IkiWiki/Plugin/htmlscrubber.pm

   1 #!/usr/bin/perl
   2 package IkiWiki::Plugin::htmlscrubber;
   3
   4 use warnings;
   5 use strict;
   6 use IkiWiki 2.00;
   7
   8 # This regexp matches urls that are in a known safe scheme.
   9 # Feel free to use it from other plugins.
  10 our $safe_url_regexp;
  11
  12 sub import { #{{{
  13         hook(type => "getsetup", id => "htmlscrubber", call => \&getsetup);
  14         hook(type => "sanitize", id => "htmlscrubber", call => \&sanitize);
  15
  16         # Only known uri schemes are allowed to avoid all the ways of
  17         # embedding javascrpt.
  18         # List at http://en.wikipedia.org/wiki/URI_scheme
  19         my $uri_schemes=join("|", map quotemeta,
  20                 # IANA registered schemes
  21                 "http", "https", "ftp", "mailto", "file", "telnet", "gopher",
  22                 "aaa", "aaas", "acap",  "cap", "cid", "crid",
  23                 "dav", "dict", "dns", "fax", "go", "h323", "im", "imap",
  24                 "ldap", "mid", "news", "nfs", "nntp", "pop", "pres",
  25                 "sip", "sips", "snmp", "tel", "urn", "wais", "xmpp",
  26                 "z39.50r", "z39.50s",
  27                 # Selected unofficial schemes
  28                 "aim", "callto", "cvs", "ed2k", "feed", "fish", "gg",
  29                 "irc", "ircs", "lastfm", "ldaps", "magnet", "mms",
  30                 "msnim", "notes", "rsync", "secondlife", "skype", "ssh",
  31                 "sftp", "smb", "sms", "snews", "webcal", "ymsgr",
  32         );
  33         # data is a special case. Allow data:image/*, but
  34         # disallow data:text/javascript and everything else.
  35         $safe_url_regexp=qr/^(?:(?:$uri_schemes):|data:image\/|[^:]+(?:$|\/))/i;
  36 } # }}}
  37
  38 sub getsetup () { #{{{
  39         return
  40                 plugin => {
  41                         safe => 1,
  42                         rebuild => undef,
  43                 },
  44                 htmlscrubber_skip => {
  45                         type => "pagespec",
  46                         example => "!*/Discussion",
  47                         description => "PageSpec specifying pages not to scrub",
  48                         link => "ikiwiki/PageSpec",
  49                         safe => 1,
  50                         rebuild => undef,
  51                 },
  52 } #}}}
  53
  54 sub sanitize (@) { #{{{
  55         my %params=@_;
  56
  57         if (exists $config{htmlscrubber_skip} &&
  58             length $config{htmlscrubber_skip} &&
  59             exists $params{destpage} &&
  60             pagespec_match($params{destpage}, $config{htmlscrubber_skip})) {
  61                 return $params{content};
  62         }
  63
  64         return scrubber()->scrub($params{content});
  65 } # }}}
  66
  67 my $_scrubber;
  68 sub scrubber { #{{{
  69         return $_scrubber if defined $_scrubber;
  70
  71         eval q{use HTML::Scrubber};
  72         error($@) if $@;
  73         # Lists based on http://feedparser.org/docs/html-sanitization.html
  74         # With html 5 video and audio tags added.
  75         $_scrubber = HTML::Scrubber->new(
  76                 allow => [qw{
  77                         a abbr acronym address area b big blockquote br br/
  78                         button caption center cite code col colgroup dd del
  79                         dfn dir div dl dt em fieldset font form h1 h2 h3 h4
  80                         h5 h6 hr hr/ i img input ins kbd label legend li map
  81                         menu ol optgroup option p p/ pre q s samp select small
  82                         span strike strong sub sup table tbody td textarea
  83                         tfoot th thead tr tt u ul var
  84                         video audio
  85                 }],
  86                 default => [undef, { (
  87                         map { $_ => 1 } qw{
  88                                 abbr accept accept-charset accesskey
  89                                 align alt axis border cellpadding cellspacing
  90                                 char charoff charset checked class
  91                                 clear cols colspan color compact coords
  92                                 datetime dir disabled enctype for frame
  93                                 headers height hreflang hspace id ismap
  94                                 label lang maxlength media method
  95                                 multiple name nohref noshade nowrap prompt
  96                                 readonly rel rev rows rowspan rules scope
  97                                 selected shape size span start summary
  98                                 tabindex target title type valign
  99                                 value vspace width
 100                                 autoplay loopstart loopend end
 101                                 playcount controls
 102                         } ),
 103                         "/" => 1, # emit proper <hr /> XHTML
 104                         href => $safe_url_regexp,
 105                         src => $safe_url_regexp,
 106                         action => $safe_url_regexp,
 107                         cite => $safe_url_regexp,
 108                         longdesc => $safe_url_regexp,
 109                         poster => $safe_url_regexp,
 110                         usemap => $safe_url_regexp,
 111                 }],
 112         );
 113         return $_scrubber;
 114 } # }}}
 115
 116 1