| 1 |
<?php
|
| 2 |
// $Id: csplitter.module,v 1.3 2006/04/29 05:18:43 zealy Exp $
|
| 3 |
// mailto: i.zealy AT gmail dot com
|
| 4 |
|
| 5 |
/**
|
| 6 |
* Implementation of hook_help
|
| 7 |
*/
|
| 8 |
function csplitter_help($section) {
|
| 9 |
$output = "";
|
| 10 |
|
| 11 |
switch ($section) {
|
| 12 |
case "admin/modules#description":
|
| 13 |
$output = t("Provides an chinese splitter for search");
|
| 14 |
break;
|
| 15 |
}
|
| 16 |
return $output;
|
| 17 |
}
|
| 18 |
|
| 19 |
/**
|
| 20 |
* Implementation of hook_menu
|
| 21 |
*/
|
| 22 |
function csplitter_menu($may_cache) {
|
| 23 |
$items = array();
|
| 24 |
|
| 25 |
if ($may_cache) {
|
| 26 |
|
| 27 |
$items[] = array('path' => 'csplitter',
|
| 28 |
'title' => t('chinese splitter'),
|
| 29 |
'callback' => 'csplitter_view',
|
| 30 |
'access' => user_access('search content'),
|
| 31 |
'type' => MENU_CALLBACK);
|
| 32 |
|
| 33 |
$items[] = array('path' => 'admin/settings/csplitter',
|
| 34 |
'title' => t('Chinese splitter'),
|
| 35 |
'description' => t('Configure relevance settings for splitter options'),
|
| 36 |
'callback' => 'drupal_get_form',
|
| 37 |
'callback arguments' => array('csplitter_admin_settings'),
|
| 38 |
'access' => user_access('administer search'),
|
| 39 |
'type' => MENU_NORMAL_ITEM);
|
| 40 |
}
|
| 41 |
return $items;
|
| 42 |
}
|
| 43 |
|
| 44 |
/**
|
| 45 |
* Menu callback; displays the module settings page.
|
| 46 |
*/
|
| 47 |
function csplitter_admin_settings() {
|
| 48 |
$form = array();
|
| 49 |
|
| 50 |
$form['weight'] = array(
|
| 51 |
'#type' => 'markup',
|
| 52 |
'#value' => t('Optional. You can test your configuration '.l('here','csplitter').'.'),
|
| 53 |
);
|
| 54 |
|
| 55 |
$form['Config Algorithm'] = array('#type' => 'fieldset',
|
| 56 |
'#title' => t('Config Arithmetic'),
|
| 57 |
'#collapsible' => TRUE,
|
| 58 |
'#description' => t('Choose which algorithm to use. If you change algorithm, you should re-index search database.'),
|
| 59 |
);
|
| 60 |
if (!function_exists('mb_strlen')) {
|
| 61 |
$form['Config Algorithm']['Error Message']= array('#type' => 'markup',
|
| 62 |
'#value' => '<font color=red>'.t('Your PHP have no mbstring support. Csplitter require it to run correctly.').'</font>',
|
| 63 |
);
|
| 64 |
return system_settings_form($form);
|
| 65 |
}
|
| 66 |
|
| 67 |
$algors = array(1 => t('Forward maximum matching'), 2 => t('Conversely maximum mathing'));
|
| 68 |
$form['Config Algorithm']['csplitter_algorithm'] = array('#type' => 'radios',
|
| 69 |
'#title' => t('Default Algorithm'),
|
| 70 |
'#default_value' => variable_get('csplitter_algorithm', '1'),
|
| 71 |
'#options' => $algors,
|
| 72 |
'#description' => t('The default algorithm for word splitting.'),
|
| 73 |
);
|
| 74 |
$form['Define Cache'] = array('#type' => 'fieldset',
|
| 75 |
'#title' => t('Define Cache'),
|
| 76 |
'#collapsible' => TRUE,
|
| 77 |
'#description' => t('Config using cache for dictionary or not. You can test performance <a href="/csplitter">here</a>.'),
|
| 78 |
);
|
| 79 |
$cache_method = array('none' => t('No caching'), 'disk' => t('Caching to disk'), 'database' => t('Caching to database'));
|
| 80 |
$form['Define Cache']['csplitter_cache'] = array('#type' => 'radios',
|
| 81 |
'#title' => t('Define Dictionary Caching'),
|
| 82 |
'#default_value' => variable_get('csplitter_cache', 'none'),
|
| 83 |
'#options' => $cache_method,
|
| 84 |
'#description' => t('In most server system, disk caching is fastest. If you use database caching, you should sure your database can accept very large sql sentence(max_allowed_packet for mysql should be > 5M possible). '),
|
| 85 |
);
|
| 86 |
|
| 87 |
return system_settings_form($form);
|
| 88 |
}
|
| 89 |
|
| 90 |
/* Menu callback; lists all nodes posted on a given date.
|
| 91 |
*/
|
| 92 |
function csplitter_view() {
|
| 93 |
global $user;
|
| 94 |
$test = variable_get('csplitter_test', NULL);
|
| 95 |
if ($test){
|
| 96 |
$output = '';
|
| 97 |
$test = trim($test);
|
| 98 |
$function = '_csplitter_splite_'. variable_get('csplitter_algorithm', '1');
|
| 99 |
|
| 100 |
if (mb_strlen($test)>0){
|
| 101 |
$cache_method = variable_get('csplitter_cache', 'none');
|
| 102 |
timer_start('csplitter');
|
| 103 |
$outstr = call_user_func($function, $test);//_csplitter_splite_1($test);
|
| 104 |
drupal_set_message($function.', cache:'.$cache_method.'. Processing used '.timer_read('csplitter')."ms");
|
| 105 |
}
|
| 106 |
variable_set('csplitter_test', NULL);
|
| 107 |
$output .= "<h4>input: ".$test."</h4>";
|
| 108 |
$output .= "<h4>output: ".$outstr."</h4>";
|
| 109 |
}
|
| 110 |
|
| 111 |
//return theme('page', $output.csplitter_form());
|
| 112 |
return $output.drupal_get_form('csplitter_form');
|
| 113 |
}
|
| 114 |
|
| 115 |
function csplitter_form(){
|
| 116 |
|
| 117 |
$form['test'] = array(
|
| 118 |
'#type' => 'textarea',
|
| 119 |
'#title' => t('Input'),
|
| 120 |
//'#default_value' => $edit['test'],
|
| 121 |
'#description' => t('Input test sentences(Chinese or other) here.')
|
| 122 |
);
|
| 123 |
$form['submit'] = array(
|
| 124 |
'#type' => 'submit',
|
| 125 |
'#value' => t('Test')
|
| 126 |
);
|
| 127 |
//return drupal_get_form('csplitter_form', $form, 'csplitter_form');
|
| 128 |
return $form;
|
| 129 |
}
|
| 130 |
|
| 131 |
function csplitter_form_submit($form_id, $form_values){
|
| 132 |
if (mb_strlen($form_values['test'])<=1000){
|
| 133 |
variable_set('csplitter_test', $form_values['test']);
|
| 134 |
}
|
| 135 |
else{
|
| 136 |
drupal_set_message('Csplitter: Test String is too long. ( >1000 characters )', 'error');
|
| 137 |
}
|
| 138 |
}
|
| 139 |
|
| 140 |
function _csplitter_readkeys($filepath){
|
| 141 |
unset($keys);
|
| 142 |
$handle = fopen($filepath, 'r');
|
| 143 |
if ($handle){
|
| 144 |
while (!feof($handle)) {
|
| 145 |
$buffer = trim(fgets($handle));
|
| 146 |
$keys[$buffer] = 0;
|
| 147 |
}
|
| 148 |
fclose($handle);
|
| 149 |
}
|
| 150 |
return $keys;
|
| 151 |
}
|
| 152 |
|
| 153 |
function _csplitter_splite_1($str){
|
| 154 |
// Search common pattern is ignored
|
| 155 |
// char value is setted by search.module
|
| 156 |
//drupal_set_message("$str|".ord($str[0])." ".ord($str[mb_strlen($str)-1]));
|
| 157 |
if (ord($str[0])==239){// && ord($str[mb_strlen($str)-1]) == 239){
|
| 158 |
return $str;
|
| 159 |
}
|
| 160 |
|
| 161 |
$filepath = drupal_get_path('module', 'csplitter')."/dict.txt";
|
| 162 |
$fdicpath = drupal_get_path('module', 'csplitter')."/filter.txt";
|
| 163 |
$arraypath = drupal_get_path('module', 'csplitter')."/cs_keys.obj";
|
| 164 |
$fobjpath = drupal_get_path('module', 'csplitter')."/filter_keys.obj";
|
| 165 |
|
| 166 |
$cs_keys = array();
|
| 167 |
$cf_keys = array();
|
| 168 |
|
| 169 |
// init dictionary keys and cache array to disk
|
| 170 |
$cache_method = variable_get('csplitter_cache', 'none');
|
| 171 |
switch ($cache_method){
|
| 172 |
case 'none':
|
| 173 |
$cs_keys = _csplitter_readkeys($filepath);
|
| 174 |
$cf_keys = _csplitter_readkeys($fdicpath);
|
| 175 |
break;
|
| 176 |
case 'disk':
|
| 177 |
if (!file_exists($arraypath)){
|
| 178 |
$cs_keys = _csplitter_readkeys($filepath);
|
| 179 |
$serialized = serialize($cs_keys);
|
| 180 |
$handle = fopen($arraypath, 'w');
|
| 181 |
if ($handle){
|
| 182 |
fwrite($handle, $serialized);
|
| 183 |
fclose($handle);
|
| 184 |
}
|
| 185 |
}
|
| 186 |
else{
|
| 187 |
$handle = fopen($arraypath, 'r');
|
| 188 |
if ($handle){
|
| 189 |
$cs_keys = unserialize(fread($handle, filesize($arraypath)));
|
| 190 |
fclose($handle);
|
| 191 |
}
|
| 192 |
}
|
| 193 |
// init dictionary keys and cache array to disk
|
| 194 |
if (!file_exists($fobjpath)){
|
| 195 |
$cf_keys = _csplitter_readkeys($fdicpath);
|
| 196 |
$serialized = serialize($cf_keys);
|
| 197 |
$handle = fopen($fobjpath, 'w');
|
| 198 |
if ($handle){
|
| 199 |
fwrite($handle, $serialized);
|
| 200 |
fclose($handle);
|
| 201 |
}
|
| 202 |
}
|
| 203 |
else{
|
| 204 |
$handle = fopen($fobjpath, 'r');
|
| 205 |
if ($handle){
|
| 206 |
$cf_keys = unserialize(fread($handle, filesize($fobjpath)));
|
| 207 |
fclose($handle);
|
| 208 |
}
|
| 209 |
}
|
| 210 |
break;
|
| 211 |
case 'database':
|
| 212 |
$cache = cache_get('csplitter:cs');
|
| 213 |
if ($cache){
|
| 214 |
$cs_keys = unserialize($cache->data);
|
| 215 |
}
|
| 216 |
else{
|
| 217 |
$cs_keys = _csplitter_readkeys($filepath);
|
| 218 |
cache_set('csplitter:cs', serialize($cs_keys), time()+86400);
|
| 219 |
}
|
| 220 |
$cache = cache_get('csplitter:cf');
|
| 221 |
if ($cache){
|
| 222 |
$cf_keys = unserialize($cache->data);
|
| 223 |
}
|
| 224 |
else{
|
| 225 |
$cf_keys = _csplitter_readkeys($fdicpath);
|
| 226 |
cache_set('csplitter:cf', serialize($cf_keys), time()+86400);
|
| 227 |
}
|
| 228 |
break;
|
| 229 |
}
|
| 230 |
|
| 231 |
//forward max match
|
| 232 |
$outstr = "";
|
| 233 |
//drupal_set_message(mb_strlen($str)."|$str|");
|
| 234 |
for ($i=0; $i<mb_strlen($str);){
|
| 235 |
$found = 1;
|
| 236 |
for ($len = 1; $len< 8 && $i - $len>= 0; $len++){
|
| 237 |
$word = mb_substr($str, $i, $len, "UTF-8");
|
| 238 |
if (ord($word[0])<176){ // not chinese
|
| 239 |
break;
|
| 240 |
}
|
| 241 |
//now process chinese
|
| 242 |
if ($len ==1 ||array_key_exists($word, $cs_keys)){
|
| 243 |
$found = $len;
|
| 244 |
continue;
|
| 245 |
}
|
| 246 |
//else{
|
| 247 |
// $len--;
|
| 248 |
// break;
|
| 249 |
//}
|
| 250 |
}
|
| 251 |
$len = $found;
|
| 252 |
$word = mb_substr($str, $i, $len, "UTF-8");
|
| 253 |
//drupal_set_message(mb_strlen($str).": $i - $found :".$word);
|
| 254 |
|
| 255 |
if ($len >1) // || ($len == 1 && ord($word[0])>=176))
|
| 256 |
$outstr .= " ".$word." ";
|
| 257 |
else{
|
| 258 |
if (ord($word[0])>=176){
|
| 259 |
if (array_key_exists($word, $cf_keys))
|
| 260 |
$outstr .= " ".$word." ";
|
| 261 |
else{
|
| 262 |
$nextchar = mb_substr($str, $i+1, 1, "UTF-8");
|
| 263 |
//Need Test: if (mb_strlen($nextchar) > 0 && ord($nextchar[0])<176){
|
| 264 |
if (ord($nextchar[0])<176){
|
| 265 |
$outstr .= $word. " ";
|
| 266 |
}
|
| 267 |
else{
|
| 268 |
$outstr .= "".$word."";
|
| 269 |
}
|
| 270 |
}
|
| 271 |
}
|
| 272 |
else{
|
| 273 |
$outstr .= $word;
|
| 274 |
}
|
| 275 |
}
|
| 276 |
|
| 277 |
$i += $len;
|
| 278 |
}
|
| 279 |
|
| 280 |
//drupal_set_message($outstr);
|
| 281 |
return $outstr;
|
| 282 |
}
|
| 283 |
|
| 284 |
function _csplitter_splite_2($str){
|
| 285 |
$filepath = drupal_get_path('module', 'csplitter')."/dict.txt";
|
| 286 |
$fdicpath = drupal_get_path('module', 'csplitter')."/filter.txt";
|
| 287 |
$arraypath = drupal_get_path('module', 'csplitter')."/cs_keys.obj";
|
| 288 |
$fobjpath = drupal_get_path('module', 'csplitter')."/filter_keys.obj";
|
| 289 |
|
| 290 |
$cs_keys = array();
|
| 291 |
$cf_keys = array();
|
| 292 |
|
| 293 |
// init dictionary keys and cache array to disk
|
| 294 |
$cache_method = variable_get('csplitter_cache', 'none');
|
| 295 |
switch ($cache_method){
|
| 296 |
case 'none':
|
| 297 |
$cs_keys = _csplitter_readkeys($filepath);
|
| 298 |
$cf_keys = _csplitter_readkeys($fdicpath);
|
| 299 |
break;
|
| 300 |
case 'disk':
|
| 301 |
if (!file_exists($arraypath)){
|
| 302 |
$cs_keys = _csplitter_readkeys($filepath);
|
| 303 |
$serialized = serialize($cs_keys);
|
| 304 |
$handle = fopen($arraypath, 'w');
|
| 305 |
if ($handle){
|
| 306 |
fwrite($handle, $serialized);
|
| 307 |
fclose($handle);
|
| 308 |
}
|
| 309 |
}
|
| 310 |
else{
|
| 311 |
$handle = fopen($arraypath, 'r');
|
| 312 |
if ($handle){
|
| 313 |
$cs_keys = unserialize(fread($handle, filesize($arraypath)));
|
| 314 |
fclose($handle);
|
| 315 |
}
|
| 316 |
}
|
| 317 |
// init dictionary keys and cache array to disk
|
| 318 |
if (!file_exists($fobjpath)){
|
| 319 |
$cf_keys = _csplitter_readkeys($fdicpath);
|
| 320 |
$serialized = serialize($cf_keys);
|
| 321 |
$handle = fopen($fobjpath, 'w');
|
| 322 |
if ($handle){
|
| 323 |
fwrite($handle, $serialized);
|
| 324 |
fclose($handle);
|
| 325 |
}
|
| 326 |
}
|
| 327 |
else{
|
| 328 |
$handle = fopen($fobjpath, 'r');
|
| 329 |
if ($handle){
|
| 330 |
$cf_keys = unserialize(fread($handle, filesize($fobjpath)));
|
| 331 |
fclose($handle);
|
| 332 |
}
|
| 333 |
}
|
| 334 |
break;
|
| 335 |
case 'database':
|
| 336 |
$cache = cache_get('csplitter:cs');
|
| 337 |
if ($cache){
|
| 338 |
$cs_keys = unserialize($cache->data);
|
| 339 |
}
|
| 340 |
else{
|
| 341 |
$cs_keys = _csplitter_readkeys($filepath);
|
| 342 |
cache_set('csplitter:cs', serialize($cs_keys), time()+86400);
|
| 343 |
}
|
| 344 |
$cache = cache_get('csplitter:cf');
|
| 345 |
if ($cache){
|
| 346 |
$cf_keys = unserialize($cache->data);
|
| 347 |
}
|
| 348 |
else{
|
| 349 |
$cf_keys = _csplitter_readkeys($fdicpath);
|
| 350 |
cache_set('csplitter:cf', serialize($cf_keys), time()+86400);
|
| 351 |
}
|
| 352 |
break;
|
| 353 |
}
|
| 354 |
|
| 355 |
//converse max match
|
| 356 |
$outstr = "";
|
| 357 |
for ($i=mb_strlen($str); $i>0;){
|
| 358 |
$found = 1;
|
| 359 |
for ($len = 1; $len< 8 && $i - $len>= 0; $len++){
|
| 360 |
$word = mb_substr($str, $i-$len, $len, "UTF-8");
|
| 361 |
|
| 362 |
//now process chinese
|
| 363 |
if ($len ==1 ||array_key_exists($word, $cs_keys)){
|
| 364 |
$found = $len;
|
| 365 |
continue;
|
| 366 |
}
|
| 367 |
//else{
|
| 368 |
// $len--;
|
| 369 |
// break;
|
| 370 |
//}
|
| 371 |
|
| 372 |
if (ord($word[0])<176){ // not chinese
|
| 373 |
break;
|
| 374 |
}
|
| 375 |
}
|
| 376 |
$len = $found;
|
| 377 |
|
| 378 |
$word = mb_substr($str, $i-$len, $len, "UTF-8");
|
| 379 |
//drupal_set_message("$i - $found :".$word);
|
| 380 |
|
| 381 |
if ($len >1){
|
| 382 |
$outstr = " " .$word. " " . $outstr;
|
| 383 |
}
|
| 384 |
else{
|
| 385 |
$nextchar = mb_substr($str, $i-$len-1, 1, "UTF-8");
|
| 386 |
if (ord($nextchar[0])>=176){
|
| 387 |
if (ord($word[0])<176){
|
| 388 |
$outstr = " " . $word . $outstr;
|
| 389 |
}
|
| 390 |
else if (array_key_exists($word, $cf_keys)){
|
| 391 |
$outstr = " " . $word . " ". $outstr;
|
| 392 |
}
|
| 393 |
else
|
| 394 |
$outstr = "" . $word . $outstr;
|
| 395 |
}
|
| 396 |
else if (ord($outstr[0])>=176)
|
| 397 |
$outstr = $word. " " .$outstr;
|
| 398 |
else {
|
| 399 |
$outstr = $word . $outstr;
|
| 400 |
//drupal_set_message($word);
|
| 401 |
}
|
| 402 |
}
|
| 403 |
|
| 404 |
$i -= $len;
|
| 405 |
}
|
| 406 |
|
| 407 |
return $outstr;
|
| 408 |
}
|
| 409 |
|
| 410 |
function csplitter_search_preprocess($text) {
|
| 411 |
$function = '_csplitter_splite_'. variable_get('csplitter_algorithm', '1');
|
| 412 |
return call_user_func($function, $text);//_csplitter_splite_1($text);
|
| 413 |
}
|
| 414 |
|
| 415 |
?>
|
Parent Directory
| 
Revision Log
| 
Revision Graph