/**
* Calculate the Daitch-Mokotoff soundex for a word.
*
* @param string $name
*
* @return string[] List of possible DM codes for the word.
*/
private static function daitchMokotoffWord($name)
{
// Apply special transformation rules to the input string
$name = I18N::strtoupper($name);
foreach (self::$transformNameTable as $transformRule) {
$name = str_replace($transformRule[0], $transformRule[1], $name);
}
// Initialize
$name_script = I18N::textScript($name);
$noVowels = $name_script == 'Hebr' || $name_script == 'Arab';
$lastPos = strlen($name) - 1;
$currPos = 0;
$state = 1;
// 1: start of input string, 2: before vowel, 3: other
$result = array();
// accumulate complete 6-digit D-M codes here
$partialResult = array();
// accumulate incomplete D-M codes here
$partialResult[] = array('!');
// initialize 1st partial result ('!' stops "duplicate sound" check)
// Loop through the input string.
// Stop when the string is exhausted or when no more partial results remain
while (count($partialResult) !== 0 && $currPos <= $lastPos) {
// Find the DM coding table entry for the chunk at the current position
$thisEntry = substr($name, $currPos, self::MAXCHAR);
// Get maximum length chunk
while ($thisEntry != '') {
if (isset(self::$dmsounds[$thisEntry])) {
break;
}
$thisEntry = substr($thisEntry, 0, -1);
// Not in table: try a shorter chunk
}
if ($thisEntry === '') {
$currPos++;
// Not in table: advance pointer to next byte
continue;
// and try again
}
$soundTableEntry = self::$dmsounds[$thisEntry];
$workingResult = $partialResult;
$partialResult = array();
$currPos += strlen($thisEntry);
// Not at beginning of input string
if ($state != 1) {
if ($currPos <= $lastPos) {
// Determine whether the next chunk is a vowel
$nextEntry = substr($name, $currPos, self::MAXCHAR);
// Get maximum length chunk
while ($nextEntry != '') {
if (isset(self::$dmsounds[$nextEntry])) {
break;
}
$nextEntry = substr($nextEntry, 0, -1);
// Not in table: try a shorter chunk
}
} else {
$nextEntry = '';
}
if ($nextEntry != '' && self::$dmsounds[$nextEntry][0] != '0') {
$state = 2;
} else {
// Next chunk is a vowel
$state = 3;
}
}
while ($state < count($soundTableEntry)) {
// empty means 'ignore this sound in this state'
if ($soundTableEntry[$state] == '') {
foreach ($workingResult as $workingEntry) {
$tempEntry = $workingEntry;
$tempEntry[count($tempEntry) - 1] .= '!';
// Prevent false 'doubles'
$partialResult[] = $tempEntry;
}
} else {
foreach ($workingResult as $workingEntry) {
if ($soundTableEntry[$state] !== $workingEntry[count($workingEntry) - 1]) {
// Incoming sound isn't a duplicate of the previous sound
$workingEntry[] = $soundTableEntry[$state];
} else {
// Incoming sound is a duplicate of the previous sound
// For Hebrew and Arabic, we need to create a pair of D-M sound codes,
// one of the pair with only a single occurrence of the duplicate sound,
// the other with both occurrences
if ($noVowels) {
$workingEntry[] = $soundTableEntry[$state];
}
}
if (count($workingEntry) < 7) {
$partialResult[] = $workingEntry;
} else {
// This is the 6th code in the sequence
// We're looking for 7 entries because the first is '!' and doesn't count
$tempResult = str_replace('!', '', implode('', $workingEntry));
// Only return codes from recognisable sounds
if ($tempResult) {
$result[] = substr($tempResult . '000000', 0, 6);
}
}
}
}
$state = $state + 3;
// Advance to next triplet while keeping the same basic state
}
}
// Zero-fill and copy all remaining partial results
foreach ($partialResult as $workingEntry) {
$tempResult = str_replace('!', '', implode('', $workingEntry));
// Only return codes from recognisable sounds
if ($tempResult) {
$result[] = substr($tempResult . '000000', 0, 6);
}
}
return $result;
}
