${(s::)VAR} vs "${(s::)VAR}"

${(s::)VAR} vs "${(s::)VAR}"

[Kynn Jones]

[-- Attachment #1: Type: text/plain, Size: 596 bytes --]

Hi!

I was surprised by the following (note that the first output consists of
3 lines while the second output consists of 4; the only difference between
the two cases is that the first one iterates over ${(s::)VAR} while the
second does so over its double-quoted variant, "${(s::)VAR}"):

% (VAR=123; for c ( ${(s::)VAR} ) echo ">$c<")
>1<
>2<
>3<
% (VAR=123; for c ( "${(s::)VAR}" ) echo ">$c<")
>1<
>2<
>3<
><

Even after-the-fact I cannot explain this difference.  I would appreciate
a description of what's going on (and that, I hope, will lead to the
sought-after explanation).

Thanks!

kj

Re: ${(s::)VAR} vs "${(s::)VAR}"

[Peter Stephenson]

On Thu, 26 Apr 2012 16:48:00 -0400
Kynn Jones <kynnjo@gmail.com> wrote:
> % (VAR=123; for c ( "${(s::)VAR}" ) echo ">$c<")
> >1<
> >2<
> >3<
> ><

I think I can say with some degree of certainty I haven't a clue what's
going on here, and it doesn't agree with the documentation "An empty
string may also be given in which case every character will be a
separate element" since you don't magically get an extra character
by sticking the expression in double quotes.

This comes from the function wordcount() in utils.c, which is called in
a couple of functions used for splitting words, none of which has any
comments apart from a few speculations from me dating from 2003.  The
function takes a mysterious argument mul which presumably has got
something to do with doing something multiple times.

It seems hard to believe it should ever be counting an extra character
when it's reached the null at the end of the string.  The following
seems to fix it without any obvious side effects.

Index: Src/utils.c
===================================================================
RCS file: /cvsroot/zsh/zsh/Src/utils.c,v
retrieving revision 1.268
diff -p -u -r1.268 utils.c
--- Src/utils.c	16 Apr 2012 11:26:10 -0000	1.268
+++ Src/utils.c	27 Apr 2012 20:35:40 -0000
@@ -3114,7 +3114,7 @@ wordcount(char *s, char *sep, int mul)
 	r = 1;
 	sl = strlen(sep);
 	for (; (c = findsep(&s, sep, 0)) >= 0; s += sl)
-	    if ((c && *(s + sl)) || mul)
+	    if ((c || mul) && *(s + sl))
 		r++;
     } else {
 	char *t = s;
Index: Test/D04parameter.ztst
===================================================================
RCS file: /cvsroot/zsh/zsh/Test/D04parameter.ztst,v
retrieving revision 1.67
diff -p -u -r1.67 D04parameter.ztst
--- Test/D04parameter.ztst	22 Apr 2012 18:39:53 -0000	1.67
+++ Test/D04parameter.ztst	27 Apr 2012 20:35:40 -0000
@@ -1304,6 +1304,19 @@
 >in
 >it
 
+  str=abcd
+  print -l ${(s..)str}
+  print -l "${(s..)str}"
+0:splitting of strings into characters
+>a
+>b
+>c
+>d
+>a
+>b
+>c
+>d
+
   array=('%' '$' 'j' '*' '$foo')
   print ${array[(i)*]} "${array[(i)*]}"
   print ${array[(ie)*]} "${array[(ie)*]}"

-- 
Peter Stephenson <p.w.stephenson@ntlworld.com>
Web page now at http://homepage.ntlworld.com/p.w.stephenson/

Re: ${(s::)VAR} vs "${(s::)VAR}"

[Bart Schaefer]

On Apr 27,  9:38pm, Peter Stephenson wrote:
}
} This comes from the function wordcount() in utils.c, which is called in
} a couple of functions used for splitting words, none of which has any
} comments apart from a few speculations from me dating from 2003.  The
} function takes a mysterious argument mul which presumably has got
} something to do with doing something multiple times.

"mul" takes values -1, 0, 1.  The only time it's less than zero is when
called from spacesplit() with allownull == 0.  So it must have to do
with handling multiple consecutive appearances of the separator.

}  	for (; (c = findsep(&s, sep, 0)) >= 0; s += sl)
} -	    if ((c && *(s + sl)) || mul)
} +	    if ((c || mul) && *(s + sl))
}  		r++;

Does it matter that the changed line dereferences s+sl in cases where
the original line does not?

Re: ${(s::)VAR} vs "${(s::)VAR}"

[Peter Stephenson]

On Fri, 27 Apr 2012 19:50:48 -0700
Bart Schaefer <schaefer@brasslantern.com> wrote:
> }  	for (; (c = findsep(&s, sep, 0)) >= 0; s += sl)
> } -	    if ((c && *(s + sl)) || mul)
> } +	    if ((c || mul) && *(s + sl))
> }  		r++;
> 
> Does it matter that the changed line dereferences s+sl in cases where
> the original line does not?

That's true only if c is 0 and mul is 1 (before c had to be 1).	 c is 0
if we're already at a separator, so I think dereferencing s+sl is still
OK.

-- 
Peter Stephenson <p.w.stephenson@ntlworld.com>
Web page now at http://homepage.ntlworld.com/p.w.stephenson/

Re: ${(s::)VAR} vs "${(s::)VAR}"

[Peter Stephenson]

Aha, I have however worked out what's wrong with what I did:

% foo=":bar:"
% print -l "${(s.:.)foo}"

foo
%

There should be an extra empty element, as there used to be.  So I think
I need the following instead...  Possibly this explains why mul was
relegated to the end in that why.

Index: Src/utils.c
===================================================================
RCS file: /cvsroot/zsh/zsh/Src/utils.c,v
retrieving revision 1.268
diff -p -u -r1.268 utils.c
--- Src/utils.c	16 Apr 2012 11:26:10 -0000	1.268
+++ Src/utils.c	29 Apr 2012 18:39:39 -0000
@@ -3114,7 +3114,7 @@ wordcount(char *s, char *sep, int mul)
 	r = 1;
 	sl = strlen(sep);
 	for (; (c = findsep(&s, sep, 0)) >= 0; s += sl)
-	    if ((c && *(s + sl)) || mul)
+	    if ((c || mul) && (sl || *(s + sl)))
 		r++;
     } else {
 	char *t = s;
Index: Test/D04parameter.ztst
===================================================================
RCS file: /cvsroot/zsh/zsh/Test/D04parameter.ztst,v
retrieving revision 1.67
diff -p -u -r1.67 D04parameter.ztst
--- Test/D04parameter.ztst	22 Apr 2012 18:39:53 -0000	1.67
+++ Test/D04parameter.ztst	29 Apr 2012 18:39:39 -0000
@@ -1292,7 +1292,7 @@
 >in
 >it
 
-  foo="line:with::missing::fields:in:it"
+  foo="line:with::missing::fields:in:it:"
   print -l "${(@s.:.)foo}"
 0:Retention of empty fields in quoted splitting with "@"
 >line
@@ -1303,6 +1303,20 @@
 >fields
 >in
 >it
+>
+
+  str=abcd
+  print -l ${(s..)str}
+  print -l "${(s..)str}"
+0:splitting of strings into characters
+>a
+>b
+>c
+>d
+>a
+>b
+>c
+>d
 
   array=('%' '$' 'j' '*' '$foo')
   print ${array[(i)*]} "${array[(i)*]}"

-- 
Peter Stephenson <p.w.stephenson@ntlworld.com>
Web page now at http://homepage.ntlworld.com/p.w.stephenson/