HTML::Restrict - Strip unwanted HTML tags and attributes


version v3.0.2


    use HTML::Restrict;

    my $hr = HTML::Restrict->new();

    # use default rules to start with (strip away all HTML)
    my $processed = $hr->process('  <b>i am bold</b>  ');

    # $processed now equals: 'i am bold'

    # Now, a less restrictive example:
    $hr = HTML::Restrict->new(
        rules => {
            b   => [],
            img => [qw( src alt / )]

    my $html = q[<body><b>hello</b> <img src="pic.jpg" alt="me" id="test" /></body>];
    $processed = $hr->process( $html );

    # $processed now equals: <b>hello</b> <img src="pic.jpg" alt="me" />


This module uses HTML::Parser to strip HTML from text in a restrictive manner. By default all HTML is restricted. You may alter the default behaviour by supplying your own tag rules.



Creates and returns a new HTML::Restrict object.

    my $hr = HTML::Restrict->new()

HTML::Restrict doesn't require any params to be passed to new. If your goal is to remove all HTML from text, then no further setup is required. Just pass your text to the process() method and you're done:

    my $plain_text = $hr->process( $html );

If you need to set up specific rules, have a look at the params which HTML::Restrict recognizes:

  • rules => \%rules

    Sets the rules which will be used to process your data. By default all HTML tags are off limits. Use this argument to define the HTML elements and corresponding attributes you'd like to use. Essentially, consider the default behaviour to be:

        rules => {}

    Rules should be passed as a HASHREF of allowed tags. Each hash value should represent the allowed attributes for the listed tag. For example, if you want to allow a fair amount of HTML, you can try something like this:

        my %rules = (
            a       => [qw( href target )],
            b       => [],
            caption => [],
            center  => [],
            em      => [],
            i       => [],
            img     => [qw( alt border height width src style )],
            li      => [],
            ol      => [],
            p       => [qw(style)],
            span    => [qw(style)],
            strong  => [],
            sub     => [],
            sup     => [],
            table   => [qw( style border cellspacing cellpadding align )],
            tbody   => [],
            td      => [],
            tr      => [],
            u       => [],
            ul      => [],
        my $hr = HTML::Restrict->new( rules => \%rules )

    Or, to allow only bolded text:

        my $hr = HTML::Restrict->new( rules => { b => [] } );

    Allow bolded text, images and some (but not all) image attributes:

        my %rules = (
            b   => [ ],
            img => [qw( src alt width height border / )
        my $hr = HTML::Restrict->new( rules => \%rules );

    Since HTML::Parser treats a closing slash as an attribute, you'll need to add "/" to your list of allowed attributes if you'd like your tags to retain closing slashes. For example:

        my $hr = HTML::Restrict->new( rules =>{ hr => [] } );
        $hr->process( "<hr />"); # returns: <hr>
        my $hr = HTML::Restrict->new( rules =>{ hr => [qw( / )] } );
        $hr->process( "<hr />"); # returns: <hr />

    HTML::Restrict strips away any tags and attributes which are not explicitly allowed. It also rebuilds your explicitly allowed tags and places their attributes in the order in which they appear in your rules.

    So, if you define the following rules:

        my %rules = (
            img => [qw( src alt title width height id / )]

    then your image tags will all be built like this:

        <img src=".." alt="..." title="..." width="..." height="..." id=".." />

    This gives you greater consistency in your tag layout. If you don't care about element order you don't need to pay any attention to this, but you should be aware that your elements are being reconstructed rather than just stripped down.

    As of 2.1.0, you can also specify a regex to be tested against the attribute value. This feature should be considered experimental for the time being:

        my $hr = HTML::Restrict->new(
            rules => {
                iframe => [
                    qw( width height allowfullscreen ),
                    {   src         => qr{^http://www\.youtube\.com},
                        frameborder => qr{^(0|1)$},
                img => [ qw( alt ), { src => qr{^/my/images/} }, ],
        my $html = '<img src="" alt="Alt Text">';
        my $processed = $hr->process( $html );
        # $processed now equals: <img alt="Alt Text">

    As of 2.3.0, the value to be tested against can also be a code reference. The code reference will be passed the value of the attribute, and should return either a string to use for the attribute value, or undef to remove the attribute.

        my $hr = HTML::Restrict->new(
            rules => {
                span => [
                    { style     => sub {
                        my $value = shift;
                        # all colors are orange
                        $value =~ s/\bcolor\s*:\s*[^;]+/color: orange/g;
                        return $value;
                    } }
        my $html = '<span style="color: #0000ff;">This is blue</span>';
        my $processed = $hr->process( $html );
        # $processed now equals: <span style="color: orange;">
  • trim => [0|1]

    By default all leading and trailing spaces will be removed when text is processed. Set this value to 0 in order to disable this behaviour.

  • uri_schemes => [undef, 'http', 'https', 'irc', ... ]

    As of version 1.0.3, URI scheme checking is performed on all href and src tag attributes. The following schemes are allowed out of the box. No action is required on your part:

        [ undef, 'http', 'https' ]

    (undef represents relative URIs). These restrictions have been put in place to prevent XSS in the form of:

        <a href="javascript:alert(document.cookie)">click for cookie!</a>

    See URI for more detailed info on scheme parsing. If, for example, you wanted to filter out every scheme barring SSL, you would do it like this:

        uri_schemes => ['https']

    This feature is new in 1.0.3. Previous to this, there was no schema checking at all. Moving forward, you'll need to whitelist explicitly all URI schemas which are not supported by default. This is in keeping with the whitelisting behaviour of this module and is also the safest possible approach. Keep in mind that changes to uri_schemes are not additive, so you'll need to include the defaults in any changes you make, should you wish to keep them:

        # defaults + irc + mailto
        uri_schemes => [ 'undef', 'http', 'https', 'irc', 'mailto' ]
  • allow_declaration => [0|1]

    Set this value to true if you'd like to allow/preserve DOCTYPE declarations in your content. Useful when cleaning up your own static files or templates. This feature is off by default.

        my $html = q[<!doctype html><body>foo</body>];
        my $hr = HTML::Restrict->new( allow_declaration => 1 );
        $html = $hr->process( $html );
        # $html is now: "<!doctype html>foo"
  • allow_comments => [0|1]

    Set this value to true if you'd like to allow/preserve HTML comments in your content. Useful when cleaning up your own static files or templates. This feature is off by default.

        my $html = q[<body><!-- comments! -->foo</body>];
        my $hr = HTML::Restrict->new( allow_comments => 1 );
        $html = $hr->process( $html );
        # $html is now: "<!-- comments! -->foo"
  • create_newlines => [0|1]

    Set the value to true if you'd like to have each br tag replaced by a newline and every p tag replaced by two newlines. If a tag is specified in the allowed HTML, it won't be replaced.

  • replace_img => [0|1|CodeRef]

    Set the value to true if you'd like to have img tags replaced with [IMAGE: ...] containing the alt attribute text. If you set it to a code reference, you can provide your own replacement (which may even contain HTML).

        sub replacer {
            my ($tagname, $attr, $text) = @_; # from HTML::Parser
            return qq{<a href="$attr->{src}">IMAGE: $attr->{alt}</a>};
        my $hr = HTML::Restrict->new( replace_img => \&replacer );

    This attribute will only take effect if the img tag is not included in the allowed HTML.

  • strip_enclosed_content => [0|1]

    The default behaviour up to 1.0.4 was to preserve the content between script and style tags, even when the tags themselves were being deleted. So, you'd be left with a bunch of JavaScript or CSS, just with the enclosing tags missing. This is almost never what you want, so starting at 1.0.5 the default will be to remove any script or style info which is enclosed in these tags, unless they have specifically been whitelisted in the rules. This will be a sane default when cleaning up content submitted via a web form. However, if you're using HTML::Restrict to purge your own HTML you can be more restrictive.

        # strip the head section, in addition to JS and CSS
        my $html = '<html><head>...</head><body>...<script>JS here</script>foo';
        my $hr = HTML::Restrict->new(
            strip_enclosed_content => [ 'script', 'style', 'head' ]
        $html = $hr->process( $html );
        # $html is now '<html><body>';

    The caveat here is that HTML::Restrict will not try to fix broken HTML. In the above example, if you have any opening script, style or head tags which don't also include matching closing tags, all following content will be stripped away, regardless of any parent tags.

    Keep in mind that changes to strip_enclosed_content are not additive, so if you are adding additional tags you'll need to include the entire list of tags whose enclosed content you'd like to remove. This feature strips script and style tags by default.

  • filter_text => [0|1|CodeRef]

    By default all text will be filtered to fix any encoding problems which may cause security issues. You may override the encoding behaviour by providing your own anonymous sub to filter_text. This first and only argument to the sub is the text which needs to be filtered. The sub should return a scalar containing the transformed text.

        filter_text => sub {
            my $text = shift;
            ... # transform text
            return $text;

    You may also this value to 0 in order to disable this behaviour entirely. Please be advised this is a security risk. Use caution when disabling this parameter or providing your own filter function.


process( $html )

This is the method which does the real work. It parses your data, removes any tags and attributes which are not specifically allowed and returns the resulting text. Requires and returns a SCALAR.


Accessor which returns a hash ref of the current rule set.


Accessor which returns an array ref of the current valid uri schemes.


Please note that all tag and attribute names passed via the rules param must be supplied in lower case.

    # correct
    my $hr = HTML::Restrict->new( rules => { body => ['onload'] } );

    # throws a fatal error
    my $hr = HTML::Restrict->new( rules => { Body => ['onLoad'] } );


There are already several modules on the CPAN which accomplish much of the same thing, but after doing a lot of poking around, I was unable to find a solution with a simple setup which I was happy with.

The most common use case might be stripping HTML from user submitted data completely or allowing just a few tags and attributes to be displayed. With the exception of URI scheme checking, this module doesn't do any validation on the actual content of the tags or attributes. If this is a requirement, you can either mess with the parser object, post-process the text yourself or have a look at one of the more feature-rich modules in the SEE ALSO section below.

My aim here is to keep things easy and, hopefully, cover a lot of the less complex use cases with just a few lines of code and some brief documentation. The idea is to be up and running quickly.


HTML::TagFilter, HTML::Defang, MojoMojo::Declaw, HTML::StripScripts, HTML::Detoxifier, HTML::Sanitizer, HTML::Scrubber


Thanks to Raybec Communications for funding my work on this module and for releasing it to the world.

Thanks also to the many other contributors.


Olaf Alders <>


This software is copyright (c) 2009 by Olaf Alders.

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.