use 5.006;
package WWW::Scripter;
our $VERSION = '0.030';
use strict; use warnings; no warnings qw 'utf8 parenthesis bareword';
use CSS'DOM'Interface;
use Encode qw'encode decode';
use Exporter 5.57 'import';
use HTML::DOM 0.045; # weaken_response
use HTML::DOM::EventTarget 0.053; # DOMAttrModified with correct type and
use HTML::DOM::Interface 0.019 ':all'; # cancellability
use HTML::DOM::View 0.018;
use HTTP::Headers::Util 'split_header_words';
use HTTP::Response;
use HTTP::Request;
use Scalar::Util 1.09 qw 'blessed weaken reftype';
use List'Util 'sum';
use LWP::UserAgent;
use Time::HiRes 'time';
BEGIN {
require constant;
require WWW::Mechanize;
VERSION WWW::Mechanize $LWP::UserAgent::VERSION >= 5.815 ? 1.52 : 1.2;
# Version 1.52 is necessary for LWP 5.815 compatibility. Version 1.2 is
# needed otherwise for its handling of cookie jars during cloning.
import constant Mech => 'WWW::Mechanize';
}
BEGIN {
if(eval { require Hash::Util::FieldHash }) {
import Hash::Util::FieldHash qw < fieldhash fieldhashes >;
} else {
require Tie::RefHash::Weak;
VERSION Tie::RefHash::Weak 0.08; # fieldhash
import Tie::RefHash::Weak qw < fieldhash fieldhashes >;
}
}
our @ISA = (Mech, qw( HTML::DOM::View HTML::DOM::EventTarget ));
eval <<'' unless exists &UNIVERSAL'DOES;
sub DOES {
goto &{$_[0]->can("SUPER::DOES")||$_[0]->can("isa")}
}
our @EXPORT_OK = qw/abort/;
our %EXPORT_TAGS = (
all => \@EXPORT_OK,
);
# Fields that we don’t want fiddled with when the page stack is
# manipulated:
fieldhashes \my( %scriptable, %script_handlers, %scrn,
%class_info, %navi );
# ~~~ Actually, most of these can be eliminated, since we can store them
# directly in the object, as we are not doing that cloning that Mech
# used to do between pages.
# Fields keyed by document:
fieldhashes \my( %timeouts, %timers, %frames, %evtg, %status, %dstatus );
fieldhash my %document; # keyed by response — we actually use
# HTML::DOM::View’s storage for the current doc,
# but this field hash is necessary when we return
# to a page.
# These are used to create a link between a WWW::Mechanize::(Image|Link)
# object and the DOM equivalent.
fieldhash my %dom_obj;
# ------------- Mech overrides (or does it?) ------------- #
sub new {
my $class = shift;
my %args = @_;
exists $args{max_docs}
and $args{stack_depth} = -1+delete$args{max_docs};
my $max_history = delete $args{max_history};
my $self = $class->SUPER::new(%args);
$$self{Scripter_max_hist} = $max_history;
$script_handlers{$self} = {};
$scriptable{$self} = 1;
$self->{page_stack} = WWW'Scripter'History->new( $self );
weaken(my $self_fc = $self); # for closures
$class_info{$self} = [
\(%HTML::DOM'Interface, %CSS'DOM'Interface, our%Interface), {
'WWW::Scripter::Image' => "Image",
Image => {
_constructor => sub {
my $i = $self_fc->document->createElement('img');
@_ and $i->attr('width',shift);
@_ and $i->attr('height',shift);
$i
}
},
}
];
unless(exists $args{agent}) {
$self->agent("WWW::Scripter/$VERSION");
}
# I would like to avoid doing this when it is not necessary, but
# the alternative would require overriding HTML::DOM::View’s
# document method, and that might slow things down more, since
# document is called more often than new Scripter objects
# are created.
_initial_page($self);
$self;
}
sub _initial_page {
my $req = new HTTP::Request 'GET', 'about:blank';
my $res = new HTTP::Response 200, OK => [
'content-length' => 0,
'content-type' => 'text/html',
], '';
$res->request($req);
shift->_update_page(
$req, $res
);
}
sub clone {
my $clone = (my $self = shift)->SUPER::clone(@_);
$$_{$clone}=$$_{$self} for \(
%scriptable,%script_handlers
);
$class_info{$clone} = [@{$class_info{$self}}];
$clone->{handlers} = $self->{handlers};
$clone->{page_stack} = WWW'Scripter'History->new($clone);
delete @$clone{<Scripter_loc Scripter_nm>};
$clone->_clone_plugins;
$clone;
}
sub title { (shift->document||return)->title(@_) }
sub content {
my $self = shift;
if($self->is_html && $self->document) {
my %parms = @_;
my $cs = (my $doc = $self->document)->charset;;
if(exists $parms{format} && $parms{format} eq 'text') {
my $text = $doc->documentElement->as_text;
return defined $cs ? encode $cs, $text : $text;
}
my $content = $doc->innerHTML;
$content = encode $cs, $content if defined $cs;
$self->{content} = $content; # banana
}
$self->SUPER::content(@_);
}
#sub discontent { ... }
# Some parts of this were taken straight from WWW::Mechanize.
sub follow_link {
no warnings 'redefine';
my $self = shift;
my %parms = ( n=>1, @_ );
if ( $parms{n} eq 'all' ) {
delete $parms{n};
$self->warn( q{follow_link(n=>"all") is not valid} );
}
my $link = $self->find_link(%parms);
if($link and tag $link =~ '^a') {
my $follow;
my $dom_link = $dom_obj{$link};
$dom_link->trigger_event('click',
# We used to have simply DOMActivate_default => ...
# but that did absolutely nothing, since the
# *_default arguments apply solely to the current
# event (which is a click event). So we have
# to override HTML::DOM::Element’s click_default
# to trigger the DOMActivate event with the
# DOMActivate_default argument. And, no, some sort
# of localisation mechanism would not do instead,
# because event handlers could click other links
# (or even this one again), which events should
# remain unaffected by this *_default override.
# ~~~ Or should they???
click_default => sub {
$dom_link->trigger_event('DOMActivate',
DOMActivate_default => sub { ++$follow }
)
}
);
return unless $follow;
return ($self->find_target($dom_link->target)||$self)
->get($link->url);
}
else {
$self->die(
'Link not found: ',
join ", ", map "$_ => '$parms{$_}'", sort keys %parms
)
if $self->{autocheck};
}
Scripter_plit:
}
sub request {
for (my $foo) { # protect against tied $_
my $self = shift;
return unless defined(my $request = shift);
$request = $self->_modify_request( $request );
my $meth = $request->method;
my $orig_uri = $request->uri;
my $new_uri;
if ((my $path = $orig_uri->path) =~ s-^(/*)/\.\./-$1||'/'-e) {
0while $path =~ s\\$1||'/'\e;
($new_uri = $orig_uri->clone)->path($path)
}
my $skip_fetch;
if(defined($orig_uri->fragment)) {
($new_uri ||= $orig_uri->clone)->fragment(undef);
# Skip fetching the URL if it is the same (and there is a fragment).
# We don’t need to strip the fragment from $self->uri before compari-
# son as that always contains the actual URL sent in the request.
$meth eq "GET" and $new_uri->eq($self->uri) and ++$skip_fetch;
}
if ($new_uri) {
$request->uri($new_uri);
}
my $response;
if($skip_fetch) {
$response = $self->response;
}
else {
Scripter_REQUEST: {
Scripter_ABORT: {
$response = $self->_make_request( $request, @_ );
last Scripter_REQUEST;
}
return 1
}
}
if ( $meth eq 'GET' || $meth eq 'POST' ) {
$self->get_event_listeners('unload') and
$self->trigger_event('unload'),
$self->{page_stack}->_delete_res;
$self->{page_stack}->${\(
$self->{Scripter_replace} ? '_replace' : '_add'
)}($request, $response, $orig_uri);
}
return $self->_update_page($request, $response);
}
}
# Protect against tied $_
sub get { return SUPER::get{@_} for my $foo }
sub put { return SUPER::put{@_} for my $foo }
sub post { return SUPER::post{@_} for my $foo }
sub head { return SUPER::head{@_} for my $foo }
# The only difference between this one and Mech is the args to
# decoded_content. I.e., this is the way Mech *used* to work.
sub _update_page {
my ($self, $request, $res) = @_;
$self->{req} = $request;
$self->{redirected_uri} = $request->uri->as_string;
$self->{res} = $res;
$self->{status} = $res->code;
$self->{base} = $res->base;
$self->{ct} = $res->content_type || '';
if ( $res->is_success ) {
$self->{uri} = $self->{redirected_uri};
$self->{last_uri} = $self->{uri};
}
if ( $res->is_error ) {
if ( $self->{autocheck} ) {
$self->die( 'Error ', $request->method, 'ing ', $request->uri, ': ', $res->message );
}
}
$self->_reset_page;
# Try to decode the content. Undef will be returned if there's nothing to decompress.
# See docs in HTTP::Message for details. Do we need to expose the options there?
my $content = $res->decoded_content(charset => "none");
$content = $res->content if (not defined $content);
$content .= &{\&{Mech."::_taintedness"}};
if (
!defined $$self{Scripter_dumb} || $$self{Scripter_dumb}
and $self->is_html
) {
$res = $self->update_html($content);
}
else {
$self->{content} = $content;
$self->document(undef);
}
return $res;
} # _update_page
sub _fetch_url {
my ($self) = @'_;
my $fetcher = $self->{Scripter_f}
||= do {
(
my $clone = $self->clone->clear_history(1)
)->dom_enabled(0);
$clone->max_history(1);
$clone;
};
$fetcher->{last_uri} = $self->{uri};
require URI;
my $base = $self->base;
$_[1] = URI->new_abs( $_[1], $base )
if $base;
$fetcher->get($_[1]);
}
sub update_html {
my ($self,$src) = @_;
# Restore an existing document (in case we are coming back from
# another page).
my $res = $self->{res};
if(my $doc = $document{$res}) {
$self->document($doc);
$self->{form} = ($self->{forms} = $doc->forms)->[0];
return $res;
}
my $life_raft = $self;
weaken($self);
$self->document($document{$res} = my $tree = new HTML::DOM
response => $res,
weaken_response => 1,
cookie_jar => $self->cookie_jar);
$tree->error_handler(sub{$self->warn($@)});
$tree->default_event_handler_for( link => sub {
my $link = shift->target;
($self->find_target($link->target)||$self)
->get($link->href)
});
$tree->default_event_handler_for( submit => sub {
my $form = shift->target;
($self->find_target($form->target)||$self)
->request($form->make_request);
});
if(%{$script_handlers{$self}}) {
my $script_type = $res->header(
'Content-Script-Type');
defined $script_type or $tree->elem_handler(meta =>
sub {
my($tree, $elem) = @_;
no warnings 'uninitialized';
return unless lc $elem->attr('http-equiv')
eq 'content-script-type';
$script_type = $elem->attr('content');
});
$tree->elem_handler(script => sub {
return unless $scriptable{$self};
my($tree, $elem) = @_;
my $lang = $elem->attr('type');
defined $lang
or $lang = $elem->attr('language');
defined $lang or $lang = $script_type;
my $uri;
my($inline, $code, $line) = 0;
if($uri = $elem->attr('src')) {
my $res = _fetch_url($self, $uri);
$res->is_success or do {
my $url = $self->uri;
my $offset = $elem->content_offset;
if (!defined $offset) {
$url .= ' (generated HTML)';
}
else {
$url .= ' line '
. _line_no($src,$offset);
}
$self->warn("couldn't get script $uri: "
. $res->status_line . " at $url"
),
return;
};
# Find out the encoding:
my $cs = {
map @$_,
split_header_words $res->header(
'Content-Type'
)
}->{charset};
$code = decode $cs||$elem->charset
||$tree->charset||'latin1',
$res->decoded_content(charset=>'none');
$line = 1;
}
else {
$code = ($elem->firstChild||return)->data;
++$inline;
$uri = $self->uri;
if(defined(
my $offset = $elem->content_offset
)) {
$line = _line_no(
$src,$elem->content_offset
);
}
else { $uri .= " (generated HTML)" }
};
length $code or return; # optimisation
my $h = $self->_handler_for_lang($lang);
$h && $h->eval($self, $code,
$uri, $line, $inline);
$@ and $self->warn($@);
});
$tree->elem_handler(noscript => sub {
return unless $scriptable{$self};
$_[1]->detach#->delete;
# ~~~ delete currently stops it from work-
# ing; I need to looook into this.
});
$tree->event_attr_handler(sub {
return unless $scriptable{$self};
my($elem, $event, $code, $offset) = @_;
my $lang = $elem->attr('language');
defined $lang or $lang = $script_type;
my $uri = $self->uri;
my $line = defined $offset ? _line_no(
$src, $offset
) : undef;
local *@;
if(my $h = $self->_handler_for_lang($lang))
{
my $ret = $h->event2sub(
$self,$elem,$event,$code,$uri,$line
);
$@ and $self->warn($@);
return $ret;
}
});
}
$tree->elem_handler(noscript => sub {
return if $scriptable{$self} && %{$script_handlers{$self}};
$_[1]->replace_with_content->delete;
# ~~~ why does this need delete?
});
if($self->{Scripter_i}){
$tree->elem_handler(img => my $img_cb = sub {
return unless defined (my $src = $_[1]->attr('src'));
my $res = _fetch_url($self, $src);
defined $self->{Scripter_ih} &&
$self->{Scripter_ih}($self,$_[1],$res);
});
$tree->elem_handler(input => sub {
return unless $_[1]->type eq 'image';
goto &$img_cb;
});
$tree->default_event_handler(sub {
return unless (my $event = shift)->type eq 'DOMAttrModified';
return unless (my $target = target $event)->tag=~/^i(mg|nput)\z/;
return if $1 eq 'nput' && $target->type ne 'image';
&$img_cb(undef, $target);
});
}
$tree->defaultView(
$self
);
$tree->event_parent($self);
$tree->set_location_object($self->location);
$tree->elem_handler(iframe => my $frame_handler = sub {
my ($doc,$elem) = @_;
my $subwin = $self->clone->clear_history(1);
if(defined(my $name = attr $elem 'name')) {
name $subwin $name
}
$elem->contentWindow($subwin);
$subwin->_set_parent(my $parent = $doc->defaultView);
length(my $src = $elem->src) or return;
$subwin->get(new_abs URI $src, $parent->base);
});
$tree->elem_handler(frame => $frame_handler);
# Find out the encoding:
my $cs = {
map @$_,
split_header_words $res->header('Content-Type')
}->{charset};
$cs or $res->can('content_charset')
and $cs = (
$LWP::UserAgent::VERSION <= 5.834 && local *_,
$res->content_charset
);
$tree->charset($cs||'iso-8859-1');
# banana
$self->{form} = undef;
$self->{forms} = $tree->forms;
$tree->write(defined $cs ? decode $cs, $src : $src);
$tree->close;
# This used to trigger the load event on the body element (which
# conformed to HTML 5 at the time [10 June 2008 draft]), but which
# was not fully compatible with any existing browser. HTML 5
# changed to what Firefox and Safari did (some time before Septem-
# ber, 2009), which is what we now have here. (It still doesn’t
# quite make sense, as the document is not actually the target.)
$self->trigger_event('load', target => $tree);
# banana
$self->{form} ||= $self->{forms}[0];
return $self->{res};
}
# Not an override, but used by update_html
sub _handler_for_lang {
my ($self,$lang) = @_;
if(defined $lang) {
while(my($lang_re,$handler) = each
%{$script_handlers{$self}}) {
next if $lang_re eq 'default';
$lang =~ $lang_re and
# reset iterator:
keys %{$script_handlers{$self}},
return $handler;
}
}
return $script_handlers{$self}{default} || ();
}
# Not an override, but used by update_html
sub _line_no {
my ($src,$offset) = @_;
defined $offset or Carp::cluck;
return 1 + (() =
substr($src,0,$offset)
=~ /\cm\cj?|[\cj\x{2028}\x{2029}]/g
);
}
my %link_tags = (
a => 'href',
area => 'href',
frame => 'src',
iframe => 'src',
link => 'href',
meta => 'content',
);
# ~~~ This ends up creating a new WSL object every time we come back to the
# same page. We need a way to make this more efficient. The same goes
# for images.
sub _extract_links {
my $self = shift;
my @links;
if (my $doc = $self->document) {
tie @links, WWW'Scripter'Links:: =>
HTML::DOM::NodeList::Magic->new(
sub { grep {
my $tag = tag $_;
no warnings 'uninitialized';
exists $link_tags{$tag}
and defined $_->attr($link_tags{$tag})
and $tag ne 'meta'
|| lc $_->attr('http-equiv') eq 'refresh'
} $doc->descendants }, $doc
);
}
# banana
$self->{links} = \@links;
$self->{_extracted_links} = 1;
return;
}
sub _extract_images {
my $doc = (my $self= shift)->document;
my $list = HTML::DOM::NodeList::Magic->new(
sub { grep tag $_ =~ /^i(?:mg|nput)\z/,
$doc->descendants },
$doc
);
tie my @images, WWW'Scripter'Images:: => $list;
# banana
$self->{images} = \@images;
$self->{_extracted_images} = 1;
return;
}
sub back {
shift->{page_stack}->go(-1)
}
sub submit {
if(defined wantarray) {
# We have to return the response object if a request was made, so we
# override the default event handler for this particular case.
my $go_for_it;
(my $form = $_[0]->current_form)->trigger_event(
'submit',
submit_default => sub { ++$go_for_it }
);
$go_for_it
? ($_[0]->find_target($form->target)||$_[0])
->request($form->make_request)
: ()
}
else {
shift->current_form->submit
}
}
sub base {
my $self = shift;
my $base = ($self->document || return SUPER'base $self @_)->base;
if($base eq 'about:blank' and (my $parent = $self->parent) != $self) {
return $parent->base;
}
length $base ? $base : undef;
}
sub click { # This duplicates a lot of code from WWW::Mechanize::click,
# HTML::DOM::Element::Form::click and HTML::DOM::Ele-
# ment::Input, but I don’t see a way around it.
if(defined wantarray) {
# We have to return the response object if a request was made, so we
# override the default event handler for this particular case.
my ($self, $button, $x, $y) = @_;
# From HTML::DOM::Element::Form (ultimately from HTML::Form):
# try to find first submit button to activate
my $input;
my $form = $self->current_form;
for ($form->inputs) {
next unless $_->type =~ /^(?:submit|image)\z/;
next if $button && $_->name ne $button;
next if $_->disabled;
$input = $_;
last;
}
Carp::croak("No clickable input with name $button")
if $button && !$input;
# From HTML::DOM::Element::Input:
# We can’t put this in multiple statements, as the ‘local’ would go out
# of scope too soon.
my $continue;
$input and
# ~~~ We are breaking encapsulation here.
local($$input{_HTML_DOM_clicked}) = [$x,$y],
$input->trigger_event(
'click',
click_default => sub {
$input->trigger_event(
'DOMActivate', DOMActivate_default => sub { ++$continue }
)
}
),
$continue || return;
my $go_for_it;
$form->trigger_event(
'submit',
submit_default => sub { ++$go_for_it }
);
$go_for_it
? ($self->find_target($form->target)||$self)
->request($form->make_request)
: ()
}
else {
# Unlike the submit method, we *can* delegate to the superclass here,
# as the form’s click method (which Mech->click calls) calls our
# default_event_handler_for submit, which chooses the right target.
shift->SUPER::click(@_);
}
}
# ------------- Window interface ------------- #
# This does not follow the same format as %HTML::DOM::Interface; this cor-
# responds to the format of hashes *within* %H:D:I. The other format does
# not apply here, since we can’t bind the class like other classes. This
# needs to be bound to the global object (at least in JavaScript).
our %WindowInterface = (
%{$HTML::DOM::Interface{AbstractView}},
%{$HTML::DOM::Interface{EventTarget}},
alert => VOID|METHOD,
confirm => BOOL|METHOD,
prompt => STR|METHOD,
location => OBJ,
setTimeout => NUM|METHOD,
clearTimeout => NUM|METHOD,
setInterval => NUM|METHOD,
clearInterval => NUM|METHOD,
open => OBJ|METHOD,
blur => VOID|METHOD,
close => VOID|METHOD,
focus => VOID|METHOD,
window => OBJ|READONLY,
self => OBJ|READONLY,
navigator => OBJ|READONLY,
screen => OBJ|READONLY,
top => OBJ|READONLY,
frames => OBJ|READONLY,
length => NUM|READONLY,
parent => OBJ|READONLY,
name => STR,
scroll => VOID|METHOD,
scrollBy => VOID|METHOD,
scrollTo => VOID|METHOD,
history => OBJ|READONLY,
# See the comment preceding the commented-out subs.
# status => STR,
# defaultStatus => STR,
);
sub alert {
my $self = shift;
&{$$self{Scripter_alert}||sub{print @_,"\n";()}}(@_);
}
sub confirm {
my $self = shift;
($$self{Scripter_confirm}||$self->die(
"There is no default confirm function"
))->(@_)
}
sub prompt {
my $self = shift;
($$self{Scripter_prompt}||$self->die(
"There is no default prompt function"
))->(@_)
}
sub location {
my $self = shift;
my $loc = $self->{Scripter_loc} ||= WWW::Scripter::Location->new(
$self
);
$loc->href(@_) if @_;
$loc;
}
sub navigator {
my $self = shift;
$navi{$self} ||=
new WWW::Scripter::Navigator:: $self;
}
sub screen {
my $self = shift;
$scrn{$self} ||=
bless \my $foo, WWW::Scripter::Screen::;
}
@WWW::Scripter::Interface{WWW::Scripter::Screen::,'Screen'} = (
'Screen', {}
);
sub setTimeout {
my $doc = shift->document;
my $time = time;
my ($code, $ms) = (shift,shift);
$ms /= 1000;
my $t_o = $timeouts{$doc}||=[];
$$t_o[my $id = @$t_o] =
[$ms+$time, $code, @_];
return $id;
}
sub clearTimeout {
delete $timeouts{shift->document}[shift];
return;
}
sub setInterval {
my $doc = shift->document;
my $time = time;
my ($code, $ms) = (shift,shift);
$ms /= 1000;
my $t_o = $timers{$doc}||=[];
$$t_o[my $id = @$t_o] =
[$ms+$time, $code, @_];
return $id;
}
sub clearInterval {
delete $timers{shift->document}[shift];
return;
}
sub open {
my($self,$url,$target,undef,$replace) = @_;
$target
= $self->find_target(defined $target ? $target : '_blank');
if(defined $url and length $url) {
if(my $base = $self->base) {
require URI;
$url = URI->new_abs( $url, $base );
}
$target||=$self->top;
$replace
? $target->location->replace($url)
: $target->get($url);
$target;
}
elsif(!$target) {
# undef or "" in single-window mode: append an ‘unbrowsed’
# history entry to simulate a new window
(my $ret = $self->top)->{page_stack}->_add();
_initial_page($ret);
$ret;
}
else {
# open("") with existing window; do nothing
$target
}
}
sub close {
if(my $g = $_[0]{Scripter_g}) {
$g->detach($_[0]);
}
else {
$_[0]->history->go(-1);
}
_:
}
sub focus {
my $g = $_[0]{Scripter_g} or return;
$g->bring_to_front(shift);
return;
}
sub blur {
my $g = $_[0]{Scripter_g} or return;
my($maybe_self,$next) = $g->windows;
$next or return;
$maybe_self == $_[0] or return;
$g->bring_to_front($next);
return;
}
sub history { $_[0]{page_stack} }
sub frames {
my $doc = $_[0]->document;
my $frames = $frames{$doc||''} # the ||'' is for non-HTML docu-
||= WWW::Scripter'Frames->new( $_[0], $doc ); # ments, which all share
wantarray ? @$frames : $frames # an empty frames
} # collection
sub window { $_[0] }
*self = *window;
sub length { $frames{$_[0]->document}->length }
sub top {
my $self = shift;
$$self{Scripter_t} || do {
my $parent = $self;
while() {
$$parent{Scripter_pa} or
weaken( $$self{Scripter_t} = $parent), last;
$parent = $$parent{Scripter_pa};
}
$$self{Scripter_t}
};
}
sub parent {
my $self = shift;
$$self{Scripter_pa} || $self;
}
sub _set_parent { weaken( $_[0]{Scripter_pa} = $_[1] ) }
sub name {
my $self = shift;
my $old = $$self{Scripter_nm};
$$self{Scripter_nm} = $_[0] if @_;
$old;
}
sub scroll{}; *scrollBy=*scrollTo=*scroll;
# ~~~ This conflicts with Mech’s method. We probably need to bite the
# bullet and provide a separate window object for scripts.
#sub status {
# my $old = $status{my $doc = shift->document};
# no warnings 'uninitialized';
# $status{$doc} = "$_[0]" if @_;
# defined $old ? $old : ''
#}
#
# ~~~ This one is commented out because it makes no sense without the
# previous one.
#sub defaultStatus {
# my $old = $dstatus{my $doc = shift->document};
# no warnings 'uninitialized';
# $dstatus{$doc} = "$_[0]" if @_;
# defined $old ? $old : ''
#}
# ------------- Window-Related Public Methods -------------- #
sub set_alert_function { ${$_[0]}{Scripter_alert} = $_[1]; }
sub set_confirm_function { ${$_[0]}{Scripter_confirm} = $_[1]; }
sub set_prompt_function { ${$_[0]}{Scripter_prompt} = $_[1]; }
sub check_timers {
my $time = time;
my $self = shift;
local *_;
my $doing_timers_now;
my $jh;
for my $timers(\%timeouts, \%timers) {
my $t_o = $$timers{$self->document}||next;
for my $id(0..$#$t_o) {
next unless $_ = $$t_o[$id];
no warnings 'uninitialized';
local *@;
$$_[0] <= $time and
reftype $$_[1] eq 'CODE' || (
exists $INC{'overload.pm'}
&& defined blessed $$_[1]
&& overload'Method($$_[1],'&{}')
)
? eval { $$_[1]->(@$_[2..$#$_]) }
: (
$jh ||= $self->_handler_for_lang('JavaScript')
and $jh->eval($self,$$_[1])
),
$@ && $self->warn($@),
$doing_timers_now ? $$_[0] = time : delete $$t_o[$id];
}
} continue { ++$doing_timers_now }
$_->check_timers for $self->frames;
# ~~~ Should we try to trigger the timers in the right order if,
# exempli gratia, an iframe’s timer was registered with 200 as
# the timeout, and then the main window with 210 immediately
# thereafter?
return
}
sub count_timers {
my $self = shift;
my $count;
for(\%timeouts, \%timers) {
if(my $t_o = $$_{$self->document}) {
#use DDS; Dump [map $_&&[map "$_", @$_], @$t_o];
for my $id(0..$#$t_o) {
next unless $$t_o[$id];
++$count
}
}
}
sum $count||(), map $_->count_timers, $self->frames or 0;
}
sub wait_for_timers {
my($self, %args) = @_;
my $start_time = time if $args{max_wait};
my $interval = $args{interval} || .1;
my $min = $args{min_timers} || 0;
$self->check_timers;
while(
$self->count_timers > $min
and !$args{max_wait} || time-$start_time < $args{max_wait}
) {
select(undef,undef,undef,$interval);
$self->check_timers;
}
_:
}
sub window_group {
my $old = (my $self = shift)->{Scripter_g};
@_ and weaken($self->{Scripter_g} = shift);
$old
}
sub find_target {
my $self = shift;
my $name = shift;
no warnings 'uninitialized';
if(!CORE::length $name and my $doc = document $self) {
if(my $base_elem = $doc->look_down(_tag => 'base', target => qr)(?:\)))){
$name = $base_elem->attr('target');
}
}
CORE::length $name or return $self;
if($name =~ /^_[Bb][Ll][Aa][Nn][Kk]\z/) {
if(my $g = $$self{Scripter_g}) {
attach $g my $neww = $self->clone->clear_history(1);
return $neww;
}
return undef;
}
$name =~ /^_[Ss][Ee][Ll][Ff]\z/ and return $self;
$name =~ /^_[Pp][Aa][Rr][Ee][Nn][Tt]\z/ and return $self->parent;
$name =~ /^_[Tt][Oo][Pp]\z/ and return $self->top;
# Search subframes, and then ancestors (including their subframes), in
# breadth-first order
my $current_ancestor = $self;
my $prev_ancestor;
while() {
$current_ancestor->name eq $name and return $current_ancestor;
my $next_level = [
$prev_ancestor
? grep $_ != $prev_ancestor, $current_ancestor->frames
: $current_ancestor->frames
];
while($next_level) {
my $tmp = $next_level; $next_level = undef;
for(@$tmp) {
if($_->name eq $name) { return $_ }
push @$next_level, $_->frames;
}
}
$prev_ancestor = $current_ancestor;
$current_ancestor = $current_ancestor->parent;
last if $prev_ancestor == $current_ancestor;
}
# If we reach this point, there are no frames named $name. Return undef
# in single-window mode, or look for a window.
my $g = $$self{Scripter_g} or return undef;
my $named = ($$self{Scripter_n}||=&fieldhash({}))->{$self->response}||={};
# The extra ${} is there since a reference in a tied hash element cannot
# be weakened directly, as the element is just temporary each time.
$$named{$name} && ${$$named{$name}}->window_group
? ${$$named{$name}}
: do {
attach $g my $neww = $self->clone->clear_history(1);
weaken(${$$named{$name}} = $neww);
$neww
}
}
# ------------- EventTarget interface ------------- #
*event_listeners_enabled = *scripts_enabled;
# What we are doing here is delegating event handler/listener storage to
# a response object (and fooling EventTarget into thinking that the
# response object is an EventTarget). This is so that each page has its own
# set of event handlers, but we still use the WWW::Scripter object as the
# event target.
for my $meth (qw b addEventListener removeEventListener event_handler
get_event_listeners b) {
no strict 'refs';
my $full_meth= "HTML::DOM::EventTarget::$meth";
*$meth = sub {
shift->response->$full_meth(@_);
}
}
# ------------- Image Hooks -------------- #
sub fetch_images {
my $old = (my $self = shift)->{Scripter_i};
@_ and $self->{Scripter_i} = shift;
$old
}
sub image_handler {
my $old = (my $self = shift)->{Scripter_ih};
@_ and $self->{Scripter_ih} = shift;
$old
}
# ------------- Scripting hooks and what-not ------------- #
sub eval {
my ($self,$code) = (shift,shift);
my $h = $self->_handler_for_lang(my $lang = shift);
my $ret = (
$h or $self->die(
defined $lang ? "No scripting handlers have been registered for $lang"
: "No scripting handlers have been registered"
)
)->eval($self,$code);
$@ and $self->warn($@);
$ret;
}
sub use_plugin {
my ($self, $plugin, @opts) = (shift, shift, @_);
my $plugins = $self->{plugins} ||= {};
$plugin = _plugin2module($plugin);
return $plugins->{$plugin} if $self->{cloning};
if(exists $plugins->{$plugin}) {
$plugins->{$plugin}->options(@opts) if @opts;
}
else {
(my $plugin_file = $plugin) =~ s-::-/-g;
require "$plugin_file.pm";
$plugins->{$plugin} = $plugin->init($self, \@opts);
$plugins->{$plugin}->options(@opts) if @opts;
}
$plugins->{$plugin};
}
sub plugin {
my $self = shift;
my $plugin = _plugin2module(shift);
return exists $self->{plugins}{$plugin}
? $self->{plugins}{$plugin} || 1 : 0;
}
sub _plugin2module { # This is NOT a method
my $name = shift;
return $name if $name =~ /::/;
$name =~ s/-/::/g;
return __PACKAGE__."::Plugin::$name";
}
sub _clone_plugins {
my $self = shift;
return unless $self->{plugins};
my $plugins = $self->{plugins} = { %{$self->{plugins}} };
while ( my($pn,$po) = each %$plugins ) {
# plugin name, plugin object
next unless $po && defined blessed $po && $po->can('clone');
$plugins->{$pn} = $po->clone($self);
}
}
sub dom_enabled {
my $old = (my $self = shift)->{Scripter_dumb};
defined $old or $old = 1; # default
if(@_) {{
$$self{Scripter_dumb} = !!$_[0]; # We don’t want undef
}} # resetting it.
$old
}
sub scripts_enabled {
my $old = $scriptable{my $self = shift};
defined $old or $old = 1; # default
if(@_) {{
$scriptable{$self} = !!$_[0]; # We don’t want undef resetting it.
($self->document ||last) ->event_listeners_enabled(shift) ;
}}
$old
}
# used by HTML::DOM::EventTarget:
*event_listeners_enabled = *scripts_enabled;
sub script_handler {
my($self,$key) = (shift,shift);
my $old = $script_handlers{$self}{$key};
@_ and $script_handlers{$self}{$key} = shift;
$old
}
sub class_info {
my $self = shift;
@_ and push @{ $class_info{$self} }, shift;
@{ $class_info{$self} } if defined wantarray;
}
# ------------- Miss Elaine E. S. ------------- #
# This function is exported upon request.
sub abort {
no warnings 'exiting';
last Scripter_ABORT;
}
sub forward {
my $self = shift;
$self->{page_stack}->go(1);
}
sub clear_history {
my $self = shift;
$$self{'page_stack'}->_clear(@_);
if (shift) {
$self->_reset_page;
# list of keys taken from _update_page
delete $self->{$_} for qw[ req redirected_url res status base ct
uri last_uri content ];
_initial_page($self);
}
return $self;
}
sub max_docs {
my $self= shift;
defined wantarray and my $old = $self->stack_depth+1;
$self->stack_depth(shift()-1) if @_;
$old;
}
sub max_history {
my $old = (my $self = shift)->{Scripter_max_hist};
@_ and $self->{Scripter_max_hist} = shift;
$old
}
# ------------- History object ------------- #
package WWW::Scripter::History;
<<'mldistwatch' if 0;
use WWW::Scripter; $VERSION = $WWW'Scripter'VERSION;
mldistwatch
our $VERSION = $WWW'Scripter'VERSION;
BEGIN { *fieldhashes = *WWW::Scripter::fieldhashes }
use HTML::DOM::Interface qw 'NUM STR READONLY METHOD VOID';
use Scalar::Util 'weaken';
=begin comment
History notes
A history object is a blessed array ref. That array ref holds the browser
history entries. Each entry is itself an array ref containing:
0 - request object
1 - response object
2 - URL
3 - state info
4 - title
The length of the array tells us whether it is a state-info entry. The URL
is used both for fragments and for state objects. The second element will
be blank if it has been erased because of max_docs.
The history object has a pointer to the ‘current’ history item
($index{$self}).
Document objects are referenced by response: $document{$response}. The
window’s ‘document’ method is inherited from HTML::DOM::View, and we set it
whenever history is browsed, retrieving it from %document.
The ‘unbrowsed’ state that used to be mentioned in HTML 5 (before it got
really convoluted) is represented by an empty array. An empty array can
exist alongside other entries, as we add one when we simulate a
new window in single-window mode.
Response objects are also listed in the array ref stored in $res{$self} in
the order in which they were accessed. Subroutines that add to this array
then call _clean($self), which then eliminates duplicate entries and
deletes from the history object itself as many of the oldest response
objects as are necessary to satisfy max_docs.
=end comment
=cut
$$_{~~__PACKAGE__} = 'History',
$$_{History} = {
length => NUM|READONLY,
index => NUM|READONLY,
userAgent => STR|READONLY,
go => METHOD|VOID,
back => METHOD|VOID,
forward => METHOD|VOID,
pushState => METHOD|VOID,
}
for \%WWW::Scripter::Interface;
fieldhashes \my ( %w, %index, %res );
sub new {
my ($pack,$mech) = @_;
my $self = bless [[]], $pack;
weaken(${$w{$self}} = $mech);
$index{$self} = 0;
$res{$self} = [];
$self
}
sub _add {
my $self = shift;
if(defined $$self[-1][0]) { # if there is no ‘undef’ entry
splice @$self, ++$index{$self};
push @$self, \@_;
$_[1] and push(@{$res{$self}}, $_[1]), _clean($self,1);
}
else {
$$self[-1] = \@_;
push @{$res{$self}}, $_[1] if $_[1];
}
}
# Called when browsing to a stale history entry and also by
# location->replace
sub _replace {
my $self = shift;
if(defined $$self[-1][0]) { # if browsing has occurred
$$self[$index{$self}] = \@_;
$_[1] and push(@{$res{$self}}, $_[1]), _clean($self);
}
else {
$$self[-1] = \@_;
push @{$res{$self}}, $_[1] if $_[1];
}
}
sub _delete_res {
delete $_[0][$index{$_[0]}][1];
}
sub _clear { # called by Scripter->clear_history
my $self = shift;
@$self = shift() ? undef : $$self[$index{$self}];
$index{$self} = 0;
}
sub length {
scalar @{+shift}
}
sub index { # ~~~ We can probably make this modifiable later.
$index{+shift}
}
sub go {
my $self = shift;
if(0==$_[0]) {
${$w{$self}}->reload;
}
else {
my $new_pos = $index{$self}+shift;
$new_pos < 0 || $new_pos > $#$self and return;
$index{$self} = $new_pos;
# ~~~ trigger popstate
# If there is a response object, we just reset the page from that. If
# there isn’t then this is a stale entry and we need to
# re-fetch the page.
my $entry = $$self[$new_pos];
if(defined $$entry[1]) { # response
${$w{$self}}->_update_page(@$entry)
}
else {
local(my $w = ${$w{$self}})->{Scripter_replace} = 1;
$w->request($$entry[0]);
}
}
return;
}
sub back { shift->go(-1) }
sub forward { shift->go(1) }
sub pushState {
my $self = shift;
my $index = $index{$self}++;
my($req,$res) = @{$$self[$index]}[0,1];
# count future entries that share the same doc
my $to_delete;
for($index+1..$#$self) {
($$self[$_][1]||0) == $res ? ++$to_delete : last;
}
# replace those future entries with the new item
splice @$self, $index+1, $to_delete||0, [ $req, $res, $_[2], @_ ];
_clean($self);
return;
}
sub _clean {
my($self, $check_max_hist) = @_;
if($check_max_hist) {
my $max = (my $w = ${$w{$self}})->{Scripter_max_hist};
if($max && @$self > $max) {
my $diff = @$self-$max;
$index{$self} -= $diff;
splice @$self, 0, $diff;
}
}
my $max = ${$w{$self}}->stack_depth + 1;
my $res = $res{$self};
my %res;
for(@$self) {
defined $$_[1] and $res{0+$$_[1]}++
}
if($max) { # ~~~ It may be more efficient if, instead of searching for
my @res; # duplicates here, we scan for the ones we know we’ve added
my %seen; # in _add and _replace.
for(reverse @$res) {
my $refaddr = 0+$_;
unshift @res, $_ if exists $res{$refaddr} && !$seen{$refaddr}++;
}
@$res = @res, return unless @res > $max;
my $diff = @res-$max;
my %to_delete;
@to_delete{map 0+$_, splice @res, 0,$diff}=();
@$res = @res;
for(@$self) {
next unless defined $$_[1];
delete $$_[1] if exists $to_delete{0+$$_[1]};
}
}
else {
@$res = grep exists $res{refaddr $_}, @$res;
}
}
sub _uri {
my $self = shift;
$$self[$index{$self}][2] || ${$w{$self}}->uri;
}
# ~~~
# ------------- Location object ------------- #
package WWW'Scripter'Location;
use HTML::DOM::Interface qw'STR METHOD VOID';
use Scalar::Util 'weaken';
use overload fallback => 1, '""' => sub{${+shift}->history->_uri};
$$_{~~__PACKAGE__} = 'Location',
$$_{Location} = {
assign => VOID|METHOD,
hash => STR,
host => STR,
hostname => STR,
href => STR,
pathname => STR,
port => STR,
protocol => STR,
search => STR,
reload => VOID|METHOD,
replace => VOID|METHOD,
}
for \%WWW::Scripter::Interface;
sub new { # usage: new .....::Location $mech
my $class = shift;
weaken (my $mech = shift);
my $self = bless \$mech, $class;
$self;
}
sub hash {
my $loc = shift;
my $old = (my $uri = $$loc->history->_uri)->fragment;
$old = "#$old" if defined $old;
if (@_){
shift() =~ /#?(.*)/s;
(my $uri_copy = $uri->clone)->fragment($1);
$uri_copy->eq($uri) or $$loc->get($uri_copy);
}
$old||''
}
sub host {
my $loc = shift;
my $uri = $$loc->history->_uri;
if (@_) {
(my $uri = $uri->clone)->port("");
$uri->host_port(shift);
$$loc->get($uri);
}
defined wantarray ? $uri->host_port : ()
}
sub hostname {
my $loc = shift;
my $uri = $$loc->history->_uri;
if (@_) {
(my $uri = $uri->clone)->host(shift);
$$loc->get($uri);
}
defined wantarray ? $uri->host : ()
}
sub href {
my $loc = shift;
my $old = $$loc->history->_uri->as_string if defined wantarray;
if (@_) {
$$loc->get(shift);
}
$old;
}
sub assign { ${$_[0]}->get($_[1]); () }
sub pathname {
my $loc = shift;
my $uri = $$loc->history->_uri;
if (@_) {
(my $uri = $uri->clone)->path(shift);
$$loc->get($uri);
}
defined wantarray ? $uri->path : ()
}
sub port {
my $loc = shift;
my $uri = $$loc->history->_uri;
if (@_) {
(my $uri = $uri->clone)->port(shift);
$$loc->get($uri);
}
defined wantarray ? $uri->port : ()
}
sub protocol {
my $loc = shift;
my $uri = $$loc->history->_uri;
if (@_) {
shift() =~ /(.*):?/s;
(my $uri = $uri->clone)->scheme($1);
$$loc->get($uri);
}
defined wantarray ? $uri->scheme . ':' : ()
}
sub search {
my $loc = shift;
my $uri = $$loc->history->_uri;
if (@_){
shift() =~ /(\??)(.*)/s;
(
my $uri_copy = $uri->clone
)->query(
$1||length$2 ? "$2" : undef
);
$$loc->get($uri_copy);
}
return unless defined wantarray;
my $q = $uri->query;
defined $q ? "?$q" : "";
}
# ~~~ Safari doesn't support forceGet. Do I need to?
sub reload { # args (forceGet)
${+shift}->reload
}
sub replace { # args (URL)
my $mech = ${+shift};
local $$mech{Scripter_replace } = 1;
$mech->get(shift);
}
# ------------- Navigator object ------------- #
package WWW::Scripter::Navigator;
use HTML::DOM::Interface qw'STR READONLY METHOD BOOL';
use Scalar::Util 'weaken';
$$_{~~__PACKAGE__} = 'Navigator',
$$_{Navigator} = {
appName => STR|READONLY,
appCodeName => STR|READONLY,
appVersion => STR|READONLY,
userAgent => STR|READONLY,
javaEnabled => METHOD|BOOL,
platform => STR|READONLY,
taintEnabled => METHOD|BOOL,
cookieEnabled => BOOL|READONLY,
}
for \%WWW::Scripter::Interface;
use constant 1.03 our $_const = {
mech => 0,
name => 1,
vers => 2,
cnam => 3,
plat => 4,
};
{ no strict; delete @{__PACKAGE__.::}{_const => keys %$_const} }
sub new {
weaken((my $self = bless[],pop)->[mech] = pop);
$self;
}
sub appName {
my $self = shift;
my $old = $self->[name];
defined $old or $old = ref $self->[mech];
@_ and $self->[name] = shift;
return $old;
}
sub appCodeName {
my $self = shift;
my $old = $self->[cnam];
defined $old or $old = ref $self->[mech];
@_ and $self->[cnam] = shift;
return $old;
}
sub appVersion {
my $self = shift;
my $old = $self->[vers];
if(!defined $old and defined wantarray) {
$old = $self->userAgent;
$old =~ /(\d.*)/s
? $old = $1
: $old = ref($self->[mech])->VERSION;
}
@_ and $self->[vers] = shift;
return $old;
}
sub userAgent {
shift->[mech]->agent;
}
sub platform {
my $self = shift;
my $old = $self->[plat];
if(!defined $old and defined wantarray) {
my $ua = $self->[mech]->agent;
no warnings 'uninitialized';
$old
= $ua =~ /\bWin(?:dows|32)?\b/ ? 'Win32'
: $ua =~ /\bMac(?:intosh)\b/ ? $ua =~ /\bIntel\b/
? 'MacIntel' : 'MacPPC'
: $ua =~ /\b(FreeBSD(?: i386)?|Linux)\b/
? $1
: $^O eq 'MSWin32' ? 'Win32'
: $^O eq 'MacOS' ? 'MacPPC'
: $^O eq 'freebsd' ? 'FreeBSD'
: $^O eq 'linux' ? 'Linux'
: $^O ne 'darwin' ? $^O
: pack "s", 28526, eq 'on' ? 'MacPPC' : 'MacIntel';
}
@_ and $self->[plat] = shift;
return $old;
}
sub javaEnabled{}
*taintEnabled=*javaEnabled;
sub cookieEnabled { defined $_[0][mech]->cookie_jar }
# ------------- about: protocol ------------- #
package WWW'Scripter'_about_protocol;
# ~~~ This method may be a bad idea if someone else wants to implement
# other aspects of the about: protocol. Maybe we should use an LWP
# handler. (Then we would, of course, require a later LWP.)
<<'mldistwatch' if 0;
use WWW::Scripter; $VERSION = $WWW'Scripter'VERSION;
mldistwatch
our $VERSION = $WWW'Scripter'VERSION;
use LWP::Protocol;
our @ISA = LWP::Protocol::;
LWP::Protocol'implementor about => __PACKAGE__;
sub request { # based on the one in LWP::Protocol::file
my($self, $request, $proxy, $arg) = @_;
if(defined $proxy) {
return new HTTP::Response 400,,
'The about: protocol does not work with proxies';
}
my $url= $request->url;
my $scheme = $url->scheme;
if ($scheme ne 'about') {
return new HTTP::Response 500,
"WWW::Scripter::_about_protocol called for $scheme";
}
return new HTTP::Response 404,
"Nothing exists at $url" unless $url eq 'about:blank';
my $response = new HTTP::Response 200, 'OK', [
Content_Length=>0,
Content_Type =>'text/html',
];
$self->collect($arg, $response, sub {\''});
}
# ------------- Link and image lists for Mech ------------- #
package WWW::Scripter::Links;
BEGIN { eval "require ".WWW'Scripter'Mech."::Link" or die $@ }
sub TIEARRAY {
bless \(my $links = pop), shift;
}
sub FETCH {
my $link = ${$_[0]}->[$_[1]];
my $mech_link = bless [], WWW'Scripter'Mech."::Link";
tie @$mech_link, WWW'Scripter'Link::, $link;
$dom_obj{$mech_link} = $link;
$mech_link;
}
sub FETCHSIZE { scalar @${$_[0]} }
sub EXISTS { exists ${$_[0]}->links->[$_[1]] }
package WWW::Scripter::Link;
sub TIEARRAY { bless \(my $x = $_[1]) }
sub FETCH {
my $self = shift;
for(shift) {
return
$_ == 0 ? $$self->tag eq 'meta' # url
? $$self->attr('content') =~ /^\d+\s*;\s*url\s*=\s*(\S+)/i
? do { my $url = $1;
$url =~ s/^"(.+)"$/$1/ or $url =~ s/^'(.+)'$/$1/;
$url }
: undef
: $$self->attr($link_tags{$$self->tag}) :
$_ == 1 ? $$self->tag eq 'a' ? $$self->as_text : undef : # text
$_ == 2 ? $$self->attr('name') : # name
$_ == 3 ? $$self->tag : # tag
$_ == 4 ? $$self->ownerDocument->base : # base
$_ == 5 ? {$$self->all_external_attr} : # attrs
undef
}
}
sub FETCHSIZE { 6 }
package WWW::Scripter::Images;
BEGIN { eval "require ".WWW'Scripter'Mech."::Image" or die $@ }
sub TIEARRAY {
bless \(my $links = pop), shift;
}
sub FETCH {
my $img = ${$_[0]}->[$_[1]];
my $mech_img = new WWW'Scripter'Image:: $img;
$dom_obj{$mech_img} = $img;
$mech_img;
}
sub FETCHSIZE { scalar @${$_[0]} }
sub EXISTS { exists ${$_[0]}->links->[$_[1]] }
package WWW::Scripter::Image;
our @ISA = WWW::Scripter::Mech."::Image";
sub new { bless \(my $frin = pop) }
sub url { ${$_[0]}->attr('src') }
sub base { ${$_[0]}-ownerDocument->base }
sub name { ${$_[0]}->attr('name') }
sub tag { ${$_[0]}->tag }
sub height { ${$_[0]}->attr('height') }
sub width { ${$_[0]}->attr('width') }
sub alt { ${$_[0]}->attr('alt') }
# ------------- Frames list ------------- #
package WWW::Scripter::Frames;
# ~~~ This is horribly inefficient and clunky. It probably needs to be
# programmed in full here, or at least the ‘Collection’ part (a tiny
# bit of copy&paste).
use HTML::DOM::Collection;
use HTML::DOM::NodeList::Magic;
our @ISA = "HTML::DOM::Collection";
{
WWW::Scripter'fieldhash my %w;
my @empty_array;
sub new {
; my($pack,$window,$doc) = @_
; my $ret = $pack->SUPER'new(
$doc
? HTML::DOM::NodeList::Magic->new(
sub { $doc->look_down(_tag => qr/^i?frame\z/) },
$doc
)
: HTML'DOM'NodeList->new(\@empty_array)
)
; Scalar'Util'weaken($_) for $doc, $window;
; $w{$ret} = \$window;
; $ret
}
sub window { ${$w{+shift}||return undef} }
}
use overload fallback => 1,'@{}' => sub {
[map $_->contentWindow, @{shift->${\'SUPER::(@{}'}}]
};
sub FETCH { (shift->SUPER::FETCH(@_)||return)->contentWindow }
!!*!*!!*!*!!*!*!!*!*!!*!*!!*!*!!*!*!!*!*!!*!*!!*!*!!*!*!!*!*!!*!*!!*!