| 1 |
<?php |
| 2 |
|
| 3 |
if (!defined('ABSPATH')) { |
| 4 |
exit; |
| 5 |
} |
| 6 |
|
| 7 |
/** |
| 8 |
* Reads observed same-site referrer evidence for the visible captured 404 rows. |
| 9 |
* |
| 10 |
* The repository is intentionally read-side only: it does not scan content, crawl |
| 11 |
* pages, or add write-path cost to 404 logging. Callers pass the captured URLs |
| 12 |
* already visible on the current table page, and this class bounds the logsv2 |
| 13 |
* aggregate to those URLs. |
| 14 |
*/ |
| 15 |
class ABJ_404_Solution_InternalSourceEvidenceRepository { |
| 16 |
|
| 17 |
/** |
| 18 |
* Hard upper bound on aggregate rows returned per queryAggregateRows() |
| 19 |
* call. Referrer cardinality for a captured URL is visitor-supplied and |
| 20 |
* unbounded (bots and open redirects can drive arbitrarily many distinct |
| 21 |
* Referer values at one captured URL); this caps the SQL-level result |
| 22 |
* set so the query itself cannot pull unbounded rows into memory before |
| 23 |
* the PHP-side $maxSources slice in getEvidenceForCapturedUrls() runs. |
| 24 |
* Matches ABJ_404_Solution_ContentKeywordsRepository::MAX_LIMIT. |
| 25 |
*/ |
| 26 |
const MAX_AGGREGATE_ROWS = 5000; |
| 27 |
|
| 28 |
/** |
| 29 |
* Hard upper bound on how many caller-supplied captured URLs this |
| 30 |
* repository will accept per call. $capturedUrls is meant to be "URLs |
| 31 |
* visible on the current admin table page" (well under 100 in practice, |
| 32 |
* matching RestApiRequestParser's per_page cap of 100), but the method |
| 33 |
* is public and does not otherwise validate cardinality; every accepted |
| 34 |
* URL drives a SQL IN() placeholder and bound parameter in |
| 35 |
* queryAggregateRows(), so an unbounded caller-supplied array would |
| 36 |
* still cost memory and query-planning work before MAX_AGGREGATE_ROWS' |
| 37 |
* LIMIT ever applies. Extra accepted URLs beyond this bound are dropped, |
| 38 |
* not errored, consistent with this repository degrading gracefully |
| 39 |
* rather than failing the admin table render. |
| 40 |
*/ |
| 41 |
const MAX_CAPTURED_URLS = 200; |
| 42 |
|
| 43 |
/** @var ABJ_404_Solution_DatabaseQueryInterface */ |
| 44 |
private $db; |
| 45 |
|
| 46 |
/** @var ABJ_404_Solution_Functions */ |
| 47 |
private $functions; |
| 48 |
|
| 49 |
/** |
| 50 |
* Same-site referrer host, in lowercase. Resolved once at construction |
| 51 |
* time rather than lazily memoized on first use: home_url() does not |
| 52 |
* change mid-request, so there is no benefit to deferring the read, and |
| 53 |
* resolving it eagerly keeps this a plain immutable value instead of a |
| 54 |
* query-shaped accessor with a mutation side effect (CQS violation -- |
| 55 |
* see homeHost() removal, c308). |
| 56 |
* |
| 57 |
* @var string |
| 58 |
*/ |
| 59 |
private $homeHost; |
| 60 |
|
| 61 |
/** |
| 62 |
* @param ABJ_404_Solution_DatabaseQueryInterface $db |
| 63 |
* @param ABJ_404_Solution_Functions|null $functions UTF-8 sanitizer source. |
| 64 |
* Defaults to the `functions` service so every call site (including |
| 65 |
* tests that omit this argument) still gets real UTF-8 sanitization |
| 66 |
* before values reach $wpdb->prepare() -- see queryAggregateRows() |
| 67 |
* and resolvePostIdFromPermalinkCache(), which both take |
| 68 |
* visitor-supplied captured 404 URLs. prepare() escapes quote and |
| 69 |
* percent characters but does not validate or repair encoding, so |
| 70 |
* sanitization must still happen before values become bound params. |
| 71 |
*/ |
| 72 |
public function __construct(ABJ_404_Solution_DatabaseQueryInterface $db, $functions = null) { |
| 73 |
$this->db = $db; |
| 74 |
$this->functions = $functions !== null ? $functions : abj_service('functions'); |
| 75 |
$this->homeHost = $this->resolveHomeHost(); |
| 76 |
} |
| 77 |
|
| 78 |
/** |
| 79 |
* Return source evidence keyed by captured requested URL. |
| 80 |
* |
| 81 |
* source_count is the number of distinct same-site source paths OBSERVED |
| 82 |
* for that captured URL. When the underlying aggregate saturates its |
| 83 |
* MAX_AGGREGATE_ROWS cap it is a floor rather than a total; that condition |
| 84 |
* is warned about in queryAggregateRows() so it is never silent. |
| 85 |
* |
| 86 |
* @param array<int, string> $capturedUrls Visible captured URLs only. |
| 87 |
* @param int $maxSources Maximum source rows to display per captured URL. |
| 88 |
* @return array<string, array{source_count:int,displayed_source_count:int,sources:array<int,array<string,mixed>>}> |
| 89 |
*/ |
| 90 |
public function getEvidenceForCapturedUrls(array $capturedUrls, int $maxSources = 5): array { |
| 91 |
$visibleUrls = $this->visibleUrlSet($capturedUrls); |
| 92 |
if (empty($visibleUrls)) { |
| 93 |
return array(); |
| 94 |
} |
| 95 |
|
| 96 |
$aggregateRows = $this->queryAggregateRows(array_keys($visibleUrls)); |
| 97 |
$grouped = $this->groupRowsByCapturedUrl($aggregateRows, $visibleUrls); |
| 98 |
|
| 99 |
$evidence = array(); |
| 100 |
foreach ($grouped as $capturedUrl => $sourcesByPath) { |
| 101 |
uasort($sourcesByPath, function (array $a, array $b): int { |
| 102 |
$hitsCompare = $this->intField($b, 'hit_count') <=> $this->intField($a, 'hit_count'); |
| 103 |
if ($hitsCompare !== 0) { |
| 104 |
return $hitsCompare; |
| 105 |
} |
| 106 |
return $this->intField($b, 'last_seen') <=> $this->intField($a, 'last_seen'); |
| 107 |
}); |
| 108 |
|
| 109 |
$sourceCount = count($sourcesByPath); |
| 110 |
if ($sourceCount === 0) { |
| 111 |
continue; |
| 112 |
} |
| 113 |
|
| 114 |
$displayed = array_slice(array_values($sourcesByPath), 0, max(1, $maxSources)); |
| 115 |
$evidence[$capturedUrl] = array( |
| 116 |
'source_count' => $sourceCount, |
| 117 |
'displayed_source_count' => count($displayed), |
| 118 |
'sources' => $displayed, |
| 119 |
); |
| 120 |
} |
| 121 |
|
| 122 |
return $evidence; |
| 123 |
} |
| 124 |
|
| 125 |
/** |
| 126 |
* @param array<int, string> $capturedUrls |
| 127 |
* @return array<string, bool> |
| 128 |
*/ |
| 129 |
private function visibleUrlSet(array $capturedUrls): array { |
| 130 |
$set = array(); |
| 131 |
foreach ($capturedUrls as $url) { |
| 132 |
if (!is_string($url) || $url === '') { |
| 133 |
continue; |
| 134 |
} |
| 135 |
// Same per-URL length bound the capture path itself enforces |
| 136 |
// (UserRequest::getPath(), SettingsRedirectPolicy, etc.). This |
| 137 |
// repository is a separate trust boundary from whatever called |
| 138 |
// it with $capturedUrls, so it must not assume the caller |
| 139 |
// already enforced the invariant: MAX_CAPTURED_URLS above only |
| 140 |
// bounds the count, not the size of each accepted string, and |
| 141 |
// every accepted URL is later bound into a SQL IN() clause. |
| 142 |
if (defined('ABJ404_MAX_URL_LENGTH') && strlen($url) > ABJ404_MAX_URL_LENGTH) { |
| 143 |
continue; |
| 144 |
} |
| 145 |
$set[$url] = true; |
| 146 |
if (count($set) >= self::MAX_CAPTURED_URLS) { |
| 147 |
break; |
| 148 |
} |
| 149 |
} |
| 150 |
return $set; |
| 151 |
} |
| 152 |
|
| 153 |
/** |
| 154 |
* @param array<int, string> $visibleUrls |
| 155 |
* @return array<int, array<string, mixed>> |
| 156 |
*/ |
| 157 |
private function queryAggregateRows(array $visibleUrls): array { |
| 158 |
$cleanUrls = array(); |
| 159 |
foreach ($visibleUrls as $url) { |
| 160 |
// Captured 404 URLs are visitor-supplied (bots routinely deliver |
| 161 |
// garbage bytes through the request path); sanitize invalid |
| 162 |
// UTF-8 before it reaches $wpdb->prepare(), which does not |
| 163 |
// validate encoding on its own. |
| 164 |
$cleanUrls[] = $this->functions->sanitizeInvalidUTF8($url); |
| 165 |
} |
| 166 |
if (empty($cleanUrls)) { |
| 167 |
return array(); |
| 168 |
} |
| 169 |
|
| 170 |
$placeholders = implode(',', array_fill(0, count($cleanUrls), '%s')); |
| 171 |
$query = "SELECT requested_url, referrer, COUNT(*) AS hit_count, MAX(timestamp) AS last_seen" |
| 172 |
. " FROM {wp_abj404_logsv2}" |
| 173 |
. " WHERE requested_url IN (" . $placeholders . ")" |
| 174 |
. " AND referrer IS NOT NULL AND referrer != ''" |
| 175 |
. " GROUP BY requested_url, referrer" |
| 176 |
. " ORDER BY requested_url ASC, hit_count DESC, last_seen DESC" |
| 177 |
. " LIMIT " . self::MAX_AGGREGATE_ROWS; |
| 178 |
|
| 179 |
$result = $this->db->queryAndGetResults($query, array('query_params' => $cleanUrls)); |
| 180 |
$rows = is_array($result['rows'] ?? null) ? $result['rows'] : array(); |
| 181 |
|
| 182 |
// A saturated result set means the LIMIT truncated the aggregate, so |
| 183 |
// every count derived from these rows is a floor rather than a total. |
| 184 |
// Say so: a capped result that reads as a complete one is how a |
| 185 |
// backlog stays invisible (see the n-gram reconciler's post-LIMIT |
| 186 |
// "Found 50 posts missing" line, i582). |
| 187 |
if (count($rows) >= self::MAX_AGGREGATE_ROWS) { |
| 188 |
$this->logger()->warn(sprintf( |
| 189 |
'InternalSourceEvidenceRepository: source-evidence aggregate hit its %d-row cap for %d captured URLs. ' |
| 190 |
. 'Reported source counts are floors, not totals, and trailing captured URLs may show no sources.', |
| 191 |
self::MAX_AGGREGATE_ROWS, |
| 192 |
count($cleanUrls) |
| 193 |
)); |
| 194 |
} |
| 195 |
|
| 196 |
$typedRows = array(); |
| 197 |
foreach ($rows as $row) { |
| 198 |
if (is_array($row)) { |
| 199 |
$typedRows[] = $this->stringKeyedRow($row); |
| 200 |
} |
| 201 |
} |
| 202 |
return $typedRows; |
| 203 |
} |
| 204 |
|
| 205 |
/** @return ABJ_404_Solution_Logging */ |
| 206 |
private function logger() { |
| 207 |
return abj_service('logging'); |
| 208 |
} |
| 209 |
|
| 210 |
/** |
| 211 |
* @param array<int, array<string, mixed>> $rows |
| 212 |
* @param array<string, bool> $visibleUrls |
| 213 |
* @return array<string, array<string, array<string, mixed>>> |
| 214 |
*/ |
| 215 |
private function groupRowsByCapturedUrl(array $rows, array $visibleUrls): array { |
| 216 |
$grouped = array(); |
| 217 |
// Call-scoped memo, not instance state: resolveSource() does a |
| 218 |
// url_to_postid()/permalink-cache lookup per unique source path, and |
| 219 |
// the same source path commonly repeats across rows within one |
| 220 |
// top-level call. Keeping the memo local (rather than an instance |
| 221 |
// property mutated by a "read" method) avoids a command-query |
| 222 |
// separation violation and stale-across-calls results if this |
| 223 |
// repository instance is ever reused for a second page render (see |
| 224 |
// homeHost() eager-resolve fix above for the same class of issue). |
| 225 |
$resolvedSources = array(); |
| 226 |
foreach ($rows as $row) { |
| 227 |
if (!is_array($row)) { |
| 228 |
continue; |
| 229 |
} |
| 230 |
|
| 231 |
$capturedUrl = $this->stringField($row, 'requested_url'); |
| 232 |
if ($capturedUrl === '' || !isset($visibleUrls[$capturedUrl])) { |
| 233 |
continue; |
| 234 |
} |
| 235 |
|
| 236 |
$sourcePath = $this->normalizeSameSiteReferrer($this->stringField($row, 'referrer')); |
| 237 |
if ($sourcePath === null) { |
| 238 |
continue; |
| 239 |
} |
| 240 |
|
| 241 |
if (!isset($grouped[$capturedUrl][$sourcePath])) { |
| 242 |
$grouped[$capturedUrl][$sourcePath] = $this->resolveSource($sourcePath, $resolvedSources); |
| 243 |
} |
| 244 |
|
| 245 |
$currentSource = $grouped[$capturedUrl][$sourcePath]; |
| 246 |
$currentSource['hit_count'] = $this->intField($currentSource, 'hit_count') |
| 247 |
+ max(0, $this->intField($row, 'hit_count')); |
| 248 |
$currentSource['last_seen'] = max( |
| 249 |
$this->intField($currentSource, 'last_seen'), |
| 250 |
$this->intField($row, 'last_seen') |
| 251 |
); |
| 252 |
$grouped[$capturedUrl][$sourcePath] = $currentSource; |
| 253 |
} |
| 254 |
return $grouped; |
| 255 |
} |
| 256 |
|
| 257 |
private function normalizeSameSiteReferrer(string $referrer): ?string { |
| 258 |
$referrer = trim($referrer); |
| 259 |
if ($referrer === '') { |
| 260 |
return null; |
| 261 |
} |
| 262 |
|
| 263 |
if (strpos($referrer, '/') === 0 && strpos($referrer, '//') !== 0) { |
| 264 |
$path = parse_url($referrer, PHP_URL_PATH); |
| 265 |
} else { |
| 266 |
$parts = parse_url($referrer); |
| 267 |
if (!is_array($parts)) { |
| 268 |
return null; |
| 269 |
} |
| 270 |
$host = isset($parts['host']) && is_string($parts['host']) ? strtolower($parts['host']) : ''; |
| 271 |
if ($host === '' || $host !== $this->homeHost) { |
| 272 |
return null; |
| 273 |
} |
| 274 |
$path = isset($parts['path']) && is_string($parts['path']) ? $parts['path'] : ''; |
| 275 |
} |
| 276 |
|
| 277 |
$path = is_string($path) ? '/' . ltrim($path, '/') : ''; |
| 278 |
if ($path === '' || $path === '/') { |
| 279 |
return null; |
| 280 |
} |
| 281 |
if ($this->isExcludedPath($path)) { |
| 282 |
return null; |
| 283 |
} |
| 284 |
|
| 285 |
return $path; |
| 286 |
} |
| 287 |
|
| 288 |
private function resolveHomeHost(): string { |
| 289 |
$home = function_exists('home_url') ? (string)home_url('/') : ''; |
| 290 |
$host = parse_url($home, PHP_URL_HOST); |
| 291 |
return is_string($host) ? strtolower($host) : ''; |
| 292 |
} |
| 293 |
|
| 294 |
private function isExcludedPath(string $path): bool { |
| 295 |
$lower = strtolower($path); |
| 296 |
foreach (array('/wp-admin', '/wp-login.php', '/wp-json', '/wp-content', '/wp-includes') as $prefix) { |
| 297 |
if ($lower === $prefix || strpos($lower, $prefix . '/') === 0) { |
| 298 |
return true; |
| 299 |
} |
| 300 |
} |
| 301 |
|
| 302 |
return preg_match('/\.(css|js|map|json|xml|jpg|jpeg|png|gif|webp|svg|ico|pdf|zip|woff|woff2|ttf|eot)$/i', $lower) === 1; |
| 303 |
} |
| 304 |
|
| 305 |
/** |
| 306 |
* Resolves raw post identity data for a same-site referrer path only: |
| 307 |
* post_id and post_title. Authorization (current_user_can('edit_post')) |
| 308 |
* and the edit_url presentation link are NOT this repository's concern |
| 309 |
* -- a data-access repository must not make authorization decisions or |
| 310 |
* build admin-link HTML (CLAUDE.md "Strict layer separation"). The |
| 311 |
* caller that renders source-evidence rows |
| 312 |
* (ABJ_404_Solution_CapturedSourceEvidenceRenderer::editLinkHtml()) owns |
| 313 |
* that decision, using the post_id returned here (c308). |
| 314 |
* |
| 315 |
* @param array<string, array<string, mixed>> $resolvedSources Call-scoped |
| 316 |
* memo, keyed and updated by reference so repeat source paths within |
| 317 |
* the same top-level call skip the post-id lookup. See |
| 318 |
* groupRowsByCapturedUrl() for why this is a parameter rather than |
| 319 |
* instance state. |
| 320 |
* @return array<string, mixed> |
| 321 |
*/ |
| 322 |
private function resolveSource(string $sourcePath, array &$resolvedSources): array { |
| 323 |
if (isset($resolvedSources[$sourcePath])) { |
| 324 |
return $resolvedSources[$sourcePath]; |
| 325 |
} |
| 326 |
|
| 327 |
$postId = $this->resolvePostId($sourcePath); |
| 328 |
$title = $postId > 0 && function_exists('get_the_title') ? (string)get_the_title($postId) : ''; |
| 329 |
|
| 330 |
$resolvedSources[$sourcePath] = array( |
| 331 |
'referrer_url' => $sourcePath, |
| 332 |
'post_id' => $postId, |
| 333 |
'post_title' => $title, |
| 334 |
'hit_count' => 0, |
| 335 |
'last_seen' => 0, |
| 336 |
); |
| 337 |
return $resolvedSources[$sourcePath]; |
| 338 |
} |
| 339 |
|
| 340 |
private function resolvePostId(string $sourcePath): int { |
| 341 |
if (function_exists('url_to_postid') && function_exists('home_url')) { |
| 342 |
$postId = (int)url_to_postid(home_url($sourcePath)); |
| 343 |
if ($postId > 0) { |
| 344 |
return $postId; |
| 345 |
} |
| 346 |
} |
| 347 |
|
| 348 |
return $this->resolvePostIdFromPermalinkCache($sourcePath); |
| 349 |
} |
| 350 |
|
| 351 |
private function resolvePostIdFromPermalinkCache(string $sourcePath): int { |
| 352 |
$trimmed = trim($sourcePath, '/'); |
| 353 |
$variants = array_values(array_unique(array($sourcePath, '/' . $trimmed, $trimmed, '/' . $trimmed . '/'))); |
| 354 |
$cleanVariants = array(); |
| 355 |
foreach ($variants as $variant) { |
| 356 |
if ($variant === '') { |
| 357 |
continue; |
| 358 |
} |
| 359 |
// $sourcePath is derived from the HTTP Referer header (see |
| 360 |
// normalizeSameSiteReferrer()), which is visitor-supplied and can |
| 361 |
// carry invalid UTF-8 byte sequences; sanitize before |
| 362 |
// $wpdb->prepare(). |
| 363 |
$cleanVariants[] = $this->functions->sanitizeInvalidUTF8($variant); |
| 364 |
} |
| 365 |
if (empty($cleanVariants)) { |
| 366 |
return 0; |
| 367 |
} |
| 368 |
|
| 369 |
$placeholders = implode(',', array_fill(0, count($cleanVariants), '%s')); |
| 370 |
// allow-unbounded-select: literal "LIMIT 1" lands in a separate concatenated string than SELECT/FROM (audit blind spot #1); $cleanVariants is also a fixed <=4-element set, not visitor-controlled cardinality. |
| 371 |
$query = "SELECT id FROM {wp_abj404_permalink_cache} WHERE url IN (" . $placeholders . ") LIMIT 1"; |
| 372 |
$result = $this->db->queryAndGetResults($query, array('query_params' => $cleanVariants)); |
| 373 |
$rows = is_array($result['rows'] ?? null) ? $result['rows'] : array(); |
| 374 |
$row = is_array($rows[0] ?? null) ? $rows[0] : array(); |
| 375 |
return $this->intField($row, 'id'); |
| 376 |
} |
| 377 |
|
| 378 |
/** @param array<string, mixed> $row */ |
| 379 |
private function stringField(array $row, string $key): string { |
| 380 |
return isset($row[$key]) && is_scalar($row[$key]) ? (string)$row[$key] : ''; |
| 381 |
} |
| 382 |
|
| 383 |
/** @param array<string, mixed> $row */ |
| 384 |
private function intField(array $row, string $key): int { |
| 385 |
return isset($row[$key]) && is_numeric($row[$key]) ? (int)$row[$key] : 0; |
| 386 |
} |
| 387 |
|
| 388 |
/** |
| 389 |
* @param array<array-key, mixed> $row |
| 390 |
* @return array<string, mixed> |
| 391 |
*/ |
| 392 |
private function stringKeyedRow(array $row): array { |
| 393 |
$stringKeyed = array(); |
| 394 |
foreach ($row as $key => $value) { |
| 395 |
if (is_string($key)) { |
| 396 |
$stringKeyed[$key] = $value; |
| 397 |
} |
| 398 |
} |
| 399 |
return $stringKeyed; |
| 400 |
} |
| 401 |
} |
| 402 |
|
