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