Recursive fetch & parse
This commit is contained in:
parent
fafb54d293
commit
eb9e9adf96
|
@ -1,16 +1,32 @@
|
|||
<?php
|
||||
|
||||
declare(strict_types = 1);
|
||||
|
||||
// Some dependencies haven't been updated for PHP 8
|
||||
// This can probably be removed for PHP 7.3
|
||||
error_reporting(E_ALL & ~E_DEPRECATED);
|
||||
|
||||
require('vendor/autoload.php');
|
||||
|
||||
use GuzzleHttp\Client as HTTPClient;
|
||||
use PHPHtmlParser\Dom as DOM;
|
||||
|
||||
class Log {
|
||||
private static $enabled = [
|
||||
'silly' => false,
|
||||
'debug' => true,
|
||||
'info' => true
|
||||
];
|
||||
|
||||
private static function emit(string $level, $message) {
|
||||
if (!Log::$enabled[$level]) {
|
||||
return;
|
||||
};
|
||||
$now = date("c");
|
||||
print("$now [$level] $message");
|
||||
print("$now [$level] $message\n");
|
||||
}
|
||||
|
||||
public static function silly($message) {
|
||||
Log::emit('silly', $message);
|
||||
}
|
||||
|
||||
public static function debug($message) {
|
||||
|
@ -28,6 +44,9 @@ function fetchUrl(string $url) {
|
|||
$res = $client->request(
|
||||
'GET', $url, [
|
||||
'http_errors' => false,
|
||||
'headers' => [
|
||||
'User-Agent' => 'Lexie\'s RSS Monster (for Cial) (lexie@alexis-marie-wright.me)'
|
||||
]
|
||||
]
|
||||
);
|
||||
|
||||
|
@ -36,7 +55,45 @@ function fetchUrl(string $url) {
|
|||
throw "Request for $url returned {$res->getStatusCode()}";
|
||||
};
|
||||
|
||||
return $res->getBody();
|
||||
return strval($res->getBody());
|
||||
};
|
||||
|
||||
Log::info(fetchUrl('http://inhuman-comic.com'));
|
||||
$baseUrl = 'http://inhuman-comic.com';
|
||||
|
||||
$nextLinkSelector = 'div.body a.prev';
|
||||
$urls = [$baseUrl];
|
||||
$fetchedUrls = [];
|
||||
|
||||
while (count($urls) > 0) {
|
||||
$url = array_shift($urls);
|
||||
|
||||
$res = fetchUrl($url);
|
||||
Log::info("$url: fetched " . strlen($res) . " bytes");
|
||||
|
||||
Log::silly($res);
|
||||
|
||||
array_push($fetchedUrls, $url);
|
||||
|
||||
$dom = new DOM;
|
||||
$dom->loadStr($res);
|
||||
|
||||
foreach (array($dom->find($nextLinkSelector)) as $el) {
|
||||
$nextUrl = $baseUrl . '/' . $el->href;
|
||||
|
||||
if (array_search($nextUrl, $fetchedUrls)) {
|
||||
Log::debug("$url: already fetched next link $nextUrl");
|
||||
continue;
|
||||
};
|
||||
|
||||
if (array_search($nextUrl, $urls)) {
|
||||
Log::debug("$url: duplicate next link $nextUrl");
|
||||
continue;
|
||||
};
|
||||
|
||||
Log::info("$url: next link $nextUrl");
|
||||
array_push($urls, $nextUrl);
|
||||
};
|
||||
|
||||
usleep(500000); // wait half a second, to be minimally polite
|
||||
};
|
||||
Log::info("Out of next links!");
|
Loading…
Reference in New Issue
Block a user