Problem with associative arrays

Problem with associative arrays

[Sven Wischnowsky]

And this time it's for real:

  % typeset -A a
  % a[x*]=foo
  zsh: bad math expression: unbalanced stack
  % echo $a[x*]
  foo
  %

The problem is that isident() uses mathevalarg() to skip (!) over the
contents of a subscript (unless the subscript contains flags).

I don't have the time to fix this now since our LAN will soon be
upgraded but if noone else gives it a try I'll have a lokk at it.

Bye
 Sven


--
Sven Wischnowsky                         wischnow@informatik.hu-berlin.de

PATCH: 3.1.5-pws-3: (subscripts) Re: Problem with associative arrays

[Bart Schaefer]

On Dec 16,  3:31pm, Sven Wischnowsky wrote:
} Subject: Problem with associative arrays
}
} And this time it's for real:
} 
}   % typeset -A a
}   % a[x*]=foo
}   zsh: bad math expression: unbalanced stack
}   % echo $a[x*]
}   foo
}   %
} 
} The problem is that isident() uses mathevalarg() to skip (!) over the
} contents of a subscript (unless the subscript contains flags).

isident() seems to be rather inconsistent about strict checking of the
text inside the [ ].  Below is the most obvious fix (#ifdef'd for now).

This has the side-effect of making associative array syntax more like
ksh, in that you can put quoted strings inside the subscript:

    zsh% typeset -A foo
    zsh% foo['a b c']=cba
    zsh% print -l ${(kv)foo}
    a b c
    cba

However, this fails (because of the "balanced brackets" test):

    zsh% foo['a ] c']=cba
    zsh: not an identifier: foo[a ] c]

So a better test might be just strchr(ss, ']') and let other code report
subscripting errors after it knows what kind of array it's accessing.  I
also don't know how widely available strcspn() is; this might break the
build on some older platforms.

Also, this is a bit strange:

    zsh% echo $foo['a b c']

    zsh% echo ${foo[a b c]}	<-- note, must omit the quotes on 'a b c'
    cba

Probably the subscript needs to be untokenized somewhere.

Index: Src/params.c
===================================================================
--- params.c	1998/12/15 06:08:15	1.16
+++ params.c	1998/12/16 17:19:05
@@ -633,6 +633,7 @@
 	if (!iident(*ss))
 	    break;
 
+#if 0
     /* If this exhaust `s' or the next two characters *
      * are [(, then it is a valid identifier.         */
     if (!*ss || (*ss == '[' && ss[1] == '('))
@@ -642,6 +643,7 @@
      * definitely not a valid identifier.              */
     if (*ss != '[')
 	return 0;
+
     noeval = 1;
     (void)mathevalarg(++ss, &ss);
     if (*ss == ',')
@@ -650,6 +652,19 @@
     if (*ss != ']' || ss[1])
 	return 0;
     return 1;
+#else
+    /* If the next character is not [, then it is *
+     * definitely not a valid identifier.              */
+    if (!*ss)
+	return 1;
+    if (*ss != '[')
+	return 0;
+
+    /* Require balanced [ ] pairs */
+    for (s = ss; *(s += strcspn(++ss, "[]")) && s > ss; ss = s)
+	;
+    return (*ss == ']' && !s[1]);
+#endif
 }
 
 static char **garr;


-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com

Re: Problem with associative arrays

[Sven Wischnowsky]

Bart Schaefer wrote:

> ...
> 
> You have to use `typeset -A a` to get an associative array.  `set -A a`
> creates an ordinary array.  I realize this is confusing, which is why I
> originally used a different option letter for typeset, but compatibility
> with ksh requires -A.

Oops. Sorry.

Bye
 Sven


--
Sven Wischnowsky                         wischnow@informatik.hu-berlin.de

Re: Problem with associative arrays

[Bart Schaefer]

On Dec 14, 11:55am, Sven Wischnowsky wrote:
} Subject: Problem with associative arrays
}
} Subscripts for associative arrays are matheval()ed

No, they are not.  The code in params.c looks like

	if (PM_TYPE(v->pm->flags) == PM_HASHED) {
	    ...
	} else
	if (!(r = mathevalarg(s, &s)) || (isset(KSHARRAYS) && r >= 0))
	    ...

} which makes some problems:
} 
} % set -A a

You have to use `typeset -A a` to get an associative array.  `set -A a`
creates an ordinary array.  I realize this is confusing, which is why I
originally used a different option letter for typeset, but compatibility
with ksh requires -A.

} % a[foo]=bar
} % echo $a[baz]
} bar
} 
} There is no variable baz, so evaluating it yields zero and since
} numeric subscripts on associative arrays work, this yields the `first' 
} value.

This is all accurate, for ordinary arrays.

-- 
Bart Schaefer                                 Brass Lantern Enterprises
http://www.well.com/user/barts              http://www.brasslantern.com

Problem with associative arrays

[Sven Wischnowsky]

Hello

Subscripts for associative arrays are matheval()ed which makes some
problems:

% set -A a
% a[foo]=bar
% echo $a[baz]
bar

There is no variable baz, so evaluating it yields zero and since
numeric subscripts on associative arrays work, this yields the `first' 
value.

% a[x*]=frob
zsh: bad math expression: unbalanced stack
zsh: bad math expression: unbalanced stack

Yes, `x*' is an invalid math expression, but...

Bye
 Sven


--
Sven Wischnowsky                         wischnow@informatik.hu-berlin.de