2 # Feed aggregation plugin.
3 package IkiWiki::Plugin::aggregate;
11 use open qw{:utf8 :std};
17 hook(type => "getopt", id => "aggregate", call => \&getopt);
18 hook(type => "getsetup", id => "aggregate", call => \&getsetup);
19 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);
32 eval q{use Getopt::Long};
34 Getopt::Long::Configure('pass_through');
36 "aggregate" => \$config{aggregate},
37 "aggregateinternal!" => \$config{aggregateinternal},
47 aggregateinternal => {
50 description => "enable aggregation to internal pages?",
51 safe => 0, # enabling needs manual transition
54 aggregate_webtrigger => {
57 description => "allow aggregation to be triggered via the web?",
64 if (! defined $config{aggregateinternal}) {
65 $config{aggregateinternal}=1;
68 # This is done here rather than in a refresh hook because it
69 # needs to run before the wiki is locked.
70 if ($config{aggregate} && ! ($config{post_commit} &&
71 IkiWiki::commit_hook_enabled())) {
79 if (defined $cgi->param('do') &&
80 $cgi->param("do") eq "aggregate_webtrigger") {
82 print "Content-Type: text/plain\n\n";
86 print gettext("Aggregation triggered via web.")."\n\n";
87 if (launchaggregation()) {
90 require IkiWiki::Render;
95 print gettext("Nothing to do right now, all feeds are up-to-date!")."\n";
101 sub launchaggregation () {
102 # See if any feeds need aggregation.
104 my @feeds=needsaggregate();
105 return unless @feeds;
106 if (! lockaggregate()) {
107 error("an aggregation process is already running");
109 # force a later rebuild of source pages
110 $IkiWiki::forcerebuild{$_->{sourcepage}}=1
113 # Fork a child process to handle the aggregation.
114 # The parent process will then handle building the
115 # result. This avoids messy code to clear state
116 # accumulated while aggregating.
117 defined(my $pid = fork) or error("Can't fork: $!");
119 IkiWiki::loadindex();
120 # Aggregation happens without the main wiki lock
121 # being held. This allows editing pages etc while
122 # aggregation is running.
126 # Merge changes, since aggregation state may have
127 # changed on disk while the aggregation was happening.
136 error "aggregation failed with code $?";
145 # Pages with extension _aggregated have plain html markup, pass through.
148 return $params{content};
151 # Used by ikiwiki-transition aggregateinternal.
152 sub migrate_to_internal {
153 if (! lockaggregate()) {
154 error("an aggregation process is currently running");
161 foreach my $data (values %guids) {
162 next unless $data->{page};
163 next if $data->{expired};
165 $config{aggregateinternal} = 0;
166 my $oldname = "$config{srcdir}/".htmlfn($data->{page});
168 $oldname = $IkiWiki::Plugin::transient::transientdir."/".htmlfn($data->{page});
171 my $oldoutput = $config{destdir}."/".IkiWiki::htmlpage($data->{page});
173 $config{aggregateinternal} = 1;
174 my $newname = $IkiWiki::Plugin::transient::transientdir."/".htmlfn($data->{page});
176 debug "moving $oldname -> $newname";
179 error("$newname already exists");
182 debug("already renamed to $newname?");
185 elsif (-e $oldname) {
186 rename($oldname, $newname) || error("$!");
189 debug("$oldname not found");
192 require IkiWiki::Render;
193 debug("removing output file $oldoutput");
194 IkiWiki::prune($oldoutput, $config{destdir});
205 my $needsbuild=shift;
209 foreach my $feed (values %feeds) {
210 if (exists $pagesources{$feed->{sourcepage}} &&
211 grep { $_ eq $pagesources{$feed->{sourcepage}} } @$needsbuild) {
212 # Mark all feeds originating on this page as
213 # not yet seen; preprocess will unmark those that
215 markunseen($feed->{sourcepage});
225 foreach my $required (qw{name url}) {
226 if (! exists $params{$required}) {
227 error sprintf(gettext("missing %s parameter"), $required)
232 my $name=$params{name};
233 if (exists $feeds{$name}) {
240 $feed->{sourcepage}=$params{page};
241 $feed->{url}=$params{url};
242 my $dir=exists $params{dir} ? $params{dir} : $params{page}."/".titlepage($params{name});
244 ($dir)=$dir=~/$config{wiki_file_regexp}/;
246 $feed->{feedurl}=defined $params{feedurl} ? $params{feedurl} : "";
247 $feed->{updateinterval}=defined $params{updateinterval} ? $params{updateinterval} * 60 : 15 * 60;
248 $feed->{expireage}=defined $params{expireage} ? $params{expireage} : 0;
249 $feed->{expirecount}=defined $params{expirecount} ? $params{expirecount} : 0;
250 if (exists $params{template}) {
251 $params{template}=~s/[^-_a-zA-Z0-9]+//g;
254 $params{template} = "aggregatepost"
256 $feed->{template}=$params{template} . ".tmpl";
257 delete $feed->{unseen};
258 $feed->{lastupdate}=0 unless defined $feed->{lastupdate};
259 $feed->{lasttry}=$feed->{lastupdate} unless defined $feed->{lasttry};
260 $feed->{numposts}=0 unless defined $feed->{numposts};
261 $feed->{newposts}=0 unless defined $feed->{newposts};
262 $feed->{message}=gettext("new feed") unless defined $feed->{message};
263 $feed->{error}=0 unless defined $feed->{error};
269 push @{$feed->{tags}}, $value;
273 return "<a href=\"".$feed->{url}."\">".$feed->{name}."</a>: ".
274 ($feed->{error} ? "<em>" : "").$feed->{message}.
275 ($feed->{error} ? "</em>" : "").
276 " (".$feed->{numposts}." ".gettext("posts").
277 ($feed->{newposts} ? "; ".$feed->{newposts}.
278 " ".gettext("new") : "").
285 # Remove feed data for removed pages.
286 foreach my $file (@files) {
287 my $page=pagename($file);
295 foreach my $id (keys %feeds) {
296 if ($feeds{$id}->{sourcepage} eq $page) {
297 $feeds{$id}->{unseen}=1;
305 return if $state_loaded;
307 if (-e "$config{wikistatedir}/aggregate") {
308 open(IN, "<", "$config{wikistatedir}/aggregate") ||
309 die "$config{wikistatedir}/aggregate: $!";
311 $_=IkiWiki::possibly_foolish_untaint($_);
314 foreach my $i (split(/ /, $_)) {
315 my ($field, $val)=split(/=/, $i, 2);
316 if ($field eq "name" || $field eq "feed" ||
317 $field eq "guid" || $field eq "message") {
318 $data->{$field}=decode_entities($val, " \t\n");
320 elsif ($field eq "tag") {
321 push @{$data->{tags}}, $val;
324 $data->{$field}=$val;
328 if (exists $data->{name}) {
329 $feeds{$data->{name}}=$data;
331 elsif (exists $data->{guid}) {
332 $guids{$data->{guid}}=$data;
341 return unless $state_loaded;
343 my $newfile="$config{wikistatedir}/aggregate.new";
344 my $cleanup = sub { unlink($newfile) };
345 open (OUT, ">", $newfile) || error("open $newfile: $!", $cleanup);
346 foreach my $data (values %feeds, values %guids) {
348 foreach my $field (keys %$data) {
349 if ($field eq "name" || $field eq "feed" ||
350 $field eq "guid" || $field eq "message") {
351 push @line, "$field=".encode_entities($data->{$field}, " \t\n");
353 elsif ($field eq "tags") {
354 push @line, "tag=$_" foreach @{$data->{tags}};
357 push @line, "$field=".$data->{$field}
358 if defined $data->{$field};
361 print OUT join(" ", @line)."\n" || error("write $newfile: $!", $cleanup);
363 close OUT || error("save $newfile: $!", $cleanup);
364 rename($newfile, "$config{wikistatedir}/aggregate") ||
365 error("rename $newfile: $!", $cleanup);
368 foreach my $feed (keys %feeds) {
369 my $t=$feeds{$feed}->{lastupdate}+$feeds{$feed}->{updateinterval};
370 if (! defined $timestamp || $timestamp > $t) {
374 $newfile=~s/\.new$/time/;
375 open (OUT, ">", $newfile) || error("open $newfile: $!", $cleanup);
376 if (defined $timestamp) {
377 print OUT $timestamp."\n";
379 close OUT || error("save $newfile: $!", $cleanup);
382 sub garbage_collect () {
383 foreach my $name (keys %feeds) {
384 # remove any feeds that were not seen while building the pages
385 # that used to contain them
386 if ($feeds{$name}->{unseen}) {
387 delete $feeds{$name};
391 foreach my $guid (values %guids) {
392 # any guid whose feed is gone should be removed
393 if (! exists $feeds{$guid->{feed}}) {
394 if (exists $guid->{page}) {
395 unlink $IkiWiki::Plugin::transient::transientdir."/".htmlfn($guid->{page})
396 || unlink "$config{srcdir}/".htmlfn($guid->{page});
398 delete $guids{$guid->{guid}};
400 # handle expired guids
401 elsif ($guid->{expired} && exists $guid->{page}) {
402 unlink "$config{srcdir}/".htmlfn($guid->{page});
403 unlink $IkiWiki::Plugin::transient::transientdir."/".htmlfn($guid->{page});
404 delete $guid->{page};
411 # Load the current state in from disk, and merge into it
412 # values from the state in memory that might have changed
413 # during aggregation.
419 # All that can change in feed state during aggregation is a few
421 foreach my $name (keys %myfeeds) {
422 if (exists $feeds{$name}) {
423 foreach my $field (qw{message lastupdate lasttry
424 numposts newposts error}) {
425 $feeds{$name}->{$field}=$myfeeds{$name}->{$field};
430 # New guids can be created during aggregation.
431 # Guids have a few fields that may be updated during aggregation.
432 # It's also possible that guids were removed from the on-disk state
433 # while the aggregation was in process. That would only happen if
434 # their feed was also removed, so any removed guids added back here
435 # will be garbage collected later.
436 foreach my $guid (keys %myguids) {
437 if (! exists $guids{$guid}) {
438 $guids{$guid}=$myguids{$guid};
441 foreach my $field (qw{md5}) {
442 $guids{$guid}->{$field}=$myguids{$guid}->{$field};
455 foreach my $feed (values %feeds) {
456 next unless $feed->{expireage} || $feed->{expirecount};
459 foreach my $item (sort { ($IkiWiki::pagectime{$b->{page}} || 0) <=> ($IkiWiki::pagectime{$a->{page}} || 0) }
460 grep { exists $_->{page} && $_->{feed} eq $feed->{name} }
462 if ($feed->{expireage}) {
463 my $days_old = (time - ($IkiWiki::pagectime{$item->{page}} || 0)) / 60 / 60 / 24;
464 if ($days_old > $feed->{expireage}) {
465 debug(sprintf(gettext("expiring %s (%s days old)"),
466 $item->{page}, int($days_old)));
470 elsif ($feed->{expirecount} &&
471 $count >= $feed->{expirecount}) {
472 debug(sprintf(gettext("expiring %s"), $item->{page}));
476 if (! $seen{$item->{page}}) {
477 $seen{$item->{page}}=1;
485 sub needsaggregate () {
486 return values %feeds if $config{rebuild};
487 return grep { time - $_->{lastupdate} >= $_->{updateinterval} } values %feeds;
491 eval q{use Net::INET6Glue::INET_is_INET6}; # may not be available
492 eval q{use XML::Feed};
494 eval q{use URI::Fetch};
497 foreach my $feed (@_) {
498 $feed->{lasttry}=time;
500 $feed->{message}=sprintf(gettext("last checked %s"),
501 displaytime($feed->{lasttry}));
504 debug(sprintf(gettext("checking feed %s ..."), $feed->{name}));
506 if (! length $feed->{feedurl}) {
507 my @urls=XML::Feed->find_feeds($feed->{url});
509 $feed->{message}=sprintf(gettext("could not find feed at %s"), $feed->{url});
511 debug($feed->{message});
514 $feed->{feedurl}=pop @urls;
517 my $res=URI::Fetch->fetch($feed->{feedurl}, UserAgent=>$ua);
519 $feed->{message}=URI::Fetch->errstr;
521 debug($feed->{message});
525 # lastupdate is only set if we were able to contact the server
526 $feed->{lastupdate}=$feed->{lasttry};
528 if ($res->status == URI::Fetch::URI_GONE()) {
529 $feed->{message}=gettext("feed not found");
531 debug($feed->{message});
534 my $content=$res->content;
535 my $f=eval{XML::Feed->parse(\$content)};
537 # One common cause of XML::Feed crashing is a feed
538 # that contains invalid UTF-8 sequences. Convert
539 # feed to ascii to try to work around.
540 $feed->{message}.=" ".sprintf(gettext("(invalid UTF-8 stripped from feed)"));
542 $content=Encode::decode_utf8($content, 0);
543 XML::Feed->parse(\$content)
547 # Another possibility is badly escaped entities.
548 $feed->{message}.=" ".sprintf(gettext("(feed entities escaped)"));
549 $content=~s/\&(?!amp)(\w+);/&$1;/g;
551 $content=Encode::decode_utf8($content, 0);
552 XML::Feed->parse(\$content)
556 $feed->{message}=gettext("feed crashed XML::Feed!")." ($@)";
558 debug($feed->{message});
562 $feed->{message}=XML::Feed->errstr;
564 debug($feed->{message});
568 foreach my $entry ($f->entries) {
569 # XML::Feed doesn't work around XML::Atom's bizarre
570 # API, so we will. Real unicode strings? Yes please.
571 # See [[bugs/Aggregated_Atom_feeds_are_double-encoded]]
572 local $XML::Atom::ForceUnicode = 1;
574 my $c=$entry->content;
575 # atom feeds may have no content, only a summary
576 if (! defined $c && ref $entry->summary) {
582 copyright => $f->copyright,
583 title => defined $entry->title ? decode_entities($entry->title) : "untitled",
584 author => defined $entry->author ? decode_entities($entry->author) : "",
585 link => $entry->link,
586 content => (defined $c && defined $c->body) ? $c->body : "",
587 guid => defined $entry->id ? $entry->id : time."_".$feed->{name},
588 ctime => $entry->issued ? ($entry->issued->epoch || time) : time,
589 base => (defined $c && $c->can("base")) ? $c->base : undef,
598 my $feed=$params{feed};
601 if (exists $guids{$params{guid}}) {
602 # updating an existing post
603 $guid=$guids{$params{guid}};
604 return if $guid->{expired};
605 write_page($feed, $guid, $mtime, \%params);
609 $guid->{guid}=$params{guid};
610 $guids{$params{guid}}=$guid;
611 $mtime=$params{ctime};
615 # assign it an unused page
616 my $page=titlepage($params{title});
617 # escape slashes and periods in title so it doesn't specify
618 # directory name or trigger ".." disallowing code.
619 $page=~s!([/.])!"__".ord($1)."__"!eg;
620 $page=$feed->{dir}."/".$page;
621 ($page)=$page=~/$config{wiki_file_regexp}/;
622 if (! defined $page || ! length $page) {
623 $page=$feed->{dir}."/item";
626 while (exists $IkiWiki::pagecase{lc $page.$c} ||
627 -e $IkiWiki::Plugin::transient::transientdir."/".htmlfn($page.$c) ||
628 -e "$config{srcdir}/".htmlfn($page.$c)) {
634 eval { write_page($feed, $guid, $mtime, \%params) };
636 # assume failure was due to a too long filename
638 $page=$feed->{dir}."/item";
639 while (exists $IkiWiki::pagecase{lc $page.$c} ||
640 -e $IkiWiki::Plugin::transient::transientdir."/".htmlfn($page.$c) ||
641 -e "$config{srcdir}/".htmlfn($page.$c)) {
647 write_page($feed, $guid, $mtime, \%params);
650 debug(sprintf(gettext("creating new page %s"), $page));
654 sub write_page ($$$$$) {
658 my %params=%{shift()};
660 $guid->{feed}=$feed->{name};
662 # To write or not to write? Need to avoid writing unchanged pages
663 # to avoid unneccessary rebuilding. The mtime from rss cannot be
664 # trusted; let's use a digest.
665 eval q{use Digest::MD5 'md5_hex'};
668 my $digest=md5_hex(Encode::encode_utf8($params{content}));
669 return unless ! exists $guid->{md5} || $guid->{md5} ne $digest || $config{rebuild};
670 $guid->{md5}=$digest;
675 $template=template($feed->{template}, blind_cache => 1);
678 print STDERR gettext("failed to process template:")." $@";
681 $template->param(title => $params{title})
682 if defined $params{title} && length($params{title});
683 $template->param(author => $params{author})
684 if defined $params{author} && length($params{author}
685 && $params{author} ne $feed->{name});
686 $template->param(content => wikiescape(htmlabs($params{content},
687 defined $params{base} ? $params{base} : $feed->{feedurl})));
688 $template->param(name => $feed->{name});
689 $template->param(url => $feed->{url});
690 $template->param(copyright => $params{copyright})
691 if defined $params{copyright} && length $params{copyright};
692 $template->param(permalink => IkiWiki::urlabs($params{link}, $feed->{feedurl}))
693 if defined $params{link};
694 if (ref $feed->{tags}) {
695 $template->param(tags => [map { tag => $_ }, @{$feed->{tags}}]);
697 writefile(htmlfn($guid->{page}),
698 $IkiWiki::Plugin::transient::transientdir, $template->output);
700 if (defined $mtime && $mtime <= time) {
701 # Set the mtime, this lets the build process get the right
702 # creation time on record for the new page.
703 utime $mtime, $mtime,
704 $IkiWiki::Plugin::transient::transientdir."/".htmlfn($guid->{page});
705 # Store it in pagectime for expiry code to use also.
706 $IkiWiki::pagectime{$guid->{page}}=$mtime
707 unless exists $IkiWiki::pagectime{$guid->{page}};
710 # Dummy value for expiry code.
711 $IkiWiki::pagectime{$guid->{page}}=time
712 unless exists $IkiWiki::pagectime{$guid->{page}};
717 # escape accidental wikilinks and preprocessor stuff
718 return encode_entities(shift, '\[\]');
722 # Convert links in html from relative to absolute.
723 # Note that this is a heuristic, which is not specified by the rss
724 # spec and may not be right for all feeds. Also, see Debian
730 my $p = HTML::Parser->new(api_version => 3);
731 $p->handler(default => sub { $ret.=join("", @_) }, "text");
732 $p->handler(start => sub {
733 my ($tagname, $pos, $text) = @_;
734 if (ref $HTML::Tagset::linkElements{$tagname}) {
736 # use attribute sets from right to left
737 # to avoid invalidating the offsets
738 # when replacing the values
739 my($k_offset, $k_len, $v_offset, $v_len) =
741 my $attrname = lc(substr($text, $k_offset, $k_len));
742 next unless grep { $_ eq $attrname } @{$HTML::Tagset::linkElements{$tagname}};
743 next unless $v_offset; # 0 v_offset means no value
744 my $v = substr($text, $v_offset, $v_len);
745 $v =~ s/^([\'\"])(.*)\1$/$2/;
746 my $new_v=IkiWiki::urlabs($v, $urlbase);
747 $new_v =~ s/\"/"/g; # since we quote with ""
748 substr($text, $v_offset, $v_len) = qq("$new_v");
752 }, "tagname, tokenpos, text");
760 return shift().".".($config{aggregateinternal} ? "_aggregated" : $config{htmlext});
765 sub lockaggregate () {
766 # Take an exclusive lock to prevent multiple concurrent aggregators.
767 # Returns true if the lock was aquired.
768 if (! -d $config{wikistatedir}) {
769 mkdir($config{wikistatedir});
771 open($aggregatelock, '>', "$config{wikistatedir}/aggregatelock") ||
772 error ("cannot open to $config{wikistatedir}/aggregatelock: $!");
773 if (! flock($aggregatelock, 2 | 4)) { # LOCK_EX | LOCK_NB
774 close($aggregatelock) || error("failed closing aggregatelock: $!");
780 sub unlockaggregate () {
781 return close($aggregatelock) if $aggregatelock;