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         error($@) if $@;
  29         Getopt::Long::Configure('pass_through');
  30         GetOptions("aggregate" => \$config{aggregate});
  31 } #}}}
  32
  33 sub checkconfig () { #{{{
  34         IkiWiki::lockwiki();
  35         loadstate();
  36         if ($config{aggregate}) {
  37                 IkiWiki::loadindex();
  38                 aggregate();
  39                 expire();
  40                 savestate();
  41         }
  42         IkiWiki::unlockwiki();
  43 } #}}}
  44
  45 sub filter (@) { #{{{
  46         my %params=@_;
  47         my $page=$params{page};
  48
  49         # Mark all feeds originating on this page as removable;
  50         # preprocess will unmark those that still exist.
  51         remove_feeds($page);
  52
  53         return $params{content};
  54 } # }}}
  55
  56 sub preprocess (@) { #{{{
  57         my %params=@_;
  58
  59         foreach my $required (qw{name url}) {
  60                 if (! exists $params{$required}) {
  61                         return "[[aggregate plugin missing $required parameter]]";
  62                 }
  63         }
  64
  65         my $feed={};
  66         my $name=$params{name};
  67         if (exists $feeds{$name}) {
  68                 $feed=$feeds{$name};
  69         }
  70         else {
  71                 $feeds{$name}=$feed;
  72         }
  73         $feed->{name}=$name;
  74         $feed->{sourcepage}=$params{page};
  75         $feed->{url}=$params{url};
  76         my $dir=exists $params{dir} ? $params{dir} : $params{page}."/".IkiWiki::titlepage($params{name});
  77         $dir=~s/^\/+//;
  78         ($dir)=$dir=~/$config{wiki_file_regexp}/;
  79         $feed->{dir}=$dir;
  80         $feed->{feedurl}=defined $params{feedurl} ? $params{feedurl} : "";
  81         $feed->{updateinterval}=defined $params{updateinterval} ? $params{updateinterval} * 60 : 15 * 60;
  82         $feed->{expireage}=defined $params{expireage} ? $params{expireage} : 0;
  83         $feed->{expirecount}=defined $params{expirecount} ? $params{expirecount} : 0;
  84         delete $feed->{remove};
  85         delete $feed->{expired};
  86         $feed->{lastupdate}=0 unless defined $feed->{lastupdate};
  87         $feed->{numposts}=0 unless defined $feed->{numposts};
  88         $feed->{newposts}=0 unless defined $feed->{newposts};
  89         $feed->{message}="new feed" unless defined $feed->{message};
  90         $feed->{error}=0 unless defined $feed->{error};
  91         $feed->{tags}=[];
  92         while (@_) {
  93                 my $key=shift;
  94                 my $value=shift;
  95                 if ($key eq 'tag') {
  96                         push @{$feed->{tags}}, $value;
  97                 }
  98         }
  99
 100         return "<a href=\"".$feed->{url}."\">".$feed->{name}."</a>: ".
 101                ($feed->{error} ? "<em>" : "").$feed->{message}.
 102                ($feed->{error} ? "</em>" : "").
 103                " (".$feed->{numposts}." posts".
 104                ($feed->{newposts} ? "; ".$feed->{newposts}." new" : "").
 105                ")";
 106 } # }}}
 107
 108 sub delete (@) { #{{{
 109         my @files=@_;
 110
 111         # Remove feed data for removed pages.
 112         foreach my $file (@files) {
 113                 my $page=pagename($file);
 114                 remove_feeds($page);
 115         }
 116 } #}}}
 117
 118 sub loadstate () { #{{{
 119         if (-e "$config{wikistatedir}/aggregate") {
 120                 open (IN, "$config{wikistatedir}/aggregate" ||
 121                         die "$config{wikistatedir}/aggregate: $!");
 122                 while (<IN>) {
 123                         $_=IkiWiki::possibly_foolish_untaint($_);
 124                         chomp;
 125                         my $data={};
 126                         foreach my $i (split(/ /, $_)) {
 127                                 my ($field, $val)=split(/=/, $i, 2);
 128                                 if ($field eq "name" || $field eq "feed" ||
 129                                     $field eq "guid" || $field eq "message") {
 130                                         $data->{$field}=decode_entities($val, " \t\n");
 131                                 }
 132                                 elsif ($field eq "tag") {
 133                                         push @{$data->{tags}}, $val;
 134                                 }
 135                                 else {
 136                                         $data->{$field}=$val;
 137                                 }
 138                         }
 139
 140                         if (exists $data->{name}) {
 141                                 $feeds{$data->{name}}=$data;
 142                         }
 143                         elsif (exists $data->{guid}) {
 144                                 $guids{$data->{guid}}=$data;
 145                         }
 146                 }
 147
 148                 close IN;
 149         }
 150 } #}}}
 151
 152 sub savestate () { #{{{
 153         eval q{use HTML::Entities};
 154         error($@) if $@;
 155         open (OUT, ">$config{wikistatedir}/aggregate" ||
 156                 die "$config{wikistatedir}/aggregate: $!");
 157         foreach my $data (values %feeds, values %guids) {
 158                 if ($data->{remove}) {
 159                         if ($data->{name}) {
 160                                 foreach my $guid (values %guids) {
 161                                         if ($guid->{feed} eq $data->{name}) {
 162                                                 $guid->{remove}=1;
 163                                         }
 164                                 }
 165                         }
 166                         else {
 167                                 unlink pagefile($data->{page});
 168                         }
 169                         next;
 170                 }
 171                 elsif ($data->{expired} && exists $data->{page}) {
 172                         unlink pagefile($data->{page});
 173                         delete $data->{page};
 174                         delete $data->{md5};
 175                 }
 176
 177                 my @line;
 178                 foreach my $field (keys %$data) {
 179                         if ($field eq "name" || $field eq "feed" ||
 180                             $field eq "guid" || $field eq "message") {
 181                                 push @line, "$field=".encode_entities($data->{$field}, " \t\n");
 182                         }
 183                         elsif ($field eq "tags") {
 184                                 push @line, "tag=$_" foreach @{$data->{tags}};
 185                         }
 186                         else {
 187                                 push @line, "$field=".$data->{$field};
 188                         }
 189                 }
 190                 print OUT join(" ", @line)."\n";
 191         }
 192         close OUT;
 193 } #}}}
 194
 195 sub expire () { #{{{
 196         foreach my $feed (values %feeds) {
 197                 next unless $feed->{expireage} || $feed->{expirecount};
 198                 my $count=0;
 199                 foreach my $item (sort { $IkiWiki::pagectime{$b->{page}} <=> $IkiWiki::pagectime{$a->{page}} }
 200                                   grep { exists $_->{page} && $_->{feed} eq $feed->{name} && $IkiWiki::pagectime{$_->{page}} }
 201                                   values %guids) {
 202                         if ($feed->{expireage}) {
 203                                 my $days_old = (time - $IkiWiki::pagectime{$item->{page}}) / 60 / 60 / 24;
 204                                 if ($days_old > $feed->{expireage}) {
 205                                         debug("expiring ".$item->{page}." ($days_old days old)");
 206                                         $item->{expired}=1;
 207                                 }
 208                         }
 209                         elsif ($feed->{expirecount} &&
 210                                $count >= $feed->{expirecount}) {
 211                                 debug("expiring ".$item->{page});
 212                                 $item->{expired}=1;
 213                         }
 214                         else {
 215                                 $count++;
 216                         }
 217                 }
 218         }
 219 } #}}}
 220
 221 sub aggregate () { #{{{
 222         eval q{use XML::Feed};
 223         error($@) if $@;
 224         eval q{use HTML::Entities};
 225         error($@) if $@;
 226
 227         foreach my $feed (values %feeds) {
 228                 next unless $config{rebuild} ||
 229                         time - $feed->{lastupdate} >= $feed->{updateinterval};
 230                 $feed->{lastupdate}=time;
 231                 $feed->{newposts}=0;
 232                 $IkiWiki::forcerebuild{$feed->{sourcepage}}=1;
 233
 234                 debug("checking feed ".$feed->{name}." ...");
 235
 236                 if (! length $feed->{feedurl}) {
 237                         my @urls=XML::Feed->find_feeds($feed->{url});
 238                         if (! @urls) {
 239                                 $feed->{message}="could not find feed at ".$feed->{feedurl};
 240                                 $feed->{error}=1;
 241                                 debug($feed->{message});
 242                                 next;
 243                         }
 244                         $feed->{feedurl}=pop @urls;
 245                 }
 246                 my $f=eval{XML::Feed->parse(URI->new($feed->{feedurl}))};
 247                 if ($@) {
 248                         $feed->{message}="feed crashed XML::Feed! $@";
 249                         $feed->{error}=1;
 250                         debug($feed->{message});
 251                         next;
 252                 }
 253                 if (! $f) {
 254                         $feed->{message}=XML::Feed->errstr;
 255                         $feed->{error}=1;
 256                         debug($feed->{message});
 257                         next;
 258                 }
 259
 260                 foreach my $entry ($f->entries) {
 261                         add_page(
 262                                 feed => $feed,
 263                                 title => defined $entry->title ? decode_entities($entry->title) : "untitled",
 264                                 link => $entry->link,
 265                                 content => $entry->content->body,
 266                                 guid => defined $entry->id ? $entry->id : time."_".$feed->name,
 267                                 ctime => $entry->issued ? ($entry->issued->epoch || time) : time,
 268                         );
 269                 }
 270
 271                 $feed->{message}="processed ok at ".
 272                         displaytime($feed->{lastupdate});
 273                 $feed->{error}=0;
 274         }
 275 } #}}}
 276
 277 sub add_page (@) { #{{{
 278         my %params=@_;
 279
 280         my $feed=$params{feed};
 281         my $guid={};
 282         my $mtime;
 283         if (exists $guids{$params{guid}}) {
 284                 # updating an existing post
 285                 $guid=$guids{$params{guid}};
 286                 return if $guid->{expired};
 287         }
 288         else {
 289                 # new post
 290                 $guid->{guid}=$params{guid};
 291                 $guids{$params{guid}}=$guid;
 292                 $mtime=$params{ctime};
 293                 $feed->{numposts}++;
 294                 $feed->{newposts}++;
 295
 296                 # assign it an unused page
 297                 my $page=IkiWiki::titlepage($params{title});
 298                 # escape slashes and periods in title so it doesn't specify
 299                 # directory name or trigger ".." disallowing code.
 300                 $page=~s!([/.])!"__".ord($1)."__"!eg;
 301                 $page=$feed->{dir}."/".$page;
 302                 ($page)=$page=~/$config{wiki_file_regexp}/;
 303                 if (! defined $page || ! length $page) {
 304                         $page=$feed->{dir}."/item";
 305                 }
 306                 my $c="";
 307                 while (exists $IkiWiki::pagecase{lc $page.$c} ||
 308                        -e pagefile($page.$c)) {
 309                         $c++
 310                 }
 311                 $guid->{page}=$page;
 312                 debug("creating new page $page");
 313         }
 314         $guid->{feed}=$feed->{name};
 315
 316         # To write or not to write? Need to avoid writing unchanged pages
 317         # to avoid unneccessary rebuilding. The mtime from rss cannot be
 318         # trusted; let's use a digest.
 319         eval q{use Digest::MD5 'md5_hex'};
 320         error($@) if $@;
 321         require Encode;
 322         my $digest=md5_hex(Encode::encode_utf8($params{content}));
 323         return unless ! exists $guid->{md5} || $guid->{md5} ne $digest || $config{rebuild};
 324         $guid->{md5}=$digest;
 325
 326         # Create the page.
 327         my $template=template("aggregatepost.tmpl", blind_cache => 1);
 328         $template->param(title => $params{title})
 329                 if defined $params{title} && length($params{title});
 330         $template->param(content => htmlescape(htmlabs($params{content}, $feed->{feedurl})));
 331         $template->param(name => $feed->{name});
 332         $template->param(url => $feed->{url});
 333         $template->param(permalink => urlabs($params{link}, $feed->{feedurl}))
 334                 if defined $params{link};
 335         if (ref $feed->{tags}) {
 336                 $template->param(tags => [map { tag => $_ }, @{$feed->{tags}}]);
 337         }
 338         writefile(htmlpage($guid->{page}), $config{srcdir},
 339                 $template->output);
 340
 341         # Set the mtime, this lets the build process get the right creation
 342         # time on record for the new page.
 343         utime $mtime, $mtime, pagefile($guid->{page}) if defined $mtime;
 344 } #}}}
 345
 346 sub htmlescape ($) { #{{{
 347         # escape accidental wikilinks and preprocessor stuff
 348         my $html=shift;
 349         $html=~s/(?<!\\)\[\[/\\\[\[/g;
 350         return $html;
 351 } #}}}
 352
 353 sub urlabs ($$) { #{{{
 354         my $url=shift;
 355         my $urlbase=shift;
 356
 357         URI->new_abs($url, $urlbase)->as_string;
 358 } #}}}
 359
 360 sub htmlabs ($$) { #{{{
 361         # Convert links in html from relative to absolute.
 362         # Note that this is a heuristic, which is not specified by the rss
 363         # spec and may not be right for all feeds. Also, see Debian
 364         # bug #381359.
 365         my $html=shift;
 366         my $urlbase=shift;
 367
 368         my $ret="";
 369         my $p = HTML::Parser->new(api_version => 3);
 370         $p->handler(default => sub { $ret.=join("", @_) }, "text");
 371         $p->handler(start => sub {
 372                 my ($tagname, $pos, $text) = @_;
 373                 if (ref $HTML::Tagset::linkElements{$tagname}) {
 374                         while (4 <= @$pos) {
 375                                 # use attribute sets from right to left
 376                                 # to avoid invalidating the offsets
 377                                 # when replacing the values
 378                                 my($k_offset, $k_len, $v_offset, $v_len) =
 379                                         splice(@$pos, -4);
 380                                 my $attrname = lc(substr($text, $k_offset, $k_len));
 381                                 next unless grep { $_ eq $attrname } @{$HTML::Tagset::linkElements{$tagname}};
 382                                 next unless $v_offset; # 0 v_offset means no value
 383                                 my $v = substr($text, $v_offset, $v_len);
 384                                 $v =~ s/^([\'\"])(.*)\1$/$2/;
 385                                 my $new_v=urlabs($v, $urlbase);
 386                                 $new_v =~ s/\"/&quot;/g; # since we quote with ""
 387                                 substr($text, $v_offset, $v_len) = qq("$new_v");
 388                         }
 389                 }
 390                 $ret.=$text;
 391         }, "tagname, tokenpos, text");
 392         $p->parse($html);
 393         $p->eof;
 394
 395         return $ret;
 396 } #}}}
 397
 398 sub remove_feeds () { #{{{
 399         my $page=shift;
 400
 401         my %removed;
 402         foreach my $id (keys %feeds) {
 403                 if ($feeds{$id}->{sourcepage} eq $page) {
 404                         $feeds{$id}->{remove}=1;
 405                         $removed{$id}=1;
 406                 }
 407         }
 408 } #}}}
 409
 410 sub pagefile ($) { #{{{
 411         my $page=shift;
 412
 413         return "$config{srcdir}/".htmlpage($page);
 414 } #}}}
 415
 416 1