Reimplement $[ as a module

This commit reimplements $[ using PL_check hooks, custom pp func-
tions and ties.

Outside of its compile-time use, $[ is now parsed as a simple varia-
ble, so function calls like foo($[) are permitted, which was not the
case with the former implementation removed by e1dccc0.  I consider
that a bug fix.

The ‘That use of $[ is unsupported’ errors are out of necessity
deferred to run-time and implemented by a tied $[.

Indices between 0 and the array base are now treated consistently, as
are indices between a negative array base and zero.  That, too, is
a bug fix.

MANIFEST

@@ -3507,6 +3507,23 @@ epoc/epocish.c		EPOC port
 epoc/epocish.h		EPOC port
 epoc/epoc_stubs.c	EPOC port
 epoc/link.pl		EPOC port link a exe
+ext/arybase/arybase.pm		For $[
+ext/arybase/arybase.xs		For $[
+ext/arybase/Makefile.PL		For $[
+ext/arybase/ptable.h		For $[
+ext/arybase/t/aeach.t		For $[
+ext/arybase/t/aelem.t		For $[
+ext/arybase/t/akeys.t		For $[
+ext/arybase/t/arybase.t		For $[
+ext/arybase/t/aslice.t		For $[
+ext/arybase/t/av2arylen.t	For $[
+ext/arybase/t/index.t		For $[
+ext/arybase/t/lslice.t		For $[
+ext/arybase/t/pos.t		For $[
+ext/arybase/t/scope_0.pm	For $[
+ext/arybase/t/scope.t		For $[
+ext/arybase/t/splice.t		For $[
+ext/arybase/t/substr.t		For $[
 ext/attributes/attributes.pm		For "sub foo : attrlist"
 ext/attributes/attributes.xs		For "sub foo : attrlist"
 ext/B/B/Concise.pm	Compiler Concise backend

Porting/Maintainers.pl

@@ -2120,6 +2120,7 @@ package Maintainers;
 	{
 	'MAINTAINER'	=> 'p5p',
 	'FILES'		=> q[
+				ext/arybase/
 				ext/XS-APItest/
 				lib/CORE.pod
 				lib/Config.t

ext/arybase/Makefile.PL

@@ -0,0 +1,16 @@
+use ExtUtils::MakeMaker;
+
+WriteMakefile(
+    NAME          => 'arybase',
+    VERSION_FROM  => 'arybase.pm',
+    ABSTRACT_FROM => 'arybase.pm',
+    realclean     => { FILES => "" },
+);
+
+# To work around nmake stupidity.  See rt.cpan.org #71847.
+package MY;
+sub ppd {
+ my $stuff = SUPER::ppd{} @_;
+ $stuff =~ s/ \$\[/ \$\$[/;
+ $stuff;
+}

ext/arybase/arybase.pm

@@ -0,0 +1,98 @@
+package arybase;
+
+our $VERSION = "0.01";
+
+require XSLoader;
+XSLoader::load(); # This returns true, which makes require happy.
+
+__END__
+
+=head1 NAME
+
+arybase - Set indexing base via $[
+
+=head1 SYNOPSIS
+
+    $[ = 1;
+    
+    @a = qw(Sun Mon Tue Wed Thu Fri Sat);
+    print $a[3], "\n";  # prints Tue
+
+=head1 DESCRIPTION
+
+This module implements Perl's C<$[> variable.  You should not use it
+directly.
+
+Assigning to C<$[> has the I<compile-time> effect of making the assigned
+value, converted to an integer, the index of the first element in an array
+and the first character in a substring, within the enclosing lexical scope.
+
+It can be written with or without C<local>:
+
+    $[ = 1;
+    local $[ = 1;
+
+It only works if the assignment can be detected at compile time and the
+value assigned is constant.
+
+It affects the following operations:
+
+    $array[$element]
+    @array[@slice]
+    $#array
+    (list())[$slice]
+    splice @array, $index, ...
+    each @array
+    keys @array
+    
+    index $string, $substring  # return value is affected
+    pos $string
+    substr $string, $offset, ...
+
+As with the default base of 0, negative bases count from the end of the
+array or string, starting with -1.  If C<$[> is a positive integer, indices
+from C<$[-1> to 0 also count from the end.  If C<$[> is negative (why would
+you do that, though?), indices from C<$[> to 0 count from the beginning of
+the string, but indices below C<$[> count from the end of the string as
+though the base were 0.
+
+Prior to Perl 5.16, indices from 0 to C<$[-1> inclusive, for positive
+values of C<$[>, behaved differently for different operations; negative
+indices equal to or greater than a negative C<$[> likewise behaved
+inconsistently.
+
+=head1 HISTORY
+
+Before Perl 5, C<$[> was a global variable that affected all array indices
+and string offsets.
+
+Starting with Perl 5, it became a file-scoped compile-time directive, which
+could be made lexically-scoped with C<local>.  "File-scoped" means that the
+C<$[> assignment could leak out of the block in which occurred:
+
+    {
+        $[ = 1;
+        # ... array base is 1 here ...
+    }
+    # ... still 1, but not in other files ...
+
+In Perl 5.10, it became strictly lexical.  The file-scoped behaviour was
+removed (perhaps inadvertently, but what's done is done).
+
+In Perl 5.16, the implementation was moved into this module, and out of the
+Perl core.  The erratic behaviour that occurred with indices between -1 and
+C<$[> was made consistent between operations, and, for negative bases,
+indices from C<$[> to -1 inclusive were made consistent between operations.
+
+=head1 BUGS
+
+Error messages that mention array indices use the 0-based index.
+
+C<keys $arrayref> and C<each $arrayref> do not respect the current value of
+C<$[>.
+
+=head1 SEE ALSO
+
+L<perlvar/"$[">, L<Array::Base> and L<String::Base>.
+
+=cut