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} && $config{aggregate_webtrigger}) {
27 hook(type => "cgi", id => "aggregate", call => \&cgi);
32 eval q{use Getopt::Long};
34 Getopt::Long::Configure('pass_through');
36 "aggregate" => \$config{aggregate},
37 "aggregateinternal!" => \$config{aggregateinternal},
41 sub getsetup () { #{{{
47 aggregateinternal => {
50 description => "enable aggregation to internal pages?",
51 safe => 0, # enabling needs manual transition
54 aggregate_webtrigger => {
57 description => "allow aggregation to be triggered via the web?",
63 sub checkconfig () { #{{{
64 if ($config{aggregate} && ! ($config{post_commit} &&
65 IkiWiki::commit_hook_enabled())) {
73 if (defined $cgi->param('do') &&
74 $cgi->param("do") eq "aggregate_webtrigger") {
76 print "Content-Type: text/plain\n\n";
80 print gettext("Aggregation triggered via web.")."\n\n";
81 if (launchaggregation()) {
84 require IkiWiki::Render;
89 print gettext("Nothing to do right now, all feeds are up-to-date!")."\n";
95 sub launchaggregation () { #{{{
96 # See if any feeds need aggregation.
98 my @feeds=needsaggregate();
100 if (! lockaggregate()) {
101 debug("an aggregation process is already running");
104 # force a later rebuild of source pages
105 $IkiWiki::forcerebuild{$_->{sourcepage}}=1
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: $!");
114 IkiWiki::loadindex();
115 # Aggregation happens without the main wiki lock
116 # being held. This allows editing pages etc while
117 # aggregation is running.
121 # Merge changes, since aggregation state may have
122 # changed on disk while the aggregation was happening.
131 error "aggregation failed with code $?";
140 # Pages with extension _aggregated have plain html markup, pass through.
141 sub htmlize (@) { #{{{
143 return $params{content};
146 # Used by ikiwiki-transition aggregateinternal.
147 sub migrate_to_internal { #{{{
148 if (! lockaggregate()) {
149 error("an aggregation process is currently running");
156 foreach my $data (values %guids) {
157 next unless $data->{page};
158 next if $data->{expired};
160 $config{aggregateinternal} = 0;
161 my $oldname = pagefile($data->{page});
162 my $oldoutput = $config{destdir}."/".IkiWiki::htmlpage($data->{page});
164 $config{aggregateinternal} = 1;
165 my $newname = pagefile($data->{page});
167 debug "moving $oldname -> $newname";
170 error("$newname already exists");
173 debug("already renamed to $newname?");
176 elsif (-e $oldname) {
177 rename($oldname, $newname) || error("$!");
180 debug("$oldname not found");
183 require IkiWiki::Render;
184 debug("removing output file $oldoutput");
185 IkiWiki::prune($oldoutput);
195 sub needsbuild (@) { #{{{
196 my $needsbuild=shift;
200 foreach my $feed (values %feeds) {
201 if (exists $pagesources{$feed->{sourcepage}} &&
202 grep { $_ eq $pagesources{$feed->{sourcepage}} } @$needsbuild) {
203 # Mark all feeds originating on this page as
204 # not yet seen; preprocess will unmark those that
206 markunseen($feed->{sourcepage});
211 sub preprocess (@) { #{{{
214 foreach my $required (qw{name url}) {
215 if (! exists $params{$required}) {
216 error sprintf(gettext("missing %s parameter"), $required)
221 my $name=$params{name};
222 if (exists $feeds{$name}) {
229 $feed->{sourcepage}=$params{page};
230 $feed->{url}=$params{url};
231 my $dir=exists $params{dir} ? $params{dir} : $params{page}."/".IkiWiki::titlepage($params{name});
233 ($dir)=$dir=~/$config{wiki_file_regexp}/;
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;
243 $params{template} = "aggregatepost"
245 $feed->{template}=$params{template} . ".tmpl";
246 delete $feed->{unseen};
247 $feed->{lastupdate}=0 unless defined $feed->{lastupdate};
248 $feed->{numposts}=0 unless defined $feed->{numposts};
249 $feed->{newposts}=0 unless defined $feed->{newposts};
250 $feed->{message}=gettext("new feed") unless defined $feed->{message};
251 $feed->{error}=0 unless defined $feed->{error};
257 push @{$feed->{tags}}, $value;
261 return "<a href=\"".$feed->{url}."\">".$feed->{name}."</a>: ".
262 ($feed->{error} ? "<em>" : "").$feed->{message}.
263 ($feed->{error} ? "</em>" : "").
264 " (".$feed->{numposts}." ".gettext("posts").
265 ($feed->{newposts} ? "; ".$feed->{newposts}.
266 " ".gettext("new") : "").
270 sub delete (@) { #{{{
273 # Remove feed data for removed pages.
274 foreach my $file (@files) {
275 my $page=pagename($file);
280 sub markunseen ($) { #{{{
283 foreach my $id (keys %feeds) {
284 if ($feeds{$id}->{sourcepage} eq $page) {
285 $feeds{$id}->{unseen}=1;
292 sub loadstate () { #{{{
293 return if $state_loaded;
295 if (-e "$config{wikistatedir}/aggregate") {
296 open(IN, "$config{wikistatedir}/aggregate") ||
297 die "$config{wikistatedir}/aggregate: $!";
299 $_=IkiWiki::possibly_foolish_untaint($_);
302 foreach my $i (split(/ /, $_)) {
303 my ($field, $val)=split(/=/, $i, 2);
304 if ($field eq "name" || $field eq "feed" ||
305 $field eq "guid" || $field eq "message") {
306 $data->{$field}=decode_entities($val, " \t\n");
308 elsif ($field eq "tag") {
309 push @{$data->{tags}}, $val;
312 $data->{$field}=$val;
316 if (exists $data->{name}) {
317 $feeds{$data->{name}}=$data;
319 elsif (exists $data->{guid}) {
320 $guids{$data->{guid}}=$data;
328 sub savestate () { #{{{
329 return unless $state_loaded;
331 my $newfile="$config{wikistatedir}/aggregate.new";
332 my $cleanup = sub { unlink($newfile) };
333 open (OUT, ">$newfile") || error("open $newfile: $!", $cleanup);
334 foreach my $data (values %feeds, values %guids) {
336 foreach my $field (keys %$data) {
337 if ($field eq "name" || $field eq "feed" ||
338 $field eq "guid" || $field eq "message") {
339 push @line, "$field=".encode_entities($data->{$field}, " \t\n");
341 elsif ($field eq "tags") {
342 push @line, "tag=$_" foreach @{$data->{tags}};
345 push @line, "$field=".$data->{$field};
348 print OUT join(" ", @line)."\n" || error("write $newfile: $!", $cleanup);
350 close OUT || error("save $newfile: $!", $cleanup);
351 rename($newfile, "$config{wikistatedir}/aggregate") ||
352 error("rename $newfile: $!", $cleanup);
355 sub garbage_collect () { #{{{
356 foreach my $name (keys %feeds) {
357 # remove any feeds that were not seen while building the pages
358 # that used to contain them
359 if ($feeds{$name}->{unseen}) {
360 delete $feeds{$name};
364 foreach my $guid (values %guids) {
365 # any guid whose feed is gone should be removed
366 if (! exists $feeds{$guid->{feed}}) {
367 unlink pagefile($guid->{page})
368 if exists $guid->{page};
369 delete $guids{$guid->{guid}};
371 # handle expired guids
372 elsif ($guid->{expired} && exists $guid->{page}) {
373 unlink pagefile($guid->{page});
374 delete $guid->{page};
380 sub mergestate () { #{{{
381 # Load the current state in from disk, and merge into it
382 # values from the state in memory that might have changed
383 # during aggregation.
389 # All that can change in feed state during aggregation is a few
391 foreach my $name (keys %myfeeds) {
392 if (exists $feeds{$name}) {
393 foreach my $field (qw{message lastupdate numposts
395 $feeds{$name}->{$field}=$myfeeds{$name}->{$field};
400 # New guids can be created during aggregation.
401 # It's also possible that guids were removed from the on-disk state
402 # while the aggregation was in process. That would only happen if
403 # their feed was also removed, so any removed guids added back here
404 # will be garbage collected later.
405 foreach my $guid (keys %myguids) {
406 if (! exists $guids{$guid}) {
407 $guids{$guid}=$myguids{$guid};
412 sub clearstate () { #{{{
419 foreach my $feed (values %feeds) {
420 next unless $feed->{expireage} || $feed->{expirecount};
423 foreach my $item (sort { $IkiWiki::pagectime{$b->{page}} <=> $IkiWiki::pagectime{$a->{page}} }
424 grep { exists $_->{page} && $_->{feed} eq $feed->{name} }
426 if ($feed->{expireage}) {
427 my $days_old = (time - $IkiWiki::pagectime{$item->{page}}) / 60 / 60 / 24;
428 if ($days_old > $feed->{expireage}) {
429 debug(sprintf(gettext("expiring %s (%s days old)"),
430 $item->{page}, int($days_old)));
434 elsif ($feed->{expirecount} &&
435 $count >= $feed->{expirecount}) {
436 debug(sprintf(gettext("expiring %s"), $item->{page}));
440 if (! $seen{$item->{page}}) {
441 $seen{$item->{page}}=1;
449 sub needsaggregate () { #{{{
450 return values %feeds if $config{rebuild};
451 return grep { time - $_->{lastupdate} >= $_->{updateinterval} } values %feeds;
454 sub aggregate (@) { #{{{
455 eval q{use XML::Feed};
457 eval q{use URI::Fetch};
460 foreach my $feed (@_) {
461 $feed->{lastupdate}=time;
463 $feed->{message}=sprintf(gettext("processed ok at %s"),
464 displaytime($feed->{lastupdate}));
467 debug(sprintf(gettext("checking feed %s ..."), $feed->{name}));
469 if (! length $feed->{feedurl}) {
470 my @urls=XML::Feed->find_feeds($feed->{url});
472 $feed->{message}=sprintf(gettext("could not find feed at %s"), $feed->{url});
474 debug($feed->{message});
477 $feed->{feedurl}=pop @urls;
479 my $res=URI::Fetch->fetch($feed->{feedurl});
481 $feed->{message}=URI::Fetch->errstr;
483 debug($feed->{message});
486 if ($res->status == URI::Fetch::URI_GONE()) {
487 $feed->{message}=gettext("feed not found");
489 debug($feed->{message});
492 my $content=$res->content;
493 my $f=eval{XML::Feed->parse(\$content)};
495 # One common cause of XML::Feed crashing is a feed
496 # that contains invalid UTF-8 sequences. Convert
497 # feed to ascii to try to work around.
498 $feed->{message}.=" ".sprintf(gettext("(invalid UTF-8 stripped from feed)"));
499 $content=Encode::decode_utf8($content, 0);
500 $f=eval{XML::Feed->parse(\$content)};
503 # Another possibility is badly escaped entities.
504 $feed->{message}.=" ".sprintf(gettext("(feed entities escaped)"));
505 $content=~s/\&(?!amp)(\w+);/&$1;/g;
506 $content=Encode::decode_utf8($content, 0);
507 $f=eval{XML::Feed->parse(\$content)};
510 $feed->{message}=gettext("feed crashed XML::Feed!")." ($@)";
512 debug($feed->{message});
516 $feed->{message}=XML::Feed->errstr;
518 debug($feed->{message});
522 foreach my $entry ($f->entries) {
523 my $content=$content=$entry->content->body;
524 # atom feeds may have no content, only a summary
525 if (! defined $content && ref $entry->summary) {
526 $content=$entry->summary->body;
531 copyright => $f->copyright,
532 title => defined $entry->title ? decode_entities($entry->title) : "untitled",
533 link => $entry->link,
534 content => defined $content ? $content : "",
535 guid => defined $entry->id ? $entry->id : time."_".$feed->{name},
536 ctime => $entry->issued ? ($entry->issued->epoch || time) : time,
542 sub add_page (@) { #{{{
545 my $feed=$params{feed};
548 if (exists $guids{$params{guid}}) {
549 # updating an existing post
550 $guid=$guids{$params{guid}};
551 return if $guid->{expired};
555 $guid->{guid}=$params{guid};
556 $guids{$params{guid}}=$guid;
557 $mtime=$params{ctime};
561 # assign it an unused page
562 my $page=IkiWiki::titlepage($params{title});
563 # escape slashes and periods in title so it doesn't specify
564 # directory name or trigger ".." disallowing code.
565 $page=~s!([/.])!"__".ord($1)."__"!eg;
566 $page=$feed->{dir}."/".$page;
567 ($page)=$page=~/$config{wiki_file_regexp}/;
568 if (! defined $page || ! length $page) {
569 $page=$feed->{dir}."/item";
572 while (exists $IkiWiki::pagecase{lc $page.$c} ||
573 -e pagefile($page.$c)) {
577 # Make sure that the file name isn't too long.
578 # NB: This doesn't check for path length limits.
579 my $max=POSIX::pathconf($config{srcdir}, &POSIX::_PC_NAME_MAX);
580 if (defined $max && length(htmlfn($page)) >= $max) {
582 $page=$feed->{dir}."/item";
583 while (exists $IkiWiki::pagecase{lc $page.$c} ||
584 -e pagefile($page.$c)) {
590 debug(sprintf(gettext("creating new page %s"), $page));
592 $guid->{feed}=$feed->{name};
594 # To write or not to write? Need to avoid writing unchanged pages
595 # to avoid unneccessary rebuilding. The mtime from rss cannot be
596 # trusted; let's use a digest.
597 eval q{use Digest::MD5 'md5_hex'};
600 my $digest=md5_hex(Encode::encode_utf8($params{content}));
601 return unless ! exists $guid->{md5} || $guid->{md5} ne $digest || $config{rebuild};
602 $guid->{md5}=$digest;
605 my $template=template($feed->{template}, blind_cache => 1);
606 $template->param(title => $params{title})
607 if defined $params{title} && length($params{title});
608 $template->param(content => htmlescape(htmlabs($params{content}, $feed->{feedurl})));
609 $template->param(name => $feed->{name});
610 $template->param(url => $feed->{url});
611 $template->param(copyright => $params{copyright})
612 if defined $params{copyright} && length $params{copyright};
613 $template->param(permalink => urlabs($params{link}, $feed->{feedurl}))
614 if defined $params{link};
615 if (ref $feed->{tags}) {
616 $template->param(tags => [map { tag => $_ }, @{$feed->{tags}}]);
618 writefile(htmlfn($guid->{page}), $config{srcdir},
621 if (defined $mtime && $mtime <= time) {
622 # Set the mtime, this lets the build process get the right
623 # creation time on record for the new page.
624 utime $mtime, $mtime, pagefile($guid->{page});
625 # Store it in pagectime for expiry code to use also.
626 $IkiWiki::pagectime{$guid->{page}}=$mtime;
630 sub htmlescape ($) { #{{{
631 # escape accidental wikilinks and preprocessor stuff
633 $html=~s/(?<!\\)\[\[/\\\[\[/g;
637 sub urlabs ($$) { #{{{
641 URI->new_abs($url, $urlbase)->as_string;
644 sub htmlabs ($$) { #{{{
645 # Convert links in html from relative to absolute.
646 # Note that this is a heuristic, which is not specified by the rss
647 # spec and may not be right for all feeds. Also, see Debian
653 my $p = HTML::Parser->new(api_version => 3);
654 $p->handler(default => sub { $ret.=join("", @_) }, "text");
655 $p->handler(start => sub {
656 my ($tagname, $pos, $text) = @_;
657 if (ref $HTML::Tagset::linkElements{$tagname}) {
659 # use attribute sets from right to left
660 # to avoid invalidating the offsets
661 # when replacing the values
662 my($k_offset, $k_len, $v_offset, $v_len) =
664 my $attrname = lc(substr($text, $k_offset, $k_len));
665 next unless grep { $_ eq $attrname } @{$HTML::Tagset::linkElements{$tagname}};
666 next unless $v_offset; # 0 v_offset means no value
667 my $v = substr($text, $v_offset, $v_len);
668 $v =~ s/^([\'\"])(.*)\1$/$2/;
669 my $new_v=urlabs($v, $urlbase);
670 $new_v =~ s/\"/"/g; # since we quote with ""
671 substr($text, $v_offset, $v_len) = qq("$new_v");
675 }, "tagname, tokenpos, text");
682 sub pagefile ($) { #{{{
685 return "$config{srcdir}/".htmlfn($page);
688 sub htmlfn ($) { #{{{
689 return shift().".".($config{aggregateinternal} ? "_aggregated" : $config{htmlext});
694 sub lockaggregate () { #{{{
695 # Take an exclusive lock to prevent multiple concurrent aggregators.
696 # Returns true if the lock was aquired.
697 if (! -d $config{wikistatedir}) {
698 mkdir($config{wikistatedir});
700 open($aggregatelock, '>', "$config{wikistatedir}/aggregatelock") ||
701 error ("cannot open to $config{wikistatedir}/aggregatelock: $!");
702 if (! flock($aggregatelock, 2 | 4)) { # LOCK_EX | LOCK_NB
703 close($aggregatelock) || error("failed closing aggregatelock: $!");
709 sub unlockaggregate () { #{{{
710 return close($aggregatelock) if $aggregatelock;