IkiWiki/Plugin/aggregate.pm

   1 #!/usr/bin/perl
   2 # Feed aggregation plugin.
   3 package IkiWiki::Plugin::aggregate;
   4
   5 use warnings;
   6 use strict;
   7 use IkiWiki 2.00;
   8 use HTML::Parser;
   9 use HTML::Tagset;
  10 use HTML::Entities;
  11 use URI;
  12 use open qw{:utf8 :std};
  13
  14 my %feeds;
  15 my %guids;
  16
  17 sub import {
  18         hook(type => "getopt", id => "aggregate", call => \&getopt);
  19         hook(type => "getsetup", id => "aggregate", call => \&getsetup);
  20         hook(type => "checkconfig", id => "aggregate", call => \&checkconfig);
  21         hook(type => "needsbuild", id => "aggregate", call => \&needsbuild);
  22         hook(type => "preprocess", id => "aggregate", call => \&preprocess);
  23         hook(type => "delete", id => "aggregate", call => \&delete);
  24         hook(type => "savestate", id => "aggregate", call => \&savestate);
  25         hook(type => "htmlize", id => "_aggregated", call => \&htmlize);
  26         if (exists $config{aggregate_webtrigger} && $config{aggregate_webtrigger}) {
  27                 hook(type => "cgi", id => "aggregate", call => \&cgi);
  28         }
  29 }
  30
  31 sub getopt () {
  32         eval q{use Getopt::Long};
  33         error($@) if $@;
  34         Getopt::Long::Configure('pass_through');
  35         GetOptions(
  36                 "aggregate" => \$config{aggregate},
  37                 "aggregateinternal!" => \$config{aggregateinternal},
  38         );
  39 }
  40
  41 sub getsetup () {
  42         return
  43                 plugin => {
  44                         safe => 1,
  45                         rebuild => undef,
  46                 },
  47                 aggregateinternal => {
  48                         type => "boolean",
  49                         example => 0,
  50                         description => "enable aggregation to internal pages?",
  51                         safe => 0, # enabling needs manual transition
  52                         rebuild => 0,
  53                 },
  54                 aggregate_webtrigger => {
  55                         type => "boolean",
  56                         example => 0,
  57                         description => "allow aggregation to be triggered via the web?",
  58                         safe => 1,
  59                         rebuild => 0,
  60                 },
  61 }
  62
  63 sub checkconfig () {
  64         if ($config{aggregate} && ! ($config{post_commit} &&
  65                                      IkiWiki::commit_hook_enabled())) {
  66                 launchaggregation();
  67         }
  68 }
  69
  70 sub cgi ($) {
  71         my $cgi=shift;
  72
  73         if (defined $cgi->param('do') &&
  74             $cgi->param("do") eq "aggregate_webtrigger") {
  75                 $|=1;
  76                 print "Content-Type: text/plain\n\n";
  77                 $config{cgi}=0;
  78                 $config{verbose}=1;
  79                 $config{syslog}=0;
  80                 print gettext("Aggregation triggered via web.")."\n\n";
  81                 if (launchaggregation()) {
  82                         IkiWiki::lockwiki();
  83                         IkiWiki::loadindex();
  84                         require IkiWiki::Render;
  85                         IkiWiki::refresh();
  86                         IkiWiki::saveindex();
  87                 }
  88                 else {
  89                         print gettext("Nothing to do right now, all feeds are up-to-date!")."\n";
  90                 }
  91                 exit 0;
  92         }
  93 }
  94
  95 sub launchaggregation () {
  96         # See if any feeds need aggregation.
  97         loadstate();
  98         my @feeds=needsaggregate();
  99         return unless @feeds;
 100         if (! lockaggregate()) {
 101                 debug("an aggregation process is already running");
 102                 return;
 103         }
 104         # force a later rebuild of source pages
 105         $IkiWiki::forcerebuild{$_->{sourcepage}}=1
 106                 foreach @feeds;
 107
 108         # Fork a child process to handle the aggregation.
 109         # The parent process will then handle building the
 110         # result. This avoids messy code to clear state
 111         # accumulated while aggregating.
 112         defined(my $pid = fork) or error("Can't fork: $!");
 113         if (! $pid) {
 114                 IkiWiki::loadindex();
 115                 # Aggregation happens without the main wiki lock
 116                 # being held. This allows editing pages etc while
 117                 # aggregation is running.
 118                 aggregate(@feeds);
 119
 120                 IkiWiki::lockwiki;
 121                 # Merge changes, since aggregation state may have
 122                 # changed on disk while the aggregation was happening.
 123                 mergestate();
 124                 expire();
 125                 savestate();
 126                 IkiWiki::unlockwiki;
 127                 exit 0;
 128         }
 129         waitpid($pid,0);
 130         if ($?) {
 131                 error "aggregation failed with code $?";
 132         }
 133
 134         clearstate();
 135         unlockaggregate();
 136
 137         return 1;
 138 }
 139
 140 #  Pages with extension _aggregated have plain html markup, pass through.
 141 sub htmlize (@) {
 142         my %params=@_;
 143         return $params{content};
 144 }
 145
 146 # Used by ikiwiki-transition aggregateinternal.
 147 sub migrate_to_internal {
 148         if (! lockaggregate()) {
 149                 error("an aggregation process is currently running");
 150         }
 151
 152         IkiWiki::lockwiki();
 153         loadstate();
 154         $config{verbose}=1;
 155
 156         foreach my $data (values %guids) {
 157                 next unless $data->{page};
 158                 next if $data->{expired};
 159
 160                 $config{aggregateinternal} = 0;
 161                 my $oldname = "$config{srcdir}/".htmlfn($data->{page});
 162                 my $oldoutput = $config{destdir}."/".IkiWiki::htmlpage($data->{page});
 163
 164                 $config{aggregateinternal} = 1;
 165                 my $newname = "$config{srcdir}/".htmlfn($data->{page});
 166
 167                 debug "moving $oldname -> $newname";
 168                 if (-e $newname) {
 169                         if (-e $oldname) {
 170                                 error("$newname already exists");
 171                         }
 172                         else {
 173                                 debug("already renamed to $newname?");
 174                         }
 175                 }
 176                 elsif (-e $oldname) {
 177                         rename($oldname, $newname) || error("$!");
 178                 }
 179                 else {
 180                         debug("$oldname not found");
 181                 }
 182                 if (-e $oldoutput) {
 183                         require IkiWiki::Render;
 184                         debug("removing output file $oldoutput");
 185                         IkiWiki::prune($oldoutput);
 186                 }
 187         }
 188
 189         savestate();
 190         IkiWiki::unlockwiki;
 191
 192         unlockaggregate();
 193 }
 194
 195 sub needsbuild (@) {
 196         my $needsbuild=shift;
 197
 198         loadstate();
 199
 200         foreach my $feed (values %feeds) {
 201                 if (exists $pagesources{$feed->{sourcepage}} &&
 202                     grep { $_ eq $pagesources{$feed->{sourcepage}} } @$needsbuild) {
 203                         # Mark all feeds originating on this page as
 204                         # not yet seen; preprocess will unmark those that
 205                         # still exist.
 206                         markunseen($feed->{sourcepage});
 207                 }
 208         }
 209 }
 210
 211 sub preprocess (@) {
 212         my %params=@_;
 213
 214         foreach my $required (qw{name url}) {
 215                 if (! exists $params{$required}) {
 216                         error sprintf(gettext("missing %s parameter"), $required)
 217                 }
 218         }
 219
 220         my $feed={};
 221         my $name=$params{name};
 222         if (exists $feeds{$name}) {
 223                 $feed=$feeds{$name};
 224         }
 225         else {
 226                 $feeds{$name}=$feed;
 227         }
 228         $feed->{name}=$name;
 229         $feed->{sourcepage}=$params{page};
 230         $feed->{url}=$params{url};
 231         my $dir=exists $params{dir} ? $params{dir} : $params{page}."/".titlepage($params{name});
 232         $dir=~s/^\/+//;
 233         ($dir)=$dir=~/$config{wiki_file_regexp}/;
 234         $feed->{dir}=$dir;
 235         $feed->{feedurl}=defined $params{feedurl} ? $params{feedurl} : "";
 236         $feed->{updateinterval}=defined $params{updateinterval} ? $params{updateinterval} * 60 : 15 * 60;
 237         $feed->{expireage}=defined $params{expireage} ? $params{expireage} : 0;
 238         $feed->{expirecount}=defined $params{expirecount} ? $params{expirecount} : 0;
 239         if (exists $params{template}) {
 240                 $params{template}=~s/[^-_a-zA-Z0-9]+//g;
 241         }
 242         else {
 243                 $params{template} = "aggregatepost"
 244         }
 245         $feed->{template}=$params{template} . ".tmpl";
 246         delete $feed->{unseen};
 247         $feed->{lastupdate}=0 unless defined $feed->{lastupdate};
 248         $feed->{lasttry}=$feed->{lastupdate} unless defined $feed->{lasttry};
 249         $feed->{numposts}=0 unless defined $feed->{numposts};
 250         $feed->{newposts}=0 unless defined $feed->{newposts};
 251         $feed->{message}=gettext("new feed") unless defined $feed->{message};
 252         $feed->{error}=0 unless defined $feed->{error};
 253         $feed->{tags}=[];
 254         while (@_) {
 255                 my $key=shift;
 256                 my $value=shift;
 257                 if ($key eq 'tag') {
 258                         push @{$feed->{tags}}, $value;
 259                 }
 260         }
 261
 262         return "<a href=\"".$feed->{url}."\">".$feed->{name}."</a>: ".
 263                ($feed->{error} ? "<em>" : "").$feed->{message}.
 264                ($feed->{error} ? "</em>" : "").
 265                " (".$feed->{numposts}." ".gettext("posts").
 266                ($feed->{newposts} ? "; ".$feed->{newposts}.
 267                                     " ".gettext("new") : "").
 268                ")";
 269 }
 270
 271 sub delete (@) {
 272         my @files=@_;
 273
 274         # Remove feed data for removed pages.
 275         foreach my $file (@files) {
 276                 my $page=pagename($file);
 277                 markunseen($page);
 278         }
 279 }
 280
 281 sub markunseen ($) {
 282         my $page=shift;
 283
 284         foreach my $id (keys %feeds) {
 285                 if ($feeds{$id}->{sourcepage} eq $page) {
 286                         $feeds{$id}->{unseen}=1;
 287                 }
 288         }
 289 }
 290
 291 my $state_loaded=0;
 292
 293 sub loadstate () {
 294         return if $state_loaded;
 295         $state_loaded=1;
 296         if (-e "$config{wikistatedir}/aggregate") {
 297                 open(IN, "$config{wikistatedir}/aggregate") ||
 298                         die "$config{wikistatedir}/aggregate: $!";
 299                 while (<IN>) {
 300                         $_=IkiWiki::possibly_foolish_untaint($_);
 301                         chomp;
 302                         my $data={};
 303                         foreach my $i (split(/ /, $_)) {
 304                                 my ($field, $val)=split(/=/, $i, 2);
 305                                 if ($field eq "name" || $field eq "feed" ||
 306                                     $field eq "guid" || $field eq "message") {
 307                                         $data->{$field}=decode_entities($val, " \t\n");
 308                                 }
 309                                 elsif ($field eq "tag") {
 310                                         push @{$data->{tags}}, $val;
 311                                 }
 312                                 else {
 313                                         $data->{$field}=$val;
 314                                 }
 315                         }
 316
 317                         if (exists $data->{name}) {
 318                                 $feeds{$data->{name}}=$data;
 319                         }
 320                         elsif (exists $data->{guid}) {
 321                                 $guids{$data->{guid}}=$data;
 322                         }
 323                 }
 324
 325                 close IN;
 326         }
 327 }
 328
 329 sub savestate () {
 330         return unless $state_loaded;
 331         garbage_collect();
 332         my $newfile="$config{wikistatedir}/aggregate.new";
 333         my $cleanup = sub { unlink($newfile) };
 334         open (OUT, ">$newfile") || error("open $newfile: $!", $cleanup);
 335         foreach my $data (values %feeds, values %guids) {
 336                 my @line;
 337                 foreach my $field (keys %$data) {
 338                         if ($field eq "name" || $field eq "feed" ||
 339                             $field eq "guid" || $field eq "message") {
 340                                 push @line, "$field=".encode_entities($data->{$field}, " \t\n");
 341                         }
 342                         elsif ($field eq "tags") {
 343                                 push @line, "tag=$_" foreach @{$data->{tags}};
 344                         }
 345                         else {
 346                                 push @line, "$field=".$data->{$field};
 347                         }
 348                 }
 349                 print OUT join(" ", @line)."\n" || error("write $newfile: $!", $cleanup);
 350         }
 351         close OUT || error("save $newfile: $!", $cleanup);
 352         rename($newfile, "$config{wikistatedir}/aggregate") ||
 353                 error("rename $newfile: $!", $cleanup);
 354 }
 355
 356 sub garbage_collect () {
 357         foreach my $name (keys %feeds) {
 358                 # remove any feeds that were not seen while building the pages
 359                 # that used to contain them
 360                 if ($feeds{$name}->{unseen}) {
 361                         delete $feeds{$name};
 362                 }
 363         }
 364
 365         foreach my $guid (values %guids) {
 366                 # any guid whose feed is gone should be removed
 367                 if (! exists $feeds{$guid->{feed}}) {
 368                         unlink "$config{srcdir}/".htmlfn($guid->{page})
 369                                 if exists $guid->{page};
 370                         delete $guids{$guid->{guid}};
 371                 }
 372                 # handle expired guids
 373                 elsif ($guid->{expired} && exists $guid->{page}) {
 374                         unlink "$config{srcdir}/".htmlfn($guid->{page});
 375                         delete $guid->{page};
 376                         delete $guid->{md5};
 377                 }
 378         }
 379 }
 380
 381 sub mergestate () {
 382         # Load the current state in from disk, and merge into it
 383         # values from the state in memory that might have changed
 384         # during aggregation.
 385         my %myfeeds=%feeds;
 386         my %myguids=%guids;
 387         clearstate();
 388         loadstate();
 389
 390         # All that can change in feed state during aggregation is a few
 391         # fields.
 392         foreach my $name (keys %myfeeds) {
 393                 if (exists $feeds{$name}) {
 394                         foreach my $field (qw{message lastupdate lasttry
 395                                               numposts newposts error}) {
 396                                 $feeds{$name}->{$field}=$myfeeds{$name}->{$field};
 397                         }
 398                 }
 399         }
 400
 401         # New guids can be created during aggregation.
 402         # It's also possible that guids were removed from the on-disk state
 403         # while the aggregation was in process. That would only happen if
 404         # their feed was also removed, so any removed guids added back here
 405         # will be garbage collected later.
 406         foreach my $guid (keys %myguids) {
 407                 if (! exists $guids{$guid}) {
 408                         $guids{$guid}=$myguids{$guid};
 409                 }
 410         }
 411 }
 412
 413 sub clearstate () {
 414         %feeds=();
 415         %guids=();
 416         $state_loaded=0;
 417 }
 418
 419 sub expire () {
 420         foreach my $feed (values %feeds) {
 421                 next unless $feed->{expireage} || $feed->{expirecount};
 422                 my $count=0;
 423                 my %seen;
 424                 foreach my $item (sort { ($IkiWiki::pagectime{$b->{page}} || 0) <=> ($IkiWiki::pagectime{$a->{page}} || 0) }
 425                                   grep { exists $_->{page} && $_->{feed} eq $feed->{name} }
 426                                   values %guids) {
 427                         if ($feed->{expireage}) {
 428                                 my $days_old = (time - ($IkiWiki::pagectime{$item->{page}} || 0)) / 60 / 60 / 24;
 429                                 if ($days_old > $feed->{expireage}) {
 430                                         debug(sprintf(gettext("expiring %s (%s days old)"),
 431                                                 $item->{page}, int($days_old)));
 432                                         $item->{expired}=1;
 433                                 }
 434                         }
 435                         elsif ($feed->{expirecount} &&
 436                                $count >= $feed->{expirecount}) {
 437                                 debug(sprintf(gettext("expiring %s"), $item->{page}));
 438                                 $item->{expired}=1;
 439                         }
 440                         else {
 441                                 if (! $seen{$item->{page}}) {
 442                                         $seen{$item->{page}}=1;
 443                                         $count++;
 444                                 }
 445                         }
 446                 }
 447         }
 448 }
 449
 450 sub needsaggregate () {
 451         return values %feeds if $config{rebuild};
 452         return grep { time - $_->{lastupdate} >= $_->{updateinterval} } values %feeds;
 453 }
 454
 455 sub aggregate (@) {
 456         eval q{use XML::Feed};
 457         error($@) if $@;
 458         eval q{use URI::Fetch};
 459         error($@) if $@;
 460
 461         foreach my $feed (@_) {
 462                 $feed->{lasttry}=time;
 463                 $feed->{newposts}=0;
 464                 $feed->{message}=sprintf(gettext("last checked %s"),
 465                         displaytime($feed->{lasttry}));
 466                 $feed->{error}=0;
 467
 468                 debug(sprintf(gettext("checking feed %s ..."), $feed->{name}));
 469
 470                 if (! length $feed->{feedurl}) {
 471                         my @urls=XML::Feed->find_feeds($feed->{url});
 472                         if (! @urls) {
 473                                 $feed->{message}=sprintf(gettext("could not find feed at %s"), $feed->{url});
 474                                 $feed->{error}=1;
 475                                 debug($feed->{message});
 476                                 next;
 477                         }
 478                         $feed->{feedurl}=pop @urls;
 479                 }
 480                 my $res=URI::Fetch->fetch($feed->{feedurl});
 481                 if (! $res) {
 482                         $feed->{message}=URI::Fetch->errstr;
 483                         $feed->{error}=1;
 484                         debug($feed->{message});
 485                         next;
 486                 }
 487
 488                 # lastupdate is only set if we were able to contact the server
 489                 $feed->{lastupdate}=$feed->{lasttry};
 490
 491                 if ($res->status == URI::Fetch::URI_GONE()) {
 492                         $feed->{message}=gettext("feed not found");
 493                         $feed->{error}=1;
 494                         debug($feed->{message});
 495                         next;
 496                 }
 497                 my $content=$res->content;
 498                 my $f=eval{XML::Feed->parse(\$content)};
 499                 if ($@) {
 500                         # One common cause of XML::Feed crashing is a feed
 501                         # that contains invalid UTF-8 sequences. Convert
 502                         # feed to ascii to try to work around.
 503                         $feed->{message}.=" ".sprintf(gettext("(invalid UTF-8 stripped from feed)"));
 504                         $f=eval {
 505                                 $content=Encode::decode_utf8($content, 0);
 506                                 XML::Feed->parse(\$content)
 507                         };
 508                 }
 509                 if ($@) {
 510                         # Another possibility is badly escaped entities.
 511                         $feed->{message}.=" ".sprintf(gettext("(feed entities escaped)"));
 512                         $content=~s/\&(?!amp)(\w+);/&amp;$1;/g;
 513                         $f=eval {
 514                                 $content=Encode::decode_utf8($content, 0);
 515                                 XML::Feed->parse(\$content)
 516                         };
 517                 }
 518                 if ($@) {
 519                         $feed->{message}=gettext("feed crashed XML::Feed!")." ($@)";
 520                         $feed->{error}=1;
 521                         debug($feed->{message});
 522                         next;
 523                 }
 524                 if (! $f) {
 525                         $feed->{message}=XML::Feed->errstr;
 526                         $feed->{error}=1;
 527                         debug($feed->{message});
 528                         next;
 529                 }
 530
 531                 foreach my $entry ($f->entries) {
 532                         my $c=$entry->content;
 533                         # atom feeds may have no content, only a summary
 534                         if (! defined $c && ref $entry->summary) {
 535                                 $c=$entry->summary;
 536                         }
 537
 538                         add_page(
 539                                 feed => $feed,
 540                                 copyright => $f->copyright,
 541                                 title => defined $entry->title ? decode_entities($entry->title) : "untitled",
 542                                 link => $entry->link,
 543                                 content => (defined $c && defined $c->body) ? $c->body : "",
 544                                 guid => defined $entry->id ? $entry->id : time."_".$feed->{name},
 545                                 ctime => $entry->issued ? ($entry->issued->epoch || time) : time,
 546                                 base => (defined $c && $c->can("base")) ? $c->base : undef,
 547                         );
 548                 }
 549         }
 550 }
 551
 552 sub add_page (@) {
 553         my %params=@_;
 554
 555         my $feed=$params{feed};
 556         my $guid={};
 557         my $mtime;
 558         if (exists $guids{$params{guid}}) {
 559                 # updating an existing post
 560                 $guid=$guids{$params{guid}};
 561                 return if $guid->{expired};
 562         }
 563         else {
 564                 # new post
 565                 $guid->{guid}=$params{guid};
 566                 $guids{$params{guid}}=$guid;
 567                 $mtime=$params{ctime};
 568                 $feed->{numposts}++;
 569                 $feed->{newposts}++;
 570
 571                 # assign it an unused page
 572                 my $page=titlepage($params{title});
 573                 # escape slashes and periods in title so it doesn't specify
 574                 # directory name or trigger ".." disallowing code.
 575                 $page=~s!([/.])!"__".ord($1)."__"!eg;
 576                 $page=$feed->{dir}."/".$page;
 577                 ($page)=$page=~/$config{wiki_file_regexp}/;
 578                 if (! defined $page || ! length $page) {
 579                         $page=$feed->{dir}."/item";
 580                 }
 581                 my $c="";
 582                 while (exists $IkiWiki::pagecase{lc $page.$c} ||
 583                        -e "$config{srcdir}/".htmlfn($page.$c)) {
 584                         $c++
 585                 }
 586
 587                 # Make sure that the file name isn't too long.
 588                 # NB: This doesn't check for path length limits.
 589                 my $max=POSIX::pathconf($config{srcdir}, &POSIX::_PC_NAME_MAX);
 590                 if (defined $max && length(htmlfn($page)) >= $max) {
 591                         $c="";
 592                         $page=$feed->{dir}."/item";
 593                         while (exists $IkiWiki::pagecase{lc $page.$c} ||
 594                                -e "$config{srcdir}/".htmlfn($page.$c)) {
 595                                 $c++
 596                         }
 597                 }
 598
 599                 $guid->{page}=$page;
 600                 debug(sprintf(gettext("creating new page %s"), $page));
 601         }
 602         $guid->{feed}=$feed->{name};
 603
 604         # To write or not to write? Need to avoid writing unchanged pages
 605         # to avoid unneccessary rebuilding. The mtime from rss cannot be
 606         # trusted; let's use a digest.
 607         eval q{use Digest::MD5 'md5_hex'};
 608         error($@) if $@;
 609         require Encode;
 610         my $digest=md5_hex(Encode::encode_utf8($params{content}));
 611         return unless ! exists $guid->{md5} || $guid->{md5} ne $digest || $config{rebuild};
 612         $guid->{md5}=$digest;
 613
 614         # Create the page.
 615         my $template=template($feed->{template}, blind_cache => 1);
 616         $template->param(title => $params{title})
 617                 if defined $params{title} && length($params{title});
 618         $template->param(content => wikiescape(htmlabs($params{content},
 619                 defined $params{base} ? $params{base} : $feed->{feedurl})));
 620         $template->param(name => $feed->{name});
 621         $template->param(url => $feed->{url});
 622         $template->param(copyright => $params{copyright})
 623                 if defined $params{copyright} && length $params{copyright};
 624         $template->param(permalink => urlabs($params{link}, $feed->{feedurl}))
 625                 if defined $params{link};
 626         if (ref $feed->{tags}) {
 627                 $template->param(tags => [map { tag => $_ }, @{$feed->{tags}}]);
 628         }
 629         writefile(htmlfn($guid->{page}), $config{srcdir},
 630                 $template->output);
 631
 632         if (defined $mtime && $mtime <= time) {
 633                 # Set the mtime, this lets the build process get the right
 634                 # creation time on record for the new page.
 635                 utime $mtime, $mtime, "$config{srcdir}/".htmlfn($guid->{page});
 636                 # Store it in pagectime for expiry code to use also.
 637                 $IkiWiki::pagectime{$guid->{page}}=$mtime;
 638         }
 639         else {
 640                 # Dummy value for expiry code.
 641                 $IkiWiki::pagectime{$guid->{page}}=time;
 642         }
 643 }
 644
 645 sub wikiescape ($) {
 646         # escape accidental wikilinks and preprocessor stuff
 647         return encode_entities(shift, '\[\]');
 648 }
 649
 650 sub urlabs ($$) {
 651         my $url=shift;
 652         my $urlbase=shift;
 653
 654         URI->new_abs($url, $urlbase)->as_string;
 655 }
 656
 657 sub htmlabs ($$) {
 658         # Convert links in html from relative to absolute.
 659         # Note that this is a heuristic, which is not specified by the rss
 660         # spec and may not be right for all feeds. Also, see Debian
 661         # bug #381359.
 662         my $html=shift;
 663         my $urlbase=shift;
 664
 665         my $ret="";
 666         my $p = HTML::Parser->new(api_version => 3);
 667         $p->handler(default => sub { $ret.=join("", @_) }, "text");
 668         $p->handler(start => sub {
 669                 my ($tagname, $pos, $text) = @_;
 670                 if (ref $HTML::Tagset::linkElements{$tagname}) {
 671                         while (4 <= @$pos) {
 672                                 # use attribute sets from right to left
 673                                 # to avoid invalidating the offsets
 674                                 # when replacing the values
 675                                 my($k_offset, $k_len, $v_offset, $v_len) =
 676                                         splice(@$pos, -4);
 677                                 my $attrname = lc(substr($text, $k_offset, $k_len));
 678                                 next unless grep { $_ eq $attrname } @{$HTML::Tagset::linkElements{$tagname}};
 679                                 next unless $v_offset; # 0 v_offset means no value
 680                                 my $v = substr($text, $v_offset, $v_len);
 681                                 $v =~ s/^([\'\"])(.*)\1$/$2/;
 682                                 my $new_v=urlabs($v, $urlbase);
 683                                 $new_v =~ s/\"/&quot;/g; # since we quote with ""
 684                                 substr($text, $v_offset, $v_len) = qq("$new_v");
 685                         }
 686                 }
 687                 $ret.=$text;
 688         }, "tagname, tokenpos, text");
 689         $p->parse($html);
 690         $p->eof;
 691
 692         return $ret;
 693 }
 694
 695 sub htmlfn ($) {
 696         return shift().".".($config{aggregateinternal} ? "_aggregated" : $config{htmlext});
 697 }
 698
 699 my $aggregatelock;
 700
 701 sub lockaggregate () {
 702         # Take an exclusive lock to prevent multiple concurrent aggregators.
 703         # Returns true if the lock was aquired.
 704         if (! -d $config{wikistatedir}) {
 705                 mkdir($config{wikistatedir});
 706         }
 707         open($aggregatelock, '>', "$config{wikistatedir}/aggregatelock") ||
 708                 error ("cannot open to $config{wikistatedir}/aggregatelock: $!");
 709         if (! flock($aggregatelock, 2 | 4)) { # LOCK_EX | LOCK_NB
 710                 close($aggregatelock) || error("failed closing aggregatelock: $!");
 711                 return 0;
 712         }
 713         return 1;
 714 }
 715
 716 sub unlockaggregate () {
 717         return close($aggregatelock) if $aggregatelock;
 718         return;
 719 }
 720
 721 1