| 1 |
<?php |
| 2 |
|
| 3 |
|
| 4 |
if (!defined('ABSPATH')) { |
| 5 |
exit; |
| 6 |
} |
| 7 |
|
| 8 |
require_once __DIR__ . '/SpellCheckerTrait_PostListeners.php'; |
| 9 |
require_once __DIR__ . '/SpellCheckerTrait_URLMatching.php'; |
| 10 |
require_once __DIR__ . '/SpellCheckerTrait_CandidateFiltering.php'; |
| 11 |
require_once __DIR__ . '/SpellCheckerTrait_LevenshteinEngine.php'; |
| 12 |
|
| 13 |
/* Finds similar pages. |
| 14 |
* Finds search suggestions. */ |
| 15 |
|
| 16 |
class ABJ_404_Solution_SpellChecker { |
| 17 |
|
| 18 |
use SpellCheckerTrait_PostListeners, |
| 19 |
SpellCheckerTrait_URLMatching, |
| 20 |
SpellCheckerTrait_CandidateFiltering, |
| 21 |
SpellCheckerTrait_LevenshteinEngine; |
| 22 |
|
| 23 |
/** @var array<int, string> */ |
| 24 |
private array $separatingCharacters = array("-","_",".","~",'%20'); |
| 25 |
|
| 26 |
/** Same as above except without the period (.) because of the extension in the file name. |
| 27 |
* @var array<int, string> */ |
| 28 |
private array $separatingCharactersForImages = array("-","_","~",'%20'); |
| 29 |
|
| 30 |
private ?ABJ_404_Solution_PublishedPostsProvider $publishedPostsProvider = null; |
| 31 |
|
| 32 |
const MAX_DIST = 2083; |
| 33 |
|
| 34 |
/** Upper bound for the length-based distance buckets used to pre-filter candidates. */ |
| 35 |
const MAX_LIKELY_DISTANCE = 300; |
| 36 |
|
| 37 |
/** Similarity threshold for N-gram prefiltering (lower = more candidates, slower but safer). */ |
| 38 |
const NGRAM_PREFILTER_THRESHOLD = 0.3; |
| 39 |
|
| 40 |
/** Maximum candidates to retrieve during N-gram prefiltering. */ |
| 41 |
const NGRAM_PREFILTER_MAX_CANDIDATES = 500; |
| 42 |
|
| 43 |
/** Minimum N-gram cache entries required to enable prefiltering. |
| 44 |
* Small sites don't need prefiltering; this also prevents use during partial cache builds. */ |
| 45 |
const NGRAM_MIN_CACHE_ENTRIES = 50; |
| 46 |
|
| 47 |
/** Similarity threshold for secondary N-gram filtering (higher = stricter, fewer candidates). |
| 48 |
* More conservative than prefilter since we're refining an already-filtered list. */ |
| 49 |
const NGRAM_SECONDARY_THRESHOLD = 0.4; |
| 50 |
|
| 51 |
/** Maximum candidates for secondary N-gram filtering. */ |
| 52 |
const NGRAM_SECONDARY_MAX_CANDIDATES = 100; |
| 53 |
|
| 54 |
/** Minimum cache coverage ratio (ngram entries / permalink entries) to trust prefiltering. |
| 55 |
* 0.8 = require at least 80% of permalink cache entries to be in N-gram cache. */ |
| 56 |
const NGRAM_MIN_COVERAGE_RATIO = 0.8; |
| 57 |
|
| 58 |
/** Minimum candidate count to trigger secondary N-gram filtering. |
| 59 |
* Below this threshold, Levenshtein on all candidates is fast enough. */ |
| 60 |
const NGRAM_SECONDARY_MIN_CANDIDATES = 50; |
| 61 |
|
| 62 |
private static ?self $instance = null; |
| 63 |
|
| 64 |
// Performance counters (for testing efficiency - disabled by default) |
| 65 |
private bool $enablePerformanceCounters = false; |
| 66 |
|
| 67 |
// When true, skip the N-gram gate 4 early return so the full Levenshtein |
| 68 |
// scan runs. The async page-suggestions worker sets this because the |
| 69 |
// 5-second scan is acceptable in a background process. |
| 70 |
private bool $skipNgramGate4 = false; |
| 71 |
private int $levenshteinCallCount = 0; |
| 72 |
private int $totalPagesConsidered = 0; |
| 73 |
|
| 74 |
/** @var string|int|null */ |
| 75 |
private $custom404PageID = null; |
| 76 |
|
| 77 |
/** Prepared regex pattern cache for the current request lifecycle. |
| 78 |
* @var array<string, string> */ |
| 79 |
private array $preparedRegexPatternCache = array(); |
| 80 |
|
| 81 |
/** @var ABJ_404_Solution_Functions */ |
| 82 |
private $f; |
| 83 |
|
| 84 |
/** @var ABJ_404_Solution_PluginLogic */ |
| 85 |
private $logic; |
| 86 |
|
| 87 |
/** @var ABJ_404_Solution_DataAccess */ |
| 88 |
private $dao; |
| 89 |
|
| 90 |
/** @var ABJ_404_Solution_Logging */ |
| 91 |
private $logger; |
| 92 |
|
| 93 |
/** @var ABJ_404_Solution_PermalinkCache */ |
| 94 |
private $permalinkCache; |
| 95 |
|
| 96 |
/** @var ABJ_404_Solution_NGramFilter */ |
| 97 |
private $ngramFilter; |
| 98 |
|
| 99 |
/** |
| 100 |
* Constructor with dependency injection. |
| 101 |
* Dependencies are now explicit and visible. |
| 102 |
* |
| 103 |
* @param ABJ_404_Solution_Functions|null $functions String manipulation utilities |
| 104 |
* @param ABJ_404_Solution_PluginLogic|null $pluginLogic Business logic service |
| 105 |
* @param ABJ_404_Solution_DataAccess|null $dataAccess Data access layer |
| 106 |
* @param ABJ_404_Solution_Logging|null $logging Logging service |
| 107 |
* @param ABJ_404_Solution_PermalinkCache|null $permalinkCache Permalink caching service |
| 108 |
* @param ABJ_404_Solution_NGramFilter|null $ngramFilter N-gram filter for optimization |
| 109 |
*/ |
| 110 |
public function __construct($functions = null, $pluginLogic = null, $dataAccess = null, $logging = null, $permalinkCache = null, $ngramFilter = null) { |
| 111 |
// Use injected dependencies or fall back to getInstance() for backward compatibility |
| 112 |
$this->f = $functions !== null ? $functions : abj_service('functions'); |
| 113 |
$this->logic = $pluginLogic !== null ? $pluginLogic : abj_service('plugin_logic'); |
| 114 |
$this->dao = $dataAccess !== null ? $dataAccess : abj_service('data_access'); |
| 115 |
$this->logger = $logging !== null ? $logging : abj_service('logging'); |
| 116 |
$this->permalinkCache = $permalinkCache !== null ? $permalinkCache : abj_service('permalink_cache'); |
| 117 |
$this->ngramFilter = $ngramFilter !== null ? $ngramFilter : abj_service('ngram_filter'); |
| 118 |
|
| 119 |
// Set the custom 404 page id if there is one |
| 120 |
$options = $this->logic->getOptions(); |
| 121 |
$custom404PageIDRaw = |
| 122 |
(is_array($options) && isset($options['dest404page']) ? |
| 123 |
$options['dest404page'] : null); |
| 124 |
$custom404PageID = is_string($custom404PageIDRaw) ? $custom404PageIDRaw : (is_int($custom404PageIDRaw) ? (string)$custom404PageIDRaw : null); |
| 125 |
if ($this->logic->thereIsAUserSpecified404Page($custom404PageID)) { |
| 126 |
$this->custom404PageID = $custom404PageID; |
| 127 |
} |
| 128 |
} |
| 129 |
|
| 130 |
public static function getInstance(): self { |
| 131 |
if (self::$instance !== null) { |
| 132 |
return self::$instance; |
| 133 |
} |
| 134 |
|
| 135 |
// If the DI container is initialized, prefer it. |
| 136 |
if (class_exists('ABJ_404_Solution_ServiceContainer')) { |
| 137 |
$resolved = ABJ_404_Solution_ServiceContainer::safeGet('spell_checker'); |
| 138 |
if ($resolved instanceof self) { |
| 139 |
self::$instance = $resolved; |
| 140 |
return self::$instance; |
| 141 |
} |
| 142 |
} |
| 143 |
|
| 144 |
self::$instance = new ABJ_404_Solution_SpellChecker(); |
| 145 |
|
| 146 |
return self::$instance; |
| 147 |
} |
| 148 |
|
| 149 |
/** |
| 150 |
* Enable performance counters for testing efficiency (disabled by default for production) |
| 151 |
*/ |
| 152 |
public function enablePerformanceCounters(bool $enable = true): void { |
| 153 |
$this->enablePerformanceCounters = $enable; |
| 154 |
if ($enable) { |
| 155 |
$this->resetPerformanceCounters(); |
| 156 |
} |
| 157 |
} |
| 158 |
|
| 159 |
/** |
| 160 |
* Skip the N-gram gate 4 early return so the full Levenshtein scan runs. |
| 161 |
* Used by the async page-suggestions worker where the scan time is acceptable. |
| 162 |
*/ |
| 163 |
public function setSkipNgramGate4(bool $skip = true): void { |
| 164 |
$this->skipNgramGate4 = $skip; |
| 165 |
} |
| 166 |
|
| 167 |
/** |
| 168 |
* Reset performance counters to zero |
| 169 |
*/ |
| 170 |
public function resetPerformanceCounters(): void { |
| 171 |
$this->levenshteinCallCount = 0; |
| 172 |
$this->totalPagesConsidered = 0; |
| 173 |
} |
| 174 |
|
| 175 |
/** |
| 176 |
* Get current performance counter values |
| 177 |
* @return array{levenshtein_calls: int, pages_considered: int, efficiency_percent: float} |
| 178 |
*/ |
| 179 |
public function getPerformanceCounters(): array { |
| 180 |
$efficiency = 0; |
| 181 |
if ($this->totalPagesConsidered > 0) { |
| 182 |
$efficiency = ($this->levenshteinCallCount / $this->totalPagesConsidered) * 100; |
| 183 |
} |
| 184 |
|
| 185 |
return [ |
| 186 |
'levenshtein_calls' => $this->levenshteinCallCount, |
| 187 |
'pages_considered' => $this->totalPagesConsidered, |
| 188 |
'efficiency_percent' => round($efficiency, 2) |
| 189 |
]; |
| 190 |
} |
| 191 |
|
| 192 |
/** |
| 193 |
* Find URL suggestions using smart caching (N-gram filtering). |
| 194 |
* This is a wrapper around findMatchingPosts() primarily for testing. |
| 195 |
* |
| 196 |
* @param string $requestedURL The 404 URL to find matches for |
| 197 |
* @param string $includeCats Whether to include categories (default '1') |
| 198 |
* @param bool $includeTags Whether to include tags (default true, converted to '1') |
| 199 |
* @return array<int, mixed> Array of matching posts/pages |
| 200 |
*/ |
| 201 |
public function findSuggestionsForURLUsingSmartCache($requestedURL, $includeCats = '1', $includeTags = true) { |
| 202 |
// Convert boolean to string for backward compatibility |
| 203 |
$includeTagsStr = $includeTags ? '1' : '0'; |
| 204 |
return $this->findMatchingPosts($requestedURL, $includeCats, $includeTagsStr); |
| 205 |
} |
| 206 |
|
| 207 |
static function init(): void { |
| 208 |
// any time a page is saved or updated, or the permalink structure changes, then we have to clear |
| 209 |
// the spelling cache because the results may have changed. |
| 210 |
$me = abj_service('spell_checker'); |
| 211 |
|
| 212 |
add_action('updated_option', array($me,'permalinkStructureChanged'), 10, 2); |
| 213 |
add_action('save_post', array($me,'save_postListener'), 10, 3); |
| 214 |
add_action('delete_post', array($me,'delete_postListener'), 10, 2); |
| 215 |
} |
| 216 |
|
| 217 |
/** Find a match using spell checking. |
| 218 |
* Use spell checking to find the correct link. Return the permalink (map) if there is one, otherwise return null. |
| 219 |
* @param string $requestedURL The URL slug to check for spelling matches |
| 220 |
* @param string|null $fullRequestedURL Optional full URL path for caching results (e.g., '/site/bad-url') |
| 221 |
* @param array<string, mixed>|null $optionsOverride |
| 222 |
* @return array<string, mixed>|null |
| 223 |
*/ |
| 224 |
function getPermalinkUsingSpelling(string $requestedURL, ?string $fullRequestedURL = null, $optionsOverride = null) { |
| 225 |
$abj404spellChecker = abj_service('spell_checker'); |
| 226 |
|
| 227 |
$options = is_array($optionsOverride) ? $optionsOverride : $this->logic->getOptions(); |
| 228 |
|
| 229 |
if (@$options['auto_redirects'] == '1') { |
| 230 |
// Site owner wants automatic redirects. |
| 231 |
$autoCats = isset($options['auto_cats']) && is_string($options['auto_cats']) ? $options['auto_cats'] : '1'; |
| 232 |
$autoTags = isset($options['auto_tags']) && is_string($options['auto_tags']) ? $options['auto_tags'] : '1'; |
| 233 |
$permalinksPacket = $abj404spellChecker->findMatchingPosts($requestedURL, |
| 234 |
$autoCats, $autoTags); |
| 235 |
|
| 236 |
$permalinks = $permalinksPacket[0]; |
| 237 |
$rowType = $permalinksPacket[1]; |
| 238 |
|
| 239 |
$minScore = $options['auto_score']; |
| 240 |
|
| 241 |
// since the links were previously sorted so that the highest score would be first, |
| 242 |
// we only use the first element of the array; |
| 243 |
if (!is_array($permalinks) || empty($permalinks)) { |
| 244 |
return null; |
| 245 |
} |
| 246 |
$linkScore = reset($permalinks); |
| 247 |
$idAndType = key($permalinks); |
| 248 |
$idAndTypeStr = is_string($idAndType) ? $idAndType : (string)$idAndType; |
| 249 |
$linkScoreInt = is_scalar($linkScore) ? (int)$linkScore : 0; |
| 250 |
$permalink = ABJ_404_Solution_Functions::permalinkInfoToArray($idAndTypeStr, $linkScoreInt, |
| 251 |
is_string($rowType) ? $rowType : null, $options); |
| 252 |
|
| 253 |
if ($permalink['score'] >= $minScore) { |
| 254 |
// We found a permalink that will work! |
| 255 |
$redirectType = $permalink['type']; |
| 256 |
if (('' . $redirectType != ABJ404_TYPE_404_DISPLAYED) && ('' . $redirectType != ABJ404_TYPE_HOME)) { |
| 257 |
return $permalink; |
| 258 |
|
| 259 |
} else { |
| 260 |
$permalinkJson = json_encode($permalink); |
| 261 |
$this->logger->errorMessage("Unhandled permalink type: " . |
| 262 |
wp_kses_post(is_string($permalinkJson) ? $permalinkJson : '{}')); |
| 263 |
return null; |
| 264 |
} |
| 265 |
} |
| 266 |
|
| 267 |
// No match met the auto-redirect threshold - cache results for shortcode |
| 268 |
// This avoids recomputing suggestions when the 404 page renders |
| 269 |
if ($fullRequestedURL !== null) { |
| 270 |
$this->cacheComputedSuggestionsForShortcode($fullRequestedURL, $permalinksPacket); |
| 271 |
} |
| 272 |
} |
| 273 |
|
| 274 |
return null; |
| 275 |
} |
| 276 |
|
| 277 |
/** |
| 278 |
* Cache computed suggestions in a transient for the shortcode to use. |
| 279 |
* This avoids duplicate computation when getPermalinkUsingSpelling() runs |
| 280 |
* but doesn't find a match above the auto-redirect threshold. |
| 281 |
* |
| 282 |
* @param string $fullRequestedURL The full URL path (e.g., '/site/bad-url') |
| 283 |
* @param array<int, mixed> $permalinksPacket The computed suggestions [permalinks, rowType] |
| 284 |
*/ |
| 285 |
private function cacheComputedSuggestionsForShortcode(string $fullRequestedURL, array $permalinksPacket): void { |
| 286 |
// Normalize URL using centralized function for consistency |
| 287 |
$normalizedURL = $this->f->normalizeURLForCacheKey($fullRequestedURL); |
| 288 |
|
| 289 |
$urlKey = md5($normalizedURL); |
| 290 |
$transientKey = 'abj404_suggest_' . $urlKey; |
| 291 |
|
| 292 |
// Don't overwrite if already set (e.g., by async trigger) |
| 293 |
$existing = get_transient($transientKey); |
| 294 |
if ($existing !== false) { |
| 295 |
return; |
| 296 |
} |
| 297 |
|
| 298 |
// Store as 'complete' so shortcode renders immediately |
| 299 |
set_transient($transientKey, array( |
| 300 |
'status' => 'complete', |
| 301 |
'suggestions' => $permalinksPacket, |
| 302 |
'url' => $normalizedURL, |
| 303 |
'completed' => time() |
| 304 |
), 300); // 5 minute TTL |
| 305 |
|
| 306 |
$this->logger->debugMessage("Cached spell-check suggestions for shortcode: " . |
| 307 |
esc_html($normalizedURL)); |
| 308 |
} |
| 309 |
|
| 310 |
/** |
| 311 |
* Trigger asynchronous suggestion computation via non-blocking HTTP request. |
| 312 |
* Uses the requested URL (MD5 hashed) as the transient key. |
| 313 |
* |
| 314 |
* @param string $requestedURL The full requested URL that caused the 404 |
| 315 |
* @return bool True if computation was triggered, false if already pending/complete |
| 316 |
*/ |
| 317 |
public function triggerAsyncSuggestionComputation($requestedURL) { |
| 318 |
$f = abj_service('functions'); |
| 319 |
|
| 320 |
// Normalize URL using centralized function for consistency |
| 321 |
$normalizedURL = $f->normalizeURLForCacheKey($requestedURL); |
| 322 |
|
| 323 |
$urlKey = md5($normalizedURL); |
| 324 |
$transientKey = 'abj404_suggest_' . $urlKey; |
| 325 |
|
| 326 |
// Check if already computing or complete - prevent duplicate work |
| 327 |
$existing = get_transient($transientKey); |
| 328 |
if ($existing !== false) { |
| 329 |
$existingStatus = (is_array($existing) && isset($existing['status']) && is_string($existing['status'])) ? $existing['status'] : 'unknown'; |
| 330 |
$this->logger->debugMessage("Async suggestions: skipping, transient already exists for " . |
| 331 |
esc_html($normalizedURL) . " (status: " . esc_html($existingStatus) . ")"); |
| 332 |
return false; |
| 333 |
} |
| 334 |
|
| 335 |
// Generate a unique token for this computation request |
| 336 |
// This prevents unauthorized direct calls to the AJAX endpoint (DoS protection) |
| 337 |
$token = wp_generate_password(32, false); |
| 338 |
|
| 339 |
// Mark as pending BEFORE firing request (race condition protection) |
| 340 |
// TTL of 120 seconds: gives slow hosts enough time to start the worker |
| 341 |
// Note: started=0 means no worker has claimed the work yet. The first worker |
| 342 |
// will set started=time() when it claims the work. This prevents the bug where |
| 343 |
// the first worker skips itself thinking another worker is already computing. |
| 344 |
set_transient($transientKey, array( |
| 345 |
'status' => 'pending', |
| 346 |
'url' => $normalizedURL, |
| 347 |
'started' => 0, // 0 = no worker has claimed yet; worker sets time() when claiming |
| 348 |
'created' => time(), // track creation time to detect worker no-show |
| 349 |
'token' => $token |
| 350 |
), 120); // 2 minute TTL (allows slow wp_remote_post) |
| 351 |
|
| 352 |
$this->logger->debugMessage("Async suggestions: triggering background computation for " . |
| 353 |
esc_html($normalizedURL)); |
| 354 |
|
| 355 |
// Fire non-blocking request to compute suggestions |
| 356 |
// Note: timeout of 5s is needed for connection establishment (TLS handshake, etc.) |
| 357 |
// even with blocking=false, a too-short timeout can prevent the request from being sent |
| 358 |
$response = wp_remote_post(admin_url('admin-ajax.php'), array( |
| 359 |
'blocking' => false, |
| 360 |
'timeout' => 5, // 5 seconds for connection establishment |
| 361 |
'sslverify' => apply_filters('https_local_ssl_verify', false), |
| 362 |
'body' => array( |
| 363 |
'action' => 'abj404_compute_suggestions', |
| 364 |
'url' => $normalizedURL, |
| 365 |
'token' => $token |
| 366 |
) |
| 367 |
)); |
| 368 |
|
| 369 |
// If dispatch failed, delete the pending transient so caller can compute synchronously |
| 370 |
if (is_wp_error($response)) { |
| 371 |
$this->logger->debugMessage("Async suggestions: dispatch failed for " . |
| 372 |
esc_html($normalizedURL) . " - " . $response->get_error_message()); |
| 373 |
delete_transient($transientKey); |
| 374 |
return false; |
| 375 |
} |
| 376 |
|
| 377 |
return true; |
| 378 |
} |
| 379 |
|
| 380 |
/** |
| 381 |
* Check if the configured 404 page contains the suggestions shortcode. |
| 382 |
* |
| 383 |
* @return bool True if 404 page has the shortcode |
| 384 |
*/ |
| 385 |
public function does404PageHaveSuggestionsShortcode() { |
| 386 |
$options = $this->logic->getOptions(); |
| 387 |
$dest404pageRaw = isset($options['dest404page']) ? $options['dest404page'] : null; |
| 388 |
$dest404page = is_string($dest404pageRaw) ? $dest404pageRaw : null; |
| 389 |
|
| 390 |
if (!$this->logic->thereIsAUserSpecified404Page($dest404page)) { |
| 391 |
return false; |
| 392 |
} |
| 393 |
|
| 394 |
// Extract page ID from dest404page (format: "123|1") |
| 395 |
$parts = explode('|', $dest404page ?? ''); |
| 396 |
$page404Id = isset($parts[0]) ? intval($parts[0]) : 0; |
| 397 |
|
| 398 |
if ($page404Id <= 0) { |
| 399 |
return false; |
| 400 |
} |
| 401 |
|
| 402 |
$page = get_post($page404Id); |
| 403 |
if (!$page) { |
| 404 |
return false; |
| 405 |
} |
| 406 |
|
| 407 |
return has_shortcode($page->post_content, ABJ404_SHORTCODE_NAME); |
| 408 |
} |
| 409 |
|
| 410 |
} |
| 411 |
|
