ambrus has asked for the wisdom of the Perl Monks concerning the following question:

At Re: How's your Perl?, I mentioned that *a=*[;$a=5;print $a,$/; prints 0.

I discovered that $a="[";$$a=5;print$$a,$/; does the same thing. I think that both should die like $a=5;$[=$a;, that is with That use of $[ is unsupported at -e line 1.

Is this a bug in Perl? What do you think?

Replies are listed 'Best First'.
Re: *a=*[; bug?
by Abigail-II (Bishop) on Jan 09, 2004 at 12:52 UTC
    Note that the die on $[=$a happens at compile time (run it with -c to see), something you cannot determine of general cases of $$a.

    I think *a=*[;$a=5;print $a,$/; is in the same ballpark as perl -wle 'eval q!$[ = 5!; print $[', which also prints 0. After compiling, $[ is read-only, but, as a best effort, Perl ignores an assignment to $[ instead of dying.

    Abigail

[reply]
[d/l]
[select]
Re: *a=*[; bug?
by Roger (Parson) on Jan 09, 2004 at 13:26 UTC
    I have to say that according to the output of the following test script, the value of $[ is definitely not read-only at run-time. You can keep assigning a new number to $[ and change the subscript behaviour at run-time. However the use of $[ = $a is not supported. The Perl documentation clearly states that $[ is treated as a compiler directive and its use is discouraged. I suspect that Perl handles this special variable differently because of efficiency reasons (or because the whole idea of $[ was a hack in the first place afterall).
    my @elements = qw/ 0 1 2 3 4 5 6 7 8 9 /;

$[ = 0;
print $elements[1], "\n";

$[ = 1;
print $elements[1], "\n";

if ($[ == 1) {
  $[ = 0;
  print $elements[1], "\n";
}

# The following will fail at compile time
#   $a = 0;
#   $[ = $a;  # reset
#   print $elements[1], "\n";

    [download]

    And the output is -
    1
0
1

    [download]
[reply]
[d/l]
[select]
      It may look like runtime, but it's not. Just change $[ to $a see what happens to the optree by running the source through B::Concise. You'll notice that the $a version has real assignments like 
      k     <2> sassign vKS/2 ->l
i        <$> const(IV 0) s ->j
-        <1> ex-rv2sv sKRM*/1 ->k
j           <$> gvsv(*a) s ->k

      [download]
      where in the $[ version all you get is a mere 
      i     <;> nextstate(main 2 x.pl:4) v ->j

      [download]

      The effect of changing the array base index is achieved through virtual machine state, which is actually independent from the runtime value of $[. The value accessible through $[ is kept synched to the internal state but not vice versa.

      Full B::Concise output follows:

      Read more... (6 kB)

      Makeshifts last the longest.

[reply]
[d/l]
[select]

Back to Seekers of Perl Wisdom