refactor mp

This commit is contained in:
quenousimporte 2024-08-26 18:15:17 +02:00
parent b97f730205
commit e5e7ab27e3
1 changed files with 56 additions and 114 deletions

170
epub.php
View File

@ -19,7 +19,61 @@
function add_mp_article($url, $context, $epub)
{
$article = file_get_contents($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(1)->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, $title);
}
}
function get_mp_cookie()
@ -330,63 +384,7 @@
foreach ($items as $item)
{
$url = $item->link;
$article = file_get_contents($url, false, $context);
$doc = new DOMDocument();
$doc->loadHTML($article);
$finder = new DomXPath($doc);
$title = $item->title;
$author = $item->xpath('dc:creator')[0];
$date = $item->pubDate;
$summary = $item->description;
// 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, $title);
}
add_mp_article($item->link, $context, $epub);
}
write_epub($epub);
@ -420,63 +418,7 @@
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(1)->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);
}
add_mp_article('https://www.mediapart.fr' . $link->value, $context, $epub);
}
write_epub($epub);