]> git.cameronkatri.com Git - mandoc.git/blobdiff - term.c
Complete eqn.7 parsing. Features all productions from the original 1975
[mandoc.git] / term.c
diff --git a/term.c b/term.c
index f5f18c45d728611c5d29f5ed2c7b47758811165d..70260bec31a613857102aa7dbd1603ac2a118d3c 100644 (file)
--- a/term.c
+++ b/term.c
@@ -1,6 +1,7 @@
-/*     $Id: term.c,v 1.102 2009/09/20 13:43:31 kristaps Exp $ */
+/*     $Id: term.c,v 1.197 2011/05/24 21:31:23 kristaps Exp $ */
 /*
- * Copyright (c) 2008, 2009 Kristaps Dzonsons <kristaps@kth.se>
+ * Copyright (c) 2008, 2009, 2010 Kristaps Dzonsons <kristaps@bsd.lv>
+ * Copyright (c) 2010, 2011 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
  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  */
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#include <sys/types.h>
+
 #include <assert.h>
-#include <err.h>
+#include <ctype.h>
+#include <stdint.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
 
-#include "chars.h"
+#include "mandoc.h"
+#include "out.h"
 #include "term.h"
-#include "man.h"
-#include "mdoc.h"
-
-extern void              man_run(struct termp *, 
-                               const struct man *);
-extern void              mdoc_run(struct termp *, 
-                               const struct mdoc *);
-
-static struct termp     *term_alloc(enum termenc);
-static void              term_free(struct termp *);
-
-static void              do_escaped(struct termp *, const char **);
-static void              do_special(struct termp *,
-                               const char *, size_t);
-static void              do_reserved(struct termp *,
-                               const char *, size_t);
-static void              buffer(struct termp *, char);
-static void              encode(struct termp *, char);
-
-
-void *
-ascii_alloc(void)
-{
-
-       return(term_alloc(TERMENC_ASCII));
-}
+#include "main.h"
 
+static void             adjbuf(struct termp *p, int);
+static void             bufferc(struct termp *, char);
+static void             encode(struct termp *, const char *, size_t);
+static void             encode1(struct termp *, int);
 
 void
-terminal_man(void *arg, const struct man *man)
+term_free(struct termp *p)
 {
-       struct termp    *p;
 
-       p = (struct termp *)arg;
-       if (NULL == p->symtab)
-               p->symtab = chars_init(CHARS_ASCII);
+       if (p->buf)
+               free(p->buf);
+       if (p->symtab)
+               mchars_free(p->symtab);
 
-       man_run(p, man);
+       free(p);
 }
 
 
 void
-terminal_mdoc(void *arg, const struct mdoc *mdoc)
+term_begin(struct termp *p, term_margin head, 
+               term_margin foot, const void *arg)
 {
-       struct termp    *p;
 
-       p = (struct termp *)arg;
-       if (NULL == p->symtab)
-               p->symtab = chars_init(CHARS_ASCII);
-
-       mdoc_run(p, mdoc);
+       p->headf = head;
+       p->footf = foot;
+       p->argf = arg;
+       (*p->begin)(p);
 }
 
 
 void
-terminal_free(void *arg)
-{
-
-       term_free((struct termp *)arg);
-}
-
-
-static void
-term_free(struct termp *p)
+term_end(struct termp *p)
 {
 
-       if (p->buf)
-               free(p->buf);
-       if (p->symtab)
-               chars_free(p->symtab);
-
-       free(p);
+       (*p->end)(p);
 }
 
-
-static struct termp *
-term_alloc(enum termenc enc)
-{
-       struct termp *p;
-
-       if (NULL == (p = malloc(sizeof(struct termp))))
-               return(NULL);
-       bzero(p, sizeof(struct termp));
-       p->maxrmargin = 78;
-       p->enc = enc;
-       return(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
  * broken apart by newlines getting there.  A line can also be a
- * fragment of a columnar list.
- *
- * Specifically, a line is whatever's in p->buf of length p->col, which
- * is zeroed after this function returns.
+ * fragment of a columnar list (`Bl -tag' or `Bl -column'), which does
+ * not have a trailing newline.
  *
- * The usage of termp:flags is as follows:
+ * 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
@@ -145,51 +102,57 @@ term_alloc(enum termenc enc)
  *  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).
+ *  writing from the right-margin, which will lead to the above scenario
+ *  upon exit.  Otherwise, the line will break at the right margin.
  */
 void
 term_flushln(struct termp *p)
 {
-       int              i, j;
-       size_t           vbl, vsz, vis, maxvis, mmax, bp;
-       static int       overstep = 0;
+       int              i;     /* current input position in p->buf */
+       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 */
+       int              j;     /* temporary loop index for p->buf */
+       int              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. 
         */
-
-       assert(p->offset < p->rmargin);
-       assert((int)(p->rmargin - p->offset) - overstep > 0);
-
-       maxvis = /* LINTED */
-               p->rmargin - p->offset - overstep;
-       mmax = /* LINTED */
-               p->maxrmargin - p->offset - 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;
-       vis = 0;
-       overstep = 0;
 
        /*
-        * If in the standard case (left-justified), then begin with our
-        * indentation, otherwise (columns, etc.) just start spitting
-        * out text.
+        * Indent the first line of a paragraph.
         */
+       vbl = p->flags & TERMP_NOLPAD ? (size_t)0 : p->offset;
+
+       vis = vend = 0;
+       i = 0;
 
-       if ( ! (p->flags & TERMP_NOLPAD))
-               /* LINTED */
-               for (j = 0; j < (int)p->offset; j++)
-                       putchar(' ');
+       while (i < p->col) {
+               /*
+                * Handle literal tab characters: collapse all
+                * subsequent tabs into a single huge set of spaces.
+                */
+               while (i < p->col && '\t' == p->buf[i]) {
+                       vend = (vis / p->tabwidth + 1) * p->tabwidth;
+                       vbl += vend - vis;
+                       vis = vend;
+                       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
@@ -197,66 +160,108 @@ term_flushln(struct termp *p)
                 * space is printed according to regular spacing rules).
                 */
 
-               /* LINTED */
-               for (j = i, vsz = 0; j < (int)p->col; j++) {
-                       if (j && ' ' == p->buf[j]) 
+               for (j = i, jhy = 0; j < p->col; j++) {
+                       if ((j && ' ' == p->buf[j]) || '\t' == p->buf[j])
                                break;
-                       else if (8 == p->buf[j])
-                               vsz--;
-                       else
-                               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.
-                */
-               vbl = (size_t)(0 == vis ? 0 : 1);
+                       /* 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])
+                               jhy = j;
+
+                       vend += (*p->width)(p, p->buf[j]);
+               }
 
                /*
                 * Find out whether we would exceed the right margin.
-                * If so, break to the next line.  (TODO: hyphenate)
-                * Otherwise, write the chosen number of blanks now.
+                * If so, break to the next line.
                 */
-               if (vis && vis + vbl + vsz > bp) {
-                       putchar('\n');
+               if (vend > bp && 0 == jhy && vis > 0) {
+                       vend -= vis;
+                       (*p->endline)(p);
                        if (TERMP_NOBREAK & p->flags) {
-                               for (j = 0; j < (int)p->rmargin; j++)
-                                       putchar(' ');
-                               vis = p->rmargin - p->offset;
+                               p->viscol = p->rmargin;
+                               (*p->advance)(p, p->rmargin);
+                               vend += p->rmargin - p->offset;
                        } else {
-                               for (j = 0; j < (int)p->offset; j++)
-                                       putchar(' ');
-                               vis = 0;
+                               p->viscol = 0;
+                               vbl = p->offset;
                        }
-               } else {
-                       for (j = 0; j < (int)vbl; j++)
-                               putchar(' ');
-                       vis += vbl;
+
+                       /* Remove the p->overstep width. */
+
+                       bp += (size_t)p->overstep;
+                       p->overstep = 0;
                }
 
-               /*
-                * Finally, write out the word.
-                */
-               for ( ; i < (int)p->col; i++) {
-                       if (' ' == p->buf[i])
+               /* 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 (' ' == p->buf[i])
+                                       i++;
+                               dv = (size_t)(i - j) * (*p->width)(p, ' ');
+                               vbl += dv;
+                               vend += dv;
+                               break;
+                       }
+                       if (ASCII_NBRSP == p->buf[i]) {
+                               vbl += (*p->width)(p, ' ');
+                               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;
+                       }
+
+                       if (ASCII_HYPH == p->buf[i]) {
+                               (*p->letter)(p, '-');
+                               p->viscol += (*p->width)(p, '-');
+                       } else {
+                               (*p->letter)(p, p->buf[i]);
+                               p->viscol += (*p->width)(p, p->buf[i]);
+                       }
                }
-               vis += vsz;
+               vis = vend;
        }
+
+       /*
+        * If there was trailing white space, it was not printed;
+        * so reset the cursor position accordingly.
+        */
+       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. */
-               overstep = /* LINTED */
-                       vis - maxvis + 1;
+               p->overstep = (int)(vis - maxvis + (*p->width)(p, ' '));
 
                /*
                 * Behave exactly the same way as groff:
@@ -268,25 +273,25 @@ term_flushln(struct termp *p)
                 * move it one step LEFT and flag the rest of the line
                 * to be longer.
                 */
-               if (overstep >= -1) {
-                       assert((int)maxvis + overstep >= 0);
-                       /* LINTED */
-                       maxvis += overstep;
+               if (p->overstep >= -1) {
+                       assert((int)maxvis + p->overstep >= 0);
+                       maxvis += (size_t)p->overstep;
                } else
-                       overstep = 0;
+                       p->overstep = 0;
 
        } 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 (maxvis > vis +
+           ((TERMP_TWOSPACE & p->flags) ? (*p->width)(p, ' ') : 0)) {
+               p->viscol += maxvis - vis;
+               (*p->advance)(p, maxvis - vis);
+               vis += (maxvis - vis);
+       } else {        /* ...or newline break. */
+               (*p->endline)(p);
+               p->viscol = p->rmargin;
+               (*p->advance)(p, p->rmargin);
        }
 }
 
@@ -301,7 +306,7 @@ term_newln(struct termp *p)
 {
 
        p->flags |= TERMP_NOSPACE;
-       if (0 == p->col) {
+       if (0 == p->col && 0 == p->viscol) {
                p->flags &= ~TERMP_NOLPAD;
                return;
        }
@@ -321,261 +326,419 @@ term_vspace(struct termp *p)
 {
 
        term_newln(p);
-       putchar('\n');
+       p->viscol = 0;
+       (*p->endline)(p);
 }
 
+void
+term_fontlast(struct termp *p)
+{
+       enum termfont    f;
 
-static void
-do_special(struct termp *p, const char *word, size_t len)
+       f = p->fontl;
+       p->fontl = p->fontq[p->fonti];
+       p->fontq[p->fonti] = f;
+}
+
+
+void
+term_fontrepl(struct termp *p, enum termfont f)
 {
-       const char      *rhs;
-       size_t           sz;
-       int              i;
-
-       rhs = chars_a2ascii(p->symtab, word, len, &sz);
-
-       if (NULL == rhs) {
-#if 0
-               fputs("Unknown special character: ", stderr);
-               for (i = 0; i < (int)len; i++)
-                       fputc(word[i], stderr);
-               fputc('\n', stderr);
-#endif
-               return;
-       }
-       for (i = 0; i < (int)sz; i++) 
-               encode(p, rhs[i]);
+
+       p->fontl = p->fontq[p->fonti];
+       p->fontq[p->fonti] = f;
 }
 
 
-static void
-do_reserved(struct termp *p, const char *word, size_t len)
+void
+term_fontpush(struct termp *p, enum termfont f)
 {
-       const char      *rhs;
-       size_t           sz;
-       int              i;
-
-       rhs = chars_a2res(p->symtab, word, len, &sz);
-
-       if (NULL == rhs) {
-#if 0
-               fputs("Unknown reserved word: ", stderr);
-               for (i = 0; i < (int)len; i++)
-                       fputc(word[i], stderr);
-               fputc('\n', stderr);
-#endif
-               return;
-       }
-       for (i = 0; i < (int)sz; i++) 
-               encode(p, rhs[i]);
+
+       assert(p->fonti + 1 < 10);
+       p->fontl = p->fontq[p->fonti];
+       p->fontq[++p->fonti] = f;
 }
 
 
+const void *
+term_fontq(struct termp *p)
+{
+
+       return(&p->fontq[p->fonti]);
+}
+
+
+enum termfont
+term_fonttop(struct termp *p)
+{
+
+       return(p->fontq[p->fonti]);
+}
+
+
+void
+term_fontpopq(struct termp *p, const void *key)
+{
+
+       while (p->fonti >= 0 && key != &p->fontq[p->fonti])
+               p->fonti--;
+       assert(p->fonti >= 0);
+}
+
+
+void
+term_fontpop(struct termp *p)
+{
+
+       assert(p->fonti);
+       p->fonti--;
+}
+
 /*
- * 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).
+ * 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
-do_escaped(struct termp *p, const char **word)
+void
+term_word(struct termp *p, const char *word)
 {
-       int              j, type;
-       const char      *wp;
+       const char      *seq, *cp;
+       char             c;
+       int              sz, uc;
+       size_t           ssz;
+       enum mandoc_esc  esc;
+
+       if ( ! (TERMP_NOSPACE & p->flags)) {
+               if ( ! (TERMP_KEEP & p->flags)) {
+                       if (TERMP_PREKEEP & p->flags)
+                               p->flags |= TERMP_KEEP;
+                       bufferc(p, ' ');
+                       if (TERMP_SENTENCE & p->flags)
+                               bufferc(p, ' ');
+               } else
+                       bufferc(p, ASCII_NBRSP);
+       }
 
-       wp = *word;
-       type = 1;
+       if ( ! (p->flags & TERMP_NONOSPACE))
+               p->flags &= ~TERMP_NOSPACE;
+       else
+               p->flags |= TERMP_NOSPACE;
 
-       if (0 == *(++wp)) {
-               *word = wp;
-               return;
-       }
+       p->flags &= ~(TERMP_SENTENCE | TERMP_IGNDELIM);
 
-       if ('(' == *wp) {
-               wp++;
-               if (0 == *wp || 0 == *(wp + 1)) {
-                       *word = 0 == *wp ? wp : wp + 1;
-                       return;
-               }
+       while ('\0' != *word) {
+               if ((ssz = strcspn(word, "\\")) > 0)
+                       encode(p, word, ssz);
 
-               do_special(p, wp, 2);
-               *word = ++wp;
-               return;
+               word += (int)ssz;
+               if ('\\' != *word)
+                       continue;
 
-       } else if ('*' == *wp) {
-               if (0 == *(++wp)) {
-                       *word = wp;
-                       return;
-               }
+               word++;
+               esc = mandoc_escape(&word, &seq, &sz);
+               if (ESCAPE_ERROR == esc)
+                       break;
 
-               switch (*wp) {
-               case ('('):
-                       wp++;
-                       if (0 == *wp || 0 == *(wp + 1)) {
-                               *word = 0 == *wp ? wp : wp + 1;
-                               return;
+               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;
                        }
 
-                       do_reserved(p, wp, 2);
-                       *word = ++wp;
-                       return;
-               case ('['):
-                       type = 0;
+               switch (esc) {
+               case (ESCAPE_UNICODE):
+                       encode1(p, '?');
                        break;
-               default:
-                       do_reserved(p, wp, 1);
-                       *word = wp;
-                       return;
-               }
-       
-       } else if ('f' == *wp) {
-               if (0 == *(++wp)) {
-                       *word = wp;
-                       return;
-               }
-
-               switch (*wp) {
-               case ('B'):
-                       p->bold++;
+               case (ESCAPE_NUMBERED):
+                       c = mchars_num2char(seq, sz);
+                       if ('\0' != c)
+                               encode(p, &c, 1);
+                       break;
+               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 (ESCAPE_FONTBOLD):
+                       term_fontrepl(p, TERMFONT_BOLD);
                        break;
-               case ('I'):
-                       p->under++;
+               case (ESCAPE_FONTITALIC):
+                       term_fontrepl(p, TERMFONT_UNDER);
                        break;
-               case ('P'):
+               case (ESCAPE_FONT):
                        /* FALLTHROUGH */
-               case ('R'):
-                       p->bold = p->under = 0;
+               case (ESCAPE_FONTROMAN):
+                       term_fontrepl(p, TERMFONT_NONE);
+                       break;
+               case (ESCAPE_FONTPREV):
+                       term_fontlast(p);
+                       break;
+               case (ESCAPE_NOSPACE):
+                       if ('\0' == *word)
+                               p->flags |= TERMP_NOSPACE;
                        break;
                default:
                        break;
                }
+       }
+}
 
-               *word = wp;
+static void
+adjbuf(struct termp *p, int sz)
+{
+
+       if (0 == p->maxcols)
+               p->maxcols = 1024;
+       while (sz >= p->maxcols)
+               p->maxcols <<= 2;
+
+       p->buf = mandoc_realloc
+               (p->buf, sizeof(int) * (size_t)p->maxcols);
+}
+
+static void
+bufferc(struct termp *p, char c)
+{
+
+       if (p->col + 1 >= p->maxcols)
+               adjbuf(p, p->col + 1);
+
+       p->buf[p->col++] = c;
+}
+
+/*
+ * See encode().
+ * Do this for a single (probably unicode) value.
+ * Does not check for non-decorated glyphs.
+ */
+static void
+encode1(struct termp *p, int c)
+{
+       enum termfont     f;
+
+       if (p->col + 4 >= p->maxcols)
+               adjbuf(p, p->col + 4);
+
+       f = term_fonttop(p);
+
+       if (TERMFONT_NONE == f) {
+               p->buf[p->col++] = c;
                return;
+       } else if (TERMFONT_UNDER == f) {
+               p->buf[p->col++] = '_';
+       } else
+               p->buf[p->col++] = c;
 
-       } else if ('[' != *wp) {
-               do_special(p, wp, 1);
-               *word = wp;
+       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, len;
+
+       /* LINTED */
+       len = sz;
+
+       /*
+        * Encode and buffer a string of characters.  If the current
+        * font mode is unset, buffer directly, else encode then buffer
+        * character by character.
+        */
+
+       if (TERMFONT_NONE == (f = term_fonttop(p))) {
+               if (p->col + len >= p->maxcols) 
+                       adjbuf(p, p->col + len);
+               for (i = 0; i < len; i++)
+                       p->buf[p->col++] = word[i];
                return;
        }
 
-       wp++;
-       for (j = 0; *wp && ']' != *wp; wp++, j++)
-               /* Loop... */ ;
+       /* Pre-buffer, assuming worst-case. */
 
-       if (0 == *wp) {
-               *word = wp;
-               return;
+       if (p->col + 1 + (len * 3) >= p->maxcols)
+               adjbuf(p, p->col + 1 + (len * 3));
+
+       for (i = 0; i < len; i++) {
+               if ( ! isgraph((unsigned char)word[i])) {
+                       p->buf[p->col++] = word[i];
+                       continue;
+               }
+
+               if (TERMFONT_UNDER == f)
+                       p->buf[p->col++] = '_';
+               else
+                       p->buf[p->col++] = word[i];
+
+               p->buf[p->col++] = 8;
+               p->buf[p->col++] = word[i];
        }
+}
 
-       if (type)
-               do_special(p, wp - j, (size_t)j);
-       else
-               do_reserved(p, wp - j, (size_t)j);
-       *word = wp;
+size_t
+term_len(const struct termp *p, size_t sz)
+{
+
+       return((*p->width)(p, ' ') * sz);
 }
 
 
-/*
- * 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)
+size_t
+term_strlen(const struct termp *p, const char *cp)
 {
-       const char       *sv;
+       size_t           sz, rsz, i;
+       int              ssz, c;
+       const char      *seq, *rhs;
+       enum mandoc_esc  esc;
+       static const char rej[] = { '\\', ASCII_HYPH, ASCII_NBRSP, '\0' };
 
-       sv = word;
+       /*
+        * 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 (word[0] && 0 == word[1])
-               switch (word[0]) {
-               case('.'):
-                       /* FALLTHROUGH */
-               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;
-               }
+       sz = 0;
+       while ('\0' != *cp) {
+               rsz = strcspn(cp, rej);
+               for (i = 0; i < rsz; i++)
+                       sz += (*p->width)(p, *cp++);
+
+               c = 0;
+               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 += (*p->width)(p, c);
+                                       continue;
+                               case (ESCAPE_SPECIAL):
+                                       c = mchars_spec2cp
+                                               (p->symtab, seq, ssz);
+                                       if (c <= 0)
+                                               break;
+                                       sz += (*p->width)(p, c);
+                                       continue;
+                               default:
+                                       break;
+                               }
+
+                       rhs = NULL;
+
+                       switch (esc) {
+                       case (ESCAPE_UNICODE):
+                               sz += (*p->width)(p, '?');
+                               break;
+                       case (ESCAPE_NUMBERED):
+                               c = mchars_num2char(seq, ssz);
+                               if ('\0' != c)
+                                       sz += (*p->width)(p, c);
+                               break;
+                       case (ESCAPE_SPECIAL):
+                               rhs = mchars_spec2str
+                                       (p->symtab, seq, ssz, &rsz);
 
-       if ( ! (TERMP_NOSPACE & p->flags))
-               buffer(p, ' ');
+                               if (ssz != 1 || rhs)
+                                       break;
 
-       if ( ! (p->flags & TERMP_NONOSPACE))
-               p->flags &= ~TERMP_NOSPACE;
+                               rhs = seq;
+                               rsz = ssz;
+                               break;
+                       default:
+                               break;
+                       }
 
-       for ( ; *word; word++)
-               if ('\\' != *word)
-                       encode(p, *word);
-               else
-                       do_escaped(p, &word);
+                       if (NULL == rhs)
+                               break;
 
-       if (sv[0] && 0 == sv[1])
-               switch (sv[0]) {
-               case('('):
-                       /* FALLTHROUGH */
-               case('['):
-                       /* FALLTHROUGH */
-               case('{'):
-                       p->flags |= TERMP_NOSPACE;
+                       for (i = 0; i < rsz; i++)
+                               sz += (*p->width)(p, *rhs++);
+                       break;
+               case (ASCII_NBRSP):
+                       sz += (*p->width)(p, ' ');
+                       cp++;
+                       break;
+               case (ASCII_HYPH):
+                       sz += (*p->width)(p, '-');
+                       cp++;
                        break;
                default:
                        break;
                }
-}
+       }
 
+       return(sz);
+}
 
-/*
- * Insert a single character into the line-buffer.  If the buffer's
- * space is exceeded, then allocate more space by doubling the buffer
- * size.
- */
-static void
-buffer(struct termp *p, char c)
+/* ARGSUSED */
+size_t
+term_vspan(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"); /* FIXME: shouldn't be here! */
-               p->maxcols = s;
+       double           r;
+
+       switch (su->unit) {
+       case (SCALE_CM):
+               r = su->scale * 2;
+               break;
+       case (SCALE_IN):
+               r = su->scale * 6;
+               break;
+       case (SCALE_PC):
+               r = su->scale;
+               break;
+       case (SCALE_PT):
+               r = su->scale / 8;
+               break;
+       case (SCALE_MM):
+               r = su->scale / 1000;
+               break;
+       case (SCALE_VS):
+               r = su->scale;
+               break;
+       default:
+               r = su->scale - 1;
+               break;
        }
-       p->buf[(int)(p->col)++] = c;
-}
 
+       if (r < 0.0)
+               r = 0.0;
+       return(/* LINTED */(size_t)
+                       r);
+}
 
-static void
-encode(struct termp *p, char c)
+size_t
+term_hspan(const struct termp *p, const struct roffsu *su)
 {
-       
-       if (' ' != c) {
-               if (p->bold) {
-                       buffer(p, c);
-                       buffer(p, 8);
-               }
-               if (p->under) {
-                       buffer(p, '_');
-                       buffer(p, 8);
-               }
-       }
-       buffer(p, c);
+       double           v;
+
+       v = ((*p->hspan)(p, su));
+       if (v < 0.0)
+               v = 0.0;
+       return((size_t) /* LINTED */
+                       v);
 }