X-Git-Url: http://git.vanrenterghem.biz/git.ikiwiki.info.git/blobdiff_plain/02b9f69ba5da59cca9ff198b731d63d643fff4db..34115a34e0593e999b7a279e07293b090012082a:/IkiWiki/Plugin/aggregate.pm

diff --git a/IkiWiki/Plugin/aggregate.pm b/IkiWiki/Plugin/aggregate.pm
index bb5edfa65..ba40ee6bc 100644
--- a/IkiWiki/Plugin/aggregate.pm
+++ b/IkiWiki/Plugin/aggregate.pm
@@ -4,7 +4,7 @@ package IkiWiki::Plugin::aggregate;
 
 use warnings;
 use strict;
-use IkiWiki;
+use IkiWiki 2.00;
 use HTML::Entities;
 use HTML::Parser;
 use HTML::Tagset;
@@ -17,7 +17,7 @@ my %guids;
 sub import { #{{{
 	hook(type => "getopt", id => "aggregate", call => \&getopt);
 	hook(type => "checkconfig", id => "aggregate", call => \&checkconfig);
-	hook(type => "filter", id => "aggregate", call => \&filter);
+	hook(type => "needsbuild", id => "aggregate", call => \&needsbuild);
 	hook(type => "preprocess", id => "aggregate", call => \&preprocess);
         hook(type => "delete", id => "aggregate", call => \&delete);
 	hook(type => "savestate", id => "aggregate", call => \&savestate);
@@ -31,27 +31,66 @@ sub getopt () { #{{{
 } #}}}
 
 sub checkconfig () { #{{{
-	my $nolock=($config{post_commit} && ! IkiWiki::commit_hook_enabled());
-	IkiWiki::lockwiki() unless $nolock;
-	loadstate();
-	if ($config{aggregate} && ! $nolock) {
-		IkiWiki::loadindex();
-		aggregate();
-		expire();
-		savestate();
+	if ($config{aggregate} && ! ($config{post_commit} && 
+	                             IkiWiki::commit_hook_enabled())) {
+		# See if any feeds need aggregation.
+		loadstate();
+		my @feeds=needsaggregate();
+		return unless @feeds;
+		if (! lockaggregate()) {
+			debug("an aggregation process is already running");
+			return;
+		}
+		# force a later rebuild of source pages
+		$IkiWiki::forcerebuild{$_->{sourcepage}}=1
+			foreach @feeds;
+
+		# Fork a child process to handle the aggregation.
+		# The parent process will then handle building the
+		# result. This avoids messy code to clear state
+		# accumulated while aggregating.
+		defined(my $pid = fork) or error("Can't fork: $!");
+		if (! $pid) {
+			IkiWiki::loadindex();
+
+			# Aggregation happens without the main wiki lock
+			# being held. This allows editing pages etc while
+			# aggregation is running.
+			aggregate(@feeds);
+
+			IkiWiki::lockwiki;
+			# Merge changes, since aggregation state may have
+			# changed on disk while the aggregation was happening.
+			mergestate();
+			expire();
+			savestate();
+			IkiWiki::unlockwiki;
+			exit 0;
+		}
+		waitpid($pid,0);
+		if ($?) {
+			error "aggregation failed with code $?";
+		}
+
+		clearstate();
+		unlockaggregate();
 	}
-	IkiWiki::unlockwiki() unless $nolock;
 } #}}}
 
-sub filter (@) { #{{{
-	my %params=@_;
-	my $page=$params{page};
-
-	# Mark all feeds originating on this page as removable;
-	# preprocess will unmark those that still exist.
-	remove_feeds($page);
+sub needsbuild (@) { #{{{
+	my $needsbuild=shift;
+	
+	loadstate();
 
-	return $params{content};
+	foreach my $feed (values %feeds) {
+		if (exists $pagesources{$feed->{sourcepage}} && 
+		    grep { $_ eq $pagesources{$feed->{sourcepage}} } @$needsbuild) {
+			# Mark all feeds originating on this page as 
+			# not yet seen; preprocess will unmark those that
+			# still exist.
+			markunseen($feed->{sourcepage});
+		}
+	}
 } # }}}
 
 sub preprocess (@) { #{{{
@@ -82,8 +121,7 @@ sub preprocess (@) { #{{{
 	$feed->{updateinterval}=defined $params{updateinterval} ? $params{updateinterval} * 60 : 15 * 60;
 	$feed->{expireage}=defined $params{expireage} ? $params{expireage} : 0;
 	$feed->{expirecount}=defined $params{expirecount} ? $params{expirecount} : 0;
-	delete $feed->{remove};
-	delete $feed->{expired};
+	delete $feed->{unseen};
 	$feed->{lastupdate}=0 unless defined $feed->{lastupdate};
 	$feed->{numposts}=0 unless defined $feed->{numposts};
 	$feed->{newposts}=0 unless defined $feed->{newposts};
@@ -113,14 +151,28 @@ sub delete (@) { #{{{
 	# Remove feed data for removed pages.
 	foreach my $file (@files) {
 		my $page=pagename($file);
-		remove_feeds($page);
+		markunseen($page);
 	}
 } #}}}
 
+sub markunseen ($) { #{{{
+	my $page=shift;
+
+	foreach my $id (keys %feeds) {
+		if ($feeds{$id}->{sourcepage} eq $page) {
+			$feeds{$id}->{unseen}=1;
+		}
+	}
+} #}}}
+
+my $state_loaded=0;
+
 sub loadstate () { #{{{
+	return if $state_loaded;
+	$state_loaded=1;
 	if (-e "$config{wikistatedir}/aggregate") {
-		open (IN, "$config{wikistatedir}/aggregate" ||
-			die "$config{wikistatedir}/aggregate: $!");
+		open(IN, "$config{wikistatedir}/aggregate") ||
+			die "$config{wikistatedir}/aggregate: $!";
 		while (<IN>) {
 			$_=IkiWiki::possibly_foolish_untaint($_);
 			chomp;
@@ -152,33 +204,14 @@ sub loadstate () { #{{{
 } #}}}
 
 sub savestate () { #{{{
+	return unless $state_loaded;
+	garbage_collect();
 	eval q{use HTML::Entities};
 	error($@) if $@;
 	my $newfile="$config{wikistatedir}/aggregate.new";
-	# TODO: This cleanup function could use improvement. Any newly
-	# aggregated files are left behind unrecorded, and should be deleted.
 	my $cleanup = sub { unlink($newfile) };
 	open (OUT, ">$newfile") || error("open $newfile: $!", $cleanup);
 	foreach my $data (values %feeds, values %guids) {
-		if ($data->{remove}) {
-			if ($data->{name}) {
-				foreach my $guid (values %guids) {
-					if ($guid->{feed} eq $data->{name}) {
-						$guid->{remove}=1;
-					}
-				}
-			}
-			else {
-				unlink pagefile($data->{page});
-			}
-			next;
-		}
-		elsif ($data->{expired} && exists $data->{page}) {
-			unlink pagefile($data->{page});
-			delete $data->{page};
-			delete $data->{md5};
-		}
-
 		my @line;
 		foreach my $field (keys %$data) {
 			if ($field eq "name" || $field eq "feed" ||
@@ -199,10 +232,74 @@ sub savestate () { #{{{
 		error("rename $newfile: $!", $cleanup);
 } #}}}
 
+sub garbage_collect () { #{{{
+	foreach my $name (keys %feeds) {
+		# remove any feeds that were not seen while building the pages
+		# that used to contain them
+		if ($feeds{$name}->{unseen}) {
+			delete $feeds{$name};
+		}
+	}
+
+	foreach my $guid (values %guids) {
+		# any guid whose feed is gone should be removed
+		if (! exists $feeds{$guid->{feed}}) {
+			unlink pagefile($guid->{page})
+				if exists $guid->{page};
+			delete $guids{$guid->{guid}};
+		}
+		# handle expired guids
+		elsif ($guid->{expired} && exists $guid->{page}) {
+			unlink pagefile($guid->{page});
+			delete $guid->{page};
+			delete $guid->{md5};
+		}
+	}
+} #}}}
+
+sub mergestate () { #{{{
+	# Load the current state in from disk, and merge into it
+	# values from the state in memory that might have changed
+	# during aggregation.
+	my %myfeeds=%feeds;
+	my %myguids=%guids;
+	clearstate();
+	loadstate();
+
+	# All that can change in feed state during aggregation is a few
+	# fields.
+	foreach my $name (keys %myfeeds) {
+		if (exists $feeds{$name}) {
+			foreach my $field (qw{message lastupdate numposts
+			                      newposts error}) {
+				$feeds{$name}->{$field}=$myfeeds{$name}->{$field};
+			}
+		}
+	}
+
+	# New guids can be created during aggregation.
+	# It's also possible that guids were removed from the on-disk state
+	# while the aggregation was in process. That would only happen if
+	# their feed was also removed, so any removed guids added back here
+	# will be garbage collected later.
+	foreach my $guid (keys %myguids) {
+		if (! exists $guids{$guid}) {
+			$guids{$guid}=$myguids{$guid};
+		}
+	}
+} #}}}
+
+sub clearstate () { #{{{
+	%feeds=();
+	%guids=();
+	$state_loaded=0;
+} #}}}
+
 sub expire () { #{{{
 	foreach my $feed (values %feeds) {
 		next unless $feed->{expireage} || $feed->{expirecount};
 		my $count=0;
+		my %seen;
 		foreach my $item (sort { $IkiWiki::pagectime{$b->{page}} <=> $IkiWiki::pagectime{$a->{page}} }
 		                  grep { exists $_->{page} && $_->{feed} eq $feed->{name} && $IkiWiki::pagectime{$_->{page}} }
 		                  values %guids) {
@@ -210,7 +307,7 @@ sub expire () { #{{{
 				my $days_old = (time - $IkiWiki::pagectime{$item->{page}}) / 60 / 60 / 24;
 				if ($days_old > $feed->{expireage}) {
 					debug(sprintf(gettext("expiring %s (%s days old)"),
-						$item->{page}, $days_old));
+						$item->{page}, int($days_old)));
 					$item->{expired}=1;
 				}
 			}
@@ -220,24 +317,34 @@ sub expire () { #{{{
 				$item->{expired}=1;
 			}
 			else {
-				$count++;
+				if (! $seen{$item->{page}}) {
+					$seen{$item->{page}}=1;
+					$count++;
+				}
 			}
 		}
 	}
 } #}}}
 
-sub aggregate () { #{{{
+sub needsaggregate () { #{{{
+	return values %feeds if $config{rebuild};
+	return grep { time - $_->{lastupdate} >= $_->{updateinterval} } values %feeds;
+} #}}}
+
+sub aggregate (@) { #{{{
 	eval q{use XML::Feed};
 	error($@) if $@;
+	eval q{use URI::Fetch};
+	error($@) if $@;
 	eval q{use HTML::Entities};
 	error($@) if $@;
 
-	foreach my $feed (values %feeds) {
-		next unless $config{rebuild} || 
-			time - $feed->{lastupdate} >= $feed->{updateinterval};
+	foreach my $feed (@_) {
 		$feed->{lastupdate}=time;
 		$feed->{newposts}=0;
-		$IkiWiki::forcerebuild{$feed->{sourcepage}}=1;
+		$feed->{message}=sprintf(gettext("processed ok at %s"),
+			displaytime($feed->{lastupdate}));
+		$feed->{error}=0;
 
 		debug(sprintf(gettext("checking feed %s ..."), $feed->{name}));
 
@@ -251,7 +358,36 @@ sub aggregate () { #{{{
 			}
 			$feed->{feedurl}=pop @urls;
 		}
-		my $f=eval{XML::Feed->parse(URI->new($feed->{feedurl}))};
+		my $res=URI::Fetch->fetch($feed->{feedurl});
+		if (! $res) {
+			$feed->{message}=URI::Fetch->errstr;
+			$feed->{error}=1;
+			debug($feed->{message});
+			next;
+		}
+		if ($res->status == URI::Fetch::URI_GONE()) {
+			$feed->{message}=gettext("feed not found");
+			$feed->{error}=1;
+			debug($feed->{message});
+			next;
+		}
+		my $content=$res->content;
+		my $f=eval{XML::Feed->parse(\$content)};
+		if ($@) {
+			# One common cause of XML::Feed crashing is a feed
+			# that contains invalid UTF-8 sequences. Convert
+			# feed to ascii to try to work around.
+			$feed->{message}.=" ".sprintf(gettext("(invalid UTF-8 stripped from feed)"));
+			$content=Encode::decode_utf8($content);
+			$f=eval{XML::Feed->parse(\$content)};
+		}
+		if ($@) {
+			# Another possibility is badly escaped entities.
+			$feed->{message}.=" ".sprintf(gettext("(feed entities escaped)"));
+			$content=~s/\&(?!amp)(\w+);/&amp;$1;/g;
+			$content=Encode::decode_utf8($content);
+			$f=eval{XML::Feed->parse(\$content)};
+		}
 		if ($@) {
 			$feed->{message}=gettext("feed crashed XML::Feed!")." ($@)";
 			$feed->{error}=1;
@@ -268,17 +404,14 @@ sub aggregate () { #{{{
 		foreach my $entry ($f->entries) {
 			add_page(
 				feed => $feed,
+				copyright => $f->copyright,
 				title => defined $entry->title ? decode_entities($entry->title) : "untitled",
 				link => $entry->link,
-				content => $entry->content->body,
+				content => defined $entry->content->body ? $entry->content->body : "",
 				guid => defined $entry->id ? $entry->id : time."_".$feed->name,
 				ctime => $entry->issued ? ($entry->issued->epoch || time) : time,
 			);
 		}
-
-		$feed->{message}=sprintf(gettext("processed ok at %s"),
-			displaytime($feed->{lastupdate}));
-		$feed->{error}=0;
 	}
 } #}}}
 
@@ -319,7 +452,6 @@ sub add_page (@) { #{{{
 
 		# Make sure that the file name isn't too long. 
 		# NB: This doesn't check for path length limits.
-		eval q{use POSIX};
 		my $max=POSIX::pathconf($config{srcdir}, &POSIX::_PC_NAME_MAX);
 		if (defined $max && length(htmlfn($page)) >= $max) {
 			$c="";
@@ -352,6 +484,8 @@ sub add_page (@) { #{{{
 	$template->param(content => htmlescape(htmlabs($params{content}, $feed->{feedurl})));
 	$template->param(name => $feed->{name});
 	$template->param(url => $feed->{url});
+	$template->param(copyright => $params{copyright})
+		if defined $params{copyright} && length $params{copyright};
 	$template->param(permalink => urlabs($params{link}, $feed->{feedurl}))
 		if defined $params{link};
 	if (ref $feed->{tags}) {
@@ -362,7 +496,8 @@ sub add_page (@) { #{{{
 
 	# Set the mtime, this lets the build process get the right creation
 	# time on record for the new page.
-	utime $mtime, $mtime, pagefile($guid->{page}) if defined $mtime;
+	utime $mtime, $mtime, pagefile($guid->{page})
+		if defined $mtime && $mtime <= time;
 } #}}}
 
 sub htmlescape ($) { #{{{
@@ -417,26 +552,36 @@ sub htmlabs ($$) { #{{{
 	return $ret;
 } #}}}
 
-sub remove_feeds () { #{{{
+sub pagefile ($) { #{{{
 	my $page=shift;
 
-	my %removed;
-	foreach my $id (keys %feeds) {
-		if ($feeds{$id}->{sourcepage} eq $page) {
-			$feeds{$id}->{remove}=1;
-			$removed{$id}=1;
-		}
-	}
+	return "$config{srcdir}/".htmlfn($page);
 } #}}}
 
-sub pagefile ($) { #{{{
-	my $page=shift;
+sub htmlfn ($) { #{{{
+	return shift().".".$config{htmlext};
+} #}}}
+
+my $aggregatelock;
 
-	return "$config{srcdir}/".htmlpage($page);
+sub lockaggregate () { #{{{
+	# Take an exclusive lock to prevent multiple concurrent aggregators.
+	# Returns true if the lock was aquired.
+	if (! -d $config{wikistatedir}) {
+		mkdir($config{wikistatedir});
+	}
+	open($aggregatelock, '>', "$config{wikistatedir}/aggregatelock") ||
+		error ("cannot open to $config{wikistatedir}/aggregatelock: $!");
+	if (! flock($aggregatelock, 2 | 4)) { # LOCK_EX | LOCK_NB
+		close($aggregatelock) || error("failed closing aggregatelock: $!");
+		return 0;
+	}
+	return 1;
 } #}}}
 
-sub htmlfn ($) { #{{{
-	return shift().".html";
+sub unlockaggregate () { #{{{
+	return close($aggregatelock) if $aggregatelock;
+	return;
 } #}}}
 
 1