my %stripper;
my %entities = (
- 'lt' => '<',
- 'gt' => '>',
- 'amp' => '&',
+ 'lt' => '<',
+ 'gt' => '>',
+ 'amp' => '&',
+ 'nbsp' => ' ', # should be => "\x{00A0}", but this can lead to problems with
+ # a non-visible character in csv-exports for example
);
sub strip {
my $value = !ref($class_or_value) && (($class_or_value // '') eq 'SL::HTML::Util') ? $_[1] : $class_or_value;
+ return '' unless defined $value;
+
+ # Remove HTML comments.
+ $value =~ s{ <!-- .*? --> }{}gx;
+
if (!%stripper) {
%stripper = ( parser => HTML::Parser->new );
- $stripper{parser}->handler(text => sub { $stripper{text} .= $_[1]; });
+ $stripper{parser}->handler(text => sub { $stripper{text} .= ' ' . $_[1]; });
}
$stripper{text} = '';
$stripper{parser}->eof;
$stripper{text} =~ s{\&([^;]+);}{ $entities{$1} || "\&$1;" }eg;
+ $stripper{text} =~ s{^ +| +$}{}g;
+ $stripper{text} =~ s{ {2,}}{ }g;
return delete $stripper{text};
}