
+			vend -= vis;
+			(*p->endline)(p);
+			p->viscol = 0;
 
-		if ( ! (TERMP_NOBREAK & p->flags)) {
-			if (vis && vis + vsz > bp) {
-				putchar('\n');
-				for (j = 0; j < (int)p->offset; j++)
-					putchar(' ');
-				vis = 0;
-			} 
-		} else if (vis && vis + vsz > bp) {
-			putchar('\n');
-			for (j = 0; j < (int)p->rmargin; j++)
-				putchar(' ');
-			vis = p->rmargin - p->offset;
-		}
+			/* Use pending tabs on the new line. */
 
-		/* 
-		 * Write out the word and a trailing space.  Omit the
-		 * space if we're the last word in the line or beyond
-		 * our breakpoint.
-		 */
+			vbl = 0;
+			while (ntab--)
+				vbl = term_tab_next(vbl);
 
-		for ( ; i < (int)p->col; i++) {
-			if (' ' == p->buf[i])
+			/* Re-establish indentation. */
+
+			if (p->flags & TERMP_BRIND) {
+				vbl += p->rmargin;
+				vend += p->rmargin - p->offset;
+			} else
+				vbl += p->offset;
+
+			/*
+			 * Remove the p->overstep width.
+			 * Again, if p->overstep is negative,
+			 * sign extension does the right thing.
+			 */
+
+			bp += (size_t)p->overstep;
+			p->overstep = 0;
+		}
+
+		/* Write out the [remaining] word. */
+		for ( ; i < p->col; i++) {
+			if (vend > bp && jhy > 0 && i > jhy)
 				break;
-			putchar(p->buf[i]);
+			if ('\t' == p->buf[i])
+				break;
+			if (' ' == p->buf[i]) {
+				j = i;
+				while (i < p->col && ' ' == p->buf[i])
+					i++;
+				dv = (i - j) * (*p->width)(p, ' ');
+				vbl += dv;
+				vend += dv;
+				break;
+			}
+			if (ASCII_NBRSP == p->buf[i]) {
+				vbl += (*p->width)(p, ' ');
+				continue;
+			}
+			if (ASCII_BREAK == p->buf[i])
+				continue;
+
+			/*
+			 * Now we definitely know there will be
+			 * printable characters to output,
+			 * so write preceding white space now.
+			 */
+			if (vbl) {
+				(*p->advance)(p, vbl);
+				p->viscol += vbl;
+				vbl = 0;
+			}
+
+			(*p->letter)(p, p->buf[i]);
+			if (8 == p->buf[i])
+				p->viscol -= (*p->width)(p, p->buf[i-1]);
+			else
+				p->viscol += (*p->width)(p, p->buf[i]);
 		}
-		vis += vsz;
-		if (i < (int)p->col && vis <= bp) {
-			putchar(' ');
-			vis++;
-		}
+		vis = vend;
 	}
 
 	/*
-	 * If we've overstepped our maximum visible no-break space, then
-	 * cause a newline and offset at the right margin.
+	 * If there was trailing white space, it was not printed;
+	 * so reset the cursor position accordingly.
 	 */
+	if (vis > vbl)
+		vis -= vbl;
+	else
+		vis = 0;
 
-	if ((TERMP_NOBREAK & p->flags) && vis >= maxvis) {
-		if ( ! (TERMP_NONOBREAK & p->flags)) {
-			putchar('\n');
-			for (i = 0; i < (int)p->rmargin; i++)
-				putchar(' ');
-		}
-		p->col = 0;
+	p->col = 0;
+	p->overstep = 0;
+	p->flags &= ~(TERMP_BACKAFTER | TERMP_BACKBEFORE);
+
+	if ( ! (TERMP_NOBREAK & p->flags)) {
+		p->viscol = 0;
+		(*p->endline)(p);
 		return;
 	}
 
-	/*
-	 * If we're not to right-marginalise it (newline), then instead
-	 * pad to the right margin and stay off.
-	 */
+	if (TERMP_HANG & p->flags) {
+		p->overstep += (int)(p->offset + vis - p->rmargin +
+		    p->trailspace * (*p->width)(p, ' '));
 
-	if (p->flags & TERMP_NOBREAK) {
-		if ( ! (TERMP_NONOBREAK & p->flags))
-			for ( ; vis < maxvis; vis++)
-				putchar(' ');
-	} else
-		putchar('\n');
+		/*
+		 * If we have overstepped the margin, temporarily move
+		 * it to the right and flag the rest of the line to be
+		 * shorter.
+		 * If there is a request to keep the columns together,
+		 * allow negative overstep when the column is not full.
+		 */
+		if (p->trailspace && p->overstep < 0)
+			p->overstep = 0;
+		return;
 
-	p->col = 0;
-}
+	} else if (TERMP_DANGLE & p->flags)
+		return;
 
+	/* Trailing whitespace is significant in some columns. */
+	if (vis && vbl && (TERMP_BRTRSP & p->flags))
+		vis += vbl;
 
-/* 
+	/* If the column was overrun, break the line. */
+	if (maxvis < vis + p->trailspace * (*p->width)(p, ' ')) {
+		(*p->endline)(p);
+		p->viscol = 0;
+	}
+}
+
+/*
  * A newline only breaks an existing line; it won't assert vertical
  * space.  All data in the output buffer is flushed prior to the newline
  * assertion.
@@ -331,15 +317,10 @@ term_newln(struct termp *p)
 {
 
 	p->flags |= TERMP_NOSPACE;
-	if (0 == p->col) {
-		p->flags &= ~TERMP_NOLPAD;
-		return;
-	}
-	term_flushln(p);
-	p->flags &= ~TERMP_NOLPAD;
+	if (p->col || p->viscol)
+		term_flushln(p);
 }
 
-
 /*
  * Asserts a vertical space (a full, empty line-break between lines).
  * Note that if used twice, this will cause two blank spaces and so on.
@@ -351,249 +332,515 @@ term_vspace(struct termp *p)
 {
 
 	term_newln(p);
-	putchar('\n');
+	p->viscol = 0;
+	if (0 < p->skipvsp)
+		p->skipvsp--;
+	else
+		(*p->endline)(p);
 }
 
+/* Swap current and previous font; for \fP and .ft P */
+void
+term_fontlast(struct termp *p)
+{
+	enum termfont	 f;
 
+	f = p->fontl;
+	p->fontl = p->fontq[p->fonti];
+	p->fontq[p->fonti] = f;
+}
+
+/* Set font, save current, discard previous; for \f, .ft, .B etc. */
+void
+term_fontrepl(struct termp *p, enum termfont f)
+{
+
+	p->fontl = p->fontq[p->fonti];
+	p->fontq[p->fonti] = f;
+}
+
+/* Set font, save previous. */
+void
+term_fontpush(struct termp *p, enum termfont f)
+{
+
+	p->fontl = p->fontq[p->fonti];
+	if (++p->fonti == p->fontsz) {
+		p->fontsz += 8;
+		p->fontq = mandoc_reallocarray(p->fontq,
+		    p->fontsz, sizeof(*p->fontq));
+	}
+	p->fontq[p->fonti] = f;
+}
+
+/* Flush to make the saved pointer current again. */
+void
+term_fontpopq(struct termp *p, int i)
+{
+
+	assert(i >= 0);
+	if (p->fonti > i)
+		p->fonti = i;
+}
+
+/* Pop one font off the stack. */
+void
+term_fontpop(struct termp *p)
+{
+
+	assert(p->fonti);
+	p->fonti--;
+}
+
 /*
- * Break apart a word into "pwords" (partial-words, usually from
- * breaking up a phrase into individual words) and, eventually, put them
- * into the output buffer.  If we're a literal word, then don't break up
- * the word and put it verbatim into the output buffer.
+ * Handle pwords, partial words, which may be either a single word or a
+ * phrase that cannot be broken down (such as a literal string).  This
+ * handles word styling.
  */
 void
 term_word(struct termp *p, const char *word)
 {
-	int 		 i, j, len;
+	const char	 nbrsp[2] = { ASCII_NBRSP, 0 };
+	const char	*seq, *cp;
+	int		 sz, uc;
+	size_t		 ssz;
+	enum mandoc_esc	 esc;
 
-	len = (int)strlen(word);
-
-	if (p->flags & TERMP_LITERAL) {
-		term_pword(p, word, len);
-		return;
+	if ( ! (TERMP_NOSPACE & p->flags)) {
+		if ( ! (TERMP_KEEP & p->flags)) {
+			bufferc(p, ' ');
+			if (TERMP_SENTENCE & p->flags)
+				bufferc(p, ' ');
+		} else
+			bufferc(p, ASCII_NBRSP);
 	}
+	if (TERMP_PREKEEP & p->flags)
+		p->flags |= TERMP_KEEP;
 
-	/* LINTED */
-	for (j = i = 0; i < len; i++) {
-		if (' ' != word[i]) {
-			j++;
+	if ( ! (p->flags & TERMP_NONOSPACE))
+		p->flags &= ~TERMP_NOSPACE;
+	else
+		p->flags |= TERMP_NOSPACE;
+
+	p->flags &= ~(TERMP_SENTENCE | TERMP_NONEWLINE);
+	p->skipvsp = 0;
+
+	while ('\0' != *word) {
+		if ('\\' != *word) {
+			if (TERMP_NBRWORD & p->flags) {
+				if (' ' == *word) {
+					encode(p, nbrsp, 1);
+					word++;
+					continue;
+				}
+				ssz = strcspn(word, "\\ ");
+			} else
+				ssz = strcspn(word, "\\");
+			encode(p, word, ssz);
+			word += (int)ssz;
 			continue;
-		} 
-		
-		/* Escaped spaces don't delimit... */
-		if (i && ' ' == word[i] && '\\' == word[i - 1]) {
-			j++;
-			continue;
 		}
 
-		if (0 == j)
+		word++;
+		esc = mandoc_escape(&word, &seq, &sz);
+		if (ESCAPE_ERROR == esc)
 			continue;
-		assert(i >= j);
-		term_pword(p, &word[i - j], j);
-		j = 0;
+
+		switch (esc) {
+		case ESCAPE_UNICODE:
+			uc = mchars_num2uc(seq + 1, sz - 1);
+			break;
+		case ESCAPE_NUMBERED:
+			uc = mchars_num2char(seq, sz);
+			if (uc < 0)
+				continue;
+			break;
+		case ESCAPE_SPECIAL:
+			if (p->enc == TERMENC_ASCII) {
+				cp = mchars_spec2str(seq, sz, &ssz);
+				if (cp != NULL)
+					encode(p, cp, ssz);
+			} else {
+				uc = mchars_spec2cp(seq, sz);
+				if (uc > 0)
+					encode1(p, uc);
+			}
+			continue;
+		case ESCAPE_FONTBOLD:
+			term_fontrepl(p, TERMFONT_BOLD);
+			continue;
+		case ESCAPE_FONTITALIC:
+			term_fontrepl(p, TERMFONT_UNDER);
+			continue;
+		case ESCAPE_FONTBI:
+			term_fontrepl(p, TERMFONT_BI);
+			continue;
+		case ESCAPE_FONT:
+		case ESCAPE_FONTROMAN:
+			term_fontrepl(p, TERMFONT_NONE);
+			continue;
+		case ESCAPE_FONTPREV:
+			term_fontlast(p);
+			continue;
+		case ESCAPE_NOSPACE:
+			if (p->flags & TERMP_BACKAFTER)
+				p->flags &= ~TERMP_BACKAFTER;
+			else if (*word == '\0')
+				p->flags |= (TERMP_NOSPACE | TERMP_NONEWLINE);
+			continue;
+		case ESCAPE_SKIPCHAR:
+			p->flags |= TERMP_BACKAFTER;
+			continue;
+		case ESCAPE_OVERSTRIKE:
+			cp = seq + sz;
+			while (seq < cp) {
+				if (*seq == '\\') {
+					mandoc_escape(&seq, NULL, NULL);
+					continue;
+				}
+				encode1(p, *seq++);
+				if (seq < cp) {
+					if (p->flags & TERMP_BACKBEFORE)
+						p->flags |= TERMP_BACKAFTER;
+					else
+						p->flags |= TERMP_BACKBEFORE;
+				}
+			}
+			/* Trim trailing backspace/blank pair. */
+			if (p->col > 2 &&
+			    (p->buf[p->col - 1] == ' ' ||
+			     p->buf[p->col - 1] == '\t'))
+				p->col -= 2;
+			continue;
+		default:
+			continue;
+		}
+
+		/*
+		 * Common handling for Unicode and numbered
+		 * character escape sequences.
+		 */
+
+		if (p->enc == TERMENC_ASCII) {
+			cp = ascii_uc2str(uc);
+			encode(p, cp, strlen(cp));
+		} else {
+			if ((uc < 0x20 && uc != 0x09) ||
+			    (uc > 0x7E && uc < 0xA0))
+				uc = 0xFFFD;
+			encode1(p, uc);
+		}
 	}
-	if (j > 0) {
-		assert(i >= j);
-		term_pword(p, &word[i - j], j);
-	}
+	p->flags &= ~TERMP_NBRWORD;
 }
 
-
-/*
- * Determine the symbol indicated by an escape sequences, that is, one
- * starting with a backslash.  Once done, we pass this value into the
- * output buffer by way of the symbol table.
- */
 static void
-term_nescape(struct termp *p, const char *word, size_t len)
+adjbuf(struct termp *p, size_t sz)
 {
-	const char	*rhs;
-	size_t		 sz;
 
-	if (NULL == (rhs = term_a2ascii(p->symtab, word, len, &sz))) 
-		return;
-	term_stringa(p, rhs, sz);
+	if (0 == p->maxcols)
+		p->maxcols = 1024;
+	while (sz >= p->maxcols)
+		p->maxcols <<= 2;
+
+	p->buf = mandoc_reallocarray(p->buf, p->maxcols, sizeof(int));
 }
 
+static void
+bufferc(struct termp *p, char c)
+{
 
+	if (p->col + 1 >= p->maxcols)
+		adjbuf(p, p->col + 1);
+
+	p->buf[p->col++] = c;
+}
+
 /*
- * Handle an escape sequence: determine its length and pass it to the
- * escape-symbol look table.  Note that we assume mdoc(3) has validated
- * the escape sequence (we assert upon badly-formed escape sequences).
+ * See encode().
+ * Do this for a single (probably unicode) value.
+ * Does not check for non-decorated glyphs.
  */
 static void
-term_pescape(struct termp *p, const char *word, int *i, int len)
+encode1(struct termp *p, int c)
 {
-	int		 j;
+	enum termfont	  f;
 
-	if (++(*i) >= len)
-		return;
+	if (p->col + 7 >= p->maxcols)
+		adjbuf(p, p->col + 7);
 
-	if ('(' == word[*i]) {
-		(*i)++;
-		if (*i + 1 >= len)
-			return;
+	f = (c == ASCII_HYPH || c > 127 || isgraph(c)) ?
+	    p->fontq[p->fonti] : TERMFONT_NONE;
 
-		term_nescape(p, &word[*i], 2);
-		(*i)++;
-		return;
+	if (p->flags & TERMP_BACKBEFORE) {
+		if (p->buf[p->col - 1] == ' ' || p->buf[p->col - 1] == '\t')
+			p->col--;
+		else
+			p->buf[p->col++] = 8;
+		p->flags &= ~TERMP_BACKBEFORE;
+	}
+	if (TERMFONT_UNDER == f || TERMFONT_BI == f) {
+		p->buf[p->col++] = '_';
+		p->buf[p->col++] = 8;
+	}
+	if (TERMFONT_BOLD == f || TERMFONT_BI == f) {
+		if (ASCII_HYPH == c)
+			p->buf[p->col++] = '-';
+		else
+			p->buf[p->col++] = c;
+		p->buf[p->col++] = 8;
+	}
+	p->buf[p->col++] = c;
+	if (p->flags & TERMP_BACKAFTER) {
+		p->flags |= TERMP_BACKBEFORE;
+		p->flags &= ~TERMP_BACKAFTER;
+	}
+}
 
-	} else if ('*' == word[*i]) { 
-		(*i)++;
-		if (*i >= len)
-			return;
+static void
+encode(struct termp *p, const char *word, size_t sz)
+{
+	size_t		  i;
 
-		switch (word[*i]) {
-		case ('('):
-			(*i)++;
-			if (*i + 1 >= len)
-				return;
+	if (p->col + 2 + (sz * 5) >= p->maxcols)
+		adjbuf(p, p->col + 2 + (sz * 5));
 
-			term_nescape(p, &word[*i], 2);
-			(*i)++;
-			return;
-		case ('['):
-			break;
-		default:
-			term_nescape(p, &word[*i], 1);
-			return;
+	for (i = 0; i < sz; i++) {
+		if (ASCII_HYPH == word[i] ||
+		    isgraph((unsigned char)word[i]))
+			encode1(p, word[i]);
+		else {
+			p->buf[p->col++] = word[i];
+
+			/*
+			 * Postpone the effect of \z while handling
+			 * an overstrike sequence from ascii_uc2str().
+			 */
+
+			if (word[i] == '\b' &&
+			    (p->flags & TERMP_BACKBEFORE)) {
+				p->flags &= ~TERMP_BACKBEFORE;
+				p->flags |= TERMP_BACKAFTER;
+			}
 		}
-	
-	} else if ('f' == word[*i]) {
-		(*i)++;
-		if (*i >= len)
-			return;
-		switch (word[*i]) {
-		case ('B'):
-			p->flags |= TERMP_BOLD;
+	}
+}
+
+void
+term_setwidth(struct termp *p, const char *wstr)
+{
+	struct roffsu	 su;
+	int		 iop, width;
+
+	iop = 0;
+	width = 0;
+	if (NULL != wstr) {
+		switch (*wstr) {
+		case '+':
+			iop = 1;
+			wstr++;
 			break;
-		case ('I'):
-			p->flags |= TERMP_UNDER;
+		case '-':
+			iop = -1;
+			wstr++;
 			break;
-		case ('P'):
-			/* FALLTHROUGH */
-		case ('R'):
-			p->flags &= ~TERMP_STYLE;
-			break;
 		default:
 			break;
 		}
-		return;
-
-	} else if ('[' != word[*i]) {
-		term_nescape(p, &word[*i], 1);
-		return;
+		if (a2roffsu(wstr, &su, SCALE_MAX))
+			width = term_hspan(p, &su);
+		else
+			iop = 0;
 	}
+	(*p->setwidth)(p, iop, width);
+}
 
-	(*i)++;
-	for (j = 0; word[*i] && ']' != word[*i]; (*i)++, j++)
-		/* Loop... */ ;
+size_t
+term_len(const struct termp *p, size_t sz)
+{
 
-	if (0 == word[*i])
-		return;
-
-	term_nescape(p, &word[*i - j], (size_t)j);
+	return (*p->width)(p, ' ') * sz;
 }
 
+static size_t
+cond_width(const struct termp *p, int c, int *skip)
+{
 
-/*
- * Handle pwords, partial words, which may be either a single word or a
- * phrase that cannot be broken down (such as a literal string).  This
- * handles word styling.
- */
-static void
-term_pword(struct termp *p, const char *word, int len)
+	if (*skip) {
+		(*skip) = 0;
+		return 0;
+	} else
+		return (*p->width)(p, c);
+}
+
+size_t
+term_strlen(const struct termp *p, const char *cp)
 {
-	int		 i;
+	size_t		 sz, rsz, i;
+	int		 ssz, skip, uc;
+	const char	*seq, *rhs;
+	enum mandoc_esc	 esc;
+	static const char rej[] = { '\\', ASCII_NBRSP, ASCII_HYPH,
+			ASCII_BREAK, '\0' };
 
-	if (term_isclosedelim(word, len))
-		if ( ! (TERMP_IGNDELIM & p->flags))
-			p->flags |= TERMP_NOSPACE;
+	/*
+	 * Account for escaped sequences within string length
+	 * calculations.  This follows the logic in term_word() as we
+	 * must calculate the width of produced strings.
+	 */
 
-	if ( ! (TERMP_NOSPACE & p->flags))
-		term_chara(p, ' ');
+	sz = 0;
+	skip = 0;
+	while ('\0' != *cp) {
+		rsz = strcspn(cp, rej);
+		for (i = 0; i < rsz; i++)
+			sz += cond_width(p, *cp++, &skip);
 
-	if ( ! (p->flags & TERMP_NONOSPACE))
-		p->flags &= ~TERMP_NOSPACE;
+		switch (*cp) {
+		case '\\':
+			cp++;
+			esc = mandoc_escape(&cp, &seq, &ssz);
+			if (ESCAPE_ERROR == esc)
+				continue;
 
-	/* 
-	 * If ANSI (word-length styling), then apply our style now,
-	 * before the word.
-	 */
+			rhs = NULL;
 
-	for (i = 0; i < len; i++) {
-		if ('\\' == word[i]) {
-			term_pescape(p, word, &i, len);
-			continue;
-		}
+			switch (esc) {
+			case ESCAPE_UNICODE:
+				uc = mchars_num2uc(seq + 1, ssz - 1);
+				break;
+			case ESCAPE_NUMBERED:
+				uc = mchars_num2char(seq, ssz);
+				if (uc < 0)
+					continue;
+				break;
+			case ESCAPE_SPECIAL:
+				if (p->enc == TERMENC_ASCII) {
+					rhs = mchars_spec2str(seq, ssz, &rsz);
+					if (rhs != NULL)
+						break;
+				} else {
+					uc = mchars_spec2cp(seq, ssz);
+					if (uc > 0)
+						sz += cond_width(p, uc, &skip);
+				}
+				continue;
+			case ESCAPE_SKIPCHAR:
+				skip = 1;
+				continue;
+			case ESCAPE_OVERSTRIKE:
+				rsz = 0;
+				rhs = seq + ssz;
+				while (seq < rhs) {
+					if (*seq == '\\') {
+						mandoc_escape(&seq, NULL, NULL);
+						continue;
+					}
+					i = (*p->width)(p, *seq++);
+					if (rsz < i)
+						rsz = i;
+				}
+				sz += rsz;
+				continue;
+			default:
+				continue;
+			}
 
-		if (TERMP_STYLE & p->flags) {
-			if (TERMP_BOLD & p->flags) {
-				term_chara(p, word[i]);
-				term_chara(p, 8);
+			/*
+			 * Common handling for Unicode and numbered
+			 * character escape sequences.
+			 */
+
+			if (rhs == NULL) {
+				if (p->enc == TERMENC_ASCII) {
+					rhs = ascii_uc2str(uc);
+					rsz = strlen(rhs);
+				} else {
+					if ((uc < 0x20 && uc != 0x09) ||
+					    (uc > 0x7E && uc < 0xA0))
+						uc = 0xFFFD;
+					sz += cond_width(p, uc, &skip);
+					continue;
+				}
 			}
-			if (TERMP_UNDER & p->flags) {
-				term_chara(p, '_');
-				term_chara(p, 8);
+
+			if (skip) {
+				skip = 0;
+				break;
 			}
-		}
 
-		term_chara(p, word[i]);
+			/*
+			 * Common handling for all escape sequences
+			 * printing more than one character.
+			 */
+
+			for (i = 0; i < rsz; i++)
+				sz += (*p->width)(p, *rhs++);
+			break;
+		case ASCII_NBRSP:
+			sz += cond_width(p, ' ', &skip);
+			cp++;
+			break;
+		case ASCII_HYPH:
+			sz += cond_width(p, '-', &skip);
+			cp++;
+			break;
+		default:
+			break;
+		}
 	}
 
-	if (term_isopendelim(word, len))
-		p->flags |= TERMP_NOSPACE;
+	return sz;
 }
 
-
-/*
- * Like term_chara() but for arbitrary-length buffers.  Resize the
- * buffer by a factor of two (if the buffer is less than that) or the
- * buffer's size.
- */
-static void
-term_stringa(struct termp *p, const char *c, size_t sz)
+int
+term_vspan(const struct termp *p, const struct roffsu *su)
 {
-	size_t		 s;
+	double		 r;
+	int		 ri;
 
-	if (0 == sz)
-		return;
-
-	assert(c);
-	if (p->col + sz >= p->maxcols) {
-		if (0 == p->maxcols)
-			p->maxcols = 256;
-		s = sz > p->maxcols * 2 ? sz : p->maxcols * 2;
-		p->buf = realloc(p->buf, s);
-		if (NULL == p->buf)
-			err(1, "realloc");
-		p->maxcols = s;
+	switch (su->unit) {
+	case SCALE_BU:
+		r = su->scale / 40.0;
+		break;
+	case SCALE_CM:
+		r = su->scale * 6.0 / 2.54;
+		break;
+	case SCALE_FS:
+		r = su->scale * 65536.0 / 40.0;
+		break;
+	case SCALE_IN:
+		r = su->scale * 6.0;
+		break;
+	case SCALE_MM:
+		r = su->scale * 0.006;
+		break;
+	case SCALE_PC:
+		r = su->scale;
+		break;
+	case SCALE_PT:
+		r = su->scale / 12.0;
+		break;
+	case SCALE_EN:
+	case SCALE_EM:
+		r = su->scale * 0.6;
+		break;
+	case SCALE_VS:
+		r = su->scale;
+		break;
+	default:
+		abort();
 	}
-
-	(void)memcpy(&p->buf[(int)p->col], c, sz);
-	p->col += sz;
+	ri = r > 0.0 ? r + 0.4995 : r - 0.4995;
+	return ri < 66 ? ri : 1;
 }
 
-
 /*
- * Insert a single character into the line-buffer.  If the buffer's
- * space is exceeded, then allocate more space by doubling the buffer
- * size.
+ * Convert a scaling width to basic units, rounding down.
  */
-static void
-term_chara(struct termp *p, char c)
+int
+term_hspan(const struct termp *p, const struct roffsu *su)
 {
-	size_t		 s;
 
-	if (p->col + 1 >= p->maxcols) {
-		if (0 == p->maxcols)
-			p->maxcols = 256;
-		s = p->maxcols * 2;
-		p->buf = realloc(p->buf, s);
-		if (NULL == p->buf)
-			err(1, "realloc");
-		p->maxcols = s;
-	}
-	p->buf[(int)(p->col)++] = c;
+	return (*p->hspan)(p, su);
 }
-