| 1 |
<?php |
| 2 |
|
| 3 |
if (!defined('ABSPATH')) { |
| 4 |
exit; |
| 5 |
} |
| 6 |
|
| 7 |
/** |
| 8 |
* Extracts N-grams (bigrams + trigrams) from a URL string. |
| 9 |
* |
| 10 |
* Pure computation. No I/O, no DB, no transients. Both production callers |
| 11 |
* (the live 404 candidate-selection pipeline via NGramFilter, and the |
| 12 |
* background term-cache rebuild loops in DatabaseUpgradeNGram) call into |
| 13 |
* this for the same purpose: take a URL, produce a deduplicated set of |
| 14 |
* character N-grams suitable for Dice-coefficient similarity. The only |
| 15 |
* collaborator needed is the polymorphic mbstring adapter (Functions), |
| 16 |
* because URLs may contain UTF-8 (i18n permalink slugs). |
| 17 |
*/ |
| 18 |
class ABJ_404_Solution_NGramExtractor { |
| 19 |
|
| 20 |
/** Cap on URL length before extraction; above this the URL is truncated. |
| 21 |
* Most real URLs are < 200 chars; 500 prevents memory blowup on 2083-char |
| 22 |
* legacy URLs (4000+ N-grams per URL) without losing matching value. */ |
| 23 |
const MAX_URL_LENGTH = 500; |
| 24 |
|
| 25 |
/** @var ABJ_404_Solution_Functions */ |
| 26 |
private $f; |
| 27 |
|
| 28 |
/** @var ABJ_404_Solution_Logging */ |
| 29 |
private $logger; |
| 30 |
|
| 31 |
/** |
| 32 |
* @param ABJ_404_Solution_Functions|null $functions |
| 33 |
* @param ABJ_404_Solution_Logging|null $logging |
| 34 |
*/ |
| 35 |
public function __construct($functions = null, $logging = null) { |
| 36 |
$this->f = $functions !== null ? $functions : abj_service('functions'); |
| 37 |
$this->logger = $logging !== null ? $logging : abj_service('logging'); |
| 38 |
} |
| 39 |
|
| 40 |
/** |
| 41 |
* Extract N-grams from a URL string. |
| 42 |
* |
| 43 |
* Generates both bigrams (n=2) and trigrams (n=3) for optimal accuracy. |
| 44 |
* Research shows using both provides better typo detection than either alone. |
| 45 |
* |
| 46 |
* @param string $url |
| 47 |
* @param array<int, int> $ngramSizes default [2, 3] |
| 48 |
* @return array{bi: array<int, string>, tri: array<int, string>} |
| 49 |
*/ |
| 50 |
public function extractNGrams($url, $ngramSizes = [2, 3]) { |
| 51 |
if (empty($url)) { |
| 52 |
return ['bi' => [], 'tri' => []]; |
| 53 |
} |
| 54 |
|
| 55 |
$url = $this->f->strtolower($url); |
| 56 |
|
| 57 |
$originalLength = $this->f->strlen($url); |
| 58 |
if ($originalLength > self::MAX_URL_LENGTH) { |
| 59 |
$this->logger->infoMessage("WARNING: URL too long for N-gram extraction: {$originalLength} chars, truncating to " . self::MAX_URL_LENGTH . ". URL: " . $this->f->substr($url, 0, 100) . "..."); |
| 60 |
$url = $this->f->substr($url, 0, self::MAX_URL_LENGTH); |
| 61 |
} |
| 62 |
|
| 63 |
$result = []; |
| 64 |
$length = $this->f->strlen($url); |
| 65 |
|
| 66 |
foreach ($ngramSizes as $n) { |
| 67 |
$ngrams = []; |
| 68 |
for ($i = 0; $i <= $length - $n; $i++) { |
| 69 |
$ngram = $this->f->substr($url, $i, $n); |
| 70 |
$ngrams[$ngram] = true; |
| 71 |
} |
| 72 |
$key = ($n == 2) ? 'bi' : 'tri'; |
| 73 |
// Convert keys to strings to prevent PHP from converting numeric strings to integers |
| 74 |
$result[$key] = array_map('strval', array_keys($ngrams)); |
| 75 |
} |
| 76 |
|
| 77 |
/** @var array{bi: array<int, string>, tri: array<int, string>} $result */ |
| 78 |
return $result; |
| 79 |
} |
| 80 |
} |
| 81 |
|
