| 1 |
<?php |
| 2 |
|
| 3 |
if (!defined('ABSPATH')) { |
| 4 |
exit; |
| 5 |
} |
| 6 |
|
| 7 |
/** |
| 8 |
* Levenshtein distance engine and candidate pre-filtering for the |
| 9 |
* spell-checking subsystem. |
| 10 |
* |
| 11 |
* Extracted from SpellCheckerTrait_LevenshteinEngine as a standalone class |
| 12 |
* with explicit dependency injection. |
| 13 |
*/ |
| 14 |
class ABJ_404_Solution_SpellLevenshteinEngine { |
| 15 |
|
| 16 |
const NGRAM_PREFILTER_THRESHOLD = 0.3; |
| 17 |
|
| 18 |
const NGRAM_PREFILTER_MAX_CANDIDATES = 500; |
| 19 |
|
| 20 |
const NGRAM_MIN_CACHE_ENTRIES = 50; |
| 21 |
|
| 22 |
const NGRAM_SECONDARY_THRESHOLD = 0.4; |
| 23 |
|
| 24 |
const NGRAM_SECONDARY_MAX_CANDIDATES = 100; |
| 25 |
|
| 26 |
const NGRAM_MIN_COVERAGE_RATIO = 0.8; |
| 27 |
|
| 28 |
const NGRAM_SECONDARY_MIN_CANDIDATES = 50; |
| 29 |
|
| 30 |
/** @var ABJ_404_Solution_Functions */ |
| 31 |
private $f; |
| 32 |
|
| 33 |
/** @var ABJ_404_Solution_Logging */ |
| 34 |
private $logger; |
| 35 |
|
| 36 |
/** @var ABJ_404_Solution_ContentRepository */ |
| 37 |
private $contentRepository; |
| 38 |
|
| 39 |
/** @var ABJ_404_Solution_SpellURLMatcher */ |
| 40 |
private $urlMatcher; |
| 41 |
|
| 42 |
/** @var ABJ_404_Solution_SpellNGramPrefilter */ |
| 43 |
private $ngramPrefilter; |
| 44 |
|
| 45 |
/** @var ABJ_404_Solution_SpellCandidatePermalinkLookup */ |
| 46 |
private $permalinkLookup; |
| 47 |
|
| 48 |
/** @var ABJ_404_Solution_SpellLevenshteinEngineDependencies the bundle reused to build a per-call distance ranker */ |
| 49 |
private $deps; |
| 50 |
|
| 51 |
private bool $enablePerformanceCounters = false; |
| 52 |
|
| 53 |
private bool $skipNgramGate4 = false; |
| 54 |
|
| 55 |
private int $levenshteinCallCount = 0; |
| 56 |
|
| 57 |
private int $totalPagesConsidered = 0; |
| 58 |
|
| 59 |
/** @var ABJ_404_Solution_PublishedPostsProvider|null */ |
| 60 |
private ?ABJ_404_Solution_PublishedPostsProvider $publishedPostsProvider = null; |
| 61 |
|
| 62 |
/** |
| 63 |
* @param ABJ_404_Solution_SpellLevenshteinEngineDependencies $deps |
| 64 |
*/ |
| 65 |
public function __construct(ABJ_404_Solution_SpellLevenshteinEngineDependencies $deps) { |
| 66 |
$this->deps = $deps; |
| 67 |
$this->f = $deps->functions; |
| 68 |
$this->logger = $deps->logger; |
| 69 |
$this->contentRepository = $deps->contentRepository; |
| 70 |
$this->urlMatcher = $deps->urlMatcher; |
| 71 |
$this->ngramPrefilter = new ABJ_404_Solution_SpellNGramPrefilter($deps->ngramFilter, $deps->logger); |
| 72 |
$this->permalinkLookup = new ABJ_404_Solution_SpellCandidatePermalinkLookup($deps->contentRepository, $deps->urlMatcher); |
| 73 |
} |
| 74 |
|
| 75 |
/** @param ABJ_404_Solution_PublishedPostsProvider|null $provider */ |
| 76 |
public function setPublishedPostsProvider(?ABJ_404_Solution_PublishedPostsProvider $provider): void { |
| 77 |
$this->publishedPostsProvider = $provider; |
| 78 |
} |
| 79 |
|
| 80 |
public function enablePerformanceCounters(bool $enable = true): void { |
| 81 |
$this->enablePerformanceCounters = $enable; |
| 82 |
if ($enable) { |
| 83 |
$this->resetPerformanceCounters(); |
| 84 |
} |
| 85 |
} |
| 86 |
|
| 87 |
public function setSkipNgramGate4(bool $skip = true): void { |
| 88 |
$this->skipNgramGate4 = $skip; |
| 89 |
} |
| 90 |
|
| 91 |
public function resetPerformanceCounters(): void { |
| 92 |
$this->levenshteinCallCount = 0; |
| 93 |
$this->totalPagesConsidered = 0; |
| 94 |
} |
| 95 |
|
| 96 |
/** |
| 97 |
* @return array{levenshtein_calls: int, pages_considered: int, efficiency_percent: float} |
| 98 |
*/ |
| 99 |
public function getPerformanceCounters(): array { |
| 100 |
$efficiency = 0; |
| 101 |
if ($this->totalPagesConsidered > 0) { |
| 102 |
$efficiency = ($this->levenshteinCallCount / $this->totalPagesConsidered) * 100; |
| 103 |
} |
| 104 |
|
| 105 |
return [ |
| 106 |
'levenshtein_calls' => $this->levenshteinCallCount, |
| 107 |
'pages_considered' => $this->totalPagesConsidered, |
| 108 |
'efficiency_percent' => round($efficiency, 2) |
| 109 |
]; |
| 110 |
} |
| 111 |
|
| 112 |
/** |
| 113 |
* @param string $requestedURLCleaned |
| 114 |
* @param string $fullURLspaces |
| 115 |
* @param string $rowType |
| 116 |
* @param array<int, array<string, mixed>>|null $rows |
| 117 |
* @return array<int|string, mixed> |
| 118 |
*/ |
| 119 |
function getLikelyMatchIDs(string $requestedURLCleaned, string $fullURLspaces, string $rowType, ?array $rows = null) { |
| 120 |
|
| 121 |
$options = abj_service('options_repository')->getOptions(); |
| 122 |
$suggestMaxLikely = isset($options['suggest_max']) && is_scalar($options['suggest_max']) ? $options['suggest_max'] : 5; |
| 123 |
$onlyNeedThisManyPages = min(5 * absint($suggestMaxLikely), 100); |
| 124 |
|
| 125 |
$ngramPrefilterResult = $this->ngramPrefilter->tryApply( |
| 126 |
$rowType, |
| 127 |
$rows, |
| 128 |
$requestedURLCleaned, |
| 129 |
$this->publishedPostsProvider, |
| 130 |
$this->skipNgramGate4 |
| 131 |
); |
| 132 |
if ($ngramPrefilterResult === 'early_return') { |
| 133 |
return array(); |
| 134 |
} |
| 135 |
$ngramPrefilterApplied = ($ngramPrefilterResult === 'applied'); |
| 136 |
|
| 137 |
$ranker = new ABJ_404_Solution_SpellCandidateDistanceRanker($this->deps, $this->ngramPrefilter); |
| 138 |
|
| 139 |
$requestedURLCleanedLength = $this->f->strlen($requestedURLCleaned); |
| 140 |
$fullURLspacesLength = $this->f->strlen($fullURLspaces); |
| 141 |
|
| 142 |
$userRequestedURLWords = explode(" ", (empty($fullURLspaces) ? $requestedURLCleaned : $fullURLspaces)); |
| 143 |
$observedPermalinksById = array(); |
| 144 |
$wasntReadyCount = 0; |
| 145 |
|
| 146 |
if ($this->publishedPostsProvider === null) { |
| 147 |
return array(); |
| 148 |
} |
| 149 |
$postsProvider = $this->publishedPostsProvider; |
| 150 |
if (!$ngramPrefilterApplied) { |
| 151 |
$postsProvider->resetBatch(); |
| 152 |
} |
| 153 |
if ($rows != null) { |
| 154 |
$postsProvider->useThisData($rows); |
| 155 |
} |
| 156 |
$currentBatch = $postsProvider->getNextBatch($requestedURLCleanedLength); |
| 157 |
|
| 158 |
$row = array_pop($currentBatch); |
| 159 |
while ($row != null) { |
| 160 |
$row = (array)$row; |
| 161 |
|
| 162 |
if ($this->enablePerformanceCounters) { |
| 163 |
$this->totalPagesConsidered++; |
| 164 |
} |
| 165 |
|
| 166 |
$id = $this->extractRowCandidateId($row, $rowType); |
| 167 |
if ($id === null) { |
| 168 |
$row = array_pop($currentBatch); |
| 169 |
continue; |
| 170 |
} |
| 171 |
$idInt = is_scalar($id) ? (int)$id : 0; |
| 172 |
|
| 173 |
$the_permalink = null; |
| 174 |
$urlPath = null; |
| 175 |
$this->resolveCandidatePermalinkParts( |
| 176 |
$row, $idInt, $rowType, $wasntReadyCount, $the_permalink, $urlPath |
| 177 |
); |
| 178 |
|
| 179 |
abj_service('request_context')->debug_info = 'Likely match IDs processing permalink: ' . |
| 180 |
$the_permalink . ', $wasntReadyCount: ' . $wasntReadyCount; |
| 181 |
|
| 182 |
if ($urlPath === null) { |
| 183 |
// Skip this candidate (no parseable path) AND advance to the next |
| 184 |
// row, exactly like the $id === null skip above. A bare `continue` |
| 185 |
// here would re-evaluate the SAME row forever -- a per-request |
| 186 |
// infinite loop that hangs the 404 response until PHP's |
| 187 |
// max_execution_time kills it (the worst "not fast" outcome on a |
| 188 |
// large/diverse site where some candidate has an unparseable URL). |
| 189 |
$row = array_pop($currentBatch); |
| 190 |
continue; |
| 191 |
} |
| 192 |
if (is_string($the_permalink)) { |
| 193 |
$observedPermalinksById[$idInt] = $the_permalink; |
| 194 |
} |
| 195 |
|
| 196 |
$ranker->score( |
| 197 |
$id, $urlPath, $requestedURLCleanedLength, |
| 198 |
$fullURLspaces, $fullURLspacesLength, $userRequestedURLWords |
| 199 |
); |
| 200 |
|
| 201 |
$row = array_pop($currentBatch); |
| 202 |
if ($row == null) { |
| 203 |
$maxAcceptableDistance = $ranker->getMaxAcceptableDistance($onlyNeedThisManyPages); |
| 204 |
|
| 205 |
$currentBatch = $postsProvider->getNextBatch( |
| 206 |
$requestedURLCleanedLength, 1000, $maxAcceptableDistance); |
| 207 |
$row = array_pop($currentBatch); |
| 208 |
} |
| 209 |
} |
| 210 |
abj_service('request_context')->debug_info = ''; |
| 211 |
|
| 212 |
if ($wasntReadyCount > 0) { |
| 213 |
$this->logger->infoMessage("The permalink cache wasn't ready for " . $wasntReadyCount . " IDs."); |
| 214 |
} |
| 215 |
|
| 216 |
$candidateIds = $ranker->prioritize( |
| 217 |
$onlyNeedThisManyPages, $ngramPrefilterApplied, $requestedURLCleaned |
| 218 |
); |
| 219 |
|
| 220 |
return $this->permalinkLookup->lookup( |
| 221 |
array_values(array_unique($candidateIds)), $rowType, $observedPermalinksById |
| 222 |
); |
| 223 |
} |
| 224 |
|
| 225 |
/** |
| 226 |
* Pull the candidate id out of one published-content row, dispatching on |
| 227 |
* the row type. Returns null when the row carries no usable id (the caller |
| 228 |
* skips it); throws for an unrecognized row type. |
| 229 |
* |
| 230 |
* @param array<mixed, mixed> $row |
| 231 |
* @param string $rowType |
| 232 |
* @return mixed the raw candidate id as stored in the row, or null when absent |
| 233 |
*/ |
| 234 |
private function extractRowCandidateId(array $row, string $rowType) { |
| 235 |
if ($rowType == 'pages') { |
| 236 |
return $row['id']; |
| 237 |
|
| 238 |
} else if ($rowType == 'tags') { |
| 239 |
return array_key_exists('term_id', $row) ? $row['term_id'] : null; |
| 240 |
|
| 241 |
} else if ($rowType == 'categories') { |
| 242 |
return array_key_exists('term_id', $row) ? $row['term_id'] : null; |
| 243 |
|
| 244 |
} else if ($rowType == 'image') { |
| 245 |
return $row['id']; |
| 246 |
} |
| 247 |
|
| 248 |
throw new \Exception("Unknown row type ... " . esc_html($rowType)); // allow-raw-error: assertion, should never reach user |
| 249 |
} |
| 250 |
|
| 251 |
/** |
| 252 |
* Resolve a candidate's permalink and the path portion of its URL. Prefers |
| 253 |
* the url carried on the row; falls back to the permalink cache (incrementing |
| 254 |
* $wasntReadyCount) when the row has no usable url or the row url failed to |
| 255 |
* parse. The permalink and url path are returned through the by-reference |
| 256 |
* out-parameters to keep this hot-path step allocation-free. $urlPath is null |
| 257 |
* when no parseable path could be resolved (the caller then skips the row). |
| 258 |
* |
| 259 |
* @param array<mixed, mixed> $row |
| 260 |
* @param int $idInt |
| 261 |
* @param string $rowType |
| 262 |
* @param int $wasntReadyCount incremented by reference on a cache fallback |
| 263 |
* @param string|null $the_permalink |
| 264 |
* @param string|null $urlPath |
| 265 |
* @param-out string|null $the_permalink |
| 266 |
* @param-out string|null $urlPath |
| 267 |
*/ |
| 268 |
private function resolveCandidatePermalinkParts( |
| 269 |
array $row, int $idInt, string $rowType, int &$wasntReadyCount, &$the_permalink, &$urlPath |
| 270 |
): void { |
| 271 |
$the_permalink = null; |
| 272 |
$urlPath = null; |
| 273 |
$urlParts = null; |
| 274 |
if (array_key_exists('url', $row)) { |
| 275 |
$the_permalink = isset($row['url']) && is_string($row['url']) ? $row['url'] : ''; |
| 276 |
$the_permalink = abj_service('sanitizer')->normalizeUrlString($the_permalink); |
| 277 |
$urlParts = parse_url($the_permalink); |
| 278 |
|
| 279 |
if (is_bool($urlParts)) { |
| 280 |
$this->contentRepository->removeFromPermalinkCache($idInt); |
| 281 |
} |
| 282 |
} |
| 283 |
if (!array_key_exists('url', $row) || (isset($urlParts) && is_bool($urlParts))) { |
| 284 |
$wasntReadyCount++; |
| 285 |
$the_permalink = $this->urlMatcher->getPermalink($idInt, $rowType); |
| 286 |
$the_permalink = abj_service('sanitizer')->normalizeUrlString($the_permalink); |
| 287 |
$urlParts = parse_url($the_permalink); |
| 288 |
} |
| 289 |
|
| 290 |
if (is_array($urlParts) && array_key_exists('path', $urlParts)) { |
| 291 |
$urlPath = $urlParts['path']; |
| 292 |
} |
| 293 |
} |
| 294 |
|
| 295 |
/** |
| 296 |
* @param string $str1 |
| 297 |
* @param string $str2 |
| 298 |
* @return int |
| 299 |
* @throws Exception |
| 300 |
*/ |
| 301 |
function customLevenshtein($str1, $str2) { |
| 302 |
if ($this->enablePerformanceCounters) { |
| 303 |
$this->levenshteinCallCount++; |
| 304 |
} |
| 305 |
abj_service('request_context')->debug_info = 'customLevenshtein. str1: ' . esc_html($str1) . ', str2: ' . esc_html($str2); |
| 306 |
|
| 307 |
$RowLen = $this->f->strlen($str1); |
| 308 |
$ColLen = $this->f->strlen($str2); |
| 309 |
$cost = 0; |
| 310 |
|
| 311 |
if (max($RowLen, $ColLen) > ABJ404_MAX_URL_LENGTH) { |
| 312 |
throw new \Exception("Maximum string length in customLevenshtein is " . // allow-raw-error: assertion, should never reach user |
| 313 |
ABJ404_MAX_URL_LENGTH . ". Yours is " . max($RowLen, $ColLen) . "."); |
| 314 |
} |
| 315 |
|
| 316 |
if (strlen($str1) <= 255 && strlen($str2) <= 255) { |
| 317 |
return levenshtein($str1, $str2); |
| 318 |
} |
| 319 |
|
| 320 |
if ($RowLen == 0) { |
| 321 |
return $ColLen; |
| 322 |
} else if ($ColLen == 0) { |
| 323 |
return $RowLen; |
| 324 |
} |
| 325 |
|
| 326 |
$chars1 = mb_str_split($str1, 1, 'UTF-8'); |
| 327 |
$chars2 = mb_str_split($str2, 1, 'UTF-8'); |
| 328 |
|
| 329 |
$v0 = array_fill(0, $RowLen + 1, 0); |
| 330 |
$v1 = array_fill(0, $RowLen + 1, 0); |
| 331 |
|
| 332 |
for ($RowIdx = 1; $RowIdx <= $RowLen; $RowIdx++) { |
| 333 |
$v0[$RowIdx] = $RowIdx; |
| 334 |
} |
| 335 |
|
| 336 |
for ($ColIdx = 1; $ColIdx <= $ColLen; $ColIdx++) { |
| 337 |
$v1[0] = $ColIdx; |
| 338 |
|
| 339 |
for ($RowIdx = 1; $RowIdx <= $RowLen; $RowIdx++) { |
| 340 |
$cost = ($chars1[$RowIdx - 1] === $chars2[$ColIdx - 1]) ? 0 : 1; |
| 341 |
$v1[$RowIdx] = min($v0[$RowIdx] + 1, $v1[$RowIdx - 1] + 1, $v0[$RowIdx - 1] + $cost); |
| 342 |
} |
| 343 |
|
| 344 |
$vTmp = $v0; |
| 345 |
$v0 = $v1; |
| 346 |
$v1 = $vTmp; |
| 347 |
} |
| 348 |
|
| 349 |
abj_service('request_context')->debug_info = 'Cleared after customLevenshtein.'; |
| 350 |
return $v0[$RowLen]; |
| 351 |
} |
| 352 |
|
| 353 |
} |
| 354 |
|
