Data::Alias 1.0 - Comprehensive set of aliasing operations
Note:
This module does not work on Win32 or AIX.
(It should however work on cygwin -- untested)
To install this module type the following:
perl Makefile.PL
make
make test
make install
Module documentation:
NAME
Data::Alias - Comprehensive set of aliasing operations
SYNOPSIS
use Data::Alias;
alias {
# aliasing instead of copying whenever possible
};
alias $x = $y; # alias $x to $y
alias @x = @y; # alias @x to @y
alias $x[0] = $y; # similar for array and hash elements
alias push @x, $y; # push alias to $y onto @x
$x = alias [ $y, $z ]; # construct array of aliases
alias my ($x, $y) = @_; # named aliases to arguments
alias { ($x, $y) = ($y, $x) }; # swap $x and $y
alias { my @t = @x; @x = @y; @y = @t }; # swap @x and @y
use Data::Alias qw/ alias copy /;
alias { copy $x = $y }; # force copying inside alias-BLOCK
use Data::Alias qw/ deref /;
my @refs = (\$x, \@y, \%z);
foo(deref @refs) # same as foo($x, @y, %z)
DESCRIPTION
Aliasing is the phenomenon where two different expressions actually
refer to the same thing. Modifying one will modify the other, and if you
take a reference to both, the two values are the same.
Aliasing occurs in Perl for example in for-loops and sub-calls:
for $var ($x) {
# here $var is an alias to $x
}
foo($y);
sub foo {
# here $_[0] is an alias to $y
}
Data::Alias is a module that allows you to apply "aliasing semantics" to
a section of code, causing aliases to be made whereever Perl would
normally make copies instead. You can use this to improve efficiency and
readability, when compared to using references.
The exact details of aliasing semantics are below under "DETAILS".
SYNTAX
alias EXPR | alias BLOCK
Exported by default.
Enables aliasing semantics within the expression or block. Returns an
alias to the expression, or the block's return value.
"alias" is context-transparent, meaning that whichever context it is
placed in (list, scalar, void), the expression/block is evaluated in the
same context.
copy EXPR | copy BLOCK
Restores normal (copying) semantics within the expression or block, and
makes a copy of the result value (unless in void context).
Like "alias", "copy" is context-transparent.
deref LIST
Accepts a list of references to scalars, arrays, or hashes. Applies the
applicable dereferencing operator to each. This means that:
deref $scalarref, $arrayref, $hashref
behaves like:
$$scalarref, @$arrayref, %$hashref
EXAMPLES
A common usage of aliasing is to make an abbreviation for an expression,
to avoid having to repeat that (possibly verbose or ugly) expression
over and over:
alias my $fi = $self->{FrobnitzIndex};
$fi = $fi > 0 ? $fi - $adj : $fi + $adj;
sub rc4 {
alias my ($i, $j, $S) = @_;
my $a = $S->[($i += 1) &= 255];
my $b = $S->[($j += $S->[$i]) &= 255];
$S->[(($S->[$j] = $a) + ($S->[$i] = $b)) & 255]
}
In the second example, the rc4 function updates its first two arguments
(two state values) in addition to returning a value.
Aliasing can also be used to avoid copying big strings. This example
would work fine without "alias" but would be much slower when passed a
big string:
sub middlesection ($) {
alias my $s = shift;
substr $s, length($s)/4, length($s)/2
}
You can also apply aliasing semantics to an entire block. Here this is
used to swap two arrays in O(1) time:
alias {
my @temp = @x;
@x = @y;
@y = @temp;
};
The "copy" function is typically used to temporarily reinstate normal
semantics, but can also be used to explicitly copy a value when perl
would normally not do so:
my $ref = \copy $x;
DETAILS
This section describes exactly what the aliasing semantics are of
operations. Anything not listed below has unaltered behaviour.
scalar assignment to variable or element.
Makes the left-side of the assignment an alias to the right-side
expression, which can be anything.
alias my $lexvar = $foo;
alias $pkgvar = $foo;
alias $array[$i] = $foo;
alias $hash{$k} = $foo;
An attempt to do alias-assignment to an element of a tied (or
"magical") array or hash will result in a "Can't put alias into tied
array/hash" error.
scalar assignment to dereference
If $ref is a reference or undef, this simply does "$ref = \$foo".
Otherwise, the indicated package variable (via glob or symbolic
reference) is made an alias to the right-side expression.
alias $$ref = $foo;
scalar assignment to glob
Works mostly the same as normal glob-assignment, however it does not
set the import-flag. (If you don't know what this means, you
probably don't care)
alias *glob = $reference;
scalar assignment to anything else
Not supported.
alias substr(...) = $foo; # ERROR!
alias lvalsub() = $foo; # ERROR!
conditional scalar assignment
Here $var (and $var2) are aliased to $foo if the applicable
condition is satisfied. $bool and $foo can be any expression. $var
and $var2 can be anything that is valid on the left-side of an
alias-assignment.
alias $bool ? $var : $var2 = $foo;
alias $var &&= $foo;
alias $var ||= $foo;
alias $var //= $foo; # (perl 5.9.x or later)
whole aggregate assignment
These alias entire aggregates (arrays or hashes), not merely their
contents. This means for example that "\@lexarray == \@foo".
alias my @lexarray = @foo;
alias my %lexhash = %foo;
alias @pkgarray = @foo;
alias %pkghash = %foo;
Making the left-side a dereference is also supported:
alias @$ref = @foo;
alias %$ref = %foo;
and analogously to assignment to scalar dereference, these will
change $ref to reference the aggregate, if $ref was undef or already
a reference. If $ref is a string or glob, the corresponding package
variable is aliased.
If the right-side expression is not an aggregate (of the same type),
then a new one is implicitly constructed. This means:
alias my @array = ($x, $y, $z);
alias my %hash = (x => $x, y => $y);
is translated to:
alias my @array = @{ [$x, $y, $z] };
alias my %hash = %{ {x => $x, y => $y} };
If you want to merely replace the contents of the aggregate with
aliases to the contents of another aggregate, but not alias the
aggregates themselves, you can force list-assignment by
parenthesizing the left side, see below.
list assignment
List assignment is any assignment where the left-side is an
array-slice, hash-slice, or list in parentheses. This behaves
essentially like many scalar assignments in parallel.
alias my (@array) = ($x, $y, $z);
alias my (%hash) = (x => $x, y => $y);
alias my ($x, $y, @rest) = @_;
alias @x[0, 1] = @x[1, 0];
Any scalars that appear on the left side must be valid targets for
scalar assignment. When an array or hash appears on the left side,
normally as the last item, its contents are replaced by the list of
all remaining right-side elements. "undef" can also appear on the
left side to skip one corresponding item in the right-side list.
Beware when putting a parenthesized list on the left side. Just like
Perl parses "print (1+2)*10" as "(print(1+2))*10", it would parse
"alias ($x, $y) = ($y, $x)" as "(alias($x, $y)) = ($y, $x)" which
does not do any aliasing, and results in the "Useless use of alias"
warning, if warnings are enabled.
To circumvent this issue, you can either one of the following:
alias +($x, $y) = ($y, $x);
alias { ($x, $y) = ($y, $x) };
Anonymous aggregate constructors
Return a reference to a new anonymous array or hash, populated with
aliases. This means that for example "\$hashref->{x} == \$x".
my $arrayref = alias [$x, $y, $z];
my $hashref = alias {x => $x, y => $y};
Note that this also works:
alias my $arrayref = [$x, $y, $z];
alias my $hashref = {x => $x, y => $y};
but this makes the lhs an alias to the temporary, and therefore
read-only, reference made by "[]" or "{}". Therefore later attempts
to assign to $arrayref or $hashref results in an error. The
anonymous aggregate that is referenced behaves the same in both
cases obviously.
Array insertions
These work as usual, except the inserted elements are aliases.
alias push @array, $foo;
alias unshift @array, $foo;
alias splice @array, 1, 2, $foo;
An attempt to do any of these on tied (or "magical") array will
result in a "Can't push/unshift/splice alias onto tied array" error.
Returning an alias
Returns aliases from the current "sub" or "eval". Normally this only
happens for lvalue subs, but "alias return" can be used in any sub.
Subroutines and evaluations
Placing a subroutine or "eval STRING" inside "alias" causes it to be
compiled with aliasing semantics entirely. Additionally, the return
from such a sub or eval, whether explicit using "return" or
implicitly the last statement, will be an alias rather than a copy.
alias { sub foo { $x } };
my $subref = alias sub { $x };
my $xref1 = \foo;
my $xref2 = \alias eval '$x';
my $xref3 = \$subref->();
Explicitly returning an alias can also be done using "alias return"
inside any subroutine or evaluation.
sub foo { alias return $x; }
my $xref = \foo;
Localization
Use of local inside "alias" usually behaves the same as local does
in general, however there is a difference if the variable is tied:
in this case, Perl doesn't localize the variable at all but instead
preserves the tie by saving a copy of the current value, and
restoring this value at end of scope.
alias local $_ = $string;
The aliasing semantics of "local" avoids copying by always
localizing the variable itself, regardless of whether it is tied.
IMPLEMENTATION
This module does not use a source filter, and is therefore safe to use
within eval STRING. Instead, Data::Alias hooks into the Perl parser, and
replaces operations within the scope of "alias" by aliasing variants.
For those familiar with perl's internals: it triggers on a ck_rv2cv
which resolves to the imported "alias" sub, and does a parser hack to
allow the "alias BLOCK" syntax. When the ck_entersub is triggered that
corresponds to it, the op is marked to be found later. The actual work
is done in a peep-hook which is installed on the ck_rv2cv trigger, and
processes the marked entersub and its children, replacing the pp_addrs
with aliasing replacements. The peep hook will also take care of any
subs defined within the lexical (but not dynamical) scope between the
ck_rv2cv and the ck_entersub. When no marked entersubs remain, the
Data::Alias-peeper stops and is deinstalled.
KNOWN ISSUES
Lexical variables
When aliasing existing lexical variables, the effect is limited in
scope to the current subroutine and any closures create after the
aliasing is done, even if the variable itself has wider scope. While
partial fixes are possible, it cannot be fixed in any reliable or
consistent way, and therefore I'm keeping the current behaviour.
Platform support
Some platforms, notably Windows and AIX, will give link errors when
attempting to build Data::Alias. This cannot be fixed. When using
Windows, you can use perl inside cygwin ()
instead, where Data::Alias does work.
ACKNOWLEDGEMENTS
Specials thanks go to Elizabeth Mattijsen, Juerd Waalboer, and other
members of the Amsterdam Perl Mongers, for their valuable feedback.
AUTHOR
Matthijs van Duin
Copyright (C) 2003, 2004, 2006 Matthijs van Duin. All rights reserved.
This program is free software; you can redistribute it and/or modify it
under the same terms as Perl itself.