/** * Add word to word-array * This function should be used to make sure CJK sequences are split up in the right way * * @param array Array of accumulated words * @param string Complete Input string from where to extract word * @param integer Start position of word in input string * @param integer The Length of the word string from start position * @return void * @todo Define visibility */ public function addWords(&$words, &$wordString, $start, $len) { // Get word out of string: $theWord = substr($wordString, $start, $len); // Get next chars unicode number and find type: $bc = 0; $cp = $this->utf8_ord($theWord, $bc); list($cType) = $this->charType($cp); // If string is a CJK sequence we follow this algorithm: /* DESCRIPTION OF (CJK) ALGORITHMContinuous letters and numbers make up words. Spaces and symbols separate letters and numbers into words. This is sufficient for all western text.CJK doesn't use spaces or separators to separate words, so the only way to really find out what constitutes a word would be to have a dictionary and advanced heuristics. Instead, we form pairs from consecutive characters, in such a way that searches will find only characters that appear more-or-less the right sequence. For example:ABCDE => AB BC CD DEThis works okay since both the index and the search query is split in the same manner, and since the set of characters is huge so the extra matches are not significant.(Hint taken from ZOPEs chinese user group)[Kasper: As far as I can see this will only work well with or-searches!] */ if ($cType == 'cjk') { // Find total string length: $strlen = $this->csObj->utf8_strlen($theWord); // Traverse string length and add words as pairs of two chars: for ($a = 0; $a < $strlen; $a++) { if ($strlen == 1 || $a < $strlen - 1) { $words[] = $this->csObj->utf8_substr($theWord, $a, 2); } } } else { // Normal "single-byte" chars: // Remove chars: foreach ($this->lexerConf['removeChars'] as $skipJoin) { $theWord = str_replace($this->csObj->UnumberToChar($skipJoin), '', $theWord); } // Add word: $words[] = $theWord; } }
/** * @test */ public function utf8_strlenForNonEmptyAsciiOnlyStringReturnsNumberOfCharacters() { $this->assertEquals(10, $this->subject->utf8_strlen('good omens')); }