Parse HTML Content

parse_html_content.php

include_once('simple_html_dom.php');
// TODO there are 38 pages, run this script in 38 loops, change the page value of page.
$html = file_get_html('https://amsterdam2014.drupal.org/attendees?field_badge_first_name_value=&field_badge_last_name_value=&field_badge_org_value=&uid=&field_badge_country_value=All&page=8');
$details = array();
$index = 0;
foreach($html->find('div.view-id-attendees ul li') as $li) {
  // User name.
  foreach($li->find('div.views-field-name') as $data) {
    $do_username = $data->plaintext;
    $details[$index]['username'] = $do_username;
  }
  // Job title.
  foreach($li->find('div.views-field-field-badge-job-title') as $data) {
    $job_title = $data->plaintext;
    $details[$index]['job'] = $job_title;
  }
  // Organization.
  foreach($li->find('div.views-field-field-badge-org') as $data) {
    $organization = $data->plaintext;
    $details[$index]['organization'] = $organization;
  }

  // Twitter URL.
  foreach($li->find('a.username') as $data) {
    $text = $data->href;
    if (trim($text)) {
      $amsterdam_prof_url = 'https://amsterdam2014.drupal.org' . $text;
      $amsterdam_prof_html = file_get_html($amsterdam_prof_url);
      //TODO fetch the data from the amsterdam page,  if twitter link not found then from the do profile.
      $twitter = 'twitter_not';
      foreach($amsterdam_prof_html->find('div.field--name-field-twitter-handle .field__items') as $data) {
        $twitter = $data->plaintext;
        $details[$index]['twitter'] = $twitter;
      }
    }
  }

  // If twitter handler is not persent in amsterdem page, find out form DO profile.
  if ($twitter == 'twitter_not') {
    $do_prof = $amsterdam_prof_html->find('div.user-profile-item div.user-profile-item__items a');
    foreach($do_prof as $data) {
      $do_url = $data->href;
      $do_html = file_get_html($do_url);
      $do_profile = $do_html->find('div#user_user_full_group_profile_main a');
      foreach($do_profile as $twiiter_do) {
        $twitter_href = $twiiter_do->href;

        if (preg_match('/twitter/',$twitter_href)) {
          preg_match("/https?:\/\/(www\.)?twitter\.com\/(#!\/)?@?([^\/]*)/", $twitter_href, $matches);
          if (isset($matches[3])) {
            $twitter = '@' . $matches[3];
            $details[$index]['twitter'] = $twitter;
          }
        }
      }
    }
  }
  $index++;
}

$users = array();
$fp = fopen ( 'drupal_users_21.csv', "a" );
foreach ($details as $user) {
  if (isset($user['username']) && isset($user['job']) && isset($user['organization']) && isset($user['twitter'])) {

    /*fwrite ($fp, trim($user['username']) . ',' );
    fwrite ($fp, trim($user['job']) . ',' );
    fwrite ($fp, trim($user['organization']) . ',' );
    fwrite ($fp, trim($user['twitter']) . "\n" );*/
    $users[] = $user;
  }
}
fclose ($fp);
dpm($users);