From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <zsh-workers-return-6070-mason-zsh=primenet.com.au@sunsite.auc.dk>
Received: (qmail 17545 invoked from network); 21 Apr 1999 14:49:54 -0000
Received: from sunsite.auc.dk (130.225.51.30)
  by ns1.primenet.com.au with SMTP; 21 Apr 1999 14:49:54 -0000
Received: (qmail 3851 invoked by alias); 21 Apr 1999 14:49:19 -0000
Mailing-List: contact zsh-workers-help@sunsite.auc.dk; run by ezmlm
Precedence: bulk
X-No-Archive: yes
X-Seq: 6070
Received: (qmail 3844 invoked from network); 21 Apr 1999 14:49:18 -0000
Message-Id: <9904211424.AA45017@ibmth.df.unipi.it>
To: zsh-workers@sunsite.auc.dk (Zsh hackers list)
Subject: PATCH: tilde in character class in switch
Date: Wed, 21 Apr 1999 16:24:58 +0200
From: Peter Stephenson <pws@ibmth.df.unipi.it>

Sven quite reasonably introduced the pattern _(|*[^~]) to match a file
beginning with _ and not ending in a tilde.  Unfortunately, it didn't work
with extendedglob.  The reason is that ~ for an exclusion has to be
searched for before the pattern is parsed, so that we can treat exclusions
specially, but character classes weren't handled by the code that did
this and so the ~ found in this pattern messed up the parsing.

This whole chunk, before and after, is a little messy, but as far as I can
see the real problem lies in the syntax; nothing flags in advance that the
pattern is going to have an exclusion, so none of those sets of initials
found in books on compiler design which apply to nicely behaved grammars
applies.

--- Misc/globtests.ksh.tilde	Tue Dec  1 15:39:32 1998
+++ Misc/globtests.ksh	Wed Apr 21 16:22:05 1999
@@ -68,6 +68,7 @@
 t moo.cow       !(*.*).!(*.*)
 f mad.moo.cow   !(*.*).!(*.*)
 f mucca.pazza   mu!(*(c))?.pa!(*(z))?
+f _foo~         _?(*[^~])
 t fff           !(f)
 t fff           *(!(f))
 t fff           +(!(f))
--- Misc/globtests.tilde	Tue Feb 23 15:29:54 1999
+++ Misc/globtests	Wed Apr 21 16:20:08 1999
@@ -75,6 +75,7 @@
 t moo.cow       (^*.*).(^*.*)
 f sane.moo.cow  (^*.*).(^*.*)
 f mucca.pazza   mu(^c#)?.pa(^z#)?
+f _foo~         _(|*[^~])
 t fff           ((^f))
 t fff           ((^f)#)
 t fff           ((^f)##)
--- Src/glob.c.tilde	Wed Apr 14 12:04:43 1999
+++ Src/glob.c	Wed Apr 21 16:18:09 1999
@@ -625,6 +625,36 @@
     return 0;
 }
 
+/**/
+static void
+parse_charset(void)
+{
+    /* Character set: brackets had better match */
+    if (pptr[1] == Outbrack)
+	*++pptr = ']';
+    else if ((pptr[1] == Hat || pptr[1] == '^' || pptr[1] == '!') &&
+	     pptr[2] == Outbrack)
+	*(pptr += 2) = ']';
+    while (*++pptr && *pptr != Outbrack) {
+	if (itok(*pptr)) {
+	    /* POSIX classes: make sure it's a real one,
+	     * leave the Inbrack tokenised if so.
+	     * We need to untokenize the Outbrack since otherwise
+	     * it might look like we got to the end of the range without
+	     * matching; we also need to accept ']' instead of
+	     * Outbrack in case this has already happened.
+	     */
+	    char *nptr;
+	    if (*pptr == Inbrack && pptr[1] == ':'
+		&& (nptr = strchr(pptr+2, ':')) && 
+		(*++nptr == Outbrack || *nptr == ']'))
+		*(pptr = nptr) = ']';
+	    else
+		*pptr = ztokens[*pptr - Pound];
+	}
+    }
+}
+
 /* enum used with ksh-like patterns, @(...) etc. */
 
 enum { KF_NONE, KF_AT, KF_QUEST, KF_STAR, KF_PLUS, KF_NOT };
@@ -853,24 +883,7 @@
 	    if (*pptr != Outang)
 		return NULL;
 	} else if (*pptr == Inbrack) {
-	    /* Character set: brackets had better match */
-	    if (pptr[1] == Outbrack)
-		*++pptr = ']';
-	    else if ((pptr[1] == Hat || pptr[1] == '^' || pptr[1] == '!') &&
-		     pptr[2] == Outbrack)
-		*(pptr += 2) = ']';
-	    while (*++pptr && *pptr != Outbrack) {
-		if (itok(*pptr)) {
-		    /* POSIX classes: make sure it's a real one, *
-		     * leave the Inbrack tokenised if so.        */
-		    char *nptr;
-		    if (*pptr == Inbrack && pptr[1] == ':'
-			&& (nptr = strchr(pptr+2, ':')) && 
-			*++nptr == Outbrack)
-			pptr = nptr;
-		    *pptr = ztokens[*pptr - Pound];
-		}
-	    }
+	    parse_charset();
 	    if (*pptr != Outbrack)
 		return NULL;
 	} else if (itok(*pptr) && *pptr != Star && *pptr != Quest)
@@ -912,7 +925,20 @@
 		break;
 	    else if (*sptr == Bar && !pct)
 		break;
-	    else if (*sptr == Tilde && !pct) {
+	    else if (*sptr == Inbrack) {
+		/*
+		 * Character classes can have tokenized characters in,
+		 * so we have to parse them properly.
+		 */
+		char *bstart = pptr;
+
+		pptr = sptr;
+		parse_charset();
+		sptr = pptr;
+		pptr = bstart;
+		if (*sptr != Outbrack)
+		    break;
+	    } else if (*sptr == Tilde && !pct) {
 		tail = NULL;
 		break;
 	    }


-- 
Peter Stephenson <pws@ibmth.df.unipi.it>       Tel: +39 050 844536
WWW:  http://www.ifh.de/~pws/
Dipartimento di Fisica, Via Buonarroti 2, 56127 Pisa, Italy