mirror of
https://github.com/RSS-Bridge/rss-bridge.git
synced 2024-11-22 17:45:40 +03:00
143 lines
5.4 KiB
PHP
143 lines
5.4 KiB
PHP
|
<?php
|
||
|
|
||
|
class Mailman2Bridge extends BridgeAbstract
|
||
|
{
|
||
|
const NAME = 'Mailman2Bridge';
|
||
|
const URI = 'https://list.org/';
|
||
|
const MAINTAINER = 'imagoiq';
|
||
|
const CACHE_TIMEOUT = 60 * 30; // 30m
|
||
|
const DESCRIPTION = 'Fetch latest messages from Mailman 2 archive (Pipermail)';
|
||
|
|
||
|
const PARAMETERS = [
|
||
|
'Mailman 2' => [
|
||
|
'url' => [
|
||
|
'name' => 'Enter web archive URL',
|
||
|
'title' => <<<"EOL"
|
||
|
Specify the URL from the archive page where all the archive are listed by month.
|
||
|
EOL
|
||
|
, 'type' => 'text',
|
||
|
'exampleValue' => 'https://mailman.nginx.org/pipermail/nginx-announce/',
|
||
|
'required' => true
|
||
|
],
|
||
|
'limit' => [
|
||
|
'name' => 'Limit',
|
||
|
'type' => 'number',
|
||
|
'title' => 'Maximum number of items to return',
|
||
|
'defaultValue' => 5,
|
||
|
],
|
||
|
],
|
||
|
];
|
||
|
|
||
|
public function collectData()
|
||
|
{
|
||
|
$mails = [];
|
||
|
$url = $this->getInput('url');
|
||
|
$limit = $this->getInput('limit');
|
||
|
|
||
|
$html = defaultLinkTo(getSimpleHTMLDOMCached($url, 1800), $url);
|
||
|
|
||
|
// Fetch archive urls from the frontpage
|
||
|
$archives = [];
|
||
|
foreach ($html->find('tr') as $key => $tr) {
|
||
|
$archiveUrl = $tr->find('a[href$="date.html"]', 0);
|
||
|
$downloadUrl = $tr->find('a[href$=".txt"], a[href$=".txt.gz"]', 0);
|
||
|
$archives[$key] = [
|
||
|
'bydate' => $archiveUrl ? $archiveUrl->getAttribute('href') : null,
|
||
|
'download' => $downloadUrl ? $downloadUrl->getAttribute('href') : null
|
||
|
];
|
||
|
}
|
||
|
|
||
|
foreach ($archives as $archive) {
|
||
|
if (!$archive['bydate']) {
|
||
|
continue;
|
||
|
}
|
||
|
|
||
|
// Fetch urls to mails
|
||
|
$parent = pathinfo($archive['bydate'], PATHINFO_DIRNAME) . '/';
|
||
|
$html = defaultLinkTo(getSimpleHTMLDOMCached($archive['bydate'], 1800), $parent);
|
||
|
$links = array_map(function ($val) {
|
||
|
return $val->getAttribute('href');
|
||
|
}, $html->find('ul', 1)->find('li a[href$=".html"]'));
|
||
|
$mailUrls = array_reverse($links);
|
||
|
|
||
|
// Parse mbox
|
||
|
$data = getContents($archive['download']);
|
||
|
if (str_ends_with($archive['download'], '.gz')) {
|
||
|
$data = \gzdecode($data, (1024 ** 2) * 25); // 25M
|
||
|
if ($data === false) {
|
||
|
throw new \Exception('Failed to gzdecode');
|
||
|
}
|
||
|
}
|
||
|
$mboxParts = preg_split('/^From /', $data);
|
||
|
// Drop the first element which is always an empty string
|
||
|
array_shift($mboxParts);
|
||
|
$mboxMails = array_reverse($mboxParts);
|
||
|
foreach ($mboxMails as $index => $content) {
|
||
|
// Match Urls with contents from txt files.
|
||
|
// Urls cannot be reconstructed from the txt content.
|
||
|
$mails[] = [
|
||
|
'url' => $mailUrls[$index],
|
||
|
'content' => $content
|
||
|
];
|
||
|
}
|
||
|
if (count($mails) > $limit) {
|
||
|
break;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
$pluck = function ($header, $mail) {
|
||
|
// Not necessary to escape the header here
|
||
|
$pattern = sprintf('/(?<=%s:).*$/m', $header);
|
||
|
if (preg_match($pattern, $mail, $m)) {
|
||
|
return trim(\mb_decode_mimeheader($m[0]));
|
||
|
}
|
||
|
return null;
|
||
|
};
|
||
|
foreach (array_slice($mails, 0, $limit) as $mail) {
|
||
|
$item = [];
|
||
|
$item['uid'] = $pluck('Message-ID', $mail['content']);
|
||
|
$item['uri'] = $mail['url'];
|
||
|
$item['title'] = $pluck('Subject', $mail['content']);
|
||
|
$item['author'] = preg_replace('/\sat\s/', '@', $pluck('From', $mail['content']));
|
||
|
$item['timestamp'] = $pluck('Date', $mail['content']);
|
||
|
$item['content'] = nl2br(self::render($mail['content']));
|
||
|
$this->items[] = $item;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parse mbox mail. Render some useful html.
|
||
|
*
|
||
|
* Based on https://gist.github.com/jbroadway/2836900
|
||
|
*/
|
||
|
private static function render($text)
|
||
|
{
|
||
|
$rules = [
|
||
|
'/[\s\S]*?^Message-ID:[\s\S]*?>\n\n/m' => '', // Metadata
|
||
|
'/-+\s+next part\s+-+[\s\S]+?(?=^$|\Z)/m' => '', // next part
|
||
|
'/(?<!href=[\'"])https?:\/\/(www\.)?[-a-zA-Z0-9@:%._\+~#=]{1,256}\.
|
||
|
[a-zA-Z0-9()]{1,6}\b([-a-zA-Z0-9()!@:%_\+.,~#?&\/\/=]*)/' => '<a href="$0">$0</a>', // links
|
||
|
'/(\w+) <(.*) at (.*)>/' => '<a href="mailto:$2@$3">$1</a>', // emails
|
||
|
'/(\*|\*\*|__)(.*?)\1/' => '<strong>\2</strong>', // bold
|
||
|
// blockquotes
|
||
|
'/(.*)\s+((?:^>.*+\n)+)/m' => function ($regs) {
|
||
|
return sprintf(
|
||
|
'<details><summary>%s</summary><blockquote style="font-style:italic;">%s</blockquote></details>',
|
||
|
$regs[1],
|
||
|
preg_replace('/^>/m', '', $regs[2])
|
||
|
);
|
||
|
},
|
||
|
];
|
||
|
|
||
|
$text = "\n" . $text . "\n";
|
||
|
foreach ($rules as $regex => $replacement) {
|
||
|
if (is_callable($replacement)) {
|
||
|
$text = preg_replace_callback($regex, $replacement, $text);
|
||
|
} else {
|
||
|
$text = preg_replace($regex, $replacement, $text);
|
||
|
}
|
||
|
}
|
||
|
return trim($text);
|
||
|
}
|
||
|
}
|