]> git.cameronkatri.com Git - mandoc.git/blobdiff - term.c
In -Tutf8 mode, make sure that hyphens get counted against the output line
[mandoc.git] / term.c
diff --git a/term.c b/term.c
index f44dacc74b7a591ba863f6315d6284f02e98a8c8..d3b217745149e2f9741ddfe43a17c530440569a2 100644 (file)
--- a/term.c
+++ b/term.c
@@ -1,6 +1,7 @@
-/*     $Id: term.c,v 1.130 2010/04/03 12:46:35 kristaps Exp $ */
+/*     $Id: term.c,v 1.217 2014/03/13 19:23:50 schwarze Exp $ */
 /*
- * Copyright (c) 2008, 2009 Kristaps Dzonsons <kristaps@kth.se>
+ * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
+ * Copyright (c) 2010-2014 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
 
 #include <assert.h>
 #include <ctype.h>
+#include <stdint.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
-#include <time.h>
 
-#include "chars.h"
+#include "mandoc.h"
 #include "out.h"
 #include "term.h"
-#include "man.h"
-#include "mdoc.h"
 #include "main.h"
 
-static struct termp     *term_alloc(enum termenc);
-static void              term_free(struct termp *);
-static void              spec(struct termp *, const char *, size_t);
-static void              res(struct termp *, const char *, size_t);
-static void              buffera(struct termp *, const char *, size_t);
-static void              bufferc(struct termp *, char);
-static void              adjbuf(struct termp *p, size_t);
-static void              encode(struct termp *, const char *, size_t);
-
-
-void *
-ascii_alloc(void)
-{
-
-       return(term_alloc(TERMENC_ASCII));
-}
-
+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);
 
 void
-terminal_free(void *arg)
-{
-
-       term_free((struct termp *)arg);
-}
-
-
-static void
 term_free(struct termp *p)
 {
 
        if (p->buf)
                free(p->buf);
        if (p->symtab)
-               chars_free(p->symtab);
+               mchars_free(p->symtab);
 
        free(p);
 }
 
 
-static struct termp *
-term_alloc(enum termenc enc)
+void
+term_begin(struct termp *p, term_margin head, 
+               term_margin foot, const void *arg)
 {
-       struct termp *p;
 
-       p = calloc(1, sizeof(struct termp));
-       if (NULL == p) {
-               perror(NULL);
-               exit(EXIT_FAILURE);
-       }
-       p->enc = enc;
-       return(p);
+       p->headf = head;
+       p->footf = foot;
+       p->argf = arg;
+       (*p->begin)(p);
 }
 
 
+void
+term_end(struct termp *p)
+{
+
+       (*p->end)(p);
+}
+
 /*
  * 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
@@ -97,22 +80,17 @@ term_alloc(enum termenc enc)
  *
  * The following flags may be specified:
  *
- *  - 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_TWOSPACE: when padding, make sure there are at least two
- *    space characters of padding.  Otherwise, rather break the line.
+ *    columns.  In short: don't print a newline and instead expect the
+ *    next call to do the padding up to the start of the next column.
+ *    p->trailspace may be set to 0, 1, or 2, depending on how many
+ *    space characters are required at the end of the column.
  *
  *  - TERMP_DANGLE: don't newline when TERMP_NOBREAK is specified and
  *    the line is overrun, and don't pad-right if it's underrun.
  *
  *  - TERMP_HANG: like TERMP_DANGLE, but doesn't newline when
- *    overruning, instead save the position and continue at that point
+ *    overrunning, instead save the position and continue at that point
  *    when the next invocation.
  *
  *  In-line line breaking:
@@ -126,194 +104,204 @@ term_alloc(enum termenc enc)
 void
 term_flushln(struct termp *p)
 {
-       int              i;     /* current input position in p->buf */
+       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           vsz;   /* visual characters to write to output */
+       size_t           vend;  /* end of word visual position on output */
        size_t           bp;    /* visual right border position */
-       size_t           hyph;  /* visible position of hyphen */
-       int              j;     /* temporary loop index */
-       size_t           maxvis, mmax;
+       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 */
+       size_t           mmax; /* used in calculating bp */
 
        /*
         * 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. 
+        * 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.
         */
-
-       assert(p->offset < p->rmargin);
-
-       maxvis = (int)(p->rmargin - p->offset) - p->overstep < 0 ?
-               /* LINTED */ 
-               0 : p->rmargin - p->offset - p->overstep;
-       mmax = (int)(p->maxrmargin - p->offset) - p->overstep < 0 ?
-               /* LINTED */
-               0 : p->maxrmargin - p->offset - p->overstep;
+       assert  (p->rmargin >= p->offset);
+       dv     = p->rmargin - p->offset;
+       maxvis = (int)dv > p->overstep ? dv - (size_t)p->overstep : 0;
+       dv     = p->maxrmargin - p->offset;
+       mmax   = (int)dv > p->overstep ? dv - (size_t)p->overstep : 0;
 
        bp = TERMP_NOBREAK & p->flags ? mmax : maxvis;
 
-       /* 
-        * FIXME: if bp is zero, we still output the first word before
-        * breaking the line.
-        */
-
-       vis = 0;
-
        /*
-        * 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;
+
+       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 && '\t' == p->buf[i]) {
+                       vend = (vis / p->tabwidth + 1) * p->tabwidth;
+                       vbl += vend - vis;
+                       vis = vend;
+                       ntab++;
+                       i++;
+               }
 
-       for (i = 0; i < (int)p->col; 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).
-                * Collect the number of printable characters until the
-                * first hyphen, if found.  Hyphens aren't included if
-                * they're the first character (so `Fl' doesn't break)
-                * or second consecutive character (`Fl -').
                 */
 
-               /* LINTED */
-               for (j = i, vsz = 0, hyph = 0; j < (int)p->col; j++) {
-                       if (j && ' ' == p->buf[j]) 
+               for (j = i, jhy = 0; j < p->col; j++) {
+                       if (' ' == p->buf[j] || '\t' == p->buf[j])
                                break;
-                       if (8 == p->buf[j])
-                               vsz--;
-                       else
-                               vsz++;
-                       if (j > i && '-' == p->buf[j] && 0 == hyph)
-                              if ('-' != p->buf[j - 1])
-                                       hyph = vsz;
-               }
 
-               /*
-                * Choose the number of blanks to prepend: no blank at the
-                * beginning of a line, one between words -- but do not
-                * actually write them yet.
-                */
+                       /* Back over the 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] = '-';
 
-               vbl = (size_t)(0 == vis ? 0 : 1);
+                       vend += (*p->width)(p, p->buf[j]);
+               }
 
                /*
                 * Find out whether we would exceed the right margin.
-                * If so, break to the next line, possibly after
-                * emittign character up to a hyphen.  Otherwise, write
-                * the chosen number of blanks.
+                * 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) {
+                               vbl = p->rmargin;
+                               vend += p->rmargin - p->offset;
+                       } else
+                               vbl = p->offset;
 
-               if (vis && vis + vbl + vsz > bp) {
-                       /*
-                        * Has a hyphen been found before the breakpoint
-                        * that we can use?
-                        */
-                       if (hyph && vis + vbl + hyph <= bp) {
-                               /* First prepend blanks. */
-                               for (j = 0; j < (int)vbl; j++)
-                                       putchar(' ');
-                               
-                               /* Emit up to the character. */
-                               do {
-                                       if (31 == p->buf[i])
-                                               putchar(' ');
-                                       else
-                                               putchar(p->buf[i]);
-                                       if (8 != p->buf[i])
-                                               vsz--;
-                               } while ('-' != p->buf[i++]);
-
-                               /* Emit trailing decoration. */
-                               if (8 == p->buf[i]) {
-                                       putchar(p->buf[i]);
-                                       putchar(p->buf[i + 1]);
-                               }
-                       } 
+                       /* use pending tabs on the new line */
 
-                       putchar('\n');
-                       if (TERMP_NOBREAK & p->flags) {
-                               for (j = 0; j < (int)p->rmargin; j++)
-                                       putchar(' ');
-                               vis = p->rmargin - p->offset;
-                       } else {
-                               for (j = 0; j < (int)p->offset; j++)
-                                       putchar(' ');
-                               vis = 0;
-                       }
+                       if (0 < ntab)
+                               vbl += ntab * p->tabwidth;
 
-                       /* Remove the p->overstep width. */
+                       /*
+                        * Remove the p->overstep width.
+                        * Again, if p->overstep is negative,
+                        * sign extension does the right thing.
+                        */
 
-                       bp += (int)/* LINTED */
-                               p->overstep;
+                       bp += (size_t)p->overstep;
                        p->overstep = 0;
-               } else {
-                       for (j = 0; j < (int)vbl; j++)
-                               putchar(' ');
-                       vis += vbl;
                }
 
                /* Write out the [remaining] word. */
-               for ( ; i < (int)p->col; i++)
-                       if (' ' == p->buf[i])
+               for ( ; i < p->col; i++) {
+                       if (vend > bp && jhy > 0 && i > jhy)
+                               break;
+                       if ('\t' == p->buf[i])
+                               break;
+                       if (' ' == p->buf[i]) {
+                               j = i;
+                               while (' ' == p->buf[i])
+                                       i++;
+                               dv = (i - j) * (*p->width)(p, ' ');
+                               vbl += dv;
+                               vend += dv;
                                break;
-                       else if (31 == p->buf[i])
-                               putchar(' ');
-                       else
-                               putchar(p->buf[i]);
+                       }
+                       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;
+                       }
 
-               vis += vsz;
+                       (*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 = vend;
        }
 
+       /*
+        * If there was trailing white space, it was not printed;
+        * so reset the cursor position accordingly.
+        */
+       if (vis)
+               vis -= vbl;
+
        p->col = 0;
        p->overstep = 0;
 
        if ( ! (TERMP_NOBREAK & p->flags)) {
-               putchar('\n');
+               p->viscol = 0;
+               (*p->endline)(p);
                return;
        }
 
        if (TERMP_HANG & p->flags) {
-               /* We need one blank after the tag. */
-               p->overstep = /* LINTED */
-                       vis - maxvis + 1;
+               p->overstep = (int)(vis - maxvis +
+                               p->trailspace * (*p->width)(p, ' '));
 
                /*
-                * Behave exactly the same way as groff:
                 * If we have overstepped the margin, temporarily move
                 * it to the right and flag the rest of the line to be
                 * shorter.
-                * If we landed right at the margin, be happy.
-                * If we are one step before the margin, temporarily
-                * move it one step LEFT and flag the rest of the line
-                * to be longer.
+                * If there is a request to keep the columns together,
+                * allow negative overstep when the column is not full.
                 */
-               if (p->overstep >= -1) {
-                       assert((int)maxvis + p->overstep >= 0);
-                       /* LINTED */
-                       maxvis += p->overstep;
-               } else
+               if (p->trailspace && p->overstep < 0)
                        p->overstep = 0;
+               return;
 
        } else if (TERMP_DANGLE & p->flags)
                return;
 
-       /* Right-pad. */
-       if (maxvis > vis + /* LINTED */
-                       ((TERMP_TWOSPACE & p->flags) ? 1 : 0))  
-               for ( ; vis < maxvis; vis++)
-                       putchar(' ');
-       else {  /* ...or newline break. */
-               putchar('\n');
-               for (i = 0; i < (int)p->rmargin; i++)
-                       putchar(' ');
+       /* If the column was overrun, break the line. */
+       if (maxvis < vis + p->trailspace * (*p->width)(p, ' ')) {
+               (*p->endline)(p);
+               p->viscol = 0;
        }
 }
 
@@ -328,12 +316,8 @@ 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);
 }
 
 
@@ -348,34 +332,13 @@ term_vspace(struct termp *p)
 {
 
        term_newln(p);
-       putchar('\n');
-}
-
-
-static void
-spec(struct termp *p, const char *word, size_t len)
-{
-       const char      *rhs;
-       size_t           sz;
-
-       rhs = chars_a2ascii(p->symtab, word, len, &sz);
-       if (rhs) 
-               encode(p, rhs, sz);
+       p->viscol = 0;
+       if (0 < p->skipvsp)
+               p->skipvsp--;
+       else
+               (*p->endline)(p);
 }
 
-
-static void
-res(struct termp *p, const char *word, size_t len)
-{
-       const char      *rhs;
-       size_t           sz;
-
-       rhs = chars_a2res(p->symtab, word, len, &sz);
-       if (rhs)
-               encode(p, rhs, sz);
-}
-
-
 void
 term_fontlast(struct termp *p)
 {
@@ -426,7 +389,7 @@ void
 term_fontpopq(struct termp *p, const void *key)
 {
 
-       while (p->fonti >= 0 && key != &p->fontq[p->fonti])
+       while (p->fonti >= 0 && key < (void *)(p->fontq + p->fonti))
                p->fonti--;
        assert(p->fonti >= 0);
 }
@@ -440,7 +403,6 @@ term_fontpop(struct termp *p)
        p->fonti--;
 }
 
-
 /*
  * 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
@@ -449,98 +411,124 @@ term_fontpop(struct termp *p)
 void
 term_word(struct termp *p, const char *word)
 {
-       const char      *sv, *seq;
-       int              sz;
+       const char       nbrsp[2] = { ASCII_NBRSP, 0 };
+       const char      *seq, *cp;
+       char             c;
+       int              sz, uc;
        size_t           ssz;
-       enum roffdeco    deco;
-
-       sv = word;
+       enum mandoc_esc  esc;
 
-       if (word[0] && '\0' == word[1])
-               switch (word[0]) {
-               case('.'):
-                       /* FALLTHROUGH */
-               case(','):
-                       /* FALLTHROUGH */
-               case(';'):
-                       /* FALLTHROUGH */
-               case(':'):
-                       /* FALLTHROUGH */
-               case('?'):
-                       /* FALLTHROUGH */
-               case('!'):
-                       /* FALLTHROUGH */
-               case(')'):
-                       /* FALLTHROUGH */
-               case(']'):
-                       if ( ! (TERMP_IGNDELIM & p->flags))
-                               p->flags |= TERMP_NOSPACE;
-                       break;
-               default:
-                       break;
-               }
-
-       if ( ! (TERMP_NOSPACE & p->flags))
-               bufferc(p, ' ');
+       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;
 
        if ( ! (p->flags & TERMP_NONOSPACE))
                p->flags &= ~TERMP_NOSPACE;
+       else
+               p->flags |= TERMP_NOSPACE;
 
-       /* FIXME: use strcspn. */
+       p->flags &= ~TERMP_SENTENCE;
 
-       while (*word) {
+       while ('\0' != *word) {
                if ('\\' != *word) {
-                       encode(p, word, 1);
-                       word++;
+                       if (TERMP_SKIPCHAR & p->flags) {
+                               p->flags &= ~TERMP_SKIPCHAR;
+                               word++;
+                               continue;
+                       }
+                       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;
                }
 
-               seq = ++word;
-               sz = a2roffdeco(&deco, &seq, &ssz);
+               word++;
+               esc = mandoc_escape(&word, &seq, &sz);
+               if (ESCAPE_ERROR == esc)
+                       break;
 
-               switch (deco) {
-               case (DECO_RESERVED):
-                       res(p, seq, ssz);
+               if (TERMENC_ASCII != p->enc)
+                       switch (esc) {
+                       case (ESCAPE_UNICODE):
+                               uc = mchars_num2uc(seq + 1, sz - 1);
+                               if ('\0' == uc)
+                                       break;
+                               encode1(p, uc);
+                               continue;
+                       case (ESCAPE_SPECIAL):
+                               uc = mchars_spec2cp(p->symtab, seq, sz);
+                               if (uc <= 0)
+                                       break;
+                               encode1(p, uc);
+                               continue;
+                       default:
+                               break;
+                       }
+
+               switch (esc) {
+               case (ESCAPE_UNICODE):
+                       encode1(p, '?');
+                       break;
+               case (ESCAPE_NUMBERED):
+                       c = mchars_num2char(seq, sz);
+                       if ('\0' != c)
+                               encode(p, &c, 1);
                        break;
-               case (DECO_SPECIAL):
-                       spec(p, seq, ssz);
+               case (ESCAPE_SPECIAL):
+                       cp = mchars_spec2str(p->symtab, seq, sz, &ssz);
+                       if (NULL != cp) 
+                               encode(p, cp, ssz);
+                       else if (1 == ssz)
+                               encode(p, seq, sz);
                        break;
-               case (DECO_BOLD):
+               case (ESCAPE_FONTBOLD):
                        term_fontrepl(p, TERMFONT_BOLD);
                        break;
-               case (DECO_ITALIC):
+               case (ESCAPE_FONTITALIC):
                        term_fontrepl(p, TERMFONT_UNDER);
                        break;
-               case (DECO_ROMAN):
+               case (ESCAPE_FONTBI):
+                       term_fontrepl(p, TERMFONT_BI);
+                       break;
+               case (ESCAPE_FONT):
+                       /* FALLTHROUGH */
+               case (ESCAPE_FONTROMAN):
                        term_fontrepl(p, TERMFONT_NONE);
                        break;
-               case (DECO_PREVIOUS):
+               case (ESCAPE_FONTPREV):
                        term_fontlast(p);
                        break;
-               default:
+               case (ESCAPE_NOSPACE):
+                       if (TERMP_SKIPCHAR & p->flags)
+                               p->flags &= ~TERMP_SKIPCHAR;
+                       else if ('\0' == *word)
+                               p->flags |= TERMP_NOSPACE;
                        break;
-               }
-
-               word += sz;
-               if (DECO_NOSPACE == deco && '\0' == *word)
-                       p->flags |= TERMP_NOSPACE;
-       }
-
-       if (sv[0] && 0 == sv[1])
-               switch (sv[0]) {
-               case('|'):
-                       /* FALLTHROUGH */
-               case('('):
-                       /* FALLTHROUGH */
-               case('['):
-                       p->flags |= TERMP_NOSPACE;
+               case (ESCAPE_SKIPCHAR):
+                       p->flags |= TERMP_SKIPCHAR;
                        break;
                default:
                        break;
                }
+       }
+       p->flags &= ~TERMP_NBRWORD;
 }
 
-
 static void
 adjbuf(struct termp *p, size_t sz)
 {
@@ -550,42 +538,62 @@ adjbuf(struct termp *p, size_t sz)
        while (sz >= p->maxcols)
                p->maxcols <<= 2;
 
-       p->buf = realloc(p->buf, p->maxcols);
-       if (NULL == p->buf) {
-               perror(NULL);
-               exit(EXIT_FAILURE);
-       }
+       p->buf = mandoc_realloc(p->buf, sizeof(int) * p->maxcols);
 }
 
-
 static void
-buffera(struct termp *p, const char *word, size_t sz)
+bufferc(struct termp *p, char c)
 {
 
-       if (p->col + sz >= p->maxcols) 
-               adjbuf(p, p->col + sz);
+       if (p->col + 1 >= p->maxcols)
+               adjbuf(p, p->col + 1);
 
-       memcpy(&p->buf[(int)p->col], word, sz);
-       p->col += sz;
+       p->buf[p->col++] = c;
 }
 
-
+/*
+ * See encode().
+ * Do this for a single (probably unicode) value.
+ * Does not check for non-decorated glyphs.
+ */
 static void
-bufferc(struct termp *p, char c)
+encode1(struct termp *p, int c)
 {
+       enum termfont     f;
 
-       if (p->col + 1 >= p->maxcols)
-               adjbuf(p, p->col + 1);
+       if (TERMP_SKIPCHAR & p->flags) {
+               p->flags &= ~TERMP_SKIPCHAR;
+               return;
+       }
 
-       p->buf[(int)p->col++] = c;
-}
+       if (p->col + 6 >= p->maxcols)
+               adjbuf(p, p->col + 6);
 
+       f = term_fonttop(p);
+
+       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;
+}
 
 static void
 encode(struct termp *p, const char *word, size_t sz)
 {
-       enum termfont     f;
-       int               i;
+       size_t            i;
+
+       if (TERMP_SKIPCHAR & p->flags) {
+               p->flags &= ~TERMP_SKIPCHAR;
+               return;
+       }
 
        /*
         * Encode and buffer a string of characters.  If the current
@@ -593,30 +601,156 @@ encode(struct termp *p, const char *word, size_t sz)
         * character by character.
         */
 
-       if (TERMFONT_NONE == (f = term_fonttop(p))) {
-               buffera(p, word, sz);
+       if (TERMFONT_NONE == term_fonttop(p)) {
+               if (p->col + sz >= p->maxcols) 
+                       adjbuf(p, p->col + sz);
+               for (i = 0; i < sz; i++)
+                       p->buf[p->col++] = word[i];
                return;
        }
 
-       for (i = 0; i < (int)sz; i++) {
-               if ( ! isgraph((u_char)word[i])) {
-                       bufferc(p, word[i]);
-                       continue;
-               }
+       /* Pre-buffer, assuming worst-case. */
 
-               if (TERMFONT_UNDER == f)
-                       bufferc(p, '_');
-               else
-                       bufferc(p, word[i]);
+       if (p->col + 1 + (sz * 5) >= p->maxcols)
+               adjbuf(p, p->col + 1 + (sz * 5));
 
-               bufferc(p, 8);
-               bufferc(p, word[i]);
+       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];
        }
 }
 
+size_t
+term_len(const struct termp *p, size_t sz)
+{
+
+       return((*p->width)(p, ' ') * sz);
+}
+
+static size_t
+cond_width(const struct termp *p, int c, int *skip)
+{
+
+       if (*skip) {
+               (*skip) = 0;
+               return(0);
+       } else
+               return((*p->width)(p, c));
+}
+
+size_t
+term_strlen(const struct termp *p, const char *cp)
+{
+       size_t           sz, rsz, i;
+       int              ssz, skip, c;
+       const char      *seq, *rhs;
+       enum mandoc_esc  esc;
+       static const char rej[] = { '\\', ASCII_NBRSP, ASCII_HYPH,
+                       ASCII_BREAK, '\0' };
+
+       /*
+        * Account for escaped sequences within string length
+        * calculations.  This follows the logic in term_word() as we
+        * must calculate the width of produced strings.
+        */
+
+       sz = 0;
+       skip = 0;
+       while ('\0' != *cp) {
+               rsz = strcspn(cp, rej);
+               for (i = 0; i < rsz; i++)
+                       sz += cond_width(p, *cp++, &skip);
+
+               switch (*cp) {
+               case ('\\'):
+                       cp++;
+                       esc = mandoc_escape(&cp, &seq, &ssz);
+                       if (ESCAPE_ERROR == esc)
+                               return(sz);
+
+                       if (TERMENC_ASCII != p->enc)
+                               switch (esc) {
+                               case (ESCAPE_UNICODE):
+                                       c = mchars_num2uc
+                                               (seq + 1, ssz - 1);
+                                       if ('\0' == c)
+                                               break;
+                                       sz += cond_width(p, c, &skip);
+                                       continue;
+                               case (ESCAPE_SPECIAL):
+                                       c = mchars_spec2cp
+                                               (p->symtab, seq, ssz);
+                                       if (c <= 0)
+                                               break;
+                                       sz += cond_width(p, c, &skip);
+                                       continue;
+                               default:
+                                       break;
+                               }
+
+                       rhs = NULL;
+
+                       switch (esc) {
+                       case (ESCAPE_UNICODE):
+                               sz += cond_width(p, '?', &skip);
+                               break;
+                       case (ESCAPE_NUMBERED):
+                               c = mchars_num2char(seq, ssz);
+                               if ('\0' != c)
+                                       sz += cond_width(p, c, &skip);
+                               break;
+                       case (ESCAPE_SPECIAL):
+                               rhs = mchars_spec2str
+                                       (p->symtab, seq, ssz, &rsz);
+
+                               if (ssz != 1 || rhs)
+                                       break;
+
+                               rhs = seq;
+                               rsz = ssz;
+                               break;
+                       case (ESCAPE_SKIPCHAR):
+                               skip = 1;
+                               break;
+                       default:
+                               break;
+                       }
+
+                       if (NULL == rhs)
+                               break;
+
+                       if (skip) {
+                               skip = 0;
+                               break;
+                       }
+
+                       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++;
+                       /* FALLTHROUGH */
+               case (ASCII_BREAK):
+                       break;
+               default:
+                       break;
+               }
+       }
+
+       return(sz);
+}
 
+/* ARGSUSED */
 size_t
-term_vspan(const struct roffsu *su)
+term_vspan(const struct termp *p, const struct roffsu *su)
 {
        double           r;
 
@@ -650,43 +784,14 @@ term_vspan(const struct roffsu *su)
                        r);
 }
 
-
 size_t
-term_hspan(const struct roffsu *su)
+term_hspan(const struct termp *p, const struct roffsu *su)
 {
-       double           r;
-
-       /* XXX: CM, IN, and PT are approximations. */
-
-       switch (su->unit) {
-       case (SCALE_CM):
-               r = 4 * su->scale;
-               break;
-       case (SCALE_IN):
-               /* XXX: this is an approximation. */
-               r = 10 * su->scale;
-               break;
-       case (SCALE_PC):
-               r = (10 * su->scale) / 6;
-               break;
-       case (SCALE_PT):
-               r = (10 * su->scale) / 72;
-               break;
-       case (SCALE_MM):
-               r = su->scale / 1000; /* FIXME: double-check. */
-               break;
-       case (SCALE_VS):
-               r = su->scale * 2 - 1; /* FIXME: double-check. */
-               break;
-       default:
-               r = su->scale;
-               break;
-       }
+       double           v;
 
-       if (r < 0.0)
-               r = 0.0;
-       return((size_t)/* LINTED */
-                       r);
+       v = ((*p->hspan)(p, su));
+       if (v < 0.0)
+               v = 0.0;
+       return((size_t) /* LINTED */
+                       v);
 }
-
-