2 # Feed aggregation plugin.
3 package IkiWiki::Plugin::aggregate;
12 use open qw{:utf8 :std};
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);
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 $config{aggregateinternal}=1;
68 if ($config{aggregate} && ! ($config{post_commit} &&
69 IkiWiki::commit_hook_enabled())) {
77 if (defined $cgi->param('do') &&
78 $cgi->param("do") eq "aggregate_webtrigger") {
80 print "Content-Type: text/plain\n\n";
84 print gettext("Aggregation triggered via web.")."\n\n";
85 if (launchaggregation()) {
88 require IkiWiki::Render;
93 print gettext("Nothing to do right now, all feeds are up-to-date!")."\n";
99 sub launchaggregation () {
100 # See if any feeds need aggregation.
102 my @feeds=needsaggregate();
103 return unless @feeds;
104 if (! lockaggregate()) {
105 debug("an aggregation process is already running");
108 # force a later rebuild of source pages
109 $IkiWiki::forcerebuild{$_->{sourcepage}}=1
112 # Fork a child process to handle the aggregation.
113 # The parent process will then handle building the
114 # result. This avoids messy code to clear state
115 # accumulated while aggregating.
116 defined(my $pid = fork) or error("Can't fork: $!");
118 IkiWiki::loadindex();
119 # Aggregation happens without the main wiki lock
120 # being held. This allows editing pages etc while
121 # aggregation is running.
125 # Merge changes, since aggregation state may have
126 # changed on disk while the aggregation was happening.
135 error "aggregation failed with code $?";
144 # Pages with extension _aggregated have plain html markup, pass through.
147 return $params{content};
150 # Used by ikiwiki-transition aggregateinternal.
151 sub migrate_to_internal {
152 if (! lockaggregate()) {
153 error("an aggregation process is currently running");
160 foreach my $data (values %guids) {
161 next unless $data->{page};
162 next if $data->{expired};
164 $config{aggregateinternal} = 0;
165 my $oldname = "$config{srcdir}/".htmlfn($data->{page});
166 my $oldoutput = $config{destdir}."/".IkiWiki::htmlpage($data->{page});
168 $config{aggregateinternal} = 1;
169 my $newname = "$config{srcdir}/".htmlfn($data->{page});
171 debug "moving $oldname -> $newname";
174 error("$newname already exists");
177 debug("already renamed to $newname?");
180 elsif (-e $oldname) {
181 rename($oldname, $newname) || error("$!");
184 debug("$oldname not found");
187 require IkiWiki::Render;
188 debug("removing output file $oldoutput");
189 IkiWiki::prune($oldoutput);
200 my $needsbuild=shift;
204 foreach my $feed (values %feeds) {
205 if (exists $pagesources{$feed->{sourcepage}} &&
206 grep { $_ eq $pagesources{$feed->{sourcepage}} } @$needsbuild) {
207 # Mark all feeds originating on this page as
208 # not yet seen; preprocess will unmark those that
210 markunseen($feed->{sourcepage});
218 foreach my $required (qw{name url}) {
219 if (! exists $params{$required}) {
220 error sprintf(gettext("missing %s parameter"), $required)
225 my $name=$params{name};
226 if (exists $feeds{$name}) {
233 $feed->{sourcepage}=$params{page};
234 $feed->{url}=$params{url};
235 my $dir=exists $params{dir} ? $params{dir} : $params{page}."/".titlepage($params{name});
237 ($dir)=$dir=~/$config{wiki_file_regexp}/;
239 $feed->{feedurl}=defined $params{feedurl} ? $params{feedurl} : "";
240 $feed->{updateinterval}=defined $params{updateinterval} ? $params{updateinterval} * 60 : 15 * 60;
241 $feed->{expireage}=defined $params{expireage} ? $params{expireage} : 0;
242 $feed->{expirecount}=defined $params{expirecount} ? $params{expirecount} : 0;
243 if (exists $params{template}) {
244 $params{template}=~s/[^-_a-zA-Z0-9]+//g;
247 $params{template} = "aggregatepost"
249 $feed->{template}=$params{template} . ".tmpl";
250 delete $feed->{unseen};
251 $feed->{lastupdate}=0 unless defined $feed->{lastupdate};
252 $feed->{lasttry}=$feed->{lastupdate} unless defined $feed->{lasttry};
253 $feed->{numposts}=0 unless defined $feed->{numposts};
254 $feed->{newposts}=0 unless defined $feed->{newposts};
255 $feed->{message}=gettext("new feed") unless defined $feed->{message};
256 $feed->{error}=0 unless defined $feed->{error};
262 push @{$feed->{tags}}, $value;
266 return "<a href=\"".$feed->{url}."\">".$feed->{name}."</a>: ".
267 ($feed->{error} ? "<em>" : "").$feed->{message}.
268 ($feed->{error} ? "</em>" : "").
269 " (".$feed->{numposts}." ".gettext("posts").
270 ($feed->{newposts} ? "; ".$feed->{newposts}.
271 " ".gettext("new") : "").
278 # Remove feed data for removed pages.
279 foreach my $file (@files) {
280 my $page=pagename($file);
288 foreach my $id (keys %feeds) {
289 if ($feeds{$id}->{sourcepage} eq $page) {
290 $feeds{$id}->{unseen}=1;
298 return if $state_loaded;
300 if (-e "$config{wikistatedir}/aggregate") {
301 open(IN, "<", "$config{wikistatedir}/aggregate") ||
302 die "$config{wikistatedir}/aggregate: $!";
304 $_=IkiWiki::possibly_foolish_untaint($_);
307 foreach my $i (split(/ /, $_)) {
308 my ($field, $val)=split(/=/, $i, 2);
309 if ($field eq "name" || $field eq "feed" ||
310 $field eq "guid" || $field eq "message") {
311 $data->{$field}=decode_entities($val, " \t\n");
313 elsif ($field eq "tag") {
314 push @{$data->{tags}}, $val;
317 $data->{$field}=$val;
321 if (exists $data->{name}) {
322 $feeds{$data->{name}}=$data;
324 elsif (exists $data->{guid}) {
325 $guids{$data->{guid}}=$data;
334 return unless $state_loaded;
336 my $newfile="$config{wikistatedir}/aggregate.new";
337 my $cleanup = sub { unlink($newfile) };
338 open (OUT, ">", $newfile) || error("open $newfile: $!", $cleanup);
339 foreach my $data (values %feeds, values %guids) {
341 foreach my $field (keys %$data) {
342 if ($field eq "name" || $field eq "feed" ||
343 $field eq "guid" || $field eq "message") {
344 push @line, "$field=".encode_entities($data->{$field}, " \t\n");
346 elsif ($field eq "tags") {
347 push @line, "tag=$_" foreach @{$data->{tags}};
350 push @line, "$field=".$data->{$field}
351 if defined $data->{$field};
354 print OUT join(" ", @line)."\n" || error("write $newfile: $!", $cleanup);
356 close OUT || error("save $newfile: $!", $cleanup);
357 rename($newfile, "$config{wikistatedir}/aggregate") ||
358 error("rename $newfile: $!", $cleanup);
361 foreach my $feed (keys %feeds) {
362 my $t=$feeds{$feed}->{lastupdate}+$feeds{$feed}->{updateinterval};
363 if (! defined $timestamp || $timestamp > $t) {
367 $newfile=~s/\.new$/time/;
368 open (OUT, ">", $newfile) || error("open $newfile: $!", $cleanup);
369 if (defined $timestamp) {
370 print OUT $timestamp."\n";
372 close OUT || error("save $newfile: $!", $cleanup);
375 sub garbage_collect () {
376 foreach my $name (keys %feeds) {
377 # remove any feeds that were not seen while building the pages
378 # that used to contain them
379 if ($feeds{$name}->{unseen}) {
380 delete $feeds{$name};
384 foreach my $guid (values %guids) {
385 # any guid whose feed is gone should be removed
386 if (! exists $feeds{$guid->{feed}}) {
387 unlink "$config{srcdir}/".htmlfn($guid->{page})
388 if exists $guid->{page};
389 delete $guids{$guid->{guid}};
391 # handle expired guids
392 elsif ($guid->{expired} && exists $guid->{page}) {
393 unlink "$config{srcdir}/".htmlfn($guid->{page});
394 delete $guid->{page};
401 # Load the current state in from disk, and merge into it
402 # values from the state in memory that might have changed
403 # during aggregation.
409 # All that can change in feed state during aggregation is a few
411 foreach my $name (keys %myfeeds) {
412 if (exists $feeds{$name}) {
413 foreach my $field (qw{message lastupdate lasttry
414 numposts newposts error}) {
415 $feeds{$name}->{$field}=$myfeeds{$name}->{$field};
420 # New guids can be created during aggregation.
421 # Guids have a few fields that may be updated during aggregation.
422 # It's also possible that guids were removed from the on-disk state
423 # while the aggregation was in process. That would only happen if
424 # their feed was also removed, so any removed guids added back here
425 # will be garbage collected later.
426 foreach my $guid (keys %myguids) {
427 if (! exists $guids{$guid}) {
428 $guids{$guid}=$myguids{$guid};
431 foreach my $field (qw{md5}) {
432 $guids{$guid}->{$field}=$myguids{$guid}->{$field};
445 foreach my $feed (values %feeds) {
446 next unless $feed->{expireage} || $feed->{expirecount};
449 foreach my $item (sort { ($IkiWiki::pagectime{$b->{page}} || 0) <=> ($IkiWiki::pagectime{$a->{page}} || 0) }
450 grep { exists $_->{page} && $_->{feed} eq $feed->{name} }
452 if ($feed->{expireage}) {
453 my $days_old = (time - ($IkiWiki::pagectime{$item->{page}} || 0)) / 60 / 60 / 24;
454 if ($days_old > $feed->{expireage}) {
455 debug(sprintf(gettext("expiring %s (%s days old)"),
456 $item->{page}, int($days_old)));
460 elsif ($feed->{expirecount} &&
461 $count >= $feed->{expirecount}) {
462 debug(sprintf(gettext("expiring %s"), $item->{page}));
466 if (! $seen{$item->{page}}) {
467 $seen{$item->{page}}=1;
475 sub needsaggregate () {
476 return values %feeds if $config{rebuild};
477 return grep { time - $_->{lastupdate} >= $_->{updateinterval} } values %feeds;
481 eval q{use XML::Feed};
483 eval q{use URI::Fetch};
486 foreach my $feed (@_) {
487 $feed->{lasttry}=time;
489 $feed->{message}=sprintf(gettext("last checked %s"),
490 displaytime($feed->{lasttry}));
493 debug(sprintf(gettext("checking feed %s ..."), $feed->{name}));
495 if (! length $feed->{feedurl}) {
496 my @urls=XML::Feed->find_feeds($feed->{url});
498 $feed->{message}=sprintf(gettext("could not find feed at %s"), $feed->{url});
500 debug($feed->{message});
503 $feed->{feedurl}=pop @urls;
505 my $res=URI::Fetch->fetch($feed->{feedurl});
507 $feed->{message}=URI::Fetch->errstr;
509 debug($feed->{message});
513 # lastupdate is only set if we were able to contact the server
514 $feed->{lastupdate}=$feed->{lasttry};
516 if ($res->status == URI::Fetch::URI_GONE()) {
517 $feed->{message}=gettext("feed not found");
519 debug($feed->{message});
522 my $content=$res->content;
523 my $f=eval{XML::Feed->parse(\$content)};
525 # One common cause of XML::Feed crashing is a feed
526 # that contains invalid UTF-8 sequences. Convert
527 # feed to ascii to try to work around.
528 $feed->{message}.=" ".sprintf(gettext("(invalid UTF-8 stripped from feed)"));
530 $content=Encode::decode_utf8($content, 0);
531 XML::Feed->parse(\$content)
535 # Another possibility is badly escaped entities.
536 $feed->{message}.=" ".sprintf(gettext("(feed entities escaped)"));
537 $content=~s/\&(?!amp)(\w+);/&$1;/g;
539 $content=Encode::decode_utf8($content, 0);
540 XML::Feed->parse(\$content)
544 $feed->{message}=gettext("feed crashed XML::Feed!")." ($@)";
546 debug($feed->{message});
550 $feed->{message}=XML::Feed->errstr;
552 debug($feed->{message});
556 foreach my $entry ($f->entries) {
557 # XML::Feed doesn't work around XML::Atom's bizarre
558 # API, so we will. Real unicode strings? Yes please.
559 # See [[bugs/Aggregated_Atom_feeds_are_double-encoded]]
560 local $XML::Atom::ForceUnicode = 1;
562 my $c=$entry->content;
563 # atom feeds may have no content, only a summary
564 if (! defined $c && ref $entry->summary) {
570 copyright => $f->copyright,
571 title => defined $entry->title ? decode_entities($entry->title) : "untitled",
572 link => $entry->link,
573 content => (defined $c && defined $c->body) ? $c->body : "",
574 guid => defined $entry->id ? $entry->id : time."_".$feed->{name},
575 ctime => $entry->issued ? ($entry->issued->epoch || time) : time,
576 base => (defined $c && $c->can("base")) ? $c->base : undef,
585 my $feed=$params{feed};
588 if (exists $guids{$params{guid}}) {
589 # updating an existing post
590 $guid=$guids{$params{guid}};
591 return if $guid->{expired};
595 $guid->{guid}=$params{guid};
596 $guids{$params{guid}}=$guid;
597 $mtime=$params{ctime};
601 # assign it an unused page
602 my $page=titlepage($params{title});
603 # escape slashes and periods in title so it doesn't specify
604 # directory name or trigger ".." disallowing code.
605 $page=~s!([/.])!"__".ord($1)."__"!eg;
606 $page=$feed->{dir}."/".$page;
607 ($page)=$page=~/$config{wiki_file_regexp}/;
608 if (! defined $page || ! length $page) {
609 $page=$feed->{dir}."/item";
612 while (exists $IkiWiki::pagecase{lc $page.$c} ||
613 -e "$config{srcdir}/".htmlfn($page.$c)) {
617 # Make sure that the file name isn't too long.
618 # NB: This doesn't check for path length limits.
619 my $max=POSIX::pathconf($config{srcdir}, &POSIX::_PC_NAME_MAX);
620 if (defined $max && length(htmlfn($page)) >= $max) {
622 $page=$feed->{dir}."/item";
623 while (exists $IkiWiki::pagecase{lc $page.$c} ||
624 -e "$config{srcdir}/".htmlfn($page.$c)) {
630 debug(sprintf(gettext("creating new page %s"), $page));
632 $guid->{feed}=$feed->{name};
634 # To write or not to write? Need to avoid writing unchanged pages
635 # to avoid unneccessary rebuilding. The mtime from rss cannot be
636 # trusted; let's use a digest.
637 eval q{use Digest::MD5 'md5_hex'};
640 my $digest=md5_hex(Encode::encode_utf8($params{content}));
641 return unless ! exists $guid->{md5} || $guid->{md5} ne $digest || $config{rebuild};
642 $guid->{md5}=$digest;
645 my $template=template($feed->{template}, blind_cache => 1);
646 $template->param(title => $params{title})
647 if defined $params{title} && length($params{title});
648 $template->param(content => wikiescape(htmlabs($params{content},
649 defined $params{base} ? $params{base} : $feed->{feedurl})));
650 $template->param(name => $feed->{name});
651 $template->param(url => $feed->{url});
652 $template->param(copyright => $params{copyright})
653 if defined $params{copyright} && length $params{copyright};
654 $template->param(permalink => urlabs($params{link}, $feed->{feedurl}))
655 if defined $params{link};
656 if (ref $feed->{tags}) {
657 $template->param(tags => [map { tag => $_ }, @{$feed->{tags}}]);
659 writefile(htmlfn($guid->{page}), $config{srcdir},
662 if (defined $mtime && $mtime <= time) {
663 # Set the mtime, this lets the build process get the right
664 # creation time on record for the new page.
665 utime $mtime, $mtime, "$config{srcdir}/".htmlfn($guid->{page});
666 # Store it in pagectime for expiry code to use also.
667 $IkiWiki::pagectime{$guid->{page}}=$mtime
668 unless exists $IkiWiki::pagectime{$guid->{page}};
671 # Dummy value for expiry code.
672 $IkiWiki::pagectime{$guid->{page}}=time
673 unless exists $IkiWiki::pagectime{$guid->{page}};
678 # escape accidental wikilinks and preprocessor stuff
679 return encode_entities(shift, '\[\]');
686 URI->new_abs($url, $urlbase)->as_string;
690 # Convert links in html from relative to absolute.
691 # Note that this is a heuristic, which is not specified by the rss
692 # spec and may not be right for all feeds. Also, see Debian
698 my $p = HTML::Parser->new(api_version => 3);
699 $p->handler(default => sub { $ret.=join("", @_) }, "text");
700 $p->handler(start => sub {
701 my ($tagname, $pos, $text) = @_;
702 if (ref $HTML::Tagset::linkElements{$tagname}) {
704 # use attribute sets from right to left
705 # to avoid invalidating the offsets
706 # when replacing the values
707 my($k_offset, $k_len, $v_offset, $v_len) =
709 my $attrname = lc(substr($text, $k_offset, $k_len));
710 next unless grep { $_ eq $attrname } @{$HTML::Tagset::linkElements{$tagname}};
711 next unless $v_offset; # 0 v_offset means no value
712 my $v = substr($text, $v_offset, $v_len);
713 $v =~ s/^([\'\"])(.*)\1$/$2/;
714 my $new_v=urlabs($v, $urlbase);
715 $new_v =~ s/\"/"/g; # since we quote with ""
716 substr($text, $v_offset, $v_len) = qq("$new_v");
720 }, "tagname, tokenpos, text");
728 return shift().".".($config{aggregateinternal} ? "_aggregated" : $config{htmlext});
733 sub lockaggregate () {
734 # Take an exclusive lock to prevent multiple concurrent aggregators.
735 # Returns true if the lock was aquired.
736 if (! -d $config{wikistatedir}) {
737 mkdir($config{wikistatedir});
739 open($aggregatelock, '>', "$config{wikistatedir}/aggregatelock") ||
740 error ("cannot open to $config{wikistatedir}/aggregatelock: $!");
741 if (! flock($aggregatelock, 2 | 4)) { # LOCK_EX | LOCK_NB
742 close($aggregatelock) || error("failed closing aggregatelock: $!");
748 sub unlockaggregate () {
749 return close($aggregatelock) if $aggregatelock;