5 require_once(CONST_LibDir.'/SimpleWordList.php');
11 private $oNormalizer = null;
13 public function __construct(&$oDB)
16 $this->oNormalizer = \Transliterator::createFromRules(CONST_Term_Normalization_Rules);
19 public function checkStatus()
21 $sStandardWord = $this->oDB->getOne("SELECT make_standard_name('a')");
22 if ($sStandardWord === false) {
23 throw new \Exception('Module failed', 701);
26 if ($sStandardWord != 'a') {
27 throw new \Exception('Module call failed', 702);
30 $sSQL = "SELECT word_id FROM word WHERE word_token IN (' a')";
31 $iWordID = $this->oDB->getOne($sSQL);
32 if ($iWordID === false) {
33 throw new \Exception('Query failed', 703);
36 throw new \Exception('No value', 704);
41 public function normalizeString($sTerm)
43 if ($this->oNormalizer === null) {
47 return $this->oNormalizer->transliterate($sTerm);
51 public function mostFrequentWords($iNum)
53 $sSQL = 'SELECT word FROM word WHERE word is not null ';
54 $sSQL .= 'ORDER BY search_name_count DESC LIMIT '.$iNum;
55 return $this->oDB->getCol($sSQL);
59 public function tokensForSpecialTerm($sTerm)
63 $sSQL = 'SELECT word_id, class, type FROM word ';
64 $sSQL .= ' WHERE word_token = \' \' || make_standard_name(:term)';
65 $sSQL .= ' AND class is not null AND class not in (\'place\')';
67 Debug::printVar('Term', $sTerm);
68 Debug::printSQL($sSQL);
69 $aSearchWords = $this->oDB->getAll($sSQL, array(':term' => $sTerm));
71 Debug::printVar('Results', $aSearchWords);
73 foreach ($aSearchWords as $aSearchTerm) {
74 $aResults[] = new \Nominatim\Token\SpecialTerm(
75 $aSearchTerm['word_id'],
76 $aSearchTerm['class'],
78 \Nominatim\Operator::TYPE
82 Debug::printVar('Special term tokens', $aResults);
88 public function extractTokensFromPhrases(&$aPhrases)
90 // First get the normalized version of all phrases
94 foreach ($aPhrases as $iPhrase => $oPhrase) {
95 $sNormQuery .= ','.$this->normalizeString($oPhrase->getPhrase());
96 $sSQL .= 'make_standard_name(:' .$iPhrase.') as p'.$iPhrase.',';
97 $aParams[':'.$iPhrase] = $oPhrase->getPhrase();
99 // Conflicts between US state abbreviations and various words
100 // for 'the' in different languages
101 switch (strtolower($oPhrase->getPhrase())) {
103 $aParams[':'.$iPhrase] = 'illinois';
106 $aParams[':'.$iPhrase] = 'alabama';
109 $aParams[':'.$iPhrase] = 'louisiana';
112 $aParams[':'.$iPhrase] = $oPhrase->getPhrase();
116 $sSQL = substr($sSQL, 0, -1);
118 Debug::printSQL($sSQL);
119 Debug::printVar('SQL parameters', $aParams);
121 $aNormPhrases = $this->oDB->getRow($sSQL, $aParams);
123 Debug::printVar('SQL result', $aNormPhrases);
125 // now compute all possible tokens
126 $aWordLists = array();
128 foreach ($aNormPhrases as $sPhrase) {
129 $oWordList = new SimpleWordList($sPhrase);
131 foreach ($oWordList->getTokens() as $sToken) {
132 $aTokens[' '.$sToken] = ' '.$sToken;
133 $aTokens[$sToken] = $sToken;
136 $aWordLists[] = $oWordList;
139 Debug::printVar('Tokens', $aTokens);
140 Debug::printVar('WordLists', $aWordLists);
142 $oValidTokens = $this->computeValidTokens($aTokens, $sNormQuery);
144 foreach ($aPhrases as $iPhrase => $oPhrase) {
145 $oPhrase->setWordSets($aWordLists[$iPhrase]->getWordSets($oValidTokens));
148 return $oValidTokens;
152 private function computeValidTokens($aTokens, $sNormQuery)
154 $oValidTokens = new TokenList();
156 if (!empty($aTokens)) {
157 $this->addTokensFromDB($oValidTokens, $aTokens, $sNormQuery);
159 // Try more interpretations for Tokens that could not be matched.
160 foreach ($aTokens as $sToken) {
161 if ($sToken[0] != ' ' && !$oValidTokens->contains($sToken)) {
162 if (preg_match('/^([0-9]{5}) [0-9]{4}$/', $sToken, $aData)) {
163 // US ZIP+4 codes - merge in the 5-digit ZIP code
164 $oValidTokens->addToken(
166 new Token\Postcode(null, $aData[1], 'us')
168 } elseif (preg_match('/^[0-9]+$/', $sToken)) {
169 // Unknown single word token with a number.
170 // Assume it is a house number.
171 $oValidTokens->addToken(
173 new Token\HouseNumber(null, trim($sToken))
180 return $oValidTokens;
184 private function addTokensFromDB(&$oValidTokens, $aTokens, $sNormQuery)
186 // Check which tokens we have, get the ID numbers
187 $sSQL = 'SELECT word_id, word_token, word, class, type, country_code,';
188 $sSQL .= ' operator, coalesce(search_name_count, 0) as count';
189 $sSQL .= ' FROM word WHERE word_token in (';
190 $sSQL .= join(',', $this->oDB->getDBQuotedList($aTokens)).')';
192 Debug::printSQL($sSQL);
194 $aDBWords = $this->oDB->getAll($sSQL, null, 'Could not get word tokens.');
196 foreach ($aDBWords as $aWord) {
198 $iId = (int) $aWord['word_id'];
200 if ($aWord['class']) {
201 // Special terms need to appear in their normalized form.
202 // (postcodes are not normalized in the word table)
203 $sNormWord = $this->normalizeString($aWord['word']);
204 if ($aWord['word'] && strpos($sNormQuery, $sNormWord) === false) {
208 if ($aWord['class'] == 'place' && $aWord['type'] == 'house') {
209 $oToken = new Token\HouseNumber($iId, trim($aWord['word_token']));
210 } elseif ($aWord['class'] == 'place' && $aWord['type'] == 'postcode') {
212 && pg_escape_string($aWord['word']) == $aWord['word']
214 $oToken = new Token\Postcode(
217 $aWord['country_code']
221 // near and in operator the same at the moment
222 $oToken = new Token\SpecialTerm(
226 $aWord['operator'] ? Operator::NEAR : Operator::NONE
229 } elseif ($aWord['country_code']) {
230 $oToken = new Token\Country($iId, $aWord['country_code']);
231 } elseif ($aWord['word_token'][0] == ' ') {
232 $oToken = new Token\Word(
234 (int) $aWord['count'],
235 substr_count($aWord['word_token'], ' ')
237 // For backward compatibility: ignore all partial tokens with more
239 } elseif (strpos($aWord['word_token'], ' ') === false) {
240 $oToken = new Token\Partial(
242 $aWord['word_token'],
243 (int) $aWord['count']
248 // remove any leading spaces
249 if ($aWord['word_token'][0] == ' ') {
250 $oValidTokens->addToken(substr($aWord['word_token'], 1), $oToken);
252 $oValidTokens->addToken($aWord['word_token'], $oToken);