Treatment of illegal indices

Treatment of illegal indices

[Bernd Eggink]

The way zsh treats illegal array indices is pretty ugly. It simply
replaces indices which are 0 (with KSH_ARRAYS unset) or < -length by 1.
This doesn't make much sense and can cause hard-to-detect errors.
Somebody who uses an index of -4711 in a 10-element array almost
certainly doesn't want to address element 1. 

Possible improvements could be:

a) Always print an error message if an illegal index is used.
b) Do nothing if an illegal index is used on the left side of an       
assignment, return an empty value otherwise.
c) Print an error message if an illegal index is used on the left      
side of an assignment, return an empty value otherwise.

Personally, I'd prefer c). And, although I think zsh has too many
options already, an additional option which toggles between a) and c)
could be useful.

What does everybody think? I'd volunteer to make a patch if I get
positive feedback.

- Bernd

-- 
Bernd Eggink
Regionales Rechenzentrum der Universitaet Hamburg
eggink@rrz.uni-hamburg.de
http://www.rrz.uni-hamburg.de/eggink/BEggink.html

Re: Treatment of illegal indices

[Bart Schaefer]

On Feb 22,  4:07pm, Bernd Eggink wrote:
} Subject: Treatment of illegal indices
}
} The way zsh treats illegal array indices is pretty ugly. It simply
} replaces indices which are 0 (with KSH_ARRAYS unset) or < -length by 1.
} This doesn't make much sense and can cause hard-to-detect errors.

I agree that this is counter-intuitive.

} c) Print an error message if an illegal index is used on the left      
} side of an assignment, return an empty value otherwise.

I believe this to be the best behavior.

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

Re: Treatment of illegal indices

[Peter Stephenson]

Bernd Eggink wrote:
> The way zsh treats illegal array indices is pretty ugly. It simply
> replaces indices which are 0 (with KSH_ARRAYS unset) or < -length by 1.
>
> Possible improvements could be:
> 
> c) Print an error message if an illegal index is used on the left      
> side of an assignment, return an empty value otherwise.

I think this is better; it's fairly standard for interpreted languages.
The `feature' that 0 is treated as 1 has been around for a long time,
but since KSH_ARRAYS it's probably time it went.  You are supposed to
be able to extend arrays by assignment, remember:  some time ago there
was an argument about whether

23456789=''

using up a lot of memory was a bug, and it was decided it wasn't.

-- 
Peter Stephenson <pws@ifh.de>       Tel: +39 50 911239
WWW:  http://www.ifh.de/~pws/
Gruppo Teorico, Dipartimento di Fisica
Piazza Torricelli 2, 56100 Pisa, Italy

Re: Treatment of illegal indices

[Andrew Main]

Bernd Eggink wrote:
>Personally, I'd prefer c). And, although I think zsh has too many
>options already, an additional option which toggles between a) and c)
>could be useful.

Sounds good.  The existing option UNSET should probably be used for
this purpose.

-zefram

Re: Treatment of illegal indices

[Peter Stephenson]

Andrew Main wrote:
> Bernd Eggink wrote:
> >Personally, I'd prefer c). And, although I think zsh has too many
> >options already, an additional option which toggles between a) and c)
> >could be useful.
> 
> Sounds good.  The existing option UNSET should probably be used for
> this purpose.

Actually, I just checked with ksh, and that's how it's implemented
there, so this is certainly the way to do it.

-- 
Peter Stephenson <pws@ifh.de>       Tel: +39 50 911239
WWW:  http://www.ifh.de/~pws/
Gruppo Teorico, Dipartimento di Fisica
Piazza Torricelli 2, 56100 Pisa, Italy

Re: Treatment of illegal indices

[Bernd Eggink]

I wrote:

> I'd volunteer to make a patch if I get positive feedback.

Arghh! This turns out to be real work. The code is somewhat messy, and I
found more bugs. This one, for example:

   x=(a b c d e)
   x[4,1]=X
   print $x

Result:
   a b c X b c d e

So we should settle on the meaning of a[i,j]=b when j < i. I suggest
that this always inserts b right before a[i] (provided j is legal).

Opinions?

- Bernd

-- 
Bernd Eggink
Regionales Rechenzentrum der Universitaet Hamburg
eggink@rrz.uni-hamburg.de
http://www.rrz.uni-hamburg.de/eggink/BEggink.html

Re: Treatment of illegal indices

[Andrew Main]

Bernd Eggink wrote:
>   x=(a b c d e)
>   x[4,1]=X
>   print $x
>
>Result:
>   a b c X b c d e

That's a deliberate feature.

-zefram

Re: Treatment of illegal indices

[Bernd Eggink]

Andrew Main wrote:
> 
> Bernd Eggink wrote:
> >   x=(a b c d e)
> >   x[4,1]=X
> >   print $x
> >
> >Result:
> >   a b c X b c d e
> 
> That's a deliberate feature.

Huh? What's the sense in that??

	Bernd

-- 
Bernd Eggink
Regionales Rechenzentrum der Universitaet Hamburg
eggink@rrz.uni-hamburg.de
http://www.rrz.uni-hamburg.de/eggink/BEggink.html

Re: Treatment of illegal indices

[Andrew Main]

Bernd Eggink wrote:
>Andrew Main wrote:
>> That's a deliberate feature.
>
>Huh? What's the sense in that??

No one has suggested anything better for it to do.  This behaviour
is consistent with the behaviour of foo[i,j]=bar for i<=j -- the new
element is immediately preceded by the old $foo[i-1], and followed by
the old $foo[j+1].

-zefram