===================================================================
RCS file: /cvs/mandoc/term.c,v
retrieving revision 1.76
retrieving revision 1.89
diff -u -p -r1.76 -r1.89
--- mandoc/term.c	2009/06/11 07:26:35	1.76
+++ mandoc/term.c	2009/07/16 13:17:51	1.89
@@ -1,4 +1,4 @@
-/*	$Id: term.c,v 1.76 2009/06/11 07:26:35 kristaps Exp $ */
+/*	$Id: term.c,v 1.89 2009/07/16 13:17:51 kristaps Exp $ */
 /*
  * Copyright (c) 2008, 2009 Kristaps Dzonsons <kristaps@kth.se>
  *
@@ -31,16 +31,13 @@ extern	int		  mdoc_run(struct termp *, 
 
 static	struct termp	 *term_alloc(enum termenc);
 static	void		  term_free(struct termp *);
-static	void		  term_pword(struct termp *, const char *, int);
-static	void		  term_pescape(struct termp *, 
-				const char *, int *, int);
+static	void		  term_pescape(struct termp *, const char **);
 static	void		  term_nescape(struct termp *,
 				const char *, size_t);
 static	void		  term_chara(struct termp *, char);
-static	void		  term_stringa(struct termp *, 
-				const char *, size_t);
-static	int		  term_isopendelim(const char *, int);
-static	int		  term_isclosedelim(const char *, int);
+static	void		  term_encodea(struct termp *, char);
+static	int		  term_isopendelim(const char *);
+static	int		  term_isclosedelim(const char *);
 
 
 void *
@@ -106,17 +103,17 @@ term_alloc(enum termenc enc)
 	if (NULL == (p = malloc(sizeof(struct termp))))
 		err(1, "malloc");
 	bzero(p, sizeof(struct termp));
-	p->maxrmargin = 80;
+	p->maxrmargin = 78;
 	p->enc = enc;
 	return(p);
 }
 
 
 static int
-term_isclosedelim(const char *p, int len)
+term_isclosedelim(const char *p)
 {
 
-	if (1 != len)
+	if ( ! (*p && 0 == *(p + 1)))
 		return(0);
 
 	switch (*p) {
@@ -147,10 +144,10 @@ term_isclosedelim(const char *p, int len)
 
 
 static int
-term_isopendelim(const char *p, int len)
+term_isopendelim(const char *p)
 {
 
-	if (1 != len)
+	if ( ! (*p && 0 == *(p + 1)))
 		return(0);
 
 	switch (*p) {
@@ -177,8 +174,7 @@ term_isopendelim(const char *p, int len)
  * 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:
+ * The usage of termp:flags is as 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
@@ -188,8 +184,13 @@ term_isopendelim(const char *p, int len)
  *    columns.  In short: don't print a newline and instead pad to the
  *    right margin.  Used in conjunction with TERMP_NOLPAD.
  *
- *  - TERMP_NONOBREAK: don't newline when TERMP_NOBREAK is specified.
+ *  - 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
+ *    when the next invocation.
+ *
  *  In-line line breaking:
  *
  *  If TERMP_NOBREAK is specified and the line overruns the right
@@ -201,12 +202,16 @@ term_isopendelim(const char *p, int len)
  *  Otherwise, the line will break at the right margin.  Extremely long
  *  lines will cause the system to emit a warning (TODO: hyphenate, if
  *  possible).
+ *
+ *  FIXME: newline breaks occur (in groff) also occur when a single
+ *  space follows a NOBREAK!
  */
 void
 term_flushln(struct termp *p)
 {
 	int		 i, j;
-	size_t		 vsz, vis, maxvis, mmax, bp;
+	size_t		 vbl, vsz, vis, maxvis, mmax, bp;
+	static int	 sv = -1;
 
 	/*
 	 * First, establish the maximum columns of "visible" content.
@@ -221,6 +226,11 @@ term_flushln(struct termp *p)
 	bp = TERMP_NOBREAK & p->flags ? mmax : maxvis;
 	vis = 0;
 
+	if (sv >= 0) {
+		vis = (size_t)sv;
+		sv = -1;
+	}
+
 	/*
 	 * If in the standard case (left-justified), then begin with our
 	 * indentation, otherwise (columns, etc.) just start spitting
@@ -242,52 +252,52 @@ term_flushln(struct termp *p)
 
 		/* LINTED */
 		for (j = i, vsz = 0; j < (int)p->col; j++) {
-			if (' ' == p->buf[j])
+			if (' ' == p->buf[j]) 
 				break;
 			else if (8 == p->buf[j])
-				j += 1;
+				vsz--;
 			else
 				vsz++;
 		}
 
 		/*
-		 * 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.
+		 * 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);
 
-		if ( ! (TERMP_NOBREAK & p->flags)) {
-			if (vis && vis + vsz > bp) {
-				putchar('\n');
+		/*
+		 * 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 (vis && vis + vbl + vsz > bp) {
+			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;
-			} 
-		} else if (vis && vis + vsz > bp) {
-			putchar('\n');
-			for (j = 0; j < (int)p->rmargin; j++)
+			}
+		} else {
+			for (j = 0; j < (int)vbl; j++)
 				putchar(' ');
-			vis = p->rmargin - p->offset;
+			vis += vbl;
 		}
 
-		/* 
-		 * Write out the word and a trailing space.  Omit the
-		 * space if we're the last word in the line or beyond
-		 * our breakpoint.
+		/*
+		 * Finally, write out the word.
 		 */
-
 		for ( ; i < (int)p->col; i++) {
 			if (' ' == p->buf[i])
 				break;
 			putchar(p->buf[i]);
 		}
 		vis += vsz;
-		if (i < (int)p->col && vis <= bp) {
-			putchar(' ');
-			vis++;
-		}
 	}
 
 	/*
@@ -296,11 +306,14 @@ term_flushln(struct termp *p)
 	 */
 
 	if ((TERMP_NOBREAK & p->flags) && vis >= maxvis) {
-		if ( ! (TERMP_NONOBREAK & p->flags)) {
+		if ( ! (TERMP_DANGLE & p->flags) &&
+				! (TERMP_HANG & p->flags)) {
 			putchar('\n');
 			for (i = 0; i < (int)p->rmargin; i++)
 				putchar(' ');
 		}
+		if (TERMP_HANG & p->flags)
+			sv = (int)(vis - maxvis);
 		p->col = 0;
 		return;
 	}
@@ -311,7 +324,7 @@ term_flushln(struct termp *p)
 	 */
 
 	if (p->flags & TERMP_NOBREAK) {
-		if ( ! (TERMP_NONOBREAK & p->flags))
+		if ( ! (TERMP_DANGLE & p->flags))
 			for ( ; vis < maxvis; vis++)
 				putchar(' ');
 	} else
@@ -356,50 +369,6 @@ term_vspace(struct termp *p)
 
 
 /*
- * 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
-term_word(struct termp *p, const char *word)
-{
-	int 		 i, j, len;
-
-	len = (int)strlen(word);
-
-	if (p->flags & TERMP_LITERAL) {
-		term_pword(p, word, len);
-		return;
-	}
-
-	/* LINTED */
-	for (j = i = 0; i < len; i++) {
-		if (' ' != word[i]) {
-			j++;
-			continue;
-		} 
-		
-		/* Escaped spaces don't delimit... */
-		if (i && ' ' == word[i] && '\\' == word[i - 1]) {
-			j++;
-			continue;
-		}
-
-		if (0 == j)
-			continue;
-		assert(i >= j);
-		term_pword(p, &word[i - j], j);
-		j = 0;
-	}
-	if (j > 0) {
-		assert(i >= j);
-		term_pword(p, &word[i - j], j);
-	}
-}
-
-
-/*
  * 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.
@@ -409,10 +378,13 @@ term_nescape(struct termp *p, const char *word, size_t
 {
 	const char	*rhs;
 	size_t		 sz;
+	int		 i;
 
-	if (NULL == (rhs = term_a2ascii(p->symtab, word, len, &sz))) 
-		return;
-	term_stringa(p, rhs, sz);
+	rhs = term_a2ascii(p->symtab, word, len, &sz);
+
+	if (rhs)
+		for (i = 0; i < (int)sz; i++) 
+			term_encodea(p, rhs[i]);
 }
 
 
@@ -422,48 +394,61 @@ term_nescape(struct termp *p, const char *word, size_t
  * the escape sequence (we assert upon badly-formed escape sequences).
  */
 static void
-term_pescape(struct termp *p, const char *word, int *i, int len)
+term_pescape(struct termp *p, const char **word)
 {
 	int		 j;
+	const char	*wp;
 
-	if (++(*i) >= len)
+	wp = *word;
+
+	if (0 == *(++wp)) {
+		*word = wp;
 		return;
+	}
 
-	if ('(' == word[*i]) {
-		(*i)++;
-		if (*i + 1 >= len)
+	if ('(' == *wp) {
+		wp++;
+		if (0 == *wp || 0 == *(wp + 1)) {
+			*word = 0 == *wp ? wp : wp + 1;
 			return;
+		}
 
-		term_nescape(p, &word[*i], 2);
-		(*i)++;
+		term_nescape(p, wp, 2);
+		*word = ++wp;
 		return;
 
-	} else if ('*' == word[*i]) { 
-		(*i)++;
-		if (*i >= len)
+	} else if ('*' == *wp) {
+		if (0 == *(++wp)) {
+			*word = wp;
 			return;
+		}
 
-		switch (word[*i]) {
+		switch (*wp) {
 		case ('('):
-			(*i)++;
-			if (*i + 1 >= len)
+			wp++;
+			if (0 == *wp || 0 == *(wp + 1)) {
+				*word = 0 == *wp ? wp : wp + 1;
 				return;
+			}
 
-			term_nescape(p, &word[*i], 2);
-			(*i)++;
+			term_nescape(p, wp, 2);
+			*word = ++wp;
 			return;
 		case ('['):
 			break;
 		default:
-			term_nescape(p, &word[*i], 1);
+			term_nescape(p, wp, 1);
+			*word = wp;
 			return;
 		}
 	
-	} else if ('f' == word[*i]) {
-		(*i)++;
-		if (*i >= len)
+	} else if ('f' == *wp) {
+		if (0 == *(++wp)) {
+			*word = wp;
 			return;
-		switch (word[*i]) {
+		}
+
+		switch (*wp) {
 		case ('B'):
 			p->flags |= TERMP_BOLD;
 			break;
@@ -478,21 +463,27 @@ term_pescape(struct termp *p, const char *word, int *i
 		default:
 			break;
 		}
+
+		*word = wp;
 		return;
 
-	} else if ('[' != word[*i]) {
-		term_nescape(p, &word[*i], 1);
+	} else if ('[' != *wp) {
+		term_nescape(p, wp, 1);
+		*word = wp;
 		return;
 	}
 
-	(*i)++;
-	for (j = 0; word[*i] && ']' != word[*i]; (*i)++, j++)
+	wp++;
+	for (j = 0; *wp && ']' != *wp; wp++, j++)
 		/* Loop... */ ;
 
-	if (0 == word[*i])
+	if (0 == *wp) {
+		*word = wp;
 		return;
+	}
 
-	term_nescape(p, &word[*i - j], (size_t)j);
+	term_nescape(p, wp - j, (size_t)j);
+	*word = wp;
 }
 
 
@@ -501,12 +492,12 @@ term_pescape(struct termp *p, const char *word, int *i
  * phrase that cannot be broken down (such as a literal string).  This
  * handles word styling.
  */
-static void
-term_pword(struct termp *p, const char *word, int len)
+void
+term_word(struct termp *p, const char *word)
 {
-	int		 i;
+	const char	 *sv;
 
-	if (term_isclosedelim(word, len))
+	if (term_isclosedelim(word))
 		if ( ! (TERMP_IGNDELIM & p->flags))
 			p->flags |= TERMP_NOSPACE;
 
@@ -521,61 +512,18 @@ term_pword(struct termp *p, const char *word, int len)
 	 * before the word.
 	 */
 
-	for (i = 0; i < len; i++) {
-		if ('\\' == word[i]) {
-			term_pescape(p, word, &i, len);
-			continue;
-		}
+	for (sv = word; *word; word++)
+		if ('\\' != *word)
+			term_encodea(p, *word);
+		else
+			term_pescape(p, &word);
 
-		if (TERMP_STYLE & p->flags) {
-			if (TERMP_BOLD & p->flags) {
-				term_chara(p, word[i]);
-				term_chara(p, 8);
-			}
-			if (TERMP_UNDER & p->flags) {
-				term_chara(p, '_');
-				term_chara(p, 8);
-			}
-		}
-
-		term_chara(p, word[i]);
-	}
-
-	if (term_isopendelim(word, len))
+	if (term_isopendelim(sv))
 		p->flags |= TERMP_NOSPACE;
 }
 
 
 /*
- * Like term_chara() but for arbitrary-length buffers.  Resize the
- * buffer by a factor of two (if the buffer is less than that) or the
- * buffer's size.
- */
-static void
-term_stringa(struct termp *p, const char *c, size_t sz)
-{
-	size_t		 s;
-
-	if (0 == sz)
-		return;
-
-	assert(c);
-	if (p->col + sz >= p->maxcols) {
-		if (0 == p->maxcols)
-			p->maxcols = 256;
-		s = sz > p->maxcols * 2 ? sz : p->maxcols * 2;
-		p->buf = realloc(p->buf, s);
-		if (NULL == p->buf)
-			err(1, "realloc");
-		p->maxcols = s;
-	}
-
-	(void)memcpy(&p->buf[(int)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.
@@ -597,3 +545,20 @@ term_chara(struct termp *p, char c)
 	p->buf[(int)(p->col)++] = c;
 }
 
+
+static void
+term_encodea(struct termp *p, char c)
+{
+	
+	if (' ' != c && TERMP_STYLE & p->flags) {
+		if (TERMP_BOLD & p->flags) {
+			term_chara(p, c);
+			term_chara(p, 8);
+		}
+		if (TERMP_UNDER & p->flags) {
+			term_chara(p, '_');
+			term_chara(p, 8);
+		}
+	}
+	term_chara(p, c);
+}