Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
97.20% |
104 / 107 |
|
71.43% |
5 / 7 |
CRAP | |
0.00% |
0 / 1 |
| LanguageNameSearch | |
97.20% |
104 / 107 |
|
71.43% |
5 / 7 |
38 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| search | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| doSearch | |
96.36% |
53 / 55 |
|
0.00% |
0 / 1 |
17 | |||
| matchNames | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
3 | |||
| getIndex | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
2 | |||
| getCodepoint | |
100.00% |
21 / 21 |
|
100.00% |
1 / 1 |
7 | |||
| levenshteinDistance | |
95.24% |
20 / 21 |
|
0.00% |
0 / 1 |
7 | |||
| 1 | <?php |
| 2 | |
| 3 | namespace MediaWiki\Language; |
| 4 | |
| 5 | use MediaWiki\MediaWikiServices; |
| 6 | use Wikimedia\LanguageData\LanguageUtil; |
| 7 | |
| 8 | /** |
| 9 | * Cross-Language Language name search |
| 10 | * |
| 11 | * FIXME: This class can be marked as readonly once AutoLoaderStructureTest can |
| 12 | * parse "readonly" annotations. |
| 13 | * |
| 14 | * Copyright (C) 2012 Alolita Sharma, Amir Aharoni, Arun Ganesh, Brandon Harris, |
| 15 | * Niklas Laxström, Pau Giner, Santhosh Thottingal, Siebrand Mazeland and other |
| 16 | * contributors. |
| 17 | * |
| 18 | * @license GPL-2.0-or-later |
| 19 | */ |
| 20 | class LanguageNameSearch { |
| 21 | private LanguageNameUtils $languageNameUtils; |
| 22 | |
| 23 | public function __construct( LanguageNameUtils $languageNameUtils ) { |
| 24 | $this->languageNameUtils = $languageNameUtils; |
| 25 | } |
| 26 | |
| 27 | /** |
| 28 | * Find languages with fuzzy matching. |
| 29 | * The order of results is following: |
| 30 | * 1: exact language code match |
| 31 | * 2: exact language name match in any language |
| 32 | * 3: prefix language name match in any language |
| 33 | * 4: infix language name match in any language |
| 34 | * |
| 35 | * The returned language name for autocompletion is the first one that |
| 36 | * matches in this list: |
| 37 | * 1: exact match in [user, autonym, any other language] |
| 38 | * 2: prefix match in [user, autonym, any other language] |
| 39 | * 3: inline match in [user, autonym, any other language] |
| 40 | * |
| 41 | * @param string $searchKey |
| 42 | * @param int $typos |
| 43 | * @param string|null $userLanguage Language tag. |
| 44 | * @return array |
| 45 | */ |
| 46 | public static function search( string $searchKey, int $typos = 0, ?string $userLanguage = null ): array { |
| 47 | $services = MediaWikiServices::getInstance(); |
| 48 | $instance = $services->getLanguageNameSearch(); |
| 49 | return $instance->doSearch( $searchKey, $typos, $userLanguage ); |
| 50 | } |
| 51 | |
| 52 | /** |
| 53 | * Find languages with fuzzy matching. |
| 54 | * The order of results is following: |
| 55 | * 1: exact language code match |
| 56 | * 2: exact language name match in any language |
| 57 | * 3: remaining languages matche by autonym and script group |
| 58 | * |
| 59 | * The returned language name for autocompletion is the first one that |
| 60 | * matches in this list: |
| 61 | * 1: exact match in [user, autonym, any other language] |
| 62 | * 2: prefix match in [user, autonym, any other language] |
| 63 | * 3: inline match in [user, autonym, any other language] |
| 64 | * |
| 65 | * @param string $searchKey |
| 66 | * @param int $typos |
| 67 | * @param string|null $userLanguage Language tag. |
| 68 | * @return array |
| 69 | */ |
| 70 | public function doSearch( string $searchKey, int $typos = 0, ?string $userLanguage = null ): array { |
| 71 | $exactMatches = []; |
| 72 | $otherMatches = []; |
| 73 | $searchKey = mb_strtolower( $searchKey ); |
| 74 | |
| 75 | if ( mb_strlen( $searchKey ) > 100 ) { |
| 76 | // Searching with long search keys for language names is not useful. So, return early. |
| 77 | return []; |
| 78 | } |
| 79 | |
| 80 | $languageNameUtils = $this->languageNameUtils; |
| 81 | |
| 82 | // Always prefer exact language code match |
| 83 | if ( $languageNameUtils->isKnownLanguageTag( $searchKey ) ) { |
| 84 | $name = mb_strtolower( $languageNameUtils->getLanguageName( $searchKey, $userLanguage ) ); |
| 85 | // Check if language code is a prefix of the name |
| 86 | if ( str_starts_with( $name, $searchKey ) ) { |
| 87 | $exactMatches[$searchKey] = $name; |
| 88 | } else { |
| 89 | $exactMatches[$searchKey] = "$searchKey – $name"; |
| 90 | } |
| 91 | } |
| 92 | |
| 93 | $index = self::getIndex( $searchKey ); |
| 94 | static $buckets = null; |
| 95 | if ( $buckets === null ) { |
| 96 | $data = json_decode( |
| 97 | file_get_contents( __DIR__ . '/../../languages/data/LanguageNameSearchData.json' ), |
| 98 | true |
| 99 | ); |
| 100 | $buckets = $data['buckets'] ?? []; |
| 101 | } |
| 102 | $bucketsForIndex = $buckets[$index] ?? []; |
| 103 | |
| 104 | // types are 'prefix', 'infix' (in this order!) |
| 105 | foreach ( $bucketsForIndex as $bucket ) { |
| 106 | foreach ( $bucket as $name => $code ) { |
| 107 | // We can skip checking languages we already have in either match list |
| 108 | if ( isset( $exactMatches[$code] ) || isset( $otherMatches[$code] ) ) { |
| 109 | continue; |
| 110 | } |
| 111 | |
| 112 | // Apply fuzzy search |
| 113 | if ( !$this->matchNames( $name, $searchKey, $typos ) ) { |
| 114 | continue; |
| 115 | } |
| 116 | |
| 117 | // Once we find a match, figure out the best name to display to the user |
| 118 | // If $userLanguage is not provided (null), it is the same as autonym |
| 119 | $candidates = [ |
| 120 | mb_strtolower( $languageNameUtils->getLanguageName( $code, $userLanguage ) ), |
| 121 | mb_strtolower( $languageNameUtils->getLanguageName( $code, LanguageNameUtils::AUTONYMS ) ), |
| 122 | $name |
| 123 | ]; |
| 124 | |
| 125 | // First check for exact name match |
| 126 | foreach ( $candidates as $candidate ) { |
| 127 | if ( $searchKey === $candidate ) { |
| 128 | $exactMatches[$code] = $candidate; |
| 129 | continue 2; |
| 130 | } |
| 131 | } |
| 132 | |
| 133 | // Otherwise, check for prefix/infix match |
| 134 | foreach ( $candidates as $candidate ) { |
| 135 | if ( $this->matchNames( $candidate, $searchKey, $typos ) ) { |
| 136 | $otherMatches[$code] = $candidate; |
| 137 | continue 2; |
| 138 | } |
| 139 | } |
| 140 | } |
| 141 | } |
| 142 | |
| 143 | // Sort the remaining matches by autonym and script group |
| 144 | $languageUtil = LanguageUtil::get(); |
| 145 | $otherCodes = array_keys( $otherMatches ); |
| 146 | |
| 147 | $knownCodes = []; |
| 148 | $unknownCodes = []; |
| 149 | foreach ( $otherCodes as $code ) { |
| 150 | if ( $languageUtil->isKnown( $code ) ) { |
| 151 | $knownCodes[] = $code; |
| 152 | } else { |
| 153 | $unknownCodes[] = $code; |
| 154 | } |
| 155 | } |
| 156 | |
| 157 | $knownCodes = $languageUtil->sortByAutonym( $knownCodes ); |
| 158 | $knownCodes = $languageUtil->sortByScriptGroup( $knownCodes ); |
| 159 | |
| 160 | asort( $unknownCodes, SORT_STRING | SORT_FLAG_CASE ); |
| 161 | |
| 162 | $sortedOtherCodes = array_merge( $knownCodes, $unknownCodes ); |
| 163 | |
| 164 | // Merge exact matches (retaining priority order at the top) and sorted matches |
| 165 | $sortedResults = $exactMatches; |
| 166 | foreach ( $sortedOtherCodes as $code ) { |
| 167 | $sortedResults[$code] = $otherMatches[$code]; |
| 168 | } |
| 169 | |
| 170 | return $sortedResults; |
| 171 | } |
| 172 | |
| 173 | private function matchNames( string $name, string $searchKey, int $typos ): bool { |
| 174 | return strrpos( $name, $searchKey, -strlen( $name ) ) !== false |
| 175 | || ( $typos > 0 && $this->levenshteinDistance( $name, $searchKey ) <= $typos ); |
| 176 | } |
| 177 | |
| 178 | public static function getIndex( string $name ): int { |
| 179 | $codepoint = self::getCodepoint( $name ); |
| 180 | |
| 181 | if ( $codepoint < 4000 ) { |
| 182 | // For latin etc. we need smaller buckets for speed |
| 183 | return $codepoint; |
| 184 | } |
| 185 | |
| 186 | // Try to group names of same script together |
| 187 | return $codepoint - ( $codepoint % 1000 ); |
| 188 | } |
| 189 | |
| 190 | /** |
| 191 | * Get the code point of first letter of string |
| 192 | * |
| 193 | * @param string $str |
| 194 | * @return int Code point of first letter of string |
| 195 | */ |
| 196 | private static function getCodepoint( string $str ): int { |
| 197 | $values = []; |
| 198 | $lookingFor = 1; |
| 199 | $strLen = strlen( $str ); |
| 200 | $number = 0; |
| 201 | |
| 202 | for ( $i = 0; $i < $strLen; $i++ ) { |
| 203 | $thisValue = ord( $str[$i] ); |
| 204 | if ( $thisValue < 128 ) { |
| 205 | $number = $thisValue; |
| 206 | |
| 207 | break; |
| 208 | } |
| 209 | |
| 210 | // Codepoints larger than 127 are represented by multi-byte sequences |
| 211 | if ( $values === [] ) { |
| 212 | // 224 is the lowest non-overlong-encoded codepoint. |
| 213 | $lookingFor = ( $thisValue < 224 ) ? 2 : 3; |
| 214 | } |
| 215 | |
| 216 | $values[] = $thisValue; |
| 217 | if ( count( $values ) === $lookingFor ) { |
| 218 | // Refer http://en.wikipedia.org/wiki/UTF-8#Description |
| 219 | if ( $lookingFor === 3 ) { |
| 220 | $number = ( $values[0] % 16 ) * 4096; |
| 221 | $number += ( $values[1] % 64 ) * 64; |
| 222 | $number += $values[2] % 64; |
| 223 | } else { |
| 224 | $number = ( $values[0] % 32 ) * 64; |
| 225 | $number += $values[1] % 64; |
| 226 | } |
| 227 | |
| 228 | break; |
| 229 | } |
| 230 | } |
| 231 | |
| 232 | return $number; |
| 233 | } |
| 234 | |
| 235 | /** |
| 236 | * Calculate the Levenshtein distance between two strings |
| 237 | */ |
| 238 | private function levenshteinDistance( string $str1, string $str2 ): int { |
| 239 | if ( $str1 === $str2 ) { |
| 240 | return 0; |
| 241 | } |
| 242 | $length1 = mb_strlen( $str1, 'UTF-8' ); |
| 243 | $length2 = mb_strlen( $str2, 'UTF-8' ); |
| 244 | if ( $length1 === 0 ) { |
| 245 | return $length2; |
| 246 | } |
| 247 | if ( $length1 < $length2 ) { |
| 248 | return $this->levenshteinDistance( $str2, $str1 ); |
| 249 | } |
| 250 | $prevRow = range( 0, $length2 ); |
| 251 | for ( $i = 0; $i < $length1; $i++ ) { |
| 252 | $currentRow = []; |
| 253 | $currentRow[0] = $i + 1; |
| 254 | $c1 = mb_substr( $str1, $i, 1, 'UTF-8' ); |
| 255 | for ( $j = 0; $j < $length2; $j++ ) { |
| 256 | $c2 = mb_substr( $str2, $j, 1, 'UTF-8' ); |
| 257 | $insertions = $prevRow[$j + 1] + 1; |
| 258 | $deletions = $currentRow[$j] + 1; |
| 259 | $substitutions = $prevRow[$j] + ( ( $c1 !== $c2 ) ? 1 : 0 ); |
| 260 | $currentRow[] = min( $insertions, $deletions, $substitutions ); |
| 261 | } |
| 262 | $prevRow = $currentRow; |
| 263 | } |
| 264 | |
| 265 | return $prevRow[$length2]; |
| 266 | } |
| 267 | } |