/**
* Dedup: Return unique IDs (control numbers)
*
* @return string[]
*/
public function getUniqueIDs()
{
$arr = [];
$nbn = $this->getField('015');
if ($nbn) {
$nr = MetadataUtils::normalize(strtok($this->getSubfield($nbn, 'a'), ' '));
$src = $this->getSubfield($nbn, '2');
if ($src && $nr) {
$arr[] = "({$src}){$nr}";
}
}
$nba = $this->getField('016');
if ($nba) {
$nr = MetadataUtils::normalize(strtok($this->getSubfield($nba, 'a'), ' '));
$src = $this->getSubfield($nba, '2');
if ($src && $nr) {
$arr[] = "({$src}){$nr}";
}
}
$id = $this->getField('024');
if ($id) {
$nr = MetadataUtils::normalize(strtok($this->getSubfield($id, 'a'), ' '));
switch ($this->getIndicator($id, 1)) {
case '0':
$src = 'istc';
break;
case '1':
$src = 'upc';
break;
case '2':
$src = 'ismn';
break;
case '3':
$src = 'ian';
break;
case '4':
$src = 'sici';
break;
case '7':
$src = $this->getSubfield($id, '2');
break;
default:
$src = '';
}
if ($src && $nr) {
$arr[] = "({$src}){$nr}";
}
}
return $arr;
}
/**
* Check if records are duplicate matches
*
* @param array $record Mongo record
* @param object $origRecord Metadata record (from $record)
* @param array $candidate Candidate Mongo record
*
* @return boolean
*/
protected function matchRecords($record, $origRecord, $candidate)
{
$cRecord = RecordFactory::createRecord($candidate['format'], MetadataUtils::getRecordData($candidate, true), $candidate['oai_id'], $candidate['source_id']);
if ($this->verbose) {
echo "\nCandidate " . $candidate['_id'] . ":\n" . MetadataUtils::getRecordData($candidate, true) . "\n";
}
// Check that the record does not have access restrictions
if ($cRecord->getAccessRestrictions()) {
if ($this->verbose) {
echo "--Candidate has access restrictions\n";
}
return false;
}
// Check format
$origFormat = $origRecord->getFormat();
$cFormat = $cRecord->getFormat();
$origMapped = $this->solrUpdater->mapFormat($record['source_id'], $origFormat);
$cMapped = $this->solrUpdater->mapFormat($candidate['source_id'], $cFormat);
if ($origFormat != $cFormat && $origMapped != $cMapped) {
if ($this->verbose) {
echo "--Format mismatch: {$origFormat} != {$cFormat} " . "and {$origMapped} != {$cMapped}\n";
}
return false;
}
// Check for common ISBN
$origISBNs = $origRecord->getISBNs();
$cISBNs = $cRecord->getISBNs();
$isect = array_intersect($origISBNs, $cISBNs);
if (!empty($isect)) {
// Shared ISBN -> match
if ($this->verbose) {
echo "++ISBN match:\n";
print_r($origISBNs);
print_r($cISBNs);
echo $origRecord->getFullTitle() . "\n";
echo $cRecord->getFullTitle() . "\n";
}
return true;
}
// Check for other common ID (e.g. NBN)
$origIDs = $origRecord->getUniqueIDs();
$cIDs = $cRecord->getUniqueIDs();
$isect = array_intersect($origIDs, $cIDs);
if (!empty($isect)) {
// Shared ID -> match
if ($this->verbose) {
echo "++ID match:\n";
print_r($origIDs);
print_r($cIDs);
echo $origRecord->getFullTitle() . "\n";
echo $cRecord->getFullTitle() . "\n";
}
return true;
}
$origISSNs = $origRecord->getISSNs();
$cISSNs = $cRecord->getISSNs();
$commonISSNs = array_intersect($origISSNs, $cISSNs);
if (!empty($origISSNs) && !empty($cISSNs) && empty($commonISSNs)) {
// Both have ISSNs but none match
if ($this->verbose) {
echo "++ISSN mismatch:\n";
print_r($origISSNs);
print_r($cISSNs);
echo $origRecord->getFullTitle() . "\n";
echo $cRecord->getFullTitle() . "\n";
}
return false;
}
$origYear = $origRecord->getPublicationYear();
$cYear = $cRecord->getPublicationYear();
if ($origYear && $cYear && $origYear != $cYear) {
if ($this->verbose) {
echo "--Year mismatch: {$origYear} != {$cYear}\n";
}
return false;
}
$pages = $origRecord->getPageCount();
$cPages = $cRecord->getPageCount();
if ($pages && $cPages && abs($pages - $cPages) > 10) {
if ($this->verbose) {
echo "--Pages mismatch ({$pages} != {$cPages})\n";
}
return false;
}
if ($origRecord->getSeriesISSN() != $cRecord->getSeriesISSN()) {
return false;
}
if ($origRecord->getSeriesNumbering() != $cRecord->getSeriesNumbering()) {
return false;
}
$origTitle = MetadataUtils::normalize($origRecord->getTitle(true));
$cTitle = MetadataUtils::normalize($cRecord->getTitle(true));
if (!$origTitle || !$cTitle) {
// No title match without title...
if ($this->verbose) {
echo "No title - no further matching\n";
}
return false;
}
$lev = levenshtein(substr($origTitle, 0, 255), substr($cTitle, 0, 255));
$lev = $lev / strlen($origTitle) * 100;
if ($lev >= 10) {
if ($this->verbose) {
echo "--Title lev discard: {$lev}\nOriginal: {$origTitle}\n" . "Candidate: {$cTitle}\n";
}
return false;
}
$origAuthor = MetadataUtils::normalize($origRecord->getMainAuthor());
$cAuthor = MetadataUtils::normalize($cRecord->getMainAuthor());
$authorLev = 0;
if ($origAuthor || $cAuthor) {
if (!$origAuthor || !$cAuthor) {
if ($this->verbose) {
echo "\nAuthor discard:\nOriginal: {$origAuthor}\n" . "Candidate: {$cAuthor}\n";
}
return false;
}
if (!MetadataUtils::authorMatch($origAuthor, $cAuthor)) {
$authorLev = levenshtein(substr($origAuthor, 0, 255), substr($cAuthor, 0, 255));
$authorLev = $authorLev / mb_strlen($origAuthor) * 100;
if ($authorLev > 20) {
if ($this->verbose) {
echo "\nAuthor lev discard (lev: {$lev}, authorLev: " . "{$authorLev}):\nOriginal: {$origAuthor}\n" . "Candidate: {$cAuthor}\n";
}
return false;
}
}
}
if ($this->verbose) {
echo "\nTitle match (lev: {$lev}, authorLev: {$authorLev}):\n";
echo $origRecord->getFullTitle() . "\n";
echo " {$origAuthor} - {$origTitle}.\n";
echo $cRecord->getFullTitle() . "\n";
echo " {$cAuthor} - {$cTitle}.\n";
}
// We have a match!
return true;
}
/**
* Create a normalized title key for dedup
*
* @param string $title Title
*
* @return string
*/
public static function createTitleKey($title)
{
$full = false;
if (isset(MetadataUtils::$fullTitlePrefixes)) {
$normalTitle = MetadataUtils::normalize($title);
foreach (MetadataUtils::$fullTitlePrefixes as $prefix) {
if (strncmp($normalTitle, $prefix, strlen($prefix)) === 0) {
$full = true;
break;
}
}
}
$words = explode(' ', $title);
$longWords = 0;
$key = '';
$keyLen = 0;
foreach ($words as $word) {
$key .= $word;
$wordLen = mb_strlen($word);
if ($wordLen > 3) {
++$longWords;
}
$keyLen += $wordLen;
// significant chars
if (!$full && ($longWords > 3 || $keyLen > 35)) {
break;
} elseif ($full && $keyLen > 100) {
break;
}
}
return MetadataUtils::normalize($key);
}
/**
* Create Solr array for the given record
*
* @param array $record Mongo record
* @param integer $mergedComponents Number of component parts merged to the
* record
*
* @return string[]
* @throws Exception
*/
protected function createSolrArray($record, &$mergedComponents)
{
global $configArray;
$metadataRecord = RecordFactory::createRecord($record['format'], MetadataUtils::getRecordData($record, true), $record['oai_id'], $record['source_id']);
$source = $record['source_id'];
if (!isset($this->settings[$source])) {
// Try to reload data source settings as they might have been updated
// during a long run
$this->loadDatasources();
if (!isset($this->settings[$source])) {
$this->log->log('createSolrArray', "No settings found for data source '{$source}', record " . "{$record['_id']}: " . $this->prettyPrint($record, true), Logger::FATAL);
throw new Exception("No settings found for data source '{$source}'");
}
}
$settings = $this->settings[$source];
$hiddenComponent = false;
if (isset($record['host_record_id'])) {
if ($settings['componentParts'] == 'merge_all') {
$hiddenComponent = true;
} elseif ($settings['componentParts'] == 'merge_non_articles' || $settings['componentParts'] == 'merge_non_earticles') {
$format = $metadataRecord->getFormat();
if (!in_array($format, $this->allArticleFormats)) {
$hiddenComponent = true;
} elseif (in_array($format, $this->articleFormats)) {
$hiddenComponent = true;
}
}
}
if ($hiddenComponent && !$settings['indexMergedParts']) {
return false;
}
$hasComponentParts = false;
$components = null;
if (!isset($record['host_record_id'])) {
// Fetch info whether component parts exist and need to be merged
if (!$record['linking_id']) {
$this->log->log('createSolrArray', "linking_id missing for record '{$record['_id']}'", Logger::ERROR);
} else {
$components = $this->db->record->find(['source_id' => $record['source_id'], 'host_record_id' => $record['linking_id'], 'deleted' => false])->timeout($this->cursorTimeout);
$hasComponentParts = $components->hasNext();
$format = $metadataRecord->getFormat();
$merge = false;
if ($settings['componentParts'] == 'merge_all') {
$merge = true;
} elseif (!in_array($format, $this->allJournalFormats)) {
$merge = true;
} elseif (in_array($format, $this->journalFormats) && $settings['componentParts'] == 'merge_non_earticles') {
$merge = true;
}
if (!$merge) {
unset($components);
}
}
}
if (isset($components)) {
$mergedComponents += $metadataRecord->mergeComponentParts($components);
}
if (isset($settings['solrTransformationXSLT'])) {
$params = ['source_id' => $source, 'institution' => $settings['institution'], 'format' => $settings['format'], 'id_prefix' => $settings['idPrefix']];
$data = $settings['solrTransformationXSLT']->transformToSolrArray($metadataRecord->toXML(), $params);
} else {
$prependTitleWithSubtitle = isset($settings['prepend_title_with_subtitle']) ? $settings['prepend_title_with_subtitle'] : true;
$data = $metadataRecord->toSolrArray($prependTitleWithSubtitle);
$this->enrich($source, $settings, $metadataRecord, $data);
}
$data['id'] = $record['_id'];
// Record links between host records and component parts
if ($metadataRecord->getIsComponentPart()) {
$hostRecord = null;
if (isset($record['host_record_id']) && $this->db) {
$hostRecord = $this->db->record->find(['source_id' => $record['source_id'], 'linking_id' => $record['host_record_id']])->limit(-1)->timeout($this->cursorTimeout)->getNext();
}
if (!$hostRecord) {
if (isset($record['host_record_id'])) {
$this->log->log('createSolrArray', "Host record '" . $record['host_record_id'] . "' not found for record '" . $record['_id'] . "'", Logger::WARNING);
}
$data['container_title'] = $metadataRecord->getContainerTitle();
} else {
$data['hierarchy_parent_id'] = $hostRecord['_id'];
$hostMetadataRecord = RecordFactory::createRecord($hostRecord['format'], MetadataUtils::getRecordData($hostRecord, true), $hostRecord['oai_id'], $hostRecord['source_id']);
$data['container_title'] = $data['hierarchy_parent_title'] = $hostMetadataRecord->getTitle();
}
$data['container_volume'] = $metadataRecord->getVolume();
$data['container_issue'] = $metadataRecord->getIssue();
$data['container_start_page'] = $metadataRecord->getStartPage();
$data['container_reference'] = $metadataRecord->getContainerReference();
} else {
// Add prefixes to hierarchy linking fields
foreach (['hierarchy_top_id', 'hierarchy_parent_id', 'is_hierarchy_id'] as $field) {
if (isset($data[$field]) && $data[$field]) {
$data[$field] = $record['source_id'] . '.' . $data[$field];
}
}
}
if ($hasComponentParts) {
$data['is_hierarchy_id'] = $record['_id'];
$data['is_hierarchy_title'] = $metadataRecord->getTitle();
}
if (!isset($data['institution'])) {
$data['institution'] = $settings['institution'];
}
foreach ($settings['extraFields'] as $extraField) {
$fieldName = key($extraField);
$fieldValue = current($extraField);
if (isset($data[$fieldName])) {
if (!is_array($data[$fieldName])) {
$data[$fieldName] = [$data[$fieldName]];
}
$data[$fieldName][] = $fieldValue;
} else {
$data[$fieldName] = $fieldValue;
}
}
// Map field values according to any mapping files
foreach ($settings['mappingFiles'] as $field => $map) {
if (isset($data[$field]) && !empty($data[$field])) {
if (is_array($data[$field])) {
$newValues = null;
foreach ($data[$field] as $value) {
if (isset($map[$value])) {
$newValues = $map[$value];
} elseif (isset($map['##default'])) {
$newValues = $map['##default'];
}
}
if (null !== $newValues) {
if (is_array($newValues)) {
$data[$field] = array_values(array_unique($newValues));
} else {
$data[$field] = $newValues;
}
}
} else {
if (isset($map[$data[$field]])) {
$data[$field] = $map[$data[$field]];
} elseif (isset($map['##default'])) {
$data[$field] = $map['##default'];
}
}
} elseif (isset($map['##empty'])) {
$data[$field] = $map['##empty'];
} elseif (isset($map['##emptyarray'])) {
$data[$field] = [$map['##emptyarray']];
}
}
// Special case: Special values for building (institution/location).
// Used by default if building is set as a hierarchical facet.
if ($this->buildingHierarchy || isset($settings['institutionInBuilding'])) {
$useInstitution = isset($settings['institutionInBuilding']) ? $settings['institutionInBuilding'] : 'institution';
switch ($useInstitution) {
case 'driver':
$institutionCode = $data['institution'];
break;
case 'none':
$institutionCode = '';
break;
case 'source':
$institutionCode = $source;
break;
case 'institution/source':
$institutionCode = $settings['institution'] . '/' . $source;
break;
default:
$institutionCode = $settings['institution'];
break;
}
if ($institutionCode) {
if (isset($data['building']) && $data['building']) {
if (is_array($data['building'])) {
foreach ($data['building'] as &$building) {
// Allow also empty values that might result from
// mapping tables
if ($building !== '') {
$building = "{$institutionCode}/{$building}";
}
}
} else {
$data['building'] = $institutionCode . '/' . $data['building'];
}
} else {
$data['building'] = [$institutionCode];
}
}
}
// Hierarchical facets
if (isset($configArray['Solr']['hierarchical_facets'])) {
foreach ($configArray['Solr']['hierarchical_facets'] as $facet) {
if (!isset($data[$facet])) {
continue;
}
$array = [];
if (!is_array($data[$facet])) {
$data[$facet] = [$data[$facet]];
}
foreach ($data[$facet] as $datavalue) {
if ($datavalue === '') {
continue;
}
$values = explode('/', $datavalue);
$hierarchyString = '';
for ($i = 0; $i < count($values); $i++) {
$hierarchyString .= '/' . $values[$i];
$array[] = $i . $hierarchyString . '/';
}
}
$data[$facet] = $array;
}
}
if (!isset($data['allfields'])) {
$all = [];
foreach ($data as $key => $field) {
if (in_array($key, ['fullrecord', 'thumbnail', 'id', 'recordtype', 'ctrlnum'])) {
continue;
}
if (is_array($field)) {
$all = array_merge($all, $field);
} else {
$all[] = $field;
}
}
$data['allfields'] = MetadataUtils::array_iunique($all);
}
$data['first_indexed'] = MetadataUtils::formatTimestamp($record['created']->sec);
$data['last_indexed'] = MetadataUtils::formatTimestamp($record['date']->sec);
$data['recordtype'] = $record['format'];
if (!isset($data['fullrecord'])) {
$data['fullrecord'] = $metadataRecord->toXML();
}
if (!is_array($data['format'])) {
$data['format'] = [$data['format']];
}
if (isset($configArray['Solr']['format_in_allfields']) && $configArray['Solr']['format_in_allfields']) {
foreach ($data['format'] as $format) {
// Replace numbers since they may be be considered word boundaries
$data['allfields'][] = str_replace(['0', '1', '2', '3', '4', '5', '6', '7', '8', '9'], ['ax', 'bx', 'cx', 'dx', 'ex', 'fx', 'gx', 'hx', 'ix', 'jx'], MetadataUtils::normalize($format));
}
}
if ($hiddenComponent) {
$data['hidden_component_boolean'] = true;
}
foreach ($data as $key => &$values) {
if (is_array($values)) {
foreach ($values as $key => &$value) {
$value = MetadataUtils::normalizeUnicode($value);
if (empty($value) || $value === 0 || $value === 0.0 || $value === '0') {
unset($values[$key]);
}
}
$values = array_values(array_unique($values));
} elseif ($key != 'fullrecord') {
$values = MetadataUtils::normalizeUnicode($values);
}
if (empty($values) || $values === 0 || $values === 0.0 || $values === '0') {
unset($data[$key]);
}
}
return $data;
}
