X-Git-Url: https://git.cameronkatri.com/mandoc.git/blobdiff_plain/4cd5b88b9ab0146c48e737ba2e83f5d7fa273d1d..1e85c3949ce76b81aa8b89f634e30797ddc81cc8:/mdocterm.c

diff --git a/mdocterm.c b/mdocterm.c
index 0995da7d..f2f05dfa 100644
--- a/mdocterm.c
+++ b/mdocterm.c
@@ -1,4 +1,4 @@
-	/* $Id: mdocterm.c,v 1.7 2009/02/23 15:34:53 kristaps Exp $ */
+/* $Id: mdocterm.c,v 1.26 2009/03/02 17:29:16 kristaps Exp $ */
 /*
  * Copyright (c) 2008 Kristaps Dzonsons <kristaps@kth.se>
  *
@@ -16,6 +16,8 @@
  * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
  * PERFORMANCE OF THIS SOFTWARE.
  */
+#include <sys/utsname.h>
+
 #include <assert.h>
 #include <ctype.h>
 #include <err.h>
@@ -23,7 +25,6 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
-
 #ifndef __OpenBSD__
 #include <time.h>
 #endif
@@ -31,19 +32,8 @@
 #include "mmain.h"
 #include "term.h"
 
-#ifdef __NetBSD__
-#define xisspace(x) isspace((int)(x))
-#else
-#define	xisspace(x) isspace((x))
-#endif
-
-enum	termstyle {
-	STYLE_CLEAR,
-	STYLE_BOLD,
-	STYLE_UNDERLINE
-};
-
 static	void		  body(struct termp *,
+				struct termpair *,
 				const struct mdoc_meta *,
 				const struct mdoc_node *);
 static	void		  header(struct termp *,
@@ -52,17 +42,63 @@ static	void		  footer(struct termp *,
 				const struct mdoc_meta *);
 
 static	void		  pword(struct termp *, const char *, size_t);
-static	void		  pescape(struct termp *, 
-				const char *, size_t *, size_t);
+static	void		  pescape(struct termp *, const char *, 
+				size_t *, size_t);
+static	void		  nescape(struct termp *,
+				const char *, size_t);
 static	void		  chara(struct termp *, char);
-static	void		  stringa(struct termp *, const char *);
-static	void		  style(struct termp *, enum termstyle);
+static	void		  stringa(struct termp *, 
+				const char *, size_t);
+static	void		  symbola(struct termp *, enum tsym);
+static	void		  stylea(struct termp *, enum tstyle);
 
 #ifdef __linux__
 extern	size_t		  strlcat(char *, const char *, size_t);
 extern	size_t		  strlcpy(char *, const char *, size_t);
 #endif
 
+static	struct termsym	  termsym_ansi[] = {
+	{ "]", 1 },		/* TERMSYM_RBRACK */
+	{ "[", 1 },		/* TERMSYM_LBRACK */
+	{ "<-", 2 },		/* TERMSYM_LARROW */
+	{ "->", 2 },		/* TERMSYM_RARROW */
+	{ "^", 1 },		/* TERMSYM_UARROW */
+	{ "v", 1 },		/* TERMSYM_DARROW */
+	{ "`", 1 },		/* TERMSYM_LSQUOTE */
+	{ "\'", 1 },		/* TERMSYM_RSQUOTE */
+	{ "\'", 1 },		/* TERMSYM_SQUOTE */
+	{ "``", 2 },		/* TERMSYM_LDQUOTE */
+	{ "\'\'", 2 },		/* TERMSYM_RDQUOTE */
+	{ "\"", 1 },		/* TERMSYM_DQUOTE */
+	{ "<", 1 },		/* TERMSYM_LT */
+	{ ">", 1 },		/* TERMSYM_GT */
+	{ "<=", 2 },		/* TERMSYM_LE */
+	{ ">=", 2 },		/* TERMSYM_GE */
+	{ "==", 2 },		/* TERMSYM_EQ */
+	{ "!=", 2 },		/* TERMSYM_NEQ */
+	{ "\'", 1 },		/* TERMSYM_ACUTE */
+	{ "`", 1 },		/* TERMSYM_GRAVE */
+	{ "pi", 2 },		/* TERMSYM_PI */
+	{ "+=", 2 },		/* TERMSYM_PLUSMINUS */
+	{ "oo", 2 },		/* TERMSYM_INF */
+	{ "infinity", 8 },	/* TERMSYM_INF2 */
+	{ "NaN", 3 },		/* TERMSYM_NAN */
+	{ "|", 1 },		/* TERMSYM_BAR */
+	{ "o", 1 },		/* TERMSYM_BULLET */
+	{ "&", 1 },		/* TERMSYM_AND */
+	{ "|", 1 },		/* TERMSYM_OR */
+};
+
+static	const char	  ansi_clear[]  = { 27, '[', '0', 'm' };
+static	const char	  ansi_bold[]  = { 27, '[', '1', 'm' };
+static	const char	  ansi_under[]  = { 27, '[', '4', 'm' };
+
+static	struct termsym	  termstyle_ansi[] = {
+	{ ansi_clear, 4 },
+	{ ansi_bold, 4 },
+	{ ansi_under, 4 }
+};
+
 
 int
 main(int argc, char *argv[])
@@ -79,17 +115,19 @@ main(int argc, char *argv[])
 	if (NULL == (mdoc = mmain_mdoc(p)))
 		mmain_exit(p, 1);
 
-	termp.maxrmargin = 80; /* XXX */
+	termp.maxrmargin = 78; /* XXX */
 	termp.rmargin = termp.maxrmargin;
 	termp.maxcols = 1024;
 	termp.offset = termp.col = 0;
 	termp.flags = TERMP_NOSPACE;
+	termp.symtab = termsym_ansi;
+	termp.styletab = termstyle_ansi;
 
 	if (NULL == (termp.buf = malloc(termp.maxcols)))
 		err(1, "malloc");
 
 	header(&termp, mdoc_meta(mdoc));
-	body(&termp, mdoc_meta(mdoc), mdoc_node(mdoc));
+	body(&termp, NULL, mdoc_meta(mdoc), mdoc_node(mdoc));
 	footer(&termp, mdoc_meta(mdoc));
 
 	free(termp.buf);
@@ -99,10 +137,45 @@ main(int argc, char *argv[])
 }
 
 
+/*
+ * 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.
+ *
+ * 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_LITERAL: don't break apart words.  Note that a long literal
+ *    word will violate the right margin.
+ *
+ *  - 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.
+ *
+ *  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).
+ */
 void
 flushln(struct termp *p)
 {
-	size_t		 i, j, vsz, vis, maxvis;
+	size_t		 i, j, vsz, vis, maxvis, mmax, bp;
 
 	/*
 	 * First, establish the maximum columns of "visible" content.
@@ -113,6 +186,8 @@ flushln(struct termp *p)
 
 	assert(p->offset < p->rmargin);
 	maxvis = p->rmargin - p->offset;
+	mmax = p->maxrmargin - p->offset;
+	bp = TERMP_NOBREAK & p->flags ? mmax : maxvis;
 	vis = 0;
 
 	/*
@@ -126,28 +201,19 @@ flushln(struct termp *p)
 		for (j = 0; j < p->offset; j++)
 			putchar(' ');
 
-	/*
-	 * If we're literal, print out verbatim.
-	 */
-	if (p->flags & TERMP_LITERAL) {
-		/* FIXME: count non-printing chars. */
-		for (i = 0; i < p->col; i++)
-			putchar(p->buf[i]);
-		putchar('\n');
-		p->col = 0;
-		return;
-	}
-
 	for (i = 0; i < p->col; i++) {
 		/*
 		 * Count up visible word characters.  Control sequences
-		 * (starting with the CSI) aren't counted. 
+		 * (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).
 		 */
-		assert( ! xisspace(p->buf[i]));
+
+		/* FIXME: make non-ANSI friendly. */
 
 		/* LINTED */
 		for (j = i, vsz = 0; j < p->col; j++) {
-			if (xisspace(p->buf[j]))
+			if (isspace((int)p->buf[j]))
 				break;
 			else if (27 == p->buf[j]) {
 				assert(j + 4 <= p->col);
@@ -155,52 +221,76 @@ flushln(struct termp *p)
 			} else
 				vsz++;
 		}
-		assert(vsz > 0);
 
 		/*
-		 * If a word is too long and we're within a line, put it
-		 * on the next line.  Puke if we're being asked to write
-		 * something that will exceed the right margin (i.e.,
-		 * from a fresh line or when we're not allowed to break
-		 * the line with TERMP_NOBREAK).
+		 * 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.
 		 */
 
-		if (vis && vis + vsz >= maxvis) {
-			/* FIXME */
-			if (p->flags & TERMP_NOBREAK)
-				errx(1, "word breaks right margin");
-			putchar('\n');
-			for (j = 0; j < p->offset; j++)
-				putchar(' ');
-			vis = 0;
-		} else if (vis + vsz >= maxvis)
-			/* FIXME */
-			errx(1, "word breaks right margin");
+		if ( ! (TERMP_NOBREAK & p->flags)) {
+			if (vis && vis + vsz > bp) {
+				putchar('\n');
+				for (j = 0; j < p->offset; j++)
+					putchar(' ');
+				vis = 0;
+			} else if (vis + vsz > bp)
+				warnx("word breaks right margin");
+
+			/* TODO: hyphenate. */
+
+		} else {
+			if (vis && vis + vsz > bp) {
+				putchar('\n');
+				for (j = 0; j < p->rmargin; j++)
+					putchar(' ');
+				vis = p->rmargin;
+			} else if (vis + vsz > bp) 
+				warnx("word breaks right margin");
+
+			/* TODO: hyphenate. */
+		}
 
 		/* 
 		 * Write out the word and a trailing space.  Omit the
-		 * space if we're the last word in the line.
+		 * space if we're the last word in the line or beyond
+		 * our breakpoint.
 		 */
 
 		for ( ; i < p->col; i++) {
-			if (xisspace(p->buf[i]))
+			if (isspace((int)p->buf[i]))
 				break;
 			putchar(p->buf[i]);
 		}
 		vis += vsz;
-		if (i < p->col) {
+		if (i < p->col && vis <= bp) {
 			putchar(' ');
 			vis++;
 		}
 	}
 
+	/*
+	 * If we've overstepped our maximum visible no-break space, then
+	 * cause a newline and offset at the right margin.
+	 */
+
+	if ((TERMP_NOBREAK & p->flags) && vis >= maxvis) {
+		putchar('\n');
+		for (i = 0; i < p->rmargin; i++)
+			putchar(' ');
+		p->col = 0;
+		return;
+	}
+
 	/*
 	 * If we're not to right-marginalise it (newline), then instead
 	 * pad to the right margin and stay off.
 	 */
 
 	if (p->flags & TERMP_NOBREAK) {
-		for ( ; vis <= maxvis; vis++)
+		for ( ; vis < maxvis; vis++)
 			putchar(' ');
 	} else
 		putchar('\n');
@@ -209,160 +299,46 @@ flushln(struct termp *p)
 }
 
 
+/* 
+ * 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.
+ */
 void
 newln(struct termp *p)
 {
 
-	/* 
-	 * A newline only breaks an existing line; it won't assert
-	 * vertical space.
-	 */
 	p->flags |= TERMP_NOSPACE;
-	if (0 == p->col) 
+	if (0 == p->col) {
+		p->flags &= ~TERMP_NOLPAD;
 		return;
+	}
 	flushln(p);
+	p->flags &= ~TERMP_NOLPAD;
 }
 
 
+/*
+ * 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.
+ * All data in the output buffer is flushed prior to the newline
+ * assertion.
+ */
 void
 vspace(struct termp *p)
 {
 
-	/*
-	 * Asserts a vertical space (a full, empty line-break between
-	 * lines).
-	 */
 	newln(p);
 	putchar('\n');
 }
 
 
-static void
-stringa(struct termp *p, const char *s)
-{
-
-	/* XXX - speed up if not passing to chara. */
-	for ( ; *s; s++)
-		chara(p, *s);
-}
-
-
-static void
-chara(struct termp *p, char c)
-{
-
-	/* TODO: dynamically expand the buffer. */
-	if (p->col + 1 >= p->maxcols)
-		errx(1, "line overrun");
-	p->buf[(p->col)++] = c;
-}
-
-
-static void
-style(struct termp *p, enum termstyle esc)
-{
-
-	if (p->col + 4 >= p->maxcols)
-		errx(1, "line overrun");
-
-	p->buf[(p->col)++] = 27;
-	p->buf[(p->col)++] = '[';
-	switch (esc) {
-	case (STYLE_CLEAR):
-		p->buf[(p->col)++] = '0';
-		break;
-	case (STYLE_BOLD):
-		p->buf[(p->col)++] = '1';
-		break;
-	case (STYLE_UNDERLINE):
-		p->buf[(p->col)++] = '4';
-		break;
-	default:
-		abort();
-		/* NOTREACHED */
-	}
-	p->buf[(p->col)++] = 'm';
-}
-
-
-static void
-pescape(struct termp *p, const char *word, size_t *i, size_t len)
-{
-
-	(*i)++;
-	assert(*i < len);
-
-	if ('(' == word[*i]) {
-		/* Two-character escapes. */
-		(*i)++;
-		assert(*i + 1 < len);
-
-		if ('r' == word[*i] && 'B' == word[*i + 1])
-			chara(p, ']');
-		else if ('l' == word[*i] && 'B' == word[*i + 1])
-			chara(p, '[');
-		else if ('<' == word[*i] && '-' == word[*i + 1])
-			stringa(p, "<-");
-		else if ('-' == word[*i] && '>' == word[*i + 1])
-			stringa(p, "->");
-
-		(*i)++;
-		return;
-
-	} else if ('[' != word[*i]) {
-		/* One-character escapes. */
-		switch (word[*i]) {
-		case ('\\'):
-			/* FALLTHROUGH */
-		case ('\''):
-			/* FALLTHROUGH */
-		case ('`'):
-			/* FALLTHROUGH */
-		case ('-'):
-			/* FALLTHROUGH */
-		case ('.'):
-			chara(p, word[*i]);
-		default:
-			break;
-		}
-		return;
-	}
-	/* n-character escapes. */
-}
-
-
-static void
-pword(struct termp *p, const char *word, size_t len)
-{
-	size_t		 i;
-
-	/*assert(len > 0);*/ /* Can be, if literal. */
-
-	if ( ! (p->flags & TERMP_NOSPACE) && 
-			! (p->flags & TERMP_LITERAL))
-		chara(p, ' ');
-
-	p->flags &= ~TERMP_NOSPACE;
-
-	if (p->flags & TERMP_BOLD)
-		style(p, STYLE_BOLD);
-	if (p->flags & TERMP_UNDERLINE)
-		style(p, STYLE_UNDERLINE);
-
-	for (i = 0; i < len; i++) {
-		if ('\\' == word[i]) {
-			pescape(p, word, &i, len);
-			continue;
-		}
-		chara(p, word[i]);
-	}
-
-	if (p->flags & TERMP_BOLD ||
-			p->flags & TERMP_UNDERLINE)
-		style(p, STYLE_CLEAR);
-}
-
-
+/*
+ * 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.
+ */
 void
 word(struct termp *p, const char *word)
 {
@@ -384,10 +360,18 @@ word(struct termp *p, const char *word)
 
 	/* LINTED */
 	for (j = i = 0; i < len; i++) {
-		if ( ! xisspace(word[i])) {
+		if ( ! isspace((int)word[i])) {
+			j++;
+			continue;
+		} 
+		
+		/* Escaped spaces don't delimit... */
+		if (i > 0 && isspace((int)word[i]) && 
+				'\\' == word[i - 1]) {
 			j++;
 			continue;
 		}
+
 		if (0 == j)
 			continue;
 		assert(i >= j);
@@ -401,38 +385,57 @@ word(struct termp *p, const char *word)
 }
 
 
+/*
+ * This is the main function for printing out nodes.  It's constituted
+ * of PRE and POST functions, which correspond to prefix and infix
+ * processing.  The termpair structure allows data to persist between
+ * prefix and postfix invocations.
+ */
 static void
-body(struct termp *p, const struct mdoc_meta *meta,
+body(struct termp *p, struct termpair *ppair,
+		const struct mdoc_meta *meta,
 		const struct mdoc_node *node)
 {
 	int		 dochild;
+	struct termpair	 pair;
 
 	/* Pre-processing. */
 
 	dochild = 1;
+	pair.ppair = ppair;
+	pair.type = 0;
+	pair.offset = pair.rmargin = 0;
+	pair.flag = 0;
+	pair.count = 0;
 
 	if (MDOC_TEXT != node->type) {
 		if (termacts[node->tok].pre)
-			if ( ! (*termacts[node->tok].pre)(p, meta, node))
+			if ( ! (*termacts[node->tok].pre)(p, &pair, meta, node))
 				dochild = 0;
 	} else /* MDOC_TEXT == node->type */
 		word(p, node->data.text.string);
 
 	/* Children. */
 
+	if (TERMPAIR_FLAG & pair.type)
+		p->flags |= pair.flag;
+
 	if (dochild && node->child)
-		body(p, meta, node->child);
+		body(p, &pair, meta, node->child);
+
+	if (TERMPAIR_FLAG & pair.type)
+		p->flags &= ~pair.flag;
 
 	/* Post-processing. */
 
 	if (MDOC_TEXT != node->type)
 		if (termacts[node->tok].post)
-			(*termacts[node->tok].post)(p, meta, node);
+			(*termacts[node->tok].post)(p, &pair, meta, node);
 
 	/* Siblings. */
 
 	if (node->next)
-		body(p, meta, node->next);
+		body(p, ppair, meta, node->next);
 }
 
 
@@ -441,7 +444,6 @@ footer(struct termp *p, const struct mdoc_meta *meta)
 {
 	struct tm	*tm;
 	char		*buf, *os;
-	size_t		 sz, osz, ssz, i;
 
 	if (NULL == (buf = malloc(p->rmargin)))
 		err(1, "malloc");
@@ -457,26 +459,31 @@ footer(struct termp *p, const struct mdoc_meta *meta)
 #endif
 		err(1, "strftime");
 
-	osz = strlcpy(os, meta->os, p->rmargin);
+	(void)strlcpy(os, meta->os, p->rmargin);
+
+	/*
+	 * This is /slightly/ different from regular groff output
+	 * because we don't have page numbers.  Print the following:
+	 *
+	 * OS                                            MDOCDATE
+	 */
+
+	vspace(p);
 
-	sz = strlen(buf);
-	ssz = sz + osz + 1;
+	p->flags |= TERMP_NOSPACE | TERMP_NOBREAK;
+	p->rmargin = p->maxrmargin - strlen(buf);
+	p->offset = 0;
 
-	if (ssz > p->rmargin) {
-		ssz -= p->rmargin;
-		assert(ssz <= osz);
-		os[osz - ssz] = 0;
-		ssz = 1;
-	} else
-		ssz = p->rmargin - ssz + 1;
+	word(p, os);
+	flushln(p);
 
-	printf("\n");
-	printf("%s", os);
-	for (i = 0; i < ssz; i++)
-		printf(" ");
+	p->flags |= TERMP_NOLPAD | TERMP_NOSPACE;
+	p->offset = p->rmargin;
+	p->rmargin = p->maxrmargin;
+	p->flags &= ~TERMP_NOBREAK;
 
-	printf("%s\n", buf);
-	fflush(stdout);
+	word(p, buf);
+	flushln(p);
 
 	free(buf);
 	free(os);
@@ -486,16 +493,21 @@ footer(struct termp *p, const struct mdoc_meta *meta)
 static void
 header(struct termp *p, const struct mdoc_meta *meta)
 {
-	char		*buf, *title;
-	const char	*pp, *msec;
-	size_t		 ssz, tsz, ttsz, i;;
+	char		*buf, *title, *bufp, *vbuf;
+	const char	*pp;
+	struct utsname	 uts;
+
+	p->rmargin = p->maxrmargin;
+	p->offset = 0;
 
 	if (NULL == (buf = malloc(p->rmargin)))
 		err(1, "malloc");
 	if (NULL == (title = malloc(p->rmargin)))
 		err(1, "malloc");
+	if (NULL == (vbuf = malloc(p->rmargin)))
+		err(1, "malloc");
 
-	if (NULL == (pp = mdoc_vol2a(meta->vol)))
+	if (NULL == (pp = mdoc_vol2a(meta->vol))) {
 		switch (meta->msec) {
 		case (MSEC_1):
 			/* FALLTHROUGH */
@@ -520,56 +532,359 @@ header(struct termp *p, const struct mdoc_meta *meta)
 			pp = mdoc_vol2a(VOL_KM);
 			break;
 		default:
-			/* FIXME: capitalise. */
-			if (NULL == (pp = mdoc_msec2a(meta->msec)))
-				pp = mdoc_msec2a(MSEC_local);
 			break;
 		}
-	assert(pp);
-
-	tsz = strlcpy(buf, pp, p->rmargin);
-	assert(tsz < p->rmargin);
-
-	if ((pp = mdoc_arch2a(meta->arch))) {
-		tsz = strlcat(buf, " (", p->rmargin);
-		assert(tsz < p->rmargin);
-		tsz = strlcat(buf, pp, p->rmargin);
-		assert(tsz < p->rmargin);
-		tsz = strlcat(buf, ")", p->rmargin);
-		assert(tsz < p->rmargin);
 	}
+	vbuf[0] = 0;
+
+	if (pp) {
+		if (-1 == uname(&uts)) 
+			err(1, "uname");
+		(void)strlcat(vbuf, uts.sysname, p->rmargin);
+		(void)strlcat(vbuf, " ", p->rmargin);
+	} else if (NULL == (pp = mdoc_msec2a(meta->msec)))
+		pp = mdoc_msec2a(MSEC_local);
 
-	ttsz = strlcpy(title, meta->title, p->rmargin);
+	(void)strlcat(vbuf, pp, p->rmargin);
 
-	if (NULL == (msec = mdoc_msec2a(meta->msec)))
-		msec = "";
+	/*
+	 * The header is strange.  It has three components, which are
+	 * really two with the first duplicated.  It goes like this:
+	 *
+	 * IDENTIFIER              TITLE                   IDENTIFIER
+	 *
+	 * The IDENTIFIER is NAME(SECTION), which is the command-name
+	 * (if given, or "unknown" if not) followed by the manual page
+	 * section.  These are given in `Dt'.  The TITLE is a free-form
+	 * string depending on the manual volume.  If not specified, it
+	 * switches on the manual section.
+	 */
 
-	ssz = (2 * (ttsz + 2 + strlen(msec))) + tsz + 2;
+	if (mdoc_arch2a(meta->arch))
+		(void)snprintf(buf, p->rmargin, "%s (%s)",
+				vbuf, mdoc_arch2a(meta->arch));
+	else
+		(void)strlcpy(buf, vbuf, p->rmargin);
 
-	if (ssz > p->rmargin) {
-		if ((ssz -= p->rmargin) % 2)
-			ssz++;
-		ssz /= 2;
+	pp = mdoc_msec2a(meta->msec);
+
+	(void)snprintf(title, p->rmargin, "%s(%s)",
+			meta->title, pp ? pp : "");
+
+	for (bufp = title; *bufp; bufp++)
+		*bufp = toupper(*bufp);
 	
-		assert(ssz <= ttsz);
-		title[ttsz - ssz] = 0;
-		ssz = 1;
-	} else
-		ssz = ((p->rmargin - ssz) / 2) + 1;
+	p->offset = 0;
+	p->rmargin = (p->maxrmargin - strlen(buf)) / 2;
+	p->flags |= TERMP_NOBREAK | TERMP_NOSPACE;
 
-	printf("%s(%s)", title, msec);
+	word(p, title);
+	flushln(p);
 
-	for (i = 0; i < ssz; i++)
-		printf(" ");
+	p->flags |= TERMP_NOLPAD | TERMP_NOSPACE;
+	p->offset = p->rmargin;
+	p->rmargin = p->maxrmargin - strlen(title);
 
-	printf("%s", buf);
+	word(p, buf);
+	flushln(p);
 
-	for (i = 0; i < ssz; i++)
-		printf(" ");
+	p->offset = p->rmargin;
+	p->rmargin = p->maxrmargin;
+	p->flags &= ~TERMP_NOBREAK;
+	p->flags |= TERMP_NOLPAD | TERMP_NOSPACE;
 
-	printf("%s(%s)\n", title, msec);
-	fflush(stdout);
+	word(p, title);
+	flushln(p);
+
+	p->rmargin = p->maxrmargin;
+	p->offset = 0;
+	p->flags &= ~TERMP_NOSPACE;
 
 	free(title);
+	free(vbuf);
 	free(buf);
 }
+
+
+/*
+ * 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
+nescape(struct termp *p, const char *word, size_t len)
+{
+
+	switch (len) {
+	case (1):
+		switch (word[0]) {
+		case ('\\'):
+			/* FALLTHROUGH */
+		case ('\''):
+			/* FALLTHROUGH */
+		case ('`'):
+			/* FALLTHROUGH */
+		case ('-'):
+			/* FALLTHROUGH */
+		case (' '):
+			/* FALLTHROUGH */
+		case ('.'):
+			chara(p, word[0]); /* FIXME */
+			break;
+		case ('&'):
+			break;
+		case ('e'):
+			chara(p, '\\'); /* FIXME */
+			break;
+		case ('q'):
+			symbola(p, TERMSYM_DQUOTE);
+			break;
+		default:
+			warnx("escape sequence not supported: %c",
+					word[0]);
+			break;
+		}
+		break;
+
+	case (2):
+		if ('r' == word[0] && 'B' == word[1])
+			symbola(p, TERMSYM_RBRACK);
+		else if ('l' == word[0] && 'B' == word[1])
+			symbola(p, TERMSYM_LBRACK);
+		else if ('l' == word[0] && 'q' == word[1])
+			symbola(p, TERMSYM_LDQUOTE);
+		else if ('r' == word[0] && 'q' == word[1])
+			symbola(p, TERMSYM_RDQUOTE);
+		else if ('o' == word[0] && 'q' == word[1])
+			symbola(p, TERMSYM_LSQUOTE);
+		else if ('a' == word[0] && 'q' == word[1])
+			symbola(p, TERMSYM_RSQUOTE);
+		else if ('<' == word[0] && '-' == word[1])
+			symbola(p, TERMSYM_LARROW);
+		else if ('-' == word[0] && '>' == word[1])
+			symbola(p, TERMSYM_RARROW);
+		else if ('b' == word[0] && 'u' == word[1])
+			symbola(p, TERMSYM_BULLET);
+		else if ('<' == word[0] && '=' == word[1])
+			symbola(p, TERMSYM_LE);
+		else if ('>' == word[0] && '=' == word[1])
+			symbola(p, TERMSYM_GE);
+		else if ('=' == word[0] && '=' == word[1])
+			symbola(p, TERMSYM_EQ);
+		else if ('+' == word[0] && '-' == word[1])
+			symbola(p, TERMSYM_PLUSMINUS);
+		else if ('u' == word[0] && 'a' == word[1])
+			symbola(p, TERMSYM_UARROW);
+		else if ('d' == word[0] && 'a' == word[1])
+			symbola(p, TERMSYM_DARROW);
+		else if ('a' == word[0] && 'a' == word[1])
+			symbola(p, TERMSYM_ACUTE);
+		else if ('g' == word[0] && 'a' == word[1])
+			symbola(p, TERMSYM_GRAVE);
+		else if ('!' == word[0] && '=' == word[1])
+			symbola(p, TERMSYM_NEQ);
+		else if ('i' == word[0] && 'f' == word[1])
+			symbola(p, TERMSYM_INF);
+		else if ('n' == word[0] && 'a' == word[1])
+			symbola(p, TERMSYM_NAN);
+		else if ('b' == word[0] && 'a' == word[1])
+			symbola(p, TERMSYM_BAR);
+
+		/* Deprecated forms. */
+		else if ('A' == word[0] && 'm' == word[1])
+			symbola(p, TERMSYM_AMP);
+		else if ('B' == word[0] && 'a' == word[1])
+			symbola(p, TERMSYM_BAR);
+		else if ('I' == word[0] && 'f' == word[1])
+			symbola(p, TERMSYM_INF2);
+		else if ('G' == word[0] && 'e' == word[1])
+			symbola(p, TERMSYM_GE);
+		else if ('G' == word[0] && 't' == word[1])
+			symbola(p, TERMSYM_GT);
+		else if ('L' == word[0] && 'e' == word[1])
+			symbola(p, TERMSYM_LE);
+		else if ('L' == word[0] && 'q' == word[1])
+			symbola(p, TERMSYM_LDQUOTE);
+		else if ('L' == word[0] && 't' == word[1])
+			symbola(p, TERMSYM_LT);
+		else if ('N' == word[0] && 'a' == word[1])
+			symbola(p, TERMSYM_NAN);
+		else if ('N' == word[0] && 'e' == word[1])
+			symbola(p, TERMSYM_NEQ);
+		else if ('P' == word[0] && 'i' == word[1])
+			symbola(p, TERMSYM_PI);
+		else if ('P' == word[0] && 'm' == word[1])
+			symbola(p, TERMSYM_PLUSMINUS);
+		else if ('R' == word[0] && 'q' == word[1])
+			symbola(p, TERMSYM_RDQUOTE);
+		else
+			warnx("escape sequence not supported: %c%c",
+					word[0], word[1]);
+		break;
+
+	default:
+		warnx("escape sequence not supported");
+		break;
+	}
+}
+
+
+/*
+ * 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).
+ */
+static void
+pescape(struct termp *p, const char *word, size_t *i, size_t len)
+{
+	size_t		 j;
+
+	(*i)++;
+	assert(*i < len);
+
+	if ('(' == word[*i]) {
+		(*i)++;
+		assert(*i + 1 < len);
+		nescape(p, &word[*i], 2);
+		(*i)++;
+		return;
+
+	} else if ('*' == word[*i]) { 
+		/* XXX - deprecated! */
+		(*i)++;
+		assert(*i < len);
+		switch (word[*i]) {
+		case ('('):
+			(*i)++;
+			assert(*i + 1 < len);
+			nescape(p, &word[*i], 2);
+			(*i)++;
+			return;
+		case ('['):
+			break;
+		default:
+			nescape(p, &word[*i], 1);
+			return;
+		}
+
+	} else if ('[' != word[*i]) {
+		nescape(p, &word[*i], 1);
+		return;
+	}
+
+	(*i)++;
+	for (j = 0; word[*i] && ']' != word[*i]; (*i)++, j++)
+		/* Loop... */ ;
+
+	assert(word[*i]);
+	nescape(p, &word[*i - j], j);
+}
+
+
+/*
+ * 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
+pword(struct termp *p, const char *word, size_t len)
+{
+	size_t		 i;
+
+	if ( ! (TERMP_NOSPACE & p->flags) && 
+			! (TERMP_LITERAL & p->flags))
+		chara(p, ' ');
+
+	if ( ! (p->flags & TERMP_NONOSPACE))
+		p->flags &= ~TERMP_NOSPACE;
+
+	/* 
+	 * XXX - if literal and underlining, this will underline the
+	 * spaces between literal words. 
+	 */
+
+	if (p->flags & TERMP_BOLD)
+		stylea(p, TERMSTYLE_BOLD);
+	if (p->flags & TERMP_UNDERLINE)
+		stylea(p, TERMSTYLE_UNDER);
+
+	for (i = 0; i < len; i++) {
+		if ('\\' == word[i]) {
+			pescape(p, word, &i, len);
+			continue;
+		}
+		chara(p, word[i]);
+	}
+
+	if (p->flags & TERMP_BOLD ||
+			p->flags & TERMP_UNDERLINE)
+		stylea(p, TERMSTYLE_CLEAR);
+}
+
+
+/*
+ * Add a symbol to the output line buffer.
+ */
+static void
+symbola(struct termp *p, enum tsym sym)
+{
+
+	assert(p->symtab[sym].sym);
+	stringa(p, p->symtab[sym].sym, p->symtab[sym].sz);
+}
+
+
+/*
+ * Add a style to the output line buffer.
+ */
+static void
+stylea(struct termp *p, enum tstyle style)
+{
+
+	assert(p->styletab[style].sym);
+	stringa(p, p->styletab[style].sym, p->styletab[style].sz);
+}
+
+
+/*
+ * Like 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
+stringa(struct termp *p, const char *c, size_t sz)
+{
+	size_t		 s;
+
+	s = sz > p->maxcols * 2 ? sz : p->maxcols * 2;
+	
+	assert(c);
+	if (p->col + sz >= p->maxcols) {
+		p->buf = realloc(p->buf, s);
+		if (NULL == p->buf)
+			err(1, "realloc");
+		p->maxcols = s;
+	}
+
+	(void)memcpy(&p->buf[p->col], c, sz);
+	p->col += 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
+chara(struct termp *p, char c)
+{
+
+	if (p->col + 1 >= p->maxcols) {
+		p->buf = realloc(p->buf, p->maxcols * 2);
+		if (NULL == p->buf)
+			err(1, "malloc");
+		p->maxcols *= 2;
+	}
+	p->buf[(p->col)++] = c;
+}