# vim: ts=2 sw=2 expandtab package Data::Transform::Reference; use strict; use Data::Transform; use vars qw($VERSION @ISA); $VERSION = '0.01'; @ISA = qw(Data::Transform); use Carp qw(croak); sub INPUT () { 0 } sub BUFFER () { 1 } sub SERIALIZE () { 2 } sub DESERIALIZE () { 3 } =pod =head1 NAME Data::Transform::Reference - freeze and thaw arbitrary Perl data =head1 SYNOPSIS use YAML; use Data::Transform::Reference; my $filter = Data::Transform::Reference->new( serialize => YAML->can('Dump'); deserialize => YAML->can('Load'); ); ... my $string = $filter->put($some_var); ... my $other_var = $filter->get($serialized_var); =head1 DESCRIPTION Data::Transform::Reference allows programs to send and receive arbitrary Perl data structures without worrying about a line protocol. Its put() method serializes Perl data into a byte stream suitable for transmission. get_one() parses the data structures back out of such a stream. =head1 METHODS Data::Transform::Reference implements the standard Data::Transform API. Only the differences are documented here. =cut =head2 new new() creates and initializes a Data::Transform::Reference object. It requires the following parameters: =over 2 =item serializer A code ref used to serialize data. Good candidates for this are nfreeze() from L or Dump() from a YAML implementation. =item deserializer A code ref used to de-serialize data. Good candidates for this are thaw() from L or Load() from a YAML implementation. =back Both code references are expected to accept a single parameter containing the data on which to act on. =cut sub new { my $type = shift; croak "$type requires an even number of arguments" if (@_ & 1); my %param = @_; croak "$type requires a serialize parameter" unless defined $param{'serialize'}; croak "$type: serialize parameter must be a CODE reference" unless (ref $param{'serialize'} eq 'CODE'); croak "$type requires a deserialize parameter" unless defined $param{'deserialize'}; croak "$type: deserialize parameter must be a CODE reference" unless (ref $param{'deserialize'} eq 'CODE'); my $self = bless [ [], # INPUT '', # BUFFER $param{'serialize'}, # FREEZE $param{'deserialize'}, # THAW ]; return bless $self, $type; } sub clone { my $self = shift; my $new = [ [], '', $self->[SERIALIZE], $self->[DESERIALIZE], ]; return bless $new, ref $self; } sub get_pending { my $self = shift; my @ret; @ret = @{$self->[INPUT]}; if (length $self->[BUFFER]) { unshift @ret, $self->[BUFFER]; } return @ret ? \@ret : undef; } sub _handle_get_data { my ($self, $data) = @_; if (defined $data) { $self->[BUFFER] .= $data; } # Need to check lengths in octets, not characters. use bytes; if ($self->[BUFFER] =~ /^(\d+)\0/ and length($self->[BUFFER]) >= $1 + length($1) + 1 ) { substr($self->[BUFFER], 0, length($1) + 1) = ""; my $return = substr($self->[BUFFER], 0, $1); substr($self->[BUFFER], 0, $1) = ""; return $self->[DESERIALIZE]->($return); } return; } sub _handle_put_data { my ($self, $reference) = @_; # Need to check lengths in octets, not characters. use bytes; my $frozen = $self->[SERIALIZE]->($reference); return length($frozen) . "\0" . $frozen; } 1; __END__ =head1 SEE ALSO Please see L for documentation regarding the base interface. =head1 CAVEATS It's important to use identical serializers on each end of a connection. Even different versions of the same serializer can break data in transit. Most (if not all) serializers will rebless data at the destination, but many of them will not load the necessary classes to make their blessings work. =head1 AUTHORS & COPYRIGHTS The original Reference filter was contributed by Artur Bergman, with changes by Philip Gwyn. Martijn van Beers simplified the API when starting Data::Transform =cut