IkiWiki/Plugin/aggregate.pm

   1 #!/usr/bin/perl
   2 # Blog aggregation plugin.
   3 package IkiWiki::Plugin::aggregate;
   4
   5 use warnings;
   6 use strict;
   7 use IkiWiki;
   8 use HTML::Entities;
   9 use HTML::Parser;
  10 use HTML::Tagset;
  11 use URI;
  12 use open qw{:utf8 :std};
  13
  14 my %feeds;
  15 my %guids;
  16
  17 sub import { #{{{
  18         hook(type => "getopt", id => "aggregate", call => \&getopt);
  19         hook(type => "checkconfig", id => "aggregate", call => \&checkconfig);
  20         hook(type => "filter", id => "aggregate", call => \&filter);
  21         hook(type => "preprocess", id => "aggregate", call => \&preprocess);
  22         hook(type => "delete", id => "aggregate", call => \&delete);
  23         hook(type => "savestate", id => "aggregate", call => \&savestate);
  24 } # }}}
  25
  26 sub getopt () { #{{{
  27         eval q{use Getopt::Long};
  28         Getopt::Long::Configure('pass_through');
  29         GetOptions("aggregate" => \$config{aggregate});
  30 } #}}}
  31
  32 sub checkconfig () { #{{{
  33         IkiWiki::lockwiki();
  34         loadstate();
  35         if ($config{aggregate}) {
  36                 IkiWiki::loadindex();
  37                 aggregate();
  38                 expire();
  39                 savestate();
  40         }
  41         IkiWiki::unlockwiki();
  42 } #}}}
  43
  44 sub filter (@) { #{{{
  45         my %params=@_;
  46         my $page=$params{page};
  47
  48         # Mark all feeds originating on this page as removable;
  49         # preprocess will unmark those that still exist.
  50         remove_feeds($page);
  51
  52         return $params{content};
  53 } # }}}
  54
  55 sub preprocess (@) { #{{{
  56         my %params=@_;
  57
  58         foreach my $required (qw{name url}) {
  59                 if (! exists $params{$required}) {
  60                         return "[[aggregate plugin missing $required parameter]]";
  61                 }
  62         }
  63
  64         my $feed={};
  65         my $name=$params{name};
  66         if (exists $feeds{$name}) {
  67                 $feed=$feeds{$name};
  68         }
  69         else {
  70                 $feeds{$name}=$feed;
  71         }
  72         $feed->{name}=$name;
  73         $feed->{sourcepage}=$params{page};
  74         $feed->{url}=$params{url};
  75         my $dir=exists $params{dir} ? $params{dir} : $params{page}."/".IkiWiki::titlepage($params{name});
  76         $dir=~s/^\/+//;
  77         ($dir)=$dir=~/$config{wiki_file_regexp}/;
  78         $feed->{dir}=$dir;
  79         $feed->{feedurl}=defined $params{feedurl} ? $params{feedurl} : "";
  80         $feed->{updateinterval}=defined $params{updateinterval} ? $params{updateinterval} * 60 : 15 * 60;
  81         $feed->{expireage}=defined $params{expireage} ? $params{expireage} : 0;
  82         $feed->{expirecount}=defined $params{expirecount} ? $params{expirecount} : 0;
  83         delete $feed->{remove};
  84         delete $feed->{expired};
  85         $feed->{lastupdate}=0 unless defined $feed->{lastupdate};
  86         $feed->{numposts}=0 unless defined $feed->{numposts};
  87         $feed->{newposts}=0 unless defined $feed->{newposts};
  88         $feed->{message}="new feed" unless defined $feed->{message};
  89         $feed->{error}=0 unless defined $feed->{error};
  90         $feed->{tags}=[];
  91         while (@_) {
  92                 my $key=shift;
  93                 my $value=shift;
  94                 if ($key eq 'tag') {
  95                         push @{$feed->{tags}}, $value;
  96                 }
  97         }
  98
  99         return "<a href=\"".$feed->{url}."\">".$feed->{name}."</a>: ".
 100                ($feed->{error} ? "<em>" : "").$feed->{message}.
 101                ($feed->{error} ? "</em>" : "").
 102                " (".$feed->{numposts}." posts".
 103                ($feed->{newposts} ? "; ".$feed->{newposts}." new" : "").
 104                ")";
 105 } # }}}
 106
 107 sub delete (@) { #{{{
 108         my @files=@_;
 109
 110         # Remove feed data for removed pages.
 111         foreach my $file (@files) {
 112                 my $page=pagename($file);
 113                 remove_feeds($page);
 114         }
 115 } #}}}
 116
 117 sub loadstate () { #{{{
 118         if (-e "$config{wikistatedir}/aggregate") {
 119                 open (IN, "$config{wikistatedir}/aggregate" ||
 120                         die "$config{wikistatedir}/aggregate: $!");
 121                 while (<IN>) {
 122                         $_=IkiWiki::possibly_foolish_untaint($_);
 123                         chomp;
 124                         my $data={};
 125                         foreach my $i (split(/ /, $_)) {
 126                                 my ($field, $val)=split(/=/, $i, 2);
 127                                 if ($field eq "name" || $field eq "feed" ||
 128                                     $field eq "guid" || $field eq "message") {
 129                                         $data->{$field}=decode_entities($val, " \t\n");
 130                                 }
 131                                 elsif ($field eq "tag") {
 132                                         push @{$data->{tags}}, $val;
 133                                 }
 134                                 else {
 135                                         $data->{$field}=$val;
 136                                 }
 137                         }
 138
 139                         if (exists $data->{name}) {
 140                                 $feeds{$data->{name}}=$data;
 141                         }
 142                         elsif (exists $data->{guid}) {
 143                                 $guids{$data->{guid}}=$data;
 144                         }
 145                 }
 146
 147                 close IN;
 148         }
 149 } #}}}
 150
 151 sub savestate () { #{{{
 152         eval q{use HTML::Entities};
 153         die $@ if $@;
 154         open (OUT, ">$config{wikistatedir}/aggregate" ||
 155                 die "$config{wikistatedir}/aggregate: $!");
 156         foreach my $data (values %feeds, values %guids) {
 157                 if ($data->{remove}) {
 158                         if ($data->{name}) {
 159                                 foreach my $guid (values %guids) {
 160                                         if ($guid->{feed} eq $data->{name}) {
 161                                                 $guid->{remove}=1;
 162                                         }
 163                                 }
 164                         }
 165                         else {
 166                                 unlink pagefile($data->{page});
 167                         }
 168                         next;
 169                 }
 170                 elsif ($data->{expired} && exists $data->{page}) {
 171                         unlink pagefile($data->{page});
 172                         delete $data->{page};
 173                         delete $data->{md5};
 174                 }
 175
 176                 my @line;
 177                 foreach my $field (keys %$data) {
 178                         if ($field eq "name" || $field eq "feed" ||
 179                             $field eq "guid" || $field eq "message") {
 180                                 push @line, "$field=".encode_entities($data->{$field}, " \t\n");
 181                         }
 182                         elsif ($field eq "tags") {
 183                                 push @line, "tag=$_" foreach @{$data->{tags}};
 184                         }
 185                         else {
 186                                 push @line, "$field=".$data->{$field};
 187                         }
 188                 }
 189                 print OUT join(" ", @line)."\n";
 190         }
 191         close OUT;
 192 } #}}}
 193
 194 sub expire () { #{{{
 195         foreach my $feed (values %feeds) {
 196                 next unless $feed->{expireage} || $feed->{expirecount};
 197                 my $count=0;
 198                 foreach my $item (sort { $IkiWiki::pagectime{$b->{page}} <=> $IkiWiki::pagectime{$a->{page}} }
 199                                   grep { exists $_->{page} && $_->{feed} eq $feed->{name} && $IkiWiki::pagectime{$_->{page}} }
 200                                   values %guids) {
 201                         if ($feed->{expireage}) {
 202                                 my $days_old = (time - $IkiWiki::pagectime{$item->{page}}) / 60 / 60 / 24;
 203                                 if ($days_old > $feed->{expireage}) {
 204                                         debug("expiring ".$item->{page}." ($days_old days old)");
 205                                         $item->{expired}=1;
 206                                 }
 207                         }
 208                         elsif ($feed->{expirecount} &&
 209                                $count >= $feed->{expirecount}) {
 210                                 debug("expiring ".$item->{page});
 211                                 $item->{expired}=1;
 212                         }
 213                         else {
 214                                 $count++;
 215                         }
 216                 }
 217         }
 218 } #}}}
 219
 220 sub aggregate () { #{{{
 221         eval q{use XML::Feed};
 222         die $@ if $@;
 223         eval q{use HTML::Entities};
 224         die $@ if $@;
 225
 226         foreach my $feed (values %feeds) {
 227                 next unless $config{rebuild} ||
 228                         time - $feed->{lastupdate} >= $feed->{updateinterval};
 229                 $feed->{lastupdate}=time;
 230                 $feed->{newposts}=0;
 231                 $IkiWiki::forcerebuild{$feed->{sourcepage}}=1;
 232
 233                 debug("checking feed ".$feed->{name}." ...");
 234
 235                 if (! length $feed->{feedurl}) {
 236                         my @urls=XML::Feed->find_feeds($feed->{url});
 237                         if (! @urls) {
 238                                 $feed->{message}="could not find feed at ".$feed->{feedurl};
 239                                 $feed->{error}=1;
 240                                 debug($feed->{message});
 241                                 next;
 242                         }
 243                         $feed->{feedurl}=pop @urls;
 244                 }
 245                 my $f=eval{XML::Feed->parse(URI->new($feed->{feedurl}))};
 246                 if ($@) {
 247                         $feed->{message}="feed crashed XML::Feed! $@";
 248                         $feed->{error}=1;
 249                         debug($feed->{message});
 250                         next;
 251                 }
 252                 if (! $f) {
 253                         $feed->{message}=XML::Feed->errstr;
 254                         $feed->{error}=1;
 255                         debug($feed->{message});
 256                         next;
 257                 }
 258
 259                 foreach my $entry ($f->entries) {
 260                         add_page(
 261                                 feed => $feed,
 262                                 title => defined $entry->title ? decode_entities($entry->title) : "untitled",
 263                                 link => $entry->link,
 264                                 content => $entry->content->body,
 265                                 guid => defined $entry->id ? $entry->id : time."_".$feed->name,
 266                                 ctime => $entry->issued ? ($entry->issued->epoch || time) : time,
 267                         );
 268                 }
 269
 270                 $feed->{message}="processed ok at ".
 271                         displaytime($feed->{lastupdate});
 272                 $feed->{error}=0;
 273         }
 274 } #}}}
 275
 276 sub add_page (@) { #{{{
 277         my %params=@_;
 278
 279         my $feed=$params{feed};
 280         my $guid={};
 281         my $mtime;
 282         if (exists $guids{$params{guid}}) {
 283                 # updating an existing post
 284                 $guid=$guids{$params{guid}};
 285                 return if $guid->{expired};
 286         }
 287         else {
 288                 # new post
 289                 $guid->{guid}=$params{guid};
 290                 $guids{$params{guid}}=$guid;
 291                 $mtime=$params{ctime};
 292                 $feed->{numposts}++;
 293                 $feed->{newposts}++;
 294
 295                 # assign it an unused page
 296                 my $page=IkiWiki::titlepage($params{title});
 297                 # escape slashes and periods in title so it doesn't specify
 298                 # directory name or trigger ".." disallowing code.
 299                 $page=~s!([/.])!"__".ord($1)."__"!eg;
 300                 $page=$feed->{dir}."/".$page;
 301                 ($page)=$page=~/$config{wiki_file_regexp}/;
 302                 if (! defined $page || ! length $page) {
 303                         $page=$feed->{dir}."/item";
 304                 }
 305                 my $c="";
 306                 while (exists $IkiWiki::pagecase{lc $page.$c} ||
 307                        -e pagefile($page.$c)) {
 308                         $c++
 309                 }
 310                 $guid->{page}=$page;
 311                 debug("creating new page $page");
 312         }
 313         $guid->{feed}=$feed->{name};
 314
 315         # To write or not to write? Need to avoid writing unchanged pages
 316         # to avoid unneccessary rebuilding. The mtime from rss cannot be
 317         # trusted; let's use a digest.
 318         eval q{use Digest::MD5 'md5_hex'};
 319         require Encode;
 320         my $digest=md5_hex(Encode::encode_utf8($params{content}));
 321         return unless ! exists $guid->{md5} || $guid->{md5} ne $digest || $config{rebuild};
 322         $guid->{md5}=$digest;
 323
 324         # Create the page.
 325         my $template=template("aggregatepost.tmpl", blind_cache => 1);
 326         $template->param(title => $params{title})
 327                 if defined $params{title} && length($params{title});
 328         $template->param(content => htmlescape(htmlabs($params{content}, $feed->{feedurl})));
 329         $template->param(name => $feed->{name});
 330         $template->param(url => $feed->{url});
 331         $template->param(permalink => urlabs($params{link}, $feed->{feedurl}))
 332                 if defined $params{link};
 333         if (ref $feed->{tags}) {
 334                 $template->param(tags => [map { tag => $_ }, @{$feed->{tags}}]);
 335         }
 336         writefile(htmlpage($guid->{page}), $config{srcdir},
 337                 $template->output);
 338
 339         # Set the mtime, this lets the build process get the right creation
 340         # time on record for the new page.
 341         utime $mtime, $mtime, pagefile($guid->{page}) if defined $mtime;
 342 } #}}}
 343
 344 sub htmlescape ($) { #{{{
 345         # escape accidental wikilinks and preprocessor stuff
 346         my $html=shift;
 347         $html=~s/(?<!\\)\[\[/\\\[\[/g;
 348         return $html;
 349 } #}}}
 350
 351 sub urlabs ($$) { #{{{
 352         my $url=shift;
 353         my $urlbase=shift;
 354
 355         URI->new_abs($url, $urlbase)->as_string;
 356 } #}}}
 357
 358 sub htmlabs ($$) { #{{{
 359         # Convert links in html from relative to absolute.
 360         # Note that this is a heuristic, which is not specified by the rss
 361         # spec and may not be right for all feeds. Also, see Debian
 362         # bug #381359.
 363         my $html=shift;
 364         my $urlbase=shift;
 365
 366         my $ret="";
 367         my $p = HTML::Parser->new(api_version => 3);
 368         $p->handler(default => sub { $ret.=join("", @_) }, "text");
 369         $p->handler(start => sub {
 370                 my ($tagname, $pos, $text) = @_;
 371                 if (ref $HTML::Tagset::linkElements{$tagname}) {
 372                         while (4 <= @$pos) {
 373                                 # use attribute sets from right to left
 374                                 # to avoid invalidating the offsets
 375                                 # when replacing the values
 376                                 my($k_offset, $k_len, $v_offset, $v_len) =
 377                                         splice(@$pos, -4);
 378                                 my $attrname = lc(substr($text, $k_offset, $k_len));
 379                                 next unless grep { $_ eq $attrname } @{$HTML::Tagset::linkElements{$tagname}};
 380                                 next unless $v_offset; # 0 v_offset means no value
 381                                 my $v = substr($text, $v_offset, $v_len);
 382                                 $v =~ s/^([\'\"])(.*)\1$/$2/;
 383                                 my $new_v=urlabs($v, $urlbase);
 384                                 $new_v =~ s/\"/&quot;/g; # since we quote with ""
 385                                 substr($text, $v_offset, $v_len) = qq("$new_v");
 386                         }
 387                 }
 388                 $ret.=$text;
 389         }, "tagname, tokenpos, text");
 390         $p->parse($html);
 391         $p->eof;
 392
 393         return $ret;
 394 } #}}}
 395
 396 sub remove_feeds () { #{{{
 397         my $page=shift;
 398
 399         my %removed;
 400         foreach my $id (keys %feeds) {
 401                 if ($feeds{$id}->{sourcepage} eq $page) {
 402                         $feeds{$id}->{remove}=1;
 403                         $removed{$id}=1;
 404                 }
 405         }
 406 } #}}}
 407
 408 sub pagefile ($) { #{{{
 409         my $page=shift;
 410
 411         return "$config{srcdir}/".htmlpage($page);
 412 } #}}}
 413
 414 1