alexandrie/parse_known_sites

#! /usr/bin/perl

# Name        : parse_known_sites
# Author      : theocrite
# Description : parses (gets author, date & title) of
#               pages grabbed by alexandrie
#

use strict;
use LWP::Simple;
use DBI;

my $dsn = 'DBI:mysql:drupal6:172.16.0.7';
my $db_user_name = 'drupal';
my $db_password = 'xxxxxxx';



sub parse_page
{
  my $url=shift;
  my $id =shift;

  print "parsing $url\n";

  my $title_patern = '';
  my $author_patern = '';
  my $date_patern = '';
  my $site = '';
  
  if ($url =~ /pcinpact\.com/)
  {
    $author_patern = '';
    $title_patern = '<title>(.*) - PC INpact<\/title>';
    $date_patern = '<\/strong> le (.*) \(<span class="or">';
    $site = 'PC INpact';
  }
  elsif ($url =~ /zdnet\.fr/)
  {
    $author_patern = '<p><cite site="Auteur">par <a href=".*">(.*)</a></cite>';
    $title_patern = '<title>(.*) - Actualit.s - ZDNet\.fr<\/title>';
    $date_patern = 'Publi&eacute; le <span class="date">(.*)</span></p>';
    $site = 'ZDNet.fr';
  }
  elsif ($url =~ /silicon\.fr/)
  {
    $author_patern = '<p class="author">\s+Par (.*?)\s+</p>';
    $title_patern = '<title>(.*) Actualit&eacute; --- Silicon\.fr --- <\/title>';
    $date_patern = '<p class="date">(.*)</p>';
    $site = 'Silicon.fr';
  }
  elsif ($url =~ /cio-online\.com/)
  {
    $author_patern = 'par <a href=".*" class="couleur">(.*)</a>';
    $title_patern = '<title>(.*) - CIO-Online.*<\/title>';
  #  $date_patern = '<\/b><\/h3><br \/><b class="rouge">\((.*)\)<\/b>';
    $date_patern = '<br \/><b class="couleur">Edition du (.*)<\/b> - par';
    $site = 'CIO-Online';
  }
  elsif ($url =~ /lemonde\.fr/)
  {
    $author_patern = '<div class=author><b>(.*)<\/b><\/div><';
    $title_patern = '<title>(.*) - LeMonde.fr</title>';
    $date_patern = '<div class="dateline">[^|]+\| (.*?) \|';
    $site = 'Le Monde.fr';
  }
  elsif ($url =~ /lemondeinformatique\.fr/)
  {
    $author_patern = 'par <a href=".*" class="rouge">(.*)<\/a>';
    $title_patern = '<title>(.*) - .* - Le Monde Informatique<\/title>';
    $date_patern = '<br \/><b class="rouge">Edition du (.*)<\/b>';
    $site = 'Le Monde Informatique';
  }
  elsif ($url =~ /ecrans\.fr/)
  {
    $author_patern = '<p class="auteur"> par <a href=".*">(.*)<\/a>';
    $title_patern = '<title>(.*)- Ecrans';
    $date_patern = '<p class="date">(.*?)\s*<\/p>';
    $site = 'Ecrans';
  }
  elsif ($url =~ /numerama\.com/)
  {
    $author_patern = '<a href="/contact/articles.html" style="text-decoration:underline;color:#929292">(.*)</a> -'; 
    $title_patern = '<title>(.*) - Numerama<\/title>';
    $date_patern = 'publi&eacute; le (.*) - ';
    $site = 'Numerama';
  }
  elsif ($url =~ /01net\.com/)
  {
    $author_patern = '<a .*href=.mailto:commentaires@01net.fr?.* style="text-decoration:.*;">(.*)<\/a>';
    $title_patern = '<title>(.*)<\/title>';
    $date_patern = '">le (.*?)<\/';
    $site = '01net informatique';
  }
  elsif ($url =~ /lemagit\.fr/)
  {
    $author_patern = '- par   <a href=".*">(.*)<\/a>';
    $title_patern = '<title>(.*)<\/title>';
    $date_patern = '<p class="auteur_date typo_auteur"> Le (.*) - par';
    $site = 'LeMagIT';
  }
  elsif ($url =~ /www\.neteco\.com/)
  {
    $author_patern = 'Publi. par <b><a href=\S+ title="(.*?)">';
    $title_patern = '<title>(.*) par Neteco.com</title>';
    $date_patern = '<\/a><\/b> le <b>(.*?)<\/b>';
    $site = 'NetEco';
  }
  # BROKEN - journaldunet devs = stupid piece of junk
  elsif ($url =~ /www\.journaldunet\.com/)
  {
    $author_patern = '<div id="signature"><a href=".*?" rel="nofollow">(.*?) Journal du Net<\/a> <\/div>';
    $title_patern = '<title>([^-]+) -';
    $date_patern = 'span class="date_publication">Publi&eacute; le (.*)<\/span>';
    $site = 'LeJournalduNet';
  }
  elsif ($url =~ /www\.generation-nt\.com/)
  {
    $author_patern = 'par <a href="/divers/contacts.php">(.*)</a> &nbsp;';
    $title_patern = '<title>(.*)<\/title>';
    $date_patern = '(.*) par <a href="/divers/contacts.php">';
    $site = 'Génération nouvelles technologies';
  }
  elsif ($url =~ /lepoint\.fr/)
  {
    $author_patern = 'La rédaction';
    $title_patern = '<title>(.*), actualit&eacute;eacute</title>';
    $date_patern = 'Publi&eacute;\s+le (.*?)\s+-';
    $site = 'Le point';
  }
#  elsif ($url =~ //)
#  {
#    $author_patern = '';
#    $title_patern = '';
#    $date_patern = '';
#    $site = '';
#  }
#  elsif ($url =~ //)
#  {
#    $author_patern = '';
#    $title_patern = '';
#    $date_patern = '';
#    $site = '';
#  }
  else
  {
    next;
  }
  
  
  my $contents = get $url;
  
  $contents =~ /$author_patern/im;
  my $author=$1;
  $contents =~ /$title_patern/im;
  my $title=$1;
  $contents =~ /$date_patern/im;
  my $date = $1;
  
  my $update='UPDATE presse SET date=?, auteur=?, titre=?, site=? WHERE id=?';
  my $dbh = DBI->connect($dsn, $db_user_name, $db_password);
  my $sth = $dbh->prepare($update);
  $sth->execute($date, $author, $title, $site, $id);
  $sth->finish(); # we're done with this query
}


#my $url=$ARGV[0];
#parse_page $url;
my $id=$ARGV[0];

my $select='SELECT url, id FROM presse'. ($id eq ''?'':" WHERE id=$id");
my $dbh = DBI->connect($dsn, $db_user_name, $db_password);
my $sth = $dbh->prepare($select);
if ( $sth->execute() )
{
  while (my @found = $sth->fetchrow_array())
    {
      parse_page $found[0], $found[1];
    }
}
$sth->finish(); # we're done with this query