libregex: implement \b and \B (word boundary, not word boundary)

This is the last of the needed GNU expressions before we can unleash bsdgrep
by default. \b is effectively an agnostic equivalent of \< and \>, while
\B will match every space that isn't making a transition from
nonchar -> char or char -> nonchar.

contrib/netbsd-tests/lib/libc/regex/data/meta.in

@@ -5,7 +5,7 @@ a\*c		&	a*c	a*c
 a\\b		&	a\b	a\b
 a\\\*b		&	a\*b	a\*b
 # Begin FreeBSD
-a\bc		&C	EESCAPE
+a\bc		&	abc
 # End FreeBSD
 a\		&C	EESCAPE
 a\\bc		&	a\bc	a\bc

lib/libc/regex/engine.c

@@ -118,6 +118,7 @@ static states step(struct re_guts *g, sopno start, sopno stop, states bef, wint_
 #define	BOW	(BOL-4)
 #define	EOW	(BOL-5)
 #define	BADCHAR	(BOL-6)
+#define	NWBND	(BOL-7)
 #define	NONCHAR(c)	((c) <= OUT)
 /* sflags */
 #define	SBOS	0x0001
@@ -463,6 +464,8 @@ dissect(struct match *m,
 		case OEOW:
 		case OBOS:
 		case OEOS:
+		case OWBND:
+		case ONWBND:
 			break;
 		case OANY:
 		case OANYOF:
@@ -691,6 +694,21 @@ backref(struct match *m,
 			else
 				return(NULL);
 			break;
+		case OWBND:
+			if (ISBOW(m, sp) || ISEOW(m, sp))
+				{ /* yes */ }
+			else
+				return(NULL);
+			break;
+		case ONWBND:
+			if (((sp == m->beginp) && !ISWORD(*sp)) ||
+			    (sp == m->endp && !ISWORD(*(sp - 1))))
+				{ /* yes, beginning/end of subject */ }
+			else if (ISWORD(*(sp - 1)) == ISWORD(*sp))
+				{ /* yes, beginning/end of subject */ }
+			else
+				return(NULL);
+			break;
 		case OBOW:
 			if (ISBOW(m, sp))
 				{ /* yes */ }
@@ -916,6 +934,17 @@ walk(struct match *m, const char *start, const char *stop, sopno startst,
 			st = step(m->g, startst, stopst, st, flagch, st, sflags);
 			SP("sboweow", st, c);
 		}
+		if (lastc != OUT && c != OUT &&
+		    ISWORD(lastc) == ISWORD(c)) {
+			flagch = NWBND;
+		} else if ((lastc == OUT && !ISWORD(c)) ||
+		    (c == OUT && !ISWORD(lastc))) {
+			flagch = NWBND;
+		}
+		if (flagch == NWBND) {
+			st = step(m->g, startst, stopst, st, flagch, st, sflags);
+			SP("snwbnd", st, c);
+		}
 
 		/* are we done? */
 		if (ISSET(st, stopst)) {
@@ -1017,6 +1046,14 @@ step(struct re_guts *g,
 			if (ch == EOW)
 				FWD(aft, bef, 1);
 			break;
+		case OWBND:
+			if (ch == BOW || ch == EOW)
+				FWD(aft, bef, 1);
+			break;
+		case ONWBND:
+			if (ch == NWBND)
+				FWD(aft, aft, 1);
+			break;
 		case OANY:
 			if (!NONCHAR(ch))
 				FWD(aft, bef, 1);

lib/libc/regex/regcomp.c

@@ -486,6 +486,12 @@ p_ere_exp(struct parse *p, struct branchc *bc)
 			case '\'':
 				EMIT(OEOS, 0);
 				break;
+			case 'B':
+				EMIT(ONWBND, 0);
+				break;
+			case 'b':
+				EMIT(OWBND, 0);
+				break;
 			case 'W':
 			case 'w':
 			case 'S':
@@ -845,6 +851,12 @@ p_simp_re(struct parse *p, struct branchc *bc)
 			case BACKSL|'\'':
 				EMIT(OEOS, 0);
 				break;
+			case BACKSL|'B':
+				EMIT(ONWBND, 0);
+				break;
+			case BACKSL|'b':
+				EMIT(OWBND, 0);
+				break;
 			case BACKSL|'W':
 			case BACKSL|'w':
 			case BACKSL|'S':
@@ -1892,6 +1904,8 @@ findmust(struct parse *p, struct re_guts *g)
 		case OEOL:
 		case OBOS:
 		case OEOS:
+		case OWBND:
+		case ONWBND:
 		case O_QUEST:
 		case O_CH:
 		case OEND:
@@ -2043,6 +2057,8 @@ altoffset(sop *scan, int offset)
 			try++;
 		case OBOW:
 		case OEOW:
+		case OWBND:
+		case ONWBND:
 		case OLPAREN:
 		case ORPAREN:
 		case OOR2:

lib/libc/regex/regex2.h

@@ -106,6 +106,8 @@ typedef unsigned long sopno;
 #define	OEOW	(20L<<OPSHIFT)	/* end word	-			*/
 #define	OBOS	(21L<<OPSHIFT)	/* begin subj.  -			*/
 #define	OEOS	(22L<<OPSHIFT)	/* end subj.	-			*/
+#define	OWBND	(23L<<OPSHIFT)	/* word bound	-			*/
+#define	ONWBND	(24L<<OPSHIFT)	/* not bound	-			*/
 
 /*
  * Structures for [] character-set representation.

lib/libregex/tests/gnuext.in

@@ -17,14 +17,12 @@ a\|b\|c	b	abc	a
 \s\+	b	aSNTb	SNT
 # Word boundaries (\b, \B, \<, \>, \`, \')
 # (is/not boundary, start/end word, start/end subject string)
-# Most of these are disabled for the moment, and will be re-enabled as
+\babc\b	&	<abc>	abc
-# we become feature complete.
-#\babc\b	&	<abc>	abc
 \<abc\> &	<abc>	abc
-#\Babc\B	&	abc
+\Babc\B	&	abc
-#\B[abc]\B	&	<abc>	b
+\B[abc]\B	&	<abc>	b
-#\B[abc]+	-	<abc>	bc
+\B[abc]+	-	<abc>	bc
-#\B[abc]\+	b	<abc>	bc
+\B[abc]\+	b	<abc>	bc
 \`abc	&	abc	abc
 abc\'	&	abc	abc
 \`abc\'	&	abc	abc