| 1 |
<?php |
| 2 |
|
| 3 |
if (!defined('ABSPATH')) { |
| 4 |
exit; |
| 5 |
} |
| 6 |
|
| 7 |
/** |
| 8 |
* Keeps the taxonomy-term half of the n-gram cache in sync with the published |
| 9 |
* categories and tags. |
| 10 |
* |
| 11 |
* Split from {@see ABJ_404_Solution_NGramCacheReconciler} because the term |
| 12 |
* half reconciles against a different source through a different writer: the |
| 13 |
* published-term set comes from the content repository (WP taxonomy) rather |
| 14 |
* than the permalink cache, and rows are written by extracting n-grams and |
| 15 |
* storing them through the cache repository rather than by handing page ids to |
| 16 |
* the bulk rebuilder. It is also bounded differently -- the published term set |
| 17 |
* is small, so it is processed in full rather than in batches, and there is no |
| 18 |
* backlog to classify. |
| 19 |
* |
| 20 |
* Sibling of the term-side collaborators already in this layer |
| 21 |
* (ABJ_404_Solution_TermCandidateSource, |
| 22 |
* ABJ_404_Solution_TermNGramCoveragePolicy). |
| 23 |
* |
| 24 |
* Lock ownership stays with the orchestrator (DatabaseUpgradeNGram): the sync |
| 25 |
* path runs under the shared 'ngram_rebuild' lock, the cleanup path without it |
| 26 |
* (it only deletes by primary key on rows the source no longer references). |
| 27 |
*/ |
| 28 |
class ABJ_404_Solution_NGramTermCacheReconciler { |
| 29 |
|
| 30 |
/** @var ABJ_404_Solution_DatabaseCore */ |
| 31 |
private $dbCore; |
| 32 |
|
| 33 |
/** @var mixed */ |
| 34 |
private $extractor; |
| 35 |
|
| 36 |
/** @var mixed */ |
| 37 |
private $repo; |
| 38 |
|
| 39 |
/** @var mixed */ |
| 40 |
private $coveragePolicy; |
| 41 |
|
| 42 |
/** @var ABJ_404_Solution_Functions */ |
| 43 |
private $f; |
| 44 |
|
| 45 |
/** @var ABJ_404_Solution_Logging */ |
| 46 |
private $logger; |
| 47 |
|
| 48 |
/** |
| 49 |
* @param ABJ_404_Solution_DatabaseCore $dbCore |
| 50 |
* @param mixed $extractor Object exposing extractNGrams(). |
| 51 |
* @param mixed $repo Object exposing storeNGrams(). |
| 52 |
* @param mixed $coveragePolicy Object exposing invalidateCoverageCaches(). |
| 53 |
* @param ABJ_404_Solution_Functions $f |
| 54 |
* @param ABJ_404_Solution_Logging $logger |
| 55 |
*/ |
| 56 |
public function __construct($dbCore, $extractor, $repo, $coveragePolicy, $f, $logger) { |
| 57 |
$this->dbCore = $dbCore; |
| 58 |
$this->extractor = $extractor; |
| 59 |
$this->repo = $repo; |
| 60 |
$this->coveragePolicy = $coveragePolicy; |
| 61 |
$this->f = $f; |
| 62 |
$this->logger = $logger; |
| 63 |
} |
| 64 |
|
| 65 |
/** |
| 66 |
* Add n-gram entries for published terms of one taxonomy type |
| 67 |
* ('category' or 'tag') that are missing from the cache. Category and tag |
| 68 |
* sync are identical apart from the type label and source rows, so they |
| 69 |
* share this one implementation. |
| 70 |
* |
| 71 |
* @param string $ngramTable |
| 72 |
* @param string $type 'category' or 'tag'. |
| 73 |
* @param array<int, object> $terms Published terms of this type (term_id, url). |
| 74 |
* @return array{added:int, failed:int} |
| 75 |
*/ |
| 76 |
public function syncMissingTerms(string $ngramTable, string $type, array $terms): array { |
| 77 |
$stats = ['added' => 0, 'failed' => 0]; |
| 78 |
|
| 79 |
if (empty($terms)) { |
| 80 |
return $stats; |
| 81 |
} |
| 82 |
|
| 83 |
$missing = []; |
| 84 |
foreach ($terms as $term) { |
| 85 |
/** @var object{term_id: int, url: string} $term */ |
| 86 |
$termId = (int)$term->term_id; |
| 87 |
$exists = $this->dbCore->queryScalarInt( |
| 88 |
"SELECT COUNT(*) AS c FROM {$ngramTable} WHERE id = %d AND type = %s", |
| 89 |
['query_params' => [$termId, $type]] |
| 90 |
); |
| 91 |
if ($exists == 0) { |
| 92 |
$missing[] = $term; |
| 93 |
} |
| 94 |
} |
| 95 |
|
| 96 |
if (empty($missing)) { |
| 97 |
$this->logger->debugMessage("No missing {$type} ngram entries found. All {$type}s are synced."); |
| 98 |
return $stats; |
| 99 |
} |
| 100 |
|
| 101 |
$this->logger->infoMessage("Found " . count($missing) . " {$type}s missing ngram entries. Adding..."); |
| 102 |
|
| 103 |
foreach ($missing as $term) { |
| 104 |
try { |
| 105 |
/** @var object{term_id: int, url: string} $term */ |
| 106 |
$termId = (int)$term->term_id; |
| 107 |
$url = (string)$term->url; |
| 108 |
|
| 109 |
if (empty($url) || $url === 'in code') { |
| 110 |
$this->logger->debugMessage("Skipping {$type} {$termId} - no valid URL"); |
| 111 |
continue; |
| 112 |
} |
| 113 |
|
| 114 |
$urlNormalized = $this->f->strtolower(trim($url)); |
| 115 |
$ngrams = $this->extractNGrams($urlNormalized); |
| 116 |
$success = $this->storeNGrams($termId, $url, $urlNormalized, $ngrams, $type); |
| 117 |
|
| 118 |
if ($success) { |
| 119 |
$stats['added']++; |
| 120 |
} else { |
| 121 |
$stats['failed']++; |
| 122 |
} |
| 123 |
} catch (Exception $e) { |
| 124 |
$this->logger->errorMessage("Failed to add ngram for {$type} {$termId}: " . $e->getMessage()); |
| 125 |
$stats['failed']++; |
| 126 |
} |
| 127 |
} |
| 128 |
|
| 129 |
return $stats; |
| 130 |
} |
| 131 |
|
| 132 |
/** |
| 133 |
* Delete n-gram rows of one taxonomy type ('category' or 'tag') whose |
| 134 |
* source term is no longer published. Category and tag cleanup are |
| 135 |
* identical apart from the type label and source rows, so they share this |
| 136 |
* one implementation. |
| 137 |
* |
| 138 |
* @param string $ngramTable |
| 139 |
* @param string $type 'category' or 'tag'. |
| 140 |
* @param array<int, object> $publishedTerms Currently-published terms of this type. |
| 141 |
* @return array{deleted:int, errors:int} |
| 142 |
*/ |
| 143 |
public function cleanupOrphanedTerms(string $ngramTable, string $type, array $publishedTerms): array { |
| 144 |
// Positive evidence is required before deleting anything. The published |
| 145 |
// list arrives from ContentRepository::getPublishedCategories() / |
| 146 |
// getPublishedTags(), which return an EMPTY ARRAY when their query |
| 147 |
// fails (PublishedContentRepository logs the error and falls through to |
| 148 |
// objectRows($result['rows'] ?? array())). An empty list is therefore |
| 149 |
// indistinguishable from a failed read, and treating it as "nothing is |
| 150 |
// published" made a single transient database error -- a Galera |
| 151 |
// failover, a dropped connection -- delete every cached n-gram row of |
| 152 |
// this type. Refusing to act costs at most some stale rows, which only |
| 153 |
// affect suggestion ranking and are cleaned up on the next run that has |
| 154 |
// real data; acting on it costs the whole cache. |
| 155 |
if (empty($publishedTerms)) { |
| 156 |
$this->logger->debugMessage("Skipping orphaned {$type} ngram cleanup: no published {$type} " |
| 157 |
. "terms were supplied, which is indistinguishable from a failed lookup. " |
| 158 |
. "Nothing is deleted without positive evidence of what is published."); |
| 159 |
return ['deleted' => 0, 'errors' => 0]; |
| 160 |
} |
| 161 |
|
| 162 |
$publishedIds = []; |
| 163 |
foreach ($publishedTerms as $term) { |
| 164 |
/** @var object{term_id: int, url: string} $term */ |
| 165 |
$publishedIds[] = (int)$term->term_id; |
| 166 |
} |
| 167 |
|
| 168 |
$entriesResult = $this->dbCore->queryAndGetResults( |
| 169 |
"SELECT DISTINCT id FROM {$ngramTable} WHERE type = %s", |
| 170 |
['query_params' => [$type], 'result_type' => OBJECT] |
| 171 |
); |
| 172 |
$ngramEntries = isset($entriesResult['rows']) && is_array($entriesResult['rows']) ? $entriesResult['rows'] : []; |
| 173 |
|
| 174 |
if (empty($ngramEntries)) { |
| 175 |
return ['deleted' => 0, 'errors' => 0]; |
| 176 |
} |
| 177 |
|
| 178 |
$orphaned = []; |
| 179 |
foreach ($ngramEntries as $entry) { |
| 180 |
if (!is_object($entry)) { |
| 181 |
continue; |
| 182 |
} |
| 183 |
/** @var object{id: int} $entry */ |
| 184 |
$entId = (int)$entry->id; |
| 185 |
if (!in_array($entId, $publishedIds)) { |
| 186 |
$orphaned[] = $entId; |
| 187 |
} |
| 188 |
} |
| 189 |
|
| 190 |
if (empty($orphaned)) { |
| 191 |
$this->logger->debugMessage("No orphaned {$type} ngram entries found."); |
| 192 |
return ['deleted' => 0, 'errors' => 0]; |
| 193 |
} |
| 194 |
|
| 195 |
$this->logger->infoMessage("Found " . count($orphaned) . " orphaned {$type} ngram entries. Deleting..."); |
| 196 |
|
| 197 |
$deleted = 0; |
| 198 |
$errors = 0; |
| 199 |
foreach ($orphaned as $termId) { |
| 200 |
$deleteResult = $this->dbCore->queryAndGetResults( |
| 201 |
"DELETE FROM {$ngramTable} WHERE id = %d AND type = %s", |
| 202 |
['query_params' => [$termId, $type]] |
| 203 |
); |
| 204 |
|
| 205 |
$deleteError = isset($deleteResult['last_error']) && is_string($deleteResult['last_error']) ? $deleteResult['last_error'] : ''; |
| 206 |
if ($deleteError !== '') { |
| 207 |
if (!$this->dbCore->errorClassifier()->classifyAndHandleInfrastructureError($deleteError)) { |
| 208 |
$this->logger->errorMessage("Failed to delete orphaned {$type} ngram entry ID {$termId}: " . $deleteError); |
| 209 |
} |
| 210 |
$errors++; |
| 211 |
} else { |
| 212 |
$deleted++; |
| 213 |
} |
| 214 |
} |
| 215 |
|
| 216 |
if ($deleted > 0) { |
| 217 |
$this->invalidateCoverageCaches(); |
| 218 |
} |
| 219 |
|
| 220 |
return ['deleted' => $deleted, 'errors' => $errors]; |
| 221 |
} |
| 222 |
|
| 223 |
/** |
| 224 |
* @param string $url |
| 225 |
* @return array{bi: array<int, string>, tri: array<int, string>} |
| 226 |
*/ |
| 227 |
private function extractNGrams(string $url): array { |
| 228 |
$extractor = $this->extractor; |
| 229 |
if (!is_object($extractor) || !method_exists($extractor, 'extractNGrams')) { |
| 230 |
throw new RuntimeException('NGramTermCacheReconciler requires an extractor with extractNGrams().'); |
| 231 |
} |
| 232 |
$ngrams = $extractor->extractNGrams($url); |
| 233 |
return $this->normalizeNGramPayload($ngrams); |
| 234 |
} |
| 235 |
|
| 236 |
|
| 237 |
/** |
| 238 |
* @param mixed $ngrams |
| 239 |
* @return array{bi: array<int, string>, tri: array<int, string>} |
| 240 |
*/ |
| 241 |
private function normalizeNGramPayload($ngrams): array { |
| 242 |
$bi = []; |
| 243 |
$tri = []; |
| 244 |
if (is_array($ngrams)) { |
| 245 |
$biRaw = isset($ngrams['bi']) && is_array($ngrams['bi']) ? $ngrams['bi'] : []; |
| 246 |
foreach ($biRaw as $ngram) { |
| 247 |
if (is_string($ngram)) { |
| 248 |
$bi[] = $ngram; |
| 249 |
} |
| 250 |
} |
| 251 |
$triRaw = isset($ngrams['tri']) && is_array($ngrams['tri']) ? $ngrams['tri'] : []; |
| 252 |
foreach ($triRaw as $ngram) { |
| 253 |
if (is_string($ngram)) { |
| 254 |
$tri[] = $ngram; |
| 255 |
} |
| 256 |
} |
| 257 |
} |
| 258 |
return ['bi' => $bi, 'tri' => $tri]; |
| 259 |
} |
| 260 |
/** |
| 261 |
* @param int $pageId |
| 262 |
* @param string $url |
| 263 |
* @param string $urlNormalized |
| 264 |
* @param array<string, mixed> $ngrams |
| 265 |
* @param string $type |
| 266 |
* @return bool |
| 267 |
*/ |
| 268 |
private function storeNGrams(int $pageId, string $url, string $urlNormalized, array $ngrams, string $type): bool { |
| 269 |
$repo = $this->repo; |
| 270 |
if (!is_object($repo) || !method_exists($repo, 'storeNGrams')) { |
| 271 |
throw new RuntimeException('NGramTermCacheReconciler requires a repository with storeNGrams().'); |
| 272 |
} |
| 273 |
return (bool)$repo->storeNGrams($pageId, $url, $urlNormalized, $ngrams, $type); |
| 274 |
} |
| 275 |
|
| 276 |
/** @return void */ |
| 277 |
private function invalidateCoverageCaches(): void { |
| 278 |
$coveragePolicy = $this->coveragePolicy; |
| 279 |
if (!is_object($coveragePolicy) || !method_exists($coveragePolicy, 'invalidateCoverageCaches')) { |
| 280 |
throw new RuntimeException('NGramTermCacheReconciler requires a coverage policy with invalidateCoverageCaches().'); |
| 281 |
} |
| 282 |
$coveragePolicy->invalidateCoverageCaches(); |
| 283 |
} |
| 284 |
} |
| 285 |
|
