add mediapart from home page

This commit is contained in:
quenousimporte 2024-08-22 12:08:39 +02:00
parent bb11d11087
commit 566287d767
1 changed files with 109 additions and 18 deletions

113
epub.php
View File

@ -291,19 +291,19 @@
write_epub($epub);
}
// Mediapart
if (isset($_POST['forcempcookies']) && $_POST['forcempcookies'])
{
$old = $mp_sessionid;
$mp_sessionid = get_mp_cookie();
$settings = file_get_contents('settings.php');
$settings = str_replace($old, $mp_sessionid, $settings);
file_put_contents('settings.php', $settings);
echo '<div>Updated mpsessid in settings.</div>';
}
// Mediapart RSS
if (isset($_POST['mediapart']) && $_POST['mediapart'])
{
if (isset($_POST['forcempcookies']) && $_POST['forcempcookies'])
{
$old = $mp_sessionid;
$mp_sessionid = get_mp_cookie();
$settings = file_get_contents('settings.php');
$settings = str_replace($old, $mp_sessionid, $settings);
file_put_contents('settings.php', $settings);
echo '<div>Updated mpsessid in settings.</div>';
}
$feedurl = 'https://www.mediapart.fr/articles/feed';
$opts = [
'http' => [
@ -386,6 +386,94 @@
write_epub($epub);
}
// Mediapart home page
if (isset($_POST['mphome']) && $_POST['mphome'])
{
$homeurl = 'https://www.mediapart.fr';
$opts = [
'http' => [
'method' => "GET",
'header' => "Accept-language: en\nCookie: MPSESSID=" . $mp_sessionid,
]
];
$context = stream_context_create($opts);
$epub = new TPEpubCreator();
$epub->temp_folder = 'temp/';
$epub->epub_file = 'epub/mediaparthome.epub';
$epub->title = 'Mediapart home ' . $date;
$home = file_get_contents($homeurl, false, $context);
$doc = new DOMDocument();
$doc->loadHTML($home);
$finder = new DomXPath($doc);
$links = $finder->query('//section[contains(@class, "_hot")]//a[not(contains(@href, "/dossier/")) and not(contains(@href, "/studio/")) and not(contains(@href, "fil-")) and (@href != "/journal/series")]/@href');
foreach ($links as $link)
{
$url = 'https://www.mediapart.fr' . $link->value;
$article = file_get_contents($url, false, $context);
$doc = new DOMDocument();
$doc->loadHTML($article);
$finder = new DomXPath($doc);
$title = $finder->query('//h1')->item(0)->textContent;
$author = $finder->query('//a[contains(@href, "biographie")]')->item(0)->textContent;
$date = $finder->query('//time')->item(0)->textContent;
$summary = $finder->query('//p[contains(@class, "news__heading__top__intro")]')->item(0)->textContent;
// strip images
$toremove = $finder->query('//svg');
foreach ($toremove as $elt)
{
$elt->parentNode->removeChild($elt);
}
$toremove = $finder->query('//figure');
foreach ($toremove as $elt)
{
$elt->parentNode->removeChild($elt);
}
$toremove = $finder->query('//span[@class="screen-reader-only"]');
foreach ($toremove as $elt)
{
$elt->parentNode->removeChild($elt);
}
$result = '<h1>' . $title . '</h1>';
$result .= '<p>' . $author . '</p>';
$result .= '<p>' . $date . '</p>';
$result .= '<p><b>' . $summary . '</b></p>';
$nodes = $finder->query('//div[contains(@class, "paywall-restricted-content")]');
if (!$nodes->length)
{
// articles accès libre
$nodes = $finder->query('//div[contains(@class, "news__body__center__article")]');
}
if (!$nodes->length)
{
echo 'warning: could not get content of "' . $title . '"<br>';
}
else
{
$node = $nodes->item(0);
$innerHTML = '';
foreach ($node->childNodes as $childNode){
$innerHTML .= $childNode->ownerDocument->saveHTML($childNode);
}
$result .= '<div>' . strip_tags($innerHTML, '<p><b><h2><i>') . '</div>';
$epub->AddPage($result, false, $title);
}
}
write_epub($epub);
}
// New York Times
if (isset($_POST['nyt']) && $_POST['nyt'])
{
@ -565,6 +653,9 @@
<input id="mediapart" name="mediapart" type="checkbox">
<label for="mediapart">From RSS</label>
<br>
<input id="mphome" name="mphome" type="checkbox">
<label for="mphome">From home page</label>
<br>
<input id="forcempcookies" name="forcempcookies" type="checkbox">
<label for="forcempcookies">Force cookies</label>