PATCH: 3.1.9-dev-8: Re: Word splitting in zsh

["Bart Schaefer" <schaefer@xxxxxxxxxxxxxxxxxxxxxxx>]

On Feb 9, 10:34am, Andrej Borsenkow wrote:
}
} > zagzig% echo $ZSH_VERSION
} > 3.0.8
} > zagzig% setopt shwordsplit
} > zagzig% print -l ${1+"$@"}
} > a1
} > a2
} > a3
} > b
} > c
} > zagzig%
} >
} 
} > Well, that's not quite right,
} 
} It is simply wrong.
} 
}                                             but 3.1.9-dev-8 is even worse:
} >
} > zagzig% echo $ZSH_VERSION
} > 3.1.9-dev-8
} > zagzig% set "a1 a2 a3" b c
} > zagzig% print -l ${1+"$@"}
} > a1 a2 a3 b c						<-- Yipes!
} 
} Well, this is "correct" *zsh* behaviour. The part after `+' is a word -
} not array. And is taken as single word and is never splitted.

On Feb 9,  6:18pm, Bart Schaefer wrote:
}
} Yes, your analysis is correct; the bug is that there should not be a scalar
} context there.  The effect of ${NAME:+WORD} and ${NAME:-WORD} should be to
} replace the context of NAME with the context of WORD in the event that NAME
} is (or not) set.  Put another way, the type of NAME should only determine
} the type of the whole ${...} when NAME really is being substituted.

This turned out to be remarkably easy; patch and explanation below.

On Feb 9, 10:34am, Andrej Borsenkow wrote:
} 
} That is just because of above. The structure of WORD in ${name+WORD} is not
} remebered. But note the same bug again:
} 
} bor@itsrm2% set 'a b c' 1 2
} bor@itsrm2% IFS=: print -l ${1+"$@"}
} a b c 1 2
} bor@itsrm2% setopt shwordsplit
} bor@itsrm2% IFS=: print -l ${1+"$@"}
} a
} b
} c
} 1
} 2
} 
} IFS value is silently ignored.

This is actually something entirely different, and is not really a bug at
all.  The problem with Andrej's example is that ${1+"$@"} is expanded
before the assignment to IFS takes place.  If you replace the `print's
above with

    args() { print -l $* ++ ${1+"$@"} }
    IFS=: args "$@"

then you'll see that IFS is handled correctly (modulo the other bug).

OK, here's the patch; I won't commit this until I get some feedback on it.  

The solution is simply (1) pass &aval into multsub() so that it returns
an array when the word being expanded would expand to an array; (2) pass
the current value of `spbreak' through multsub() in case of ${=...}; (3)
turn off word splitting after multsub() returns, because multsub() will
have already performed quoting and splitting as appropriate.

This applies only to ${...+...} and ${...-...}, with or without the `:'.
As I mentioned, there would be some benefit to also doing this for the
${...=...} forms, because then things like ${(A)foo="$bar[@]"} would work
independently of the setting of shwordplit.

There's some question in my mind of whether ${~...} and ${^...} should be
treated this way, too; e.g. (this is what does NOT happen):

zsh% print -l ${^1+"x$@x"}
xa1:a2 a3x
xbx
xcx

That `^' and `~' are not propagated now could even be an argument for not
propagating `=' either, but that would cause a larger change in behavior
in the instances where the right-hand-side of the `+'/`-' is not quoted
(and the shwordsplit option is not set).

I've also left unchanged the behavior of ${(s( ))...}, ${(z)...}, etc.,
because those constructs aren't possible in Bourne shell and in general
the expansion flags take precedence over everything else -- and because
there's a lot more potential for breakage when messing with those.

Index: Src/subst.c
===================================================================
--- Src/subst.c	2001/01/16 17:18:12	1.72
+++ Src/subst.c	2001/02/10 18:31:20
@@ -749,7 +749,9 @@
     int chkset = 0;
     int vunset = 0;
     int wantt = 0;
-    int spbreak = isset(SHWORDSPLIT) && !ssub && !qt;
+    static int mult_spbreak, mult_shwsplit;
+    int spbreak = (!ssub && !qt &&
+		   (mult_spbreak? mult_shwsplit : isset(SHWORDSPLIT)));
     char *val = NULL, **aval = NULL;
     unsigned int fwidth = 0;
     struct value vbuf;
@@ -1378,8 +1380,14 @@
 	/* Fall Through! */
 	case '-':
 	    if (vunset) {
+		int shws = mult_shwsplit, spbr = mult_spbreak;
+		mult_shwsplit = spbreak;
+		mult_spbreak = 1;
 		val = dupstring(s);
-		multsub(&val, NULL, &isarr, NULL);
+		multsub(&val, (aspar ? NULL : &aval), &isarr, NULL);
+		mult_shwsplit = shws;
+		mult_spbreak = spbr;
+		spbreak = 0;
 		copied = 1;
 	    }
 	    break;

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

Zsh: http://www.zsh.org | PHPerl Project: http://phperl.sourceforge.net