protected function getPhotoFromWikipedia(Entity $person)
{
if ($this->imageExists($person)) {
return true;
}
//construct search query
$wikipedia = new LsWikipedia();
$response = $wikipedia->requestImages($person->name);
if (!$response) {
return false;
}
$matches = $response->query->pages->page;
$this->printDebug("Images on Wikipedia page: " . count($matches));
foreach ($matches as $match) {
$ii = (array) $match->imageinfo->ii;
$attributes = $ii['@attributes'];
$photo_url = $attributes['url'];
$photo_alt = $attributes['comment'];
//var_dump($photo_alt);
$this->printDebug("Checking: " . $photo_url);
if (preg_match('/^http.*' . $person->name_last . '.*(jpg|jpeg|gif)$/i', $photo_url)) {
$this->printDebug("Photo found on URL: " . $photo_url);
return $this->attachImage($person, $photo_url, 'Photograph');
}
}
$this->printDebug("Photo not found");
return false;
}
protected function import($url)
{
$person = null;
$this->printDebug($url);
if (!$this->browser->get($url)->responseIsError()) {
$text = $this->browser->getResponseText();
$bio = null;
$name = null;
$netWorth = null;
$birthYear = null;
$schools = null;
$schools = null;
$imageUrl = null;
$rank = null;
//get name & rank
if ($this->year > 2005 && preg_match('/<b>#(\\d+) ([^<]+)<\\/b>/', $text, $match)) {
$name = trim($match[2]);
$rank = $match[1];
}
if ($this->year == 2005 && preg_match('/<h2>#(\\d+) ([^<]+)<\\/h2>/', $text, $match)) {
$name = trim($match[2]);
$rank = $match[1];
}
//get net worth
if (preg_match('/Net Worth<\\/span> <span class="red">\\$([\\S]+) billion/', $text, $match)) {
$netWorth = $match[1] * 1000000000;
}
//get birth year
if (preg_match('/>Age<\\/span> (\\d+)/', $text, $match)) {
$birthYear = date("Y") - $match[1] . "-00-00";
}
//get schools
if (preg_match('/Education<\\/span>(.*)<\\/td>/isU', $text, $match)) {
$schools = array();
$schoolParts = explode('<br>', $match[1]);
while ($schoolPart = current($schoolParts)) {
if (preg_match('/^([^,]+),\\s+<b>([^<]+)<\\/b>/is', trim($schoolPart), $match)) {
$schoolOrg = trim($match[1]);
if ($schoolOrg == 'High School') {
next($schoolParts);
continue;
}
$schoolDegree = trim($match[2]);
$schools[] = array('org' => $schoolOrg, 'degree' => $schoolDegree);
}
next($schoolParts);
}
}
if (preg_match('#<br>[\\n\\s]<br>(.+?)<br>[\\n\\s]<br>[\\n\\s]<img#isU', $text, $match)) {
$bio = strip_tags(trim($match[1]));
} else {
$wikipedia = new LsWikipedia();
if ($wikipedia->request($name)) {
$bio = $wikipedia->getIntroduction();
}
}
//get image
$regexp = '#([A-Z1-9]{4}).html#';
if (preg_match($regexp, $url, $match)) {
$imageFilename = $match[1] . ".jpg";
$imageUrl = $this->list_urls[$this->year]['img_src'] . $imageFilename;
}
//echo "Rank: " . $rank . "\n";
$this->printDebug("Rank: " . $rank);
$this->printDebug("Name: " . $name);
$this->printDebug("Image: " . $imageUrl);
$this->printDebug("Net worth: " . $netWorth);
$this->printDebug("Birth year: " . $birthYear);
$this->printDebug("Bio: " . $bio);
$person = $this->generatePerson($name, $bio);
$person_exists = $this->getBusinessPersonQuery()->addWhere("person.name_first = ? AND person.name_last = ?", array($person->name_first, $person->name_last))->fetchOne();
if ($person_exists != false) {
$this->printDebug('Person exists');
$person = $person_exists;
} else {
$this->printDebug('Saving new person');
}
//parse name and create person object
$person->addExtension('BusinessPerson');
$person->start_date = $person->start_date == null ? $birthYear : $person->start_date;
$person->summary = $person->summary == null ? $bio : $person->summary;
$person->net_worth = $person->net_worth == null ? $netWorth : $person->net_worth;
//go through schools person attended
foreach ($schools as $school) {
//does the current school exist?
$current_school = EntityTable::getByExtensionQuery('Org')->addWhere("org.name = ?", $school['org'])->fetchOne();
if ($current_school) {
$this->printDebug(" Found School " . $school['org']);
} else {
//clear cache
Doctrine::getTable('ExtensionDefinition')->clear();
$current_school = new Entity();
$current_school->addExtension('Org');
$current_school->addExtension('School');
$current_school->name = LsLanguage::titleize($school['org']);
$current_school->save();
$current_school->addReference($source = $url, $excerpt = null, $fields = array('name'), $name = 'Forbes.com', $detail = null, $date = null);
$this->printDebug(" Adding new school: " . $school['org']);
}
//if there is no relationship between person and school. connect them!
if (!$person->getRelationshipsWithQuery($current_school, RelationshipTable::EDUCATION_CATEGORY)->fetchOne()) {
$this->printDebug(" Creating Relation between " . $current_school->name . " and " . $person->name);
$education = new Relationship();
$education->Entity1 = $person;
$education->Entity2 = $current_school;
$education->setCategory('Education');
$education->description1 = $school['degree'];
$education->is_current = 1;
$education->save();
$education->addReference($source = $url, $excerpt = null, $fields = array('description1'), $name = 'Forbes.com', $detail = null, $date = null);
}
}
$person->save();
$person->addReference($source = $url, $excerpt = null, $fields = array('name_prefix', 'name_first', 'name_middle', 'name_last', 'name_suffix', 'name_nick', 'summary', 'net_worth', 'start_date'), $name = 'Forbes.com', $detail = null, $date = null);
$this->saveToList($person, $rank);
$this->attachImage($person, $imageUrl);
} else {
echo "Couldn't get person: " . $url . "\n";
}
}
protected function getLogoFromWikipedia(Entity $org)
{
if ($this->imageExists($org)) {
return true;
}
//construct search query
$wikipedia = new LsWikipedia();
$response = $wikipedia->requestImages($org->name);
if (!$response) {
return false;
}
$matches = $response->query->pages->page;
$this->printDebug("Images on Wikipedia page: " . count($matches));
foreach ($matches as $match) {
$ii = (array) $match->imageinfo->ii;
$attributes = $ii['@attributes'];
$logo_image_url = $attributes['url'];
$logo_image_alt = $attributes['comment'];
if (preg_match('/^http.*(Wikinews|Wikiversity|Wikisource|Wikiquote|Wikibooks|Wiktionary|Commons-logo)/i', $logo_image_url)) {
continue;
}
$this->printDebug("Checking: " . $logo_image_url);
$org_name_parts = split("[ \\.\\_\\-]", strtolower($org->name));
$org_name_parts[] = "logo";
$org_name_parts[] = "seal";
$image_name_parts = split("[ \\.\\_\\-]", preg_replace("/[0-9]/", "", strtolower(basename(urldecode($logo_image_url)))));
$intersect = array_intersect($image_name_parts, $org_name_parts);
if (count($intersect) >= 2) {
if ($this->attachImage($org, $logo_image_url, 'Organization logo')) {
$this->printDebug("Saved");
return true;
}
}
/*
if((preg_match('/^http.*(logo|seal).*(png|gif|svg|jpg)$/i', $logo_image_url)) )
{
$this->printDebug( "Found 1");
if($this->attachImage($org, $logo_image_url, 'Organization logo'))
{
$this->printDebug( "Committed");
return true;
}
}
if( (preg_match('/(png|gif|svg|jpg)$/i',$logo_image_url) && preg_match('/\b(logo|seal)\b/i',$logo_image_alt)) )
{
$this->printDebug( "Found 2");
if($this->attachImage($org, $logo_image_url, 'Organization logo'))
{
$this->printDebug( "Committed");
return true;
}
}
if(preg_match('/^http.*'.str_replace(LsLanguage::$punctuations, '_', $org->name).'.+\.(png|gif|svg)$/i', $logo_image_url))
{
$this->printDebug( "Found 3");
if( $this->attachImage($org, $logo_image_url, 'Organization logo'))
{
$this->printDebug( "Committed");
return true;
}
}
*/
}
$this->printDebug("Logo not found on Wikipedia");
return false;
}
function import(Entity $person, $possible_persons)
{
//loop through the people we found. usually just one.
foreach ($possible_persons as $possible_person) {
$this->printDebug('Query returned ' . count($possible_person) . ' person named ' . $possible_person->name);
//this person does not provide education. we skip
if (count($possible_person->education)) {
$this->printDebug('Education found');
} else {
$this->printDebug('No education history found');
continue;
}
//get employement info for this possible match
$possible_person_bio = $possible_person->summary;
if (count($possible_person->employment_history)) {
foreach ($possible_person->employment_history as $employment) {
$possible_person_bio .= ' ' . $employment->company . " ";
}
$this->printDebug('Employment found');
} else {
$this->printDebug('No employment history found');
continue;
}
//get employment info for the person in our database
$relationship_orgs = $person->getRelatedEntitiesQuery('Org', RelationshipTable::POSITION_CATEGORY, null, null, null, false, 1)->execute();
$person_bio = $person->summary;
foreach ($relationship_orgs as $org) {
$person_bio .= ' ' . $org->name;
}
//lets see how many matches we get
$matches = LsLanguage::getCommonPronouns($person_bio, trim($possible_person_bio), LsLanguage::$business);
if (count($matches)) {
foreach ($possible_person->education as $school) {
$school->institution = mb_convert_encoding($school->institution, 'UTF-8');
$school->institution = preg_replace('//isu', ' ', $school->institution);
$this->printDebug('Looking for the school: ' . $school->institution);
$current_school = EntityTable::findByAlias($school->institution, $context = 'bw_school');
//find school
if ($current_school) {
$this->printDebug('Found school');
} else {
$current_school = EntityTable::getByExtensionQuery(array('Org', 'School'))->addWhere('LOWER(org.name) LIKE ?', '%' . strtolower($school->institution) . "%")->fetchOne();
if (!$current_school) {
$new_school = new Entity();
$new_school->addExtension('Org');
$new_school->addExtension('School');
$new_school->name = $school->institution;
$wikipedia = new LsWikipedia();
$wikipedia->request($school->institution);
if ($wikipedia->execute() && !$wikipedia->isDisambiguation()) {
$info_box = $wikipedia->getInfoBox();
if (isset($info_box['students']) && preg_match('/([\\d\\,]{2,})/isu', $info_box['students']['clean'], $match)) {
$new_school->students = LsNumber::clean($match[1]);
} else {
$student_types = array('undergrad', 'postgrad', 'grad', 'doctoral');
$num_students = 0;
foreach ($student_types as $st) {
if (isset($info_box[$st]) && preg_match('/([\\d\\,]{2,})/isu', $info_box[$st]['clean'], $match)) {
$num_students += LsNumber::clean($match[1]);
}
}
if ($num_students > 0) {
$new_school->students = $num_students;
}
}
if (isset($info_box['faculty']) && preg_match('/([\\d\\,]{2,})/isu', $info_box['faculty']['clean'], $match)) {
$new_school->faculty = LsNumber::clean($match[1]);
}
if (isset($info_box['type'])) {
if (stristr($info_box['type']['clean'], 'public')) {
$new_school->is_private = 0;
} else {
if (stristr($info_box['type']['clean'], 'private')) {
$new_school->is_private = 1;
}
}
}
if (isset($info_box['endowment'])) {
if (preg_match('/(\\$[\\d\\,\\.\\s]+)(million|billion)/isu', $info_box['endowment']['clean'], $match)) {
if (strtolower($match[2]) == 'billion') {
$factor = 1000000000;
} else {
$factor = 1000000;
}
$new_school->endowment = LsNumber::formatDollarAmountAsNumber($match[1], $factor);
}
}
if (isset($info_box['established'])) {
$year = null;
if ($date = LsDate::convertDate($info_box['established']['clean'])) {
$new_school->start_date = $date;
} else {
if (preg_match('/\\b(\\d\\d\\d\\d)\\b/isu', $info_box['established']['clean'], $match)) {
$new_school->start_date = $match[1];
}
}
}
$summary = trim($wikipedia->getIntroduction());
$summary = preg_replace('/\\n\\s*\\n/isu', '', $summary);
if (strlen($summary) > 10) {
$new_school->summary = $summary;
}
$new_school->save();
$new_school->addReference($source = $wikipedia->getUrl(), $excerpt = null, $fields = array('summary'), $name = 'Wikipedia');
} else {
$new_school->save();
}
$current_school = $new_school;
$this->printDebug('Adding new school');
}
$alias = new Alias();
$alias->name = $school->institution;
$alias->context = 'bw_school';
$alias->Entity = $current_school;
$alias->save();
}
//find degree
$degree = null;
if (!($degree = DegreeTable::getByText($school->degree))) {
$degree = DegreeTable::addDegree($school->degree);
$this->printDebug('Adding new degree');
}
//find relationship
$relationship = null;
$relationships = $person->getRelationshipsWithQuery($current_school, RelationshipTable::EDUCATION_CATEGORY)->execute();
foreach ($relationships as $existing_relationship) {
if ($existing_relationship->degree_id == $degree->id) {
$relationship = $existing_relationship;
break;
}
}
if ($relationship) {
$this->printDebug('Relationship between person and school exists');
} else {
$relationship = new Relationship();
$relationship->Entity1 = $person;
$relationship->Entity2 = $current_school;
$relationship->description1 = 'student';
$relationship->is_current = 0;
if ($school->year) {
$relationship->end_date = $school->year;
}
$relationship->setCategory('Education');
$this->printDebug('Creating new relationship between person and school');
}
//save
$relationship->save();
//add degree and reference
if ($relationship->degree_id == null) {
$reference_name = strstr($school->source, 'wikipedia') ? "Wikipedia" : "BusinessWeek";
$relationship->Degree = $degree;
$relationship->save();
$relationship->addReference($source = $school->source, $excerpt = null, $fields = array('degree_id'), $name = $reference_name, $detail = null, $date = null);
$this->printDebug('Adding degree and reference');
}
}
} else {
$this->printDebug('No organization matches');
return false;
}
}
return true;
}
