Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
84.67% |
116 / 137 |
|
36.36% |
4 / 11 |
CRAP | |
0.00% |
0 / 1 |
| WebCrawlCommand | |
84.67% |
116 / 137 |
|
36.36% |
4 / 11 |
61.74 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| configure | |
100.00% |
20 / 20 |
|
100.00% |
1 / 1 |
1 | |||
| downloadFile | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
2 | |||
| removeTempFile | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| getTransformCachePath | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
2 | |||
| updateLastIndexed | |
100.00% |
6 / 6 |
|
100.00% |
1 / 1 |
1 | |||
| readFromTransformCache | |
60.00% |
9 / 15 |
|
0.00% |
0 / 1 |
14.18 | |||
| indexFromTransformCache | |
87.50% |
7 / 8 |
|
0.00% |
0 / 1 |
3.02 | |||
| updateTransformCache | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
6 | |||
| harvestSitemap | |
87.50% |
42 / 48 |
|
0.00% |
0 / 1 |
18.63 | |||
| execute | |
96.55% |
28 / 29 |
|
0.00% |
0 / 1 |
13 | |||
| 1 | <?php |
| 2 | |
| 3 | /** |
| 4 | * Console command: web crawler |
| 5 | * |
| 6 | * PHP version 8 |
| 7 | * |
| 8 | * Copyright (C) Villanova University 2020. |
| 9 | * |
| 10 | * This program is free software; you can redistribute it and/or modify |
| 11 | * it under the terms of the GNU General Public License version 2, |
| 12 | * as published by the Free Software Foundation. |
| 13 | * |
| 14 | * This program is distributed in the hope that it will be useful, |
| 15 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 17 | * GNU General Public License for more details. |
| 18 | * |
| 19 | * You should have received a copy of the GNU General Public License |
| 20 | * along with this program; if not, write to the Free Software |
| 21 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 22 | * |
| 23 | * @category VuFind |
| 24 | * @package Console |
| 25 | * @author Demian Katz <demian.katz@villanova.edu> |
| 26 | * @license http://opensource.org/licenses/gpl-2.0.php GNU General Public License |
| 27 | * @link https://vufind.org/wiki/development Wiki |
| 28 | */ |
| 29 | |
| 30 | namespace VuFindConsole\Command\Import; |
| 31 | |
| 32 | use Laminas\Config\Config; |
| 33 | use Symfony\Component\Console\Attribute\AsCommand; |
| 34 | use Symfony\Component\Console\Command\Command; |
| 35 | use Symfony\Component\Console\Input\InputInterface; |
| 36 | use Symfony\Component\Console\Input\InputOption; |
| 37 | use Symfony\Component\Console\Output\OutputInterface; |
| 38 | use VuFind\Solr\Writer; |
| 39 | use VuFind\XSLT\Importer; |
| 40 | use VuFindSearch\Backend\Solr\Document\RawXMLDocument; |
| 41 | |
| 42 | use function is_string; |
| 43 | |
| 44 | /** |
| 45 | * Console command: web crawler |
| 46 | * |
| 47 | * @category VuFind |
| 48 | * @package Console |
| 49 | * @author Demian Katz <demian.katz@villanova.edu> |
| 50 | * @license http://opensource.org/licenses/gpl-2.0.php GNU General Public License |
| 51 | * @link https://vufind.org/wiki/development Wiki |
| 52 | */ |
| 53 | #[AsCommand( |
| 54 | name: 'import/webcrawl', |
| 55 | description: 'Web crawler' |
| 56 | )] |
| 57 | class WebCrawlCommand extends Command |
| 58 | { |
| 59 | /** |
| 60 | * Should we bypass cache expiration? |
| 61 | * |
| 62 | * @var bool |
| 63 | */ |
| 64 | protected bool $bypassCacheExpiration = false; |
| 65 | |
| 66 | /** |
| 67 | * Constructor |
| 68 | * |
| 69 | * @param Importer $importer XSLT importer |
| 70 | * @param Writer $solr Solr writer |
| 71 | * @param Config $config Configuration from webcrawl.ini |
| 72 | * @param string|null $name The name of the command; passing null means it |
| 73 | * must be set in configure() |
| 74 | */ |
| 75 | public function __construct( |
| 76 | protected Importer $importer, |
| 77 | protected Writer $solr, |
| 78 | protected Config $config, |
| 79 | $name = null |
| 80 | ) { |
| 81 | parent::__construct($name); |
| 82 | } |
| 83 | |
| 84 | /** |
| 85 | * Configure the command. |
| 86 | * |
| 87 | * @return void |
| 88 | */ |
| 89 | protected function configure() |
| 90 | { |
| 91 | $this |
| 92 | ->setHelp('Crawls websites to populate VuFind\'s web index.') |
| 93 | ->addOption( |
| 94 | 'test-only', |
| 95 | null, |
| 96 | InputOption::VALUE_NONE, |
| 97 | 'activates test mode, which displays output without updating Solr' |
| 98 | )->addOption( |
| 99 | 'use-expired-cache', |
| 100 | null, |
| 101 | InputOption::VALUE_NONE, |
| 102 | 'use cached data, even if expired; useful when the index needs to be quickly rebuilt, ' |
| 103 | . 'e.g. after a Solr upgrade' |
| 104 | )->addOption( |
| 105 | 'index', |
| 106 | null, |
| 107 | InputOption::VALUE_OPTIONAL, |
| 108 | 'name of search backend to index content into', |
| 109 | 'SolrWeb' |
| 110 | ); |
| 111 | } |
| 112 | |
| 113 | /** |
| 114 | * Download a URL to a temporary file. |
| 115 | * |
| 116 | * @param string $url URL to download |
| 117 | * |
| 118 | * @return string Filename of downloaded content |
| 119 | */ |
| 120 | protected function downloadFile($url) |
| 121 | { |
| 122 | $file = tempnam('/tmp', 'sitemap'); |
| 123 | file_put_contents($file, file_get_contents($url)); |
| 124 | return $file; |
| 125 | } |
| 126 | |
| 127 | /** |
| 128 | * Remove a temporary file. |
| 129 | * |
| 130 | * @param string $file Name of file to delete |
| 131 | * |
| 132 | * @return void |
| 133 | */ |
| 134 | protected function removeTempFile($file) |
| 135 | { |
| 136 | unlink($file); |
| 137 | } |
| 138 | |
| 139 | /** |
| 140 | * Given a URL, get the transform cache path (or null if the cache |
| 141 | * is disabled). |
| 142 | * |
| 143 | * @param string $url URL to cache |
| 144 | * |
| 145 | * @return ?string |
| 146 | */ |
| 147 | protected function getTransformCachePath(string $url): ?string |
| 148 | { |
| 149 | if ($dir = $this->config->Cache->transform_cache_dir ?? null) { |
| 150 | return $dir . '/' . md5($url); |
| 151 | } |
| 152 | return null; |
| 153 | } |
| 154 | |
| 155 | /** |
| 156 | * Update the last_indexed dates in a cached XML document to the current |
| 157 | * time so reindexing cached documents works correctly. |
| 158 | * |
| 159 | * @param string $xml XML to update |
| 160 | * |
| 161 | * @return string |
| 162 | */ |
| 163 | protected function updateLastIndexed(string $xml): string |
| 164 | { |
| 165 | $newDate = date('Y-m-d\TH:i:s\Z'); |
| 166 | return preg_replace( |
| 167 | '|<field name="last_indexed">([^<]+)</field>|', |
| 168 | '<field name="last_indexed">' . $newDate . '</field>', |
| 169 | $xml |
| 170 | ); |
| 171 | } |
| 172 | |
| 173 | /** |
| 174 | * Fetch transform cache data for the specified URL; return null if the cache is disabled, |
| 175 | * the data is expired, or something goes wrong. |
| 176 | * |
| 177 | * @param OutputInterface $output Output object |
| 178 | * @param string $url URL of sitemap to read. |
| 179 | * @param string $lastMod Last modification date of URL. |
| 180 | * @param bool $verbose Are we in verbose mode? |
| 181 | * |
| 182 | * @return ?string |
| 183 | */ |
| 184 | protected function readFromTransformCache( |
| 185 | OutputInterface $output, |
| 186 | string $url, |
| 187 | string $lastMod, |
| 188 | bool $verbose |
| 189 | ): ?string { |
| 190 | // If cache is write-only, don't retrieve data! |
| 191 | if ($this->config->Cache->transform_cache_write_only ?? false) { |
| 192 | return null; |
| 193 | } |
| 194 | // If we can't find the data in the cache, we can't proceed. |
| 195 | if (!($path = $this->getTransformCachePath($url)) || !file_exists($path)) { |
| 196 | return null; |
| 197 | } |
| 198 | if (strtotime($lastMod) > filemtime($path) && !$this->bypassCacheExpiration) { |
| 199 | if ($verbose) { |
| 200 | $output->writeln("Cached data for $url ($path) has expired."); |
| 201 | } |
| 202 | return null; |
| 203 | } |
| 204 | $rawXml = file_get_contents($path); |
| 205 | if (!is_string($rawXml)) { |
| 206 | $output->writeln("WARNING: Problem reading cached data for $url ($path)"); |
| 207 | return null; |
| 208 | } |
| 209 | if ($verbose) { |
| 210 | $output->writeln("Found $url in cache: $path"); |
| 211 | } |
| 212 | return $rawXml; |
| 213 | } |
| 214 | |
| 215 | /** |
| 216 | * Check the cache and configuration to see if the provided URL can |
| 217 | * be loaded from cache, and load it to Solr if possible. |
| 218 | * |
| 219 | * @param OutputInterface $output Output object |
| 220 | * @param string $url URL of sitemap to read. |
| 221 | * @param string $lastMod Last modification date of URL. |
| 222 | * @param bool $verbose Are we in verbose mode? |
| 223 | * @param string $index Solr index to update |
| 224 | * @param bool $testMode Are we in test mode? |
| 225 | * |
| 226 | * @return bool True if loaded from cache, false if not. |
| 227 | */ |
| 228 | protected function indexFromTransformCache( |
| 229 | OutputInterface $output, |
| 230 | string $url, |
| 231 | string $lastMod, |
| 232 | bool $verbose = false, |
| 233 | string $index = 'SolrWeb', |
| 234 | bool $testMode = false |
| 235 | ): bool { |
| 236 | $rawXml = $this->readFromTransformCache($output, $url, $lastMod, $verbose); |
| 237 | if ($rawXml === null) { |
| 238 | return false; |
| 239 | } |
| 240 | $xml = $this->updateLastIndexed($rawXml); |
| 241 | if ($testMode) { |
| 242 | $output->writeln($xml); |
| 243 | } else { |
| 244 | $this->solr->save($index, new RawXMLDocument($xml)); |
| 245 | } |
| 246 | return true; |
| 247 | } |
| 248 | |
| 249 | /** |
| 250 | * Update the transform cache (if activated). Returns true if the cache was updated, |
| 251 | * false otherwise. |
| 252 | * |
| 253 | * @param string $url URL to use for cache key |
| 254 | * @param string $result Result of transforming the URL |
| 255 | * |
| 256 | * @return bool |
| 257 | */ |
| 258 | protected function updateTransformCache(string $url, string $result): bool |
| 259 | { |
| 260 | if ($transformCachePath = $this->getTransformCachePath($url)) { |
| 261 | return false !== file_put_contents($transformCachePath, $result); |
| 262 | } |
| 263 | return false; |
| 264 | } |
| 265 | |
| 266 | /** |
| 267 | * Process a sitemap URL, either harvesting its contents directly or recursively |
| 268 | * reading in child sitemaps. |
| 269 | * |
| 270 | * @param OutputInterface $output Output object |
| 271 | * @param string $url URL of sitemap to read. |
| 272 | * @param bool $verbose Are we in verbose mode? |
| 273 | * @param string $index Solr index to update |
| 274 | * @param bool $testMode Are we in test mode? |
| 275 | * |
| 276 | * @return bool True on success, false on error. |
| 277 | */ |
| 278 | protected function harvestSitemap( |
| 279 | OutputInterface $output, |
| 280 | $url, |
| 281 | $verbose = false, |
| 282 | $index = 'SolrWeb', |
| 283 | $testMode = false |
| 284 | ) { |
| 285 | // Date to use as a default "last modification" date in scenarios where we |
| 286 | // don't care about cache invalidation. |
| 287 | $pastDate = '1980-01-01'; |
| 288 | |
| 289 | // If we're not concerned about cache expiration, we can potentially |
| 290 | // short-circuit the process with the cache up front. Otherwise, we'll |
| 291 | // need to wait until we can get last modification dates to know whether |
| 292 | // it's safe to rely on cached data. |
| 293 | if ( |
| 294 | $this->bypassCacheExpiration |
| 295 | && $this->indexFromTransformCache($output, $url, $pastDate, $verbose, $index, $testMode) |
| 296 | ) { |
| 297 | return true; |
| 298 | } |
| 299 | |
| 300 | if ($verbose) { |
| 301 | $output->writeln("Harvesting $url..."); |
| 302 | } |
| 303 | |
| 304 | $retVal = true; |
| 305 | |
| 306 | $file = $this->downloadFile($url); |
| 307 | $xml = simplexml_load_file($file); |
| 308 | if ($xml) { |
| 309 | // Are there any child sitemaps? If so, pull them in: |
| 310 | $results = $xml->sitemap ?? []; |
| 311 | foreach ($results as $current) { |
| 312 | if (isset($current->loc)) { |
| 313 | // If there's a last modification date (or we're forcing a |
| 314 | // reindex from the cache) and we can retrieve data from the |
| 315 | // cache, we can bypass the harvest. |
| 316 | if ( |
| 317 | (!isset($current->lastmod) && !$this->bypassCacheExpiration) |
| 318 | || !$this->indexFromTransformCache( |
| 319 | $output, |
| 320 | (string)$current->loc, |
| 321 | (string)($current->lastmod ?? $pastDate), |
| 322 | $verbose, |
| 323 | $index, |
| 324 | $testMode |
| 325 | ) |
| 326 | ) { |
| 327 | $success = $this->harvestSitemap( |
| 328 | $output, |
| 329 | (string)$current->loc, |
| 330 | $verbose, |
| 331 | $index, |
| 332 | $testMode |
| 333 | ); |
| 334 | if (!$success) { |
| 335 | $retVal = false; |
| 336 | } |
| 337 | } |
| 338 | } |
| 339 | } |
| 340 | // Only import the current sitemap if it contains URLs! |
| 341 | if (isset($xml->url)) { |
| 342 | try { |
| 343 | $result = $this->importer->save( |
| 344 | $file, |
| 345 | 'sitemap.properties', |
| 346 | $index, |
| 347 | $testMode |
| 348 | ); |
| 349 | if ($result && $this->updateTransformCache($url, $result) && $verbose) { |
| 350 | $output->writeln('Wrote results to transform cache.'); |
| 351 | } |
| 352 | if ($testMode) { |
| 353 | $output->writeln($result); |
| 354 | } |
| 355 | } catch (\Exception $e) { |
| 356 | if ($verbose) { |
| 357 | $output->writeln($e::class . ': ' . $e->getMessage()); |
| 358 | } |
| 359 | $retVal = false; |
| 360 | } |
| 361 | } |
| 362 | } |
| 363 | $this->removeTempFile($file); |
| 364 | return $retVal; |
| 365 | } |
| 366 | |
| 367 | /** |
| 368 | * Run the command. |
| 369 | * |
| 370 | * @param InputInterface $input Input object |
| 371 | * @param OutputInterface $output Output object |
| 372 | * |
| 373 | * @return int 0 for success |
| 374 | */ |
| 375 | protected function execute(InputInterface $input, OutputInterface $output) |
| 376 | { |
| 377 | // Get command line parameters: |
| 378 | $testMode = $input->getOption('test-only') ? true : false; |
| 379 | $this->bypassCacheExpiration = $input->getOption('use-expired-cache') ? true : false; |
| 380 | $index = $input->getOption('index'); |
| 381 | |
| 382 | // Get the time we started indexing -- we'll delete records older than this |
| 383 | // date after everything is finished. Note that we subtract a few seconds |
| 384 | // for safety. |
| 385 | $startTime = date('Y-m-d\TH:i:s\Z', time() - 5); |
| 386 | |
| 387 | // Are we in verbose mode? |
| 388 | $verbose = ($this->config->General->verbose ?? false) |
| 389 | || ($input->hasOption('verbose') && $input->getOption('verbose')); |
| 390 | |
| 391 | // Loop through sitemap URLs in the config file. |
| 392 | $error = false; |
| 393 | foreach ($this->config->Sitemaps->url as $current) { |
| 394 | $error = $error || !$this->harvestSitemap( |
| 395 | $output, |
| 396 | $current, |
| 397 | $verbose, |
| 398 | $index, |
| 399 | $testMode |
| 400 | ); |
| 401 | } |
| 402 | if ($error) { |
| 403 | $output->writeln('Error encountered during harvest.'); |
| 404 | } |
| 405 | |
| 406 | // Skip Solr operations if we're in test mode. |
| 407 | if (!$testMode) { |
| 408 | if ($verbose) { |
| 409 | $output->writeln("Deleting old records (prior to $startTime)..."); |
| 410 | } |
| 411 | // Perform the delete of outdated records: |
| 412 | $this->solr |
| 413 | ->deleteByQuery($index, 'last_indexed:[* TO ' . $startTime . ']'); |
| 414 | if ($verbose) { |
| 415 | $output->writeln('Committing...'); |
| 416 | } |
| 417 | $this->solr->commit($index); |
| 418 | if ($verbose) { |
| 419 | $output->writeln('Optimizing...'); |
| 420 | } |
| 421 | $this->solr->optimize($index); |
| 422 | } |
| 423 | return $error ? 1 : 0; |
| 424 | } |
| 425 | } |