search.pm

   1 #!/usr/bin/perl
   2 # xapian-omega search engine plugin
   3 package IkiWiki::Plugin::search;
   4
   5 use warnings;
   6 use strict;
   7 use IkiWiki 3.00;
   8
   9 sub import {
  10         hook(type => "getsetup", id => "search", call => \&getsetup);
  11         hook(type => "checkconfig", id => "search", call => \&checkconfig);
  12         hook(type => "pagetemplate", id => "search", call => \&pagetemplate);
  13         hook(type => "indexhtml", id => "search", call => \&indexhtml);
  14         hook(type => "delete", id => "search", call => \&delete);
  15         hook(type => "cgi", id => "search", call => \&cgi);
  16         hook(type => "disable", id => "search", call => \&disable);
  17         hook(type => "needsbuild", id => "search", call => \&needsbuild);
  18
  19         eval q{ use Search::Xapian }; # load early to work around #622591
  20 }
  21
  22 sub getsetup () {
  23         return
  24                 plugin => {
  25                         safe => 1,
  26                         rebuild => 1,
  27                         section => "web",
  28                 },
  29                 omega_cgi => {
  30                         type => "string",
  31                         example => "/usr/lib/cgi-bin/omega/omega",
  32                         description => "path to the omega cgi program",
  33                         safe => 0, # external program
  34                         rebuild => 0,
  35                 },
  36                 google_search => {
  37                         type => "booblean",
  38                         example => 1,
  39                         description => "use google site search rather than internal xapian index?",
  40                         safe => 1,
  41                         rebuild => 0,
  42                 },
  43 }
  44
  45 sub checkconfig () {
  46         foreach my $required (qw(url cgiurl)) {
  47                 if (! length $config{$required}) {
  48                         error(sprintf(gettext("Must specify %s when using the %s plugin"), $required, 'search'));
  49                 }
  50         }
  51
  52         if (! defined $config{omega_cgi}) {
  53                 $config{omega_cgi}="/usr/lib/cgi-bin/omega/omega";
  54         }
  55
  56         # This is a mass dependency, so if the search form template
  57         # changes, every page is rebuilt.
  58         add_depends("", "templates/searchform.tmpl");
  59 }
  60
  61 my $form;
  62 sub pagetemplate (@) {
  63         my %params=@_;
  64         my $page=$params{page};
  65         my $template=$params{template};
  66
  67         # Add search box to page header.
  68         if ($template->query(name => "searchform")) {
  69                 if (! defined $form) {
  70                         my $searchform = template("searchform.tmpl", blind_cache => 1);
  71                         $searchform->param(searchaction => IkiWiki::cgiurl());
  72                         $searchform->param(html5 => $config{html5});
  73                         $form=$searchform->output;
  74                 }
  75
  76                 $template->param(searchform => $form);
  77         }
  78 }
  79
  80 my $scrubber;
  81 my $stemmer;
  82 sub indexhtml (@) {
  83         my %params=@_;
  84
  85         return if $config{google_search};
  86
  87         setupfiles();
  88
  89         # A unique pageterm is used to identify the document for a page.
  90         my $pageterm=pageterm($params{page});
  91         return unless defined $pageterm;
  92
  93         my $db=xapiandb();
  94         my $doc=Search::Xapian::Document->new();
  95         my $caption=pagetitle($params{page});
  96         my $title;
  97         if (exists $pagestate{$params{page}}{meta} &&
  98                 exists $pagestate{$params{page}}{meta}{title}) {
  99                 $title=$pagestate{$params{page}}{meta}{title};
 100         }
 101         else {
 102                 $title=$caption;
 103         }
 104
 105         # Remove html from text to be indexed.
 106         if (! defined $scrubber) {
 107                 eval q{use HTML::Scrubber};
 108                 if (! $@) {
 109                         $scrubber=HTML::Scrubber->new(allow => []);
 110                 }
 111         }
 112         my $toindex = defined $scrubber ? $scrubber->scrub($params{content}) : $params{content};
 113
 114         # Take 512 characters for a sample, then extend it out
 115         # if it stopped in the middle of a word.
 116         my $size=512;
 117         my ($sample)=substr($toindex, 0, $size);
 118         if (length($sample) == $size) {
 119                 my $max=length($toindex);
 120                 my $next;
 121                 while ($size < $max &&
 122                        ($next=substr($toindex, $size++, 1)) !~ /\s/) {
 123                         $sample.=$next;
 124                 }
 125         }
 126         $sample=~s/\n/ /g;
 127
 128         my $url=urlto($params{destpage}, "");
 129         if (defined $pagestate{$params{page}}{meta}{permalink}) {
 130                 $url=$pagestate{$params{page}}{meta}{permalink}
 131         }
 132
 133         # data used by omega
 134         # Decode html entities in it, since omega re-encodes them.
 135         eval q{use HTML::Entities};
 136         error $@ if $@;
 137         $doc->set_data(
 138                 "url=".$url."\n".
 139                 "sample=".decode_entities($sample)."\n".
 140                 "caption=".decode_entities($caption)."\n".
 141                 "modtime=$IkiWiki::pagemtime{$params{page}}\n".
 142                 "size=".length($params{content})."\n"
 143         );
 144
 145         # Index document and add terms for other metadata.
 146         my $tg = Search::Xapian::TermGenerator->new();
 147         if (! $stemmer) {
 148                 my $langcode=$ENV{LANG} || "en";
 149                 $langcode=~s/_.*//;
 150
 151                 # This whitelist is here to work around a xapian bug (#486138)
 152                 my @whitelist=qw{da de en es fi fr hu it no pt ru ro sv tr};
 153
 154                 if (grep { $_ eq $langcode } @whitelist) {
 155                         $stemmer=Search::Xapian::Stem->new($langcode);
 156                 }
 157                 else {
 158                         $stemmer=Search::Xapian::Stem->new("english");
 159                 }
 160         }
 161         $tg->set_stemmer($stemmer);
 162         $tg->set_document($doc);
 163         $tg->index_text($params{page}, 2);
 164         $tg->index_text($caption, 2);
 165         $tg->index_text($title, 2) if $title ne $caption;
 166         $tg->index_text($toindex);
 167         $tg->index_text(lc($title), 1, "S"); # for title:foo
 168         foreach my $link (@{$links{$params{page}}}) {
 169                 $tg->index_text(lc($link), 1, "XLINK"); # for link:bar
 170         }
 171
 172         $doc->add_term($pageterm);
 173         $db->replace_document_by_term($pageterm, $doc);
 174 }
 175
 176 sub delete (@) {
 177         return if $config{google_search};
 178
 179         my $db=xapiandb();
 180         foreach my $page (@_) {
 181                 my $pageterm=pageterm(pagename($page));
 182                 $db->delete_document_by_term($pageterm) if defined $pageterm;
 183         }
 184 }
 185
 186 sub cgi ($) {
 187         my $cgi=shift;
 188
 189         if (defined $cgi->param('P')) {
 190                 if ($config{google_search}) {
 191                         print $cgi->redirect("https://www.google.com/search?sitesearch=$config{url}&q=".$cgi->param('P'));
 192                         exit 0;
 193                 }
 194                 else {
 195                         # only works for GET requests
 196                         chdir("$config{wikistatedir}/xapian") || error("chdir: $!");
 197                         $ENV{OMEGA_CONFIG_FILE}="./omega.conf";
 198                         $ENV{CGIURL}=IkiWiki::cgiurl();
 199                         IkiWiki::loadindex();
 200                         $ENV{HELPLINK}=htmllink("", "", "ikiwiki/searching",
 201                                 noimageinline => 1, linktext => "Help");
 202                         exec($config{omega_cgi}) || error("$config{omega_cgi} failed: $!");
 203                 }
 204         }
 205 }
 206
 207 sub pageterm ($) {
 208         my $page=shift;
 209
 210         # 240 is the number used by omindex to decide when to hash an
 211         # overlong term. This does not use a compatible hash method though.
 212         eval q{use Encode};
 213         if (length encode_utf8($page) > 240) {
 214                 eval q{use Digest::SHA};
 215                 if ($@) {
 216                         debug("search: ".sprintf(gettext("need Digest::SHA to index %s"), $page)) if $@;
 217                         return undef;
 218                 }
 219
 220                 # Note no colon, therefore it's guaranteed to not overlap
 221                 # with a page with the same name as the hash..
 222                 return "U".lc(Digest::SHA::sha1_hex($page));
 223         }
 224         else {
 225                 return "U:".$page;
 226         }
 227 }
 228
 229 my $db;
 230 sub xapiandb () {
 231         if (! defined $db) {
 232                 eval q{
 233                         use Search::Xapian;
 234                         use Search::Xapian::WritableDatabase;
 235                 };
 236                 error($@) if $@;
 237                 $db=Search::Xapian::WritableDatabase->new($config{wikistatedir}."/xapian/default",
 238                         Search::Xapian::DB_CREATE_OR_OPEN());
 239         }
 240         return $db;
 241 }
 242
 243 {
 244 my $setup=0;
 245 sub setupfiles () {
 246         if (! $setup and (! -e $config{wikistatedir}."/xapian" || $config{rebuild})) {
 247                 writefile("omega.conf", $config{wikistatedir}."/xapian",
 248                         "database_dir .\n".
 249                         "template_dir ./templates\n");
 250                 omega_template();
 251                 $setup=1;
 252         }
 253 }
 254 }
 255
 256 sub needsbuild {
 257         my $list=shift;
 258         if (grep {
 259                 $_ eq "templates/page.tmpl" ||
 260                 $_ eq "templates/searchquery.tmpl"
 261         } @$list) {
 262                 omega_template();
 263         }
 264 }
 265
 266 sub omega_template {
 267         # Avoid omega interpreting anything in the cgitemplate
 268         # as an omegascript command.
 269         eval q{use IkiWiki::CGI};
 270         my $template=IkiWiki::cgitemplate(undef, gettext("search"), "\0",
 271                 searchform => "", # avoid showing the small search form
 272         );
 273         eval q{use HTML::Entities};
 274         error $@ if $@;
 275         $template=encode_entities($template, '\$');
 276
 277         my $querytemplate=readfile(IkiWiki::template_file("searchquery.tmpl"));
 278         $template=~s/\0/$querytemplate/;
 279         writefile("query", $config{wikistatedir}."/xapian/templates",
 280                 $template);
 281 }
 282
 283 sub disable () {
 284         if (-d $config{wikistatedir}."/xapian") {
 285                 system("rm", "-rf", $config{wikistatedir}."/xapian");
 286         }
 287 }
 288
 289 1