| 1 |
<?php |
| 2 |
|
| 3 |
if (!defined('ABSPATH')) { |
| 4 |
exit; |
| 5 |
} |
| 6 |
|
| 7 |
/** |
| 8 |
* Candidate filtering, scoring, and matching on posts/tags/categories |
| 9 |
* for ABJ_404_Solution_SpellChecker. |
| 10 |
*/ |
| 11 |
trait SpellCheckerTrait_CandidateFiltering { |
| 12 |
|
| 13 |
/** Returns a list of matching posts. |
| 14 |
* @param string $requestedURLRaw |
| 15 |
* @param string $includeCats |
| 16 |
* @param string $includeTags |
| 17 |
* @return array<int, mixed> |
| 18 |
*/ |
| 19 |
function findMatchingPosts(string $requestedURLRaw, string $includeCats = '1', string $includeTags = '1') { |
| 20 |
|
| 21 |
$options = $this->logic->getOptions(); |
| 22 |
// the number of pages to cache is (max suggestions) + (the number of exclude pages). |
| 23 |
// (if either of these numbers increases then we need to clear the spelling cache.) |
| 24 |
$excludePagesCount = 0; |
| 25 |
$excludePagesRaw = isset($options['excludePages[]']) && is_string($options['excludePages[]']) ? $options['excludePages[]'] : ''; |
| 26 |
if (trim($excludePagesRaw) !== '') { |
| 27 |
$jsonResult = json_decode($excludePagesRaw); |
| 28 |
if (!is_array($jsonResult)) { |
| 29 |
$jsonResult = array($jsonResult); |
| 30 |
} |
| 31 |
$excludePagesCount = count($jsonResult); |
| 32 |
} |
| 33 |
$suggestMaxRaw = isset($options['suggest_max']) && is_scalar($options['suggest_max']) ? $options['suggest_max'] : 5; |
| 34 |
$maxCacheCount = absint($suggestMaxRaw) + $excludePagesCount; |
| 35 |
|
| 36 |
$requestedURLSpaces = $this->f->str_replace($this->separatingCharacters, " ", $requestedURLRaw); |
| 37 |
$requestedURLCleaned = $this->getLastURLPart($requestedURLSpaces); |
| 38 |
$fullURLspacesCleaned = $this->f->str_replace('/', " ", $requestedURLSpaces); |
| 39 |
// if there is no extra stuff in the path then we ignore this to save time. |
| 40 |
if ($fullURLspacesCleaned == $requestedURLCleaned) { |
| 41 |
$fullURLspacesCleaned = ''; |
| 42 |
} |
| 43 |
|
| 44 |
// prepare to get some posts. |
| 45 |
$this->initializePublishedPostsProvider(); |
| 46 |
|
| 47 |
$rowType = 'pages'; |
| 48 |
$permalinks = array(); |
| 49 |
// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - match on posts |
| 50 |
$permalinks = $this->matchOnPosts($permalinks, $requestedURLRaw, $requestedURLCleaned, |
| 51 |
$fullURLspacesCleaned, $rowType); |
| 52 |
|
| 53 |
// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - match on tags |
| 54 |
// search for a similar tag. |
| 55 |
if ($includeTags == "1") { |
| 56 |
$permalinks = $this->matchOnTags($permalinks, $requestedURLCleaned, $fullURLspacesCleaned, 'tags'); |
| 57 |
} |
| 58 |
|
| 59 |
// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - match on categories |
| 60 |
// search for a similar category. |
| 61 |
if ($includeCats == "1") { |
| 62 |
$permalinks = $this->matchOnCats($permalinks, $requestedURLCleaned, $fullURLspacesCleaned, 'categories'); |
| 63 |
} |
| 64 |
|
| 65 |
// remove excluded pages |
| 66 |
$permalinks = $this->removeExcludedPages($options, $permalinks); |
| 67 |
|
| 68 |
// This is sorted so that the link with the highest score will be first when iterating through. |
| 69 |
arsort($permalinks); |
| 70 |
|
| 71 |
$permalinks = $this->removeExcludedPagesWithRegex($options, $permalinks, $maxCacheCount); |
| 72 |
|
| 73 |
// only keep what we need. store them for later if necessary. |
| 74 |
$permalinks = array_splice($permalinks, 0, $maxCacheCount); |
| 75 |
|
| 76 |
$returnValue = array($permalinks,$rowType); |
| 77 |
$this->dao->storeSpellingPermalinksToCache($requestedURLRaw, $returnValue); |
| 78 |
$ctx = abj_service('request_context'); |
| 79 |
$ctx->permalinks_found = (string)json_encode($returnValue); |
| 80 |
$ctx->permalinks_kept = (string)json_encode($permalinks); |
| 81 |
|
| 82 |
return $returnValue; |
| 83 |
} |
| 84 |
|
| 85 |
/** |
| 86 |
* @param array<string, mixed> $options |
| 87 |
* @param array<string, string> $permalinks |
| 88 |
* @return array<string, string> |
| 89 |
*/ |
| 90 |
function removeExcludedPages(array $options, array $permalinks): array { |
| 91 |
$excludePagesJsonRaw = isset($options['excludePages[]']) ? $options['excludePages[]'] : ''; |
| 92 |
$excludePagesJson = is_string($excludePagesJsonRaw) ? $excludePagesJsonRaw : ''; |
| 93 |
if (trim($excludePagesJson) == '' && $this->custom404PageID == null) { |
| 94 |
return $permalinks; |
| 95 |
} |
| 96 |
|
| 97 |
// look at every ID to exclude. |
| 98 |
$excludePages = json_decode($excludePagesJson); |
| 99 |
if (!is_array($excludePages)) { |
| 100 |
$excludePages = array($excludePages); |
| 101 |
} |
| 102 |
|
| 103 |
// don't include the user specified 404 page in the spelling results.. |
| 104 |
if ($this->custom404PageID != null) { |
| 105 |
array_push($excludePages, $this->custom404PageID); |
| 106 |
} |
| 107 |
|
| 108 |
for ($i = 0; $i < count($excludePages); $i++) { |
| 109 |
$excludePage = $excludePages[$i]; |
| 110 |
if ($excludePage == null || trim($excludePage) == '') { |
| 111 |
continue; |
| 112 |
} |
| 113 |
// remove it from the results list. |
| 114 |
// Entry format matches permalink key format: "id|type" (e.g. "42|1"). |
| 115 |
unset($permalinks[(string)$excludePage]); |
| 116 |
} |
| 117 |
|
| 118 |
return $permalinks; |
| 119 |
} |
| 120 |
|
| 121 |
/** |
| 122 |
* Removes permalink suggestions if their URL path matches exclusion regex patterns. |
| 123 |
* |
| 124 |
* @param array<string, mixed> $options Plugin options containing 'suggest_regex_exclusions_usable'. |
| 125 |
* @param array<string, string> $permalinks An array where keys are "ID|TYPE_CONSTANT" and values are scores. |
| 126 |
* @param int $maxCacheCount |
| 127 |
* @return array<string, string> The filtered $permalinks array. |
| 128 |
*/ |
| 129 |
function removeExcludedPagesWithRegex(array $options, array $permalinks, int $maxCacheCount): array { |
| 130 |
// Check if usable regex patterns exist and are in an array format |
| 131 |
if (!isset($options['suggest_regex_exclusions_usable']) || |
| 132 |
!is_array($options['suggest_regex_exclusions_usable']) || |
| 133 |
empty($options['suggest_regex_exclusions_usable'])) { |
| 134 |
// No patterns to apply, return original list |
| 135 |
return $permalinks; |
| 136 |
} |
| 137 |
|
| 138 |
$suggestionsKeptSoFar = 0; |
| 139 |
$regexExclusions = $options['suggest_regex_exclusions_usable']; |
| 140 |
|
| 141 |
// Iterate through each permalink entry using keys directly |
| 142 |
// Modifying array while iterating requires careful handling, using keys is safer. |
| 143 |
$keys_to_check = array_keys($permalinks); |
| 144 |
|
| 145 |
foreach ($keys_to_check as $key) { |
| 146 |
// Skip if the key somehow got removed in a previous iteration (shouldn't happen here) |
| 147 |
if (!array_key_exists($key, $permalinks)) { |
| 148 |
continue; |
| 149 |
} |
| 150 |
|
| 151 |
// Split the key into ID and Type Constant |
| 152 |
$keyParts = explode('|', $key); |
| 153 |
if (count($keyParts) !== 2 || !is_numeric($keyParts[0])) { |
| 154 |
$this->logger->debugMessage("Skipping invalid key format in removeExcludedPagesWithRegex: " . $key); |
| 155 |
continue; // Skip invalid keys |
| 156 |
} |
| 157 |
|
| 158 |
$id = (int)$keyParts[0]; |
| 159 |
$typeConstant = $keyParts[1]; // Keep as string/int as needed by mapTypeConstantToString |
| 160 |
|
| 161 |
// Map the type constant (e.g., '1') to the string type ('pages', 'tags', etc.) |
| 162 |
$rowTypeString = $this->mapTypeConstantToString($typeConstant); |
| 163 |
if ($rowTypeString === null) { |
| 164 |
$this->logger->debugMessage("Skipping unknown type constant in removeExcludedPagesWithRegex: " . $typeConstant . " for key: " . $key); |
| 165 |
continue; // Skip unknown types |
| 166 |
} |
| 167 |
|
| 168 |
// Get the full URL using the class's method (handles cache) |
| 169 |
$urlOfPage = $this->getPermalink($id, $rowTypeString); |
| 170 |
if ($urlOfPage === null || trim($urlOfPage) === '') { |
| 171 |
$this->logger->debugMessage("Skipping null/empty URL for key in removeExcludedPagesWithRegex: " . $key); |
| 172 |
continue; // Skip if URL couldn't be retrieved |
| 173 |
} |
| 174 |
|
| 175 |
// Parse the URL and get the path, remove home directory if needed (consistency) |
| 176 |
$urlParts = parse_url($urlOfPage); |
| 177 |
if (!is_array($urlParts) || !isset($urlParts['path'])) { |
| 178 |
$this->logger->debugMessage("Skipping URL that failed parse_url for key in removeExcludedPagesWithRegex: " . $key . ", URL: " . esc_url($urlOfPage)); |
| 179 |
continue; // Skip invalid URLs |
| 180 |
} |
| 181 |
$pathOnly = $this->logic->removeHomeDirectory($urlParts['path']); |
| 182 |
// Ensure path starts with / for consistency if it's not empty |
| 183 |
if ( $pathOnly !== '' && substr($pathOnly, 0, 1) !== '/' ) { |
| 184 |
$pathOnly = '/' . $pathOnly; |
| 185 |
} |
| 186 |
// Handle case where path might be empty (e.g., homepage) which results in '/' |
| 187 |
if ( $pathOnly === '' ) { |
| 188 |
$pathOnly = '/'; |
| 189 |
} |
| 190 |
|
| 191 |
$stringToMatch = $pathOnly; // The string we will match the regex against |
| 192 |
|
| 193 |
$kept = true; |
| 194 |
// Check against each exclusion pattern |
| 195 |
foreach ($regexExclusions as $pattern) { |
| 196 |
// Remove slashes like in the example provided for folders_files_ignore |
| 197 |
$patternToExcludeNoSlashes = stripslashes($pattern); |
| 198 |
$matches = array(); // Variable for the match results |
| 199 |
|
| 200 |
// Use the class's regexMatch function |
| 201 |
if ($this->f->regexMatch($patternToExcludeNoSlashes, $stringToMatch, $matches)) { |
| 202 |
// Pattern matched, remove this permalink from the list |
| 203 |
unset($permalinks[$key]); |
| 204 |
$this->logger->debugMessage("Regex excluded suggestion. Key: " . $key . |
| 205 |
", Path: '" . esc_html($stringToMatch) . "', Pattern: '" . esc_html($patternToExcludeNoSlashes) . "'"); |
| 206 |
$kept = false; |
| 207 |
// Break the inner loop (patterns), move to the next permalink key |
| 208 |
break; |
| 209 |
} |
| 210 |
} |
| 211 |
|
| 212 |
// track how many suggestions we actually need and stop filtering after we reach that count |
| 213 |
if ($kept) { |
| 214 |
$suggestionsKeptSoFar++; |
| 215 |
} |
| 216 |
if ($suggestionsKeptSoFar >= $maxCacheCount) { |
| 217 |
break; |
| 218 |
} |
| 219 |
} |
| 220 |
|
| 221 |
return $permalinks; |
| 222 |
} |
| 223 |
|
| 224 |
/** |
| 225 |
* Maps internal type constants to string identifiers used by getPermalink. |
| 226 |
* NOTE: Requires ABJ404_TYPE_* constants to be defined correctly. |
| 227 |
* |
| 228 |
* @param mixed $typeConstant The type constant (e.g., ABJ404_TYPE_POST). |
| 229 |
* @return string|null The string identifier ('pages', 'tags', 'categories') or null if not found. |
| 230 |
*/ |
| 231 |
private function mapTypeConstantToString($typeConstant) { |
| 232 |
// Define these constants if they are not globally available or use their actual values |
| 233 |
if (!defined('ABJ404_TYPE_POST')) define('ABJ404_TYPE_POST', 1); |
| 234 |
if (!defined('ABJ404_TYPE_CAT')) define('ABJ404_TYPE_CAT', 2); |
| 235 |
if (!defined('ABJ404_TYPE_TAG')) define('ABJ404_TYPE_TAG', 3); |
| 236 |
// Add other types like ABJ404_TYPE_IMAGE if needed |
| 237 |
|
| 238 |
$typeConstantStr = is_scalar($typeConstant) ? (string)$typeConstant : ''; |
| 239 |
switch ($typeConstantStr) { // Cast to string for reliable comparison if needed |
| 240 |
case ABJ404_TYPE_POST: |
| 241 |
return 'pages'; // Based on getPermalink implementation which uses 'pages' for posts |
| 242 |
case ABJ404_TYPE_TAG: |
| 243 |
return 'tags'; |
| 244 |
case ABJ404_TYPE_CAT: |
| 245 |
return 'categories'; |
| 246 |
// Add 'image' case if ABJ404_TYPE_IMAGE exists and is used in $permalinks keys |
| 247 |
// case ABJ404_TYPE_IMAGE: |
| 248 |
// return 'image'; |
| 249 |
default: |
| 250 |
// Log or handle unknown type |
| 251 |
return null; |
| 252 |
} |
| 253 |
} |
| 254 |
|
| 255 |
/** |
| 256 |
* @param array<string, string> $permalinks |
| 257 |
* @param string $requestedURLCleaned |
| 258 |
* @param string $fullURLspacesCleaned |
| 259 |
* @param string $rowType |
| 260 |
* @return array<string, string> |
| 261 |
*/ |
| 262 |
function matchOnCats(array $permalinks, string $requestedURLCleaned, string $fullURLspacesCleaned, string $rowType): array { |
| 263 |
|
| 264 |
$rows = $this->dao->getPublishedCategories(); |
| 265 |
$rows = $this->getOnlyIDandTermID($rows); |
| 266 |
|
| 267 |
// pre-filter some pages based on the min and max possible levenshtein distances. |
| 268 |
$likelyMatchIDsAndPermalinks = $this->getLikelyMatchIDs($requestedURLCleaned, $fullURLspacesCleaned, 'categories', $rows); |
| 269 |
$likelyMatchIDs = array_keys($likelyMatchIDsAndPermalinks); |
| 270 |
|
| 271 |
// Early termination optimization |
| 272 |
$options = $this->logic->getOptions(); |
| 273 |
$suggestMaxRaw = isset($options['suggest_max']) && is_scalar($options['suggest_max']) ? $options['suggest_max'] : 5; |
| 274 |
$suggestMax = absint($suggestMaxRaw); |
| 275 |
$topKScores = new SplMinHeap(); |
| 276 |
$requestedURLCleanedLength = $this->f->strlen($requestedURLCleaned); |
| 277 |
|
| 278 |
// access the array directly instead of using a foreach loop so we can remove items |
| 279 |
// from the end of the array in the middle of the loop. |
| 280 |
foreach ($likelyMatchIDs as $id) { |
| 281 |
// use the levenshtein distance formula here. |
| 282 |
$the_permalink = $this->getPermalink((int)$id, 'categories'); |
| 283 |
$urlParts = parse_url(is_string($the_permalink) ? $the_permalink : ''); |
| 284 |
if (!is_array($urlParts) || !isset($urlParts['path'])) { |
| 285 |
continue; |
| 286 |
} |
| 287 |
$pathOnly = $this->logic->removeHomeDirectory($urlParts['path']); |
| 288 |
$scoreBasis = $this->f->strlen($pathOnly); |
| 289 |
if ($scoreBasis == 0) { |
| 290 |
continue; |
| 291 |
} |
| 292 |
|
| 293 |
// EARLY TERMINATION: Check if this candidate can possibly beat our worst current match |
| 294 |
if ($topKScores->count() >= $suggestMax) { |
| 295 |
$worstAcceptableScore = $topKScores->top(); |
| 296 |
|
| 297 |
// OPTIMIZATION 3: Levenshtein distance threshold pruning |
| 298 |
$maxAllowedLevenshtein = ((100 - $worstAcceptableScore) * $scoreBasis) / 100; |
| 299 |
$pathOnlyLength = $this->f->strlen($pathOnly); |
| 300 |
$minPossibleDistance = abs($requestedURLCleanedLength - $pathOnlyLength); |
| 301 |
|
| 302 |
if ($minPossibleDistance > $maxAllowedLevenshtein) { |
| 303 |
continue; // Can't possibly beat worst score in heap |
| 304 |
} |
| 305 |
} |
| 306 |
|
| 307 |
$levscore = $this->customLevenshtein($requestedURLCleaned, $pathOnly); |
| 308 |
|
| 309 |
// OPTIMIZATION 2: Lazy evaluation of fullURLspacesCleaned |
| 310 |
if ($fullURLspacesCleaned != '') { |
| 311 |
$tentativeScore = 100 - (($levscore / $scoreBasis) * 100); |
| 312 |
if ($tentativeScore < 95) { |
| 313 |
$pathOnlySpaces = $this->f->str_replace($this->separatingCharacters, " ", $pathOnly); |
| 314 |
$pathOnlySpaces = trim($this->f->str_replace('/', " ", $pathOnlySpaces)); |
| 315 |
$levscore = min($levscore, $this->customLevenshtein($fullURLspacesCleaned, $pathOnlySpaces)); |
| 316 |
} |
| 317 |
} |
| 318 |
|
| 319 |
$onlyLastPart = $this->getLastURLPart($pathOnly); |
| 320 |
if ($onlyLastPart != '' && $onlyLastPart != $pathOnly) { |
| 321 |
$levscore = min($levscore, $this->customLevenshtein($requestedURLCleaned, $onlyLastPart)); |
| 322 |
} |
| 323 |
|
| 324 |
$score = 100 - (($levscore / $scoreBasis) * 100); |
| 325 |
$permalinks[$id . "|" . ABJ404_TYPE_CAT] = number_format($score, 4, '.', ''); |
| 326 |
|
| 327 |
// Update top-K heap |
| 328 |
$topKScores->insert($score); |
| 329 |
if ($topKScores->count() > $suggestMax) { |
| 330 |
$topKScores->extract(); |
| 331 |
} |
| 332 |
} |
| 333 |
|
| 334 |
return $permalinks; |
| 335 |
} |
| 336 |
|
| 337 |
/** |
| 338 |
* @param array<string, string> $permalinks |
| 339 |
* @param string $requestedURLCleaned |
| 340 |
* @param string $fullURLspacesCleaned |
| 341 |
* @param string $rowType |
| 342 |
* @return array<string, string> |
| 343 |
*/ |
| 344 |
function matchOnTags(array $permalinks, string $requestedURLCleaned, string $fullURLspacesCleaned, string $rowType): array { |
| 345 |
|
| 346 |
$rows = $this->dao->getPublishedTags(); |
| 347 |
$rows = $this->getOnlyIDandTermID($rows); |
| 348 |
|
| 349 |
// pre-filter some pages based on the min and max possible levenshtein distances. |
| 350 |
$likelyMatchIDsAndPermalinks = $this->getLikelyMatchIDs($requestedURLCleaned, $fullURLspacesCleaned, 'tags', $rows); |
| 351 |
$likelyMatchIDs = array_keys($likelyMatchIDsAndPermalinks); |
| 352 |
|
| 353 |
// Early termination optimization |
| 354 |
$options = $this->logic->getOptions(); |
| 355 |
$suggestMaxRawT = isset($options['suggest_max']) && is_scalar($options['suggest_max']) ? $options['suggest_max'] : 5; |
| 356 |
$suggestMax = absint($suggestMaxRawT); |
| 357 |
$topKScores = new SplMinHeap(); |
| 358 |
$requestedURLCleanedLength = $this->f->strlen($requestedURLCleaned); |
| 359 |
|
| 360 |
// access the array directly instead of using a foreach loop so we can remove items |
| 361 |
// from the end of the array in the middle of the loop. |
| 362 |
foreach ($likelyMatchIDs as $id) { |
| 363 |
// use the levenshtein distance formula here. |
| 364 |
$the_permalink = $this->getPermalink((int)$id, 'tags'); |
| 365 |
$urlParts = parse_url(is_string($the_permalink) ? $the_permalink : ''); |
| 366 |
if (!is_array($urlParts) || !isset($urlParts['path'])) { |
| 367 |
continue; |
| 368 |
} |
| 369 |
$pathOnly = $this->logic->removeHomeDirectory($urlParts['path']); |
| 370 |
$scoreBasis = $this->f->strlen($pathOnly); |
| 371 |
if ($scoreBasis == 0) { |
| 372 |
continue; |
| 373 |
} |
| 374 |
|
| 375 |
// EARLY TERMINATION: Check if this candidate can possibly beat our worst current match |
| 376 |
if ($topKScores->count() >= $suggestMax) { |
| 377 |
$worstAcceptableScore = $topKScores->top(); |
| 378 |
|
| 379 |
// OPTIMIZATION 3: Levenshtein distance threshold pruning |
| 380 |
$maxAllowedLevenshtein = ((100 - $worstAcceptableScore) * $scoreBasis) / 100; |
| 381 |
$pathOnlyLength = $this->f->strlen($pathOnly); |
| 382 |
$minPossibleDistance = abs($requestedURLCleanedLength - $pathOnlyLength); |
| 383 |
|
| 384 |
if ($minPossibleDistance > $maxAllowedLevenshtein) { |
| 385 |
continue; // Can't possibly beat worst score in heap |
| 386 |
} |
| 387 |
} |
| 388 |
|
| 389 |
$levscore = $this->customLevenshtein($requestedURLCleaned, $pathOnly); |
| 390 |
|
| 391 |
// OPTIMIZATION 2: Lazy evaluation of fullURLspacesCleaned |
| 392 |
if ($fullURLspacesCleaned != '') { |
| 393 |
$tentativeScore = 100 - (($levscore / $scoreBasis) * 100); |
| 394 |
if ($tentativeScore < 95) { |
| 395 |
$pathOnlySpaces = $this->f->str_replace($this->separatingCharacters, " ", $pathOnly); |
| 396 |
$pathOnlySpaces = trim($this->f->str_replace('/', " ", $pathOnlySpaces)); |
| 397 |
$levscore = min($levscore, $this->customLevenshtein($fullURLspacesCleaned, $pathOnlySpaces)); |
| 398 |
} |
| 399 |
} |
| 400 |
$score = 100 - (($levscore / $scoreBasis) * 100); |
| 401 |
$permalinks[$id . "|" . ABJ404_TYPE_TAG] = number_format($score, 4, '.', ''); |
| 402 |
|
| 403 |
// Update top-K heap |
| 404 |
$topKScores->insert($score); |
| 405 |
if ($topKScores->count() > $suggestMax) { |
| 406 |
$topKScores->extract(); |
| 407 |
} |
| 408 |
} |
| 409 |
|
| 410 |
return $permalinks; |
| 411 |
} |
| 412 |
|
| 413 |
/** |
| 414 |
* @param array<string, string> $permalinks |
| 415 |
* @param string $requestedURLRaw |
| 416 |
* @param string $requestedURLCleaned |
| 417 |
* @param string $fullURLspacesCleaned |
| 418 |
* @param string $rowType |
| 419 |
* @return array<string, string> |
| 420 |
*/ |
| 421 |
function matchOnPosts(array $permalinks, string $requestedURLRaw, string $requestedURLCleaned, string $fullURLspacesCleaned, string $rowType): array { |
| 422 |
|
| 423 |
// pre-filter some pages based on the min and max possible levenshtein distances. |
| 424 |
$likelyMatchIDsAndPermalinks = $this->getLikelyMatchIDs($requestedURLCleaned, $fullURLspacesCleaned, $rowType); |
| 425 |
$likelyMatchIDs = array_keys($likelyMatchIDsAndPermalinks); |
| 426 |
|
| 427 |
$this->logger->debugMessage("Found " . count($likelyMatchIDs) . " likely match IDs."); |
| 428 |
|
| 429 |
// Early termination optimization: maintain a min-heap of top-K scores |
| 430 |
// Once we have K matches, we can skip candidates that can't beat the worst in heap |
| 431 |
$options = $this->logic->getOptions(); |
| 432 |
$suggestMaxRawP = isset($options['suggest_max']) && is_scalar($options['suggest_max']) ? $options['suggest_max'] : 5; |
| 433 |
$suggestMax = absint($suggestMaxRawP); |
| 434 |
$topKScores = new SplMinHeap(); // Min-heap: smallest score at top |
| 435 |
$requestedURLCleanedLength = $this->f->strlen($requestedURLCleaned); |
| 436 |
|
| 437 |
// Process candidates in order of best match first (smallest minDist first) |
| 438 |
// This is critical for early termination: filling the heap with good scores early |
| 439 |
// allows us to skip more candidates later |
| 440 |
while (count($likelyMatchIDs) > 0) { |
| 441 |
$id = array_shift($likelyMatchIDs); // Take from beginning (best matches first) |
| 442 |
|
| 443 |
// use the levenshtein distance formula here. |
| 444 |
$the_permalink = $likelyMatchIDsAndPermalinks[$id]; |
| 445 |
$thePermalinkStr = is_string($the_permalink) ? $the_permalink : ''; |
| 446 |
$urlParts = parse_url($thePermalinkStr); |
| 447 |
if (!is_array($urlParts) || !isset($urlParts['path'])) { |
| 448 |
continue; |
| 449 |
} |
| 450 |
$existingPageURL = $this->logic->removeHomeDirectory($urlParts['path']); |
| 451 |
$existingPageURLSpaces = $this->f->str_replace($this->separatingCharacters, " ", $existingPageURL); |
| 452 |
|
| 453 |
$existingPageURLCleaned = $this->getLastURLPart($existingPageURLSpaces); |
| 454 |
$scoreBasis = $this->f->strlen($existingPageURLCleaned) * 3; |
| 455 |
if ($scoreBasis == 0) { |
| 456 |
continue; |
| 457 |
} |
| 458 |
|
| 459 |
// EARLY TERMINATION: Check if this candidate can possibly beat our worst current match |
| 460 |
if ($topKScores->count() >= $suggestMax) { |
| 461 |
$worstAcceptableScore = $topKScores->top(); |
| 462 |
|
| 463 |
// OPTIMIZATION 3: Levenshtein distance threshold pruning |
| 464 |
// Calculate maximum Levenshtein distance that could still beat worstAcceptableScore |
| 465 |
// Formula: score = 100 - ((lev / scoreBasis) * 100) |
| 466 |
// Solving for lev: lev = (100 - score) * scoreBasis / 100 |
| 467 |
$maxAllowedLevenshtein = ((100 - $worstAcceptableScore) * $scoreBasis) / 100; |
| 468 |
|
| 469 |
// Calculate minimum possible distance based on length difference |
| 470 |
$existingURLCleanedLength = $this->f->strlen($existingPageURLCleaned); |
| 471 |
$minPossibleDistance = abs($requestedURLCleanedLength - $existingURLCleanedLength); |
| 472 |
|
| 473 |
// If minimum possible distance already exceeds threshold, skip |
| 474 |
if ($minPossibleDistance > $maxAllowedLevenshtein) { |
| 475 |
continue; // Can't possibly beat worst score in heap |
| 476 |
} |
| 477 |
} |
| 478 |
|
| 479 |
$levscore = $this->customLevenshtein($requestedURLCleaned, $existingPageURLCleaned); |
| 480 |
|
| 481 |
// OPTIMIZATION 2: Lazy evaluation of fullURLspacesCleaned (10-20% reduction) |
| 482 |
// Only try the second comparison if the first score isn't already excellent (>95) |
| 483 |
if ($fullURLspacesCleaned != '') { |
| 484 |
$tentativeScore = 100 - (($levscore / $scoreBasis) * 100); |
| 485 |
if ($tentativeScore < 95) { |
| 486 |
$levscore = min($levscore, $this->customLevenshtein($fullURLspacesCleaned, $existingPageURLCleaned)); |
| 487 |
} |
| 488 |
} |
| 489 |
|
| 490 |
if ($rowType == 'image') { |
| 491 |
// strip the image size from the file name and try again. |
| 492 |
// the image size is at the end of the file in the format of -640x480 |
| 493 |
$strippedImageName = $this->f->regexReplace('(.+)([-]\d{1,5}[x]\d{1,5})([.].+)', |
| 494 |
'\\1\\3', $requestedURLRaw); |
| 495 |
|
| 496 |
if (($strippedImageName != null) && ($strippedImageName != $requestedURLRaw)) { |
| 497 |
$strippedImageName = $this->f->str_replace($this->separatingCharactersForImages, " ", $strippedImageName); |
| 498 |
$levscore = min($levscore, $this->customLevenshtein($strippedImageName, $existingPageURL)); |
| 499 |
|
| 500 |
$strippedImageName = $this->getLastURLPart($strippedImageName); |
| 501 |
$levscore = min($levscore, $this->customLevenshtein($strippedImageName, $existingPageURLCleaned)); |
| 502 |
} |
| 503 |
} |
| 504 |
$score = 100 - (($levscore / $scoreBasis) * 100); |
| 505 |
$permalinks[$id . "|" . ABJ404_TYPE_POST] = number_format($score, 4, '.', ''); |
| 506 |
|
| 507 |
// Update top-K heap with this score |
| 508 |
$topKScores->insert($score); |
| 509 |
// Keep heap size at most suggestMax (remove worst if exceeded) |
| 510 |
if ($topKScores->count() > $suggestMax) { |
| 511 |
$topKScores->extract(); // Remove the smallest (worst) score |
| 512 |
} |
| 513 |
} |
| 514 |
|
| 515 |
return $permalinks; |
| 516 |
} |
| 517 |
|
| 518 |
} |
| 519 |
|
