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 sub garbage_collect () {
362 foreach my $name (keys %feeds) {
363 # remove any feeds that were not seen while building the pages
364 # that used to contain them
365 if ($feeds{$name}->{unseen}) {
366 delete $feeds{$name};
370 foreach my $guid (values %guids) {
371 # any guid whose feed is gone should be removed
372 if (! exists $feeds{$guid->{feed}}) {
373 unlink "$config{srcdir}/".htmlfn($guid->{page})
374 if exists $guid->{page};
375 delete $guids{$guid->{guid}};
377 # handle expired guids
378 elsif ($guid->{expired} && exists $guid->{page}) {
379 unlink "$config{srcdir}/".htmlfn($guid->{page});
380 delete $guid->{page};
387 # Load the current state in from disk, and merge into it
388 # values from the state in memory that might have changed
389 # during aggregation.
395 # All that can change in feed state during aggregation is a few
397 foreach my $name (keys %myfeeds) {
398 if (exists $feeds{$name}) {
399 foreach my $field (qw{message lastupdate lasttry
400 numposts newposts error}) {
401 $feeds{$name}->{$field}=$myfeeds{$name}->{$field};
406 # New guids can be created during aggregation.
407 # It's also possible that guids were removed from the on-disk state
408 # while the aggregation was in process. That would only happen if
409 # their feed was also removed, so any removed guids added back here
410 # will be garbage collected later.
411 foreach my $guid (keys %myguids) {
412 if (! exists $guids{$guid}) {
413 $guids{$guid}=$myguids{$guid};
425 foreach my $feed (values %feeds) {
426 next unless $feed->{expireage} || $feed->{expirecount};
429 foreach my $item (sort { ($IkiWiki::pagectime{$b->{page}} || 0) <=> ($IkiWiki::pagectime{$a->{page}} || 0) }
430 grep { exists $_->{page} && $_->{feed} eq $feed->{name} }
432 if ($feed->{expireage}) {
433 my $days_old = (time - ($IkiWiki::pagectime{$item->{page}} || 0)) / 60 / 60 / 24;
434 if ($days_old > $feed->{expireage}) {
435 debug(sprintf(gettext("expiring %s (%s days old)"),
436 $item->{page}, int($days_old)));
440 elsif ($feed->{expirecount} &&
441 $count >= $feed->{expirecount}) {
442 debug(sprintf(gettext("expiring %s"), $item->{page}));
446 if (! $seen{$item->{page}}) {
447 $seen{$item->{page}}=1;
455 sub needsaggregate () {
456 return values %feeds if $config{rebuild};
457 return grep { time - $_->{lastupdate} >= $_->{updateinterval} } values %feeds;
461 eval q{use XML::Feed};
463 eval q{use URI::Fetch};
466 foreach my $feed (@_) {
467 $feed->{lasttry}=time;
469 $feed->{message}=sprintf(gettext("last checked %s"),
470 displaytime($feed->{lasttry}));
473 debug(sprintf(gettext("checking feed %s ..."), $feed->{name}));
475 if (! length $feed->{feedurl}) {
476 my @urls=XML::Feed->find_feeds($feed->{url});
478 $feed->{message}=sprintf(gettext("could not find feed at %s"), $feed->{url});
480 debug($feed->{message});
483 $feed->{feedurl}=pop @urls;
485 my $res=URI::Fetch->fetch($feed->{feedurl});
487 $feed->{message}=URI::Fetch->errstr;
489 debug($feed->{message});
493 # lastupdate is only set if we were able to contact the server
494 $feed->{lastupdate}=$feed->{lasttry};
496 if ($res->status == URI::Fetch::URI_GONE()) {
497 $feed->{message}=gettext("feed not found");
499 debug($feed->{message});
502 my $content=$res->content;
503 my $f=eval{XML::Feed->parse(\$content)};
505 # One common cause of XML::Feed crashing is a feed
506 # that contains invalid UTF-8 sequences. Convert
507 # feed to ascii to try to work around.
508 $feed->{message}.=" ".sprintf(gettext("(invalid UTF-8 stripped from feed)"));
510 $content=Encode::decode_utf8($content, 0);
511 XML::Feed->parse(\$content)
515 # Another possibility is badly escaped entities.
516 $feed->{message}.=" ".sprintf(gettext("(feed entities escaped)"));
517 $content=~s/\&(?!amp)(\w+);/&$1;/g;
519 $content=Encode::decode_utf8($content, 0);
520 XML::Feed->parse(\$content)
524 $feed->{message}=gettext("feed crashed XML::Feed!")." ($@)";
526 debug($feed->{message});
530 $feed->{message}=XML::Feed->errstr;
532 debug($feed->{message});
536 foreach my $entry ($f->entries) {
537 my $c=$entry->content;
538 # atom feeds may have no content, only a summary
539 if (! defined $c && ref $entry->summary) {
545 copyright => $f->copyright,
546 title => defined $entry->title ? decode_entities($entry->title) : "untitled",
547 link => $entry->link,
548 content => (defined $c && defined $c->body) ? $c->body : "",
549 guid => defined $entry->id ? $entry->id : time."_".$feed->{name},
550 ctime => $entry->issued ? ($entry->issued->epoch || time) : time,
551 base => (defined $c && $c->can("base")) ? $c->base : undef,
560 my $feed=$params{feed};
563 if (exists $guids{$params{guid}}) {
564 # updating an existing post
565 $guid=$guids{$params{guid}};
566 return if $guid->{expired};
570 $guid->{guid}=$params{guid};
571 $guids{$params{guid}}=$guid;
572 $mtime=$params{ctime};
576 # assign it an unused page
577 my $page=titlepage($params{title});
578 # escape slashes and periods in title so it doesn't specify
579 # directory name or trigger ".." disallowing code.
580 $page=~s!([/.])!"__".ord($1)."__"!eg;
581 $page=$feed->{dir}."/".$page;
582 ($page)=$page=~/$config{wiki_file_regexp}/;
583 if (! defined $page || ! length $page) {
584 $page=$feed->{dir}."/item";
587 while (exists $IkiWiki::pagecase{lc $page.$c} ||
588 -e "$config{srcdir}/".htmlfn($page.$c)) {
592 # Make sure that the file name isn't too long.
593 # NB: This doesn't check for path length limits.
594 my $max=POSIX::pathconf($config{srcdir}, &POSIX::_PC_NAME_MAX);
595 if (defined $max && length(htmlfn($page)) >= $max) {
597 $page=$feed->{dir}."/item";
598 while (exists $IkiWiki::pagecase{lc $page.$c} ||
599 -e "$config{srcdir}/".htmlfn($page.$c)) {
605 debug(sprintf(gettext("creating new page %s"), $page));
607 $guid->{feed}=$feed->{name};
609 # To write or not to write? Need to avoid writing unchanged pages
610 # to avoid unneccessary rebuilding. The mtime from rss cannot be
611 # trusted; let's use a digest.
612 eval q{use Digest::MD5 'md5_hex'};
615 my $digest=md5_hex(Encode::encode_utf8($params{content}));
616 return unless ! exists $guid->{md5} || $guid->{md5} ne $digest || $config{rebuild};
617 $guid->{md5}=$digest;
620 my $template=template($feed->{template}, blind_cache => 1);
621 $template->param(title => $params{title})
622 if defined $params{title} && length($params{title});
623 $template->param(content => wikiescape(htmlabs($params{content},
624 defined $params{base} ? $params{base} : $feed->{feedurl})));
625 $template->param(name => $feed->{name});
626 $template->param(url => $feed->{url});
627 $template->param(copyright => $params{copyright})
628 if defined $params{copyright} && length $params{copyright};
629 $template->param(permalink => urlabs($params{link}, $feed->{feedurl}))
630 if defined $params{link};
631 if (ref $feed->{tags}) {
632 $template->param(tags => [map { tag => $_ }, @{$feed->{tags}}]);
634 writefile(htmlfn($guid->{page}), $config{srcdir},
637 if (defined $mtime && $mtime <= time) {
638 # Set the mtime, this lets the build process get the right
639 # creation time on record for the new page.
640 utime $mtime, $mtime, "$config{srcdir}/".htmlfn($guid->{page});
641 # Store it in pagectime for expiry code to use also.
642 $IkiWiki::pagectime{$guid->{page}}=$mtime;
645 # Dummy value for expiry code.
646 $IkiWiki::pagectime{$guid->{page}}=time;
651 # escape accidental wikilinks and preprocessor stuff
652 return encode_entities(shift, '\[\]');
659 URI->new_abs($url, $urlbase)->as_string;
663 # Convert links in html from relative to absolute.
664 # Note that this is a heuristic, which is not specified by the rss
665 # spec and may not be right for all feeds. Also, see Debian
671 my $p = HTML::Parser->new(api_version => 3);
672 $p->handler(default => sub { $ret.=join("", @_) }, "text");
673 $p->handler(start => sub {
674 my ($tagname, $pos, $text) = @_;
675 if (ref $HTML::Tagset::linkElements{$tagname}) {
677 # use attribute sets from right to left
678 # to avoid invalidating the offsets
679 # when replacing the values
680 my($k_offset, $k_len, $v_offset, $v_len) =
682 my $attrname = lc(substr($text, $k_offset, $k_len));
683 next unless grep { $_ eq $attrname } @{$HTML::Tagset::linkElements{$tagname}};
684 next unless $v_offset; # 0 v_offset means no value
685 my $v = substr($text, $v_offset, $v_len);
686 $v =~ s/^([\'\"])(.*)\1$/$2/;
687 my $new_v=urlabs($v, $urlbase);
688 $new_v =~ s/\"/"/g; # since we quote with ""
689 substr($text, $v_offset, $v_len) = qq("$new_v");
693 }, "tagname, tokenpos, text");
701 return shift().".".($config{aggregateinternal} ? "_aggregated" : $config{htmlext});
706 sub lockaggregate () {
707 # Take an exclusive lock to prevent multiple concurrent aggregators.
708 # Returns true if the lock was aquired.
709 if (! -d $config{wikistatedir}) {
710 mkdir($config{wikistatedir});
712 open($aggregatelock, '>', "$config{wikistatedir}/aggregatelock") ||
713 error ("cannot open to $config{wikistatedir}/aggregatelock: $!");
714 if (! flock($aggregatelock, 2 | 4)) { # LOCK_EX | LOCK_NB
715 close($aggregatelock) || error("failed closing aggregatelock: $!");
721 sub unlockaggregate () {
722 return close($aggregatelock) if $aggregatelock;