|
<?php
mb_internal_encoding("UTF-8"); // encoding for mbstring
include 'phpQuery-onefile.php';
?>
<!doctype html>
<html>
<head>
<meta charset="utf-8">
<title>DOM</title>
</head>
<body>
<?php
$doc = phpQuery::newDocumentFileHTML("wiki.html");
$doc["table"]->remove(); // remove tables as well as their contents
$doc[":header"]->remove(); // remove <h1>-<h6>
$doc["#coordinates"]->remove(); // remove the float layer 'coordinates'
$doc["span.editsection"]->remove(); // remove Links 'edit'
$doc["small"]->remove();
// remove <span> but keep its content
foreach ($doc["span"] as $span) {
$span = pq($span);
$span->after($span->text())->remove();
}
$doc["div"]->remove();
$doc["script"]->remove();
// remove [number]
foreach ($doc["a"] as $a) {
$a = pq($a);
if (preg_match("/^\\[\\d+\\]$/", trim($a->text()))) {
$a->remove();
}
}
$doc->html($doc->text()); // remove all other html labels
$html = $doc->html();
$html = preg_replace("/\\n{2,}/", "\n", $html); // \n\n\n\n\n => \n
define("MAXLEN", 400);
$shortened = false;
if (mb_strlen($html) > MAXLEN) {
$html = mb_substr($html, 0, MAXLEN);
$shortened = true;
}
$html = preg_replace("/^[\\n\\s]+/", "", $html); // remove \n\s\n\s\n\s at the beginning
$html = preg_replace("/[\\n\\s]+$/", "", $html); // remove \n\s\n\s\n\s at the end
$html = preg_replace("/\\n/", "<br>", $html); // \n => <br>
if ($shortened) {
$html .= "...";
}
$doc->html($html);
echo $doc->html();
?>
</body>
</html>
|
