/**
* Truncate a given HTML fragment to the desired length (measured as character
* count), additionally performing some cleanup.
*
* @param string $html The HTML fragment to clean up
* @param int $desiredLength The desired number of characters, or NULL to do
* just the cleanup (but no truncating).
* @param string $ellipseStr If non-empty, this string will be appended to the
* last collected node when the document gets
* truncated.
* @param bool $stripTags If TRUE, remove *all* HTML tags. Otherwise, keep a
* whitelisted 'safe' set.
* @param bool $nbsp If TRUE, convert all whitespace runs to non-breaking
* spaces (' ' entities).
*/
function trimToHTML($html, $desiredLength = null, $ellipseStr = "…", $stripTags = false, $nbsp = false)
{
// We'll use htmlmaid to clean up the HTML, but because we also have to
// step through the DOM ourselves to perform the trimming, so we'll do
// the DOM loading ourselves, rather than leave it to Maid.
// Do not load external entities - this would be a security risk.
$prevEntityLoaderDisabled = libxml_disable_entity_loader(true);
// Don't crash on invalid HTML, but recover gracefully
$prevInternalErrors = libxml_use_internal_errors(true);
$doc = new \DOMDocument();
// We need a bit of wrapping here to keep DOMDocument from adding rogue nodes
// around our HTML. By doing it explicitly, we keep things under control.
$doc->loadHTML('<!DOCTYPE html><html>' . '<head><meta http-equiv="Content-type" content="text/html;charset=utf-8"/></head>' . '<body><div>' . $html . '</div></body>' . '</html>');
$options = array();
if ($stripTags) {
$options['allowed-tags'] = array();
} else {
$options['allowed-tags'] = array('a', 'div', 'p', 'b', 'i', 'hr', 'br', 'strong', 'em');
}
$options['allowed-attribs'] = array('href', 'src', 'id', 'class', 'style');
$maid = new Maid($options);
$cleanedNodes = $maid->clean($doc->documentElement->firstChild->nextSibling->firstChild);
// To collect the cleaned nodes from a node list into a containing node,
// we have to create yet another document, because cloning nodes inside
// the same ownerDocument for some reason modifies our node list.
// I have no idea why, but it does.
$cleanedDoc = new \DOMDocument();
$cleanedNode = $cleanedDoc->createElement('div');
$length = $cleanedNodes->length;
for ($i = 0; $i < $length; ++$i) {
$node = $cleanedNodes->item($i);
$cnode = $cleanedDoc->importNode($node, true);
$cleanedNode->appendChild($cnode);
}
// And now we'll create yet another document (who's keeping count?) to
// collect our trimmed nodes.
$newDoc = new \DOMDocument();
// Again, some wrapping is necessary here...
$newDoc->loadHTML('<html><body><div></div></body></html>');
$newNode = $newDoc->documentElement->firstChild->firstChild;
$length = $desiredLength;
_collectNodesUpToLength($cleanedNode, $newNode, $length, $ellipseStr);
// Convert spaces inside text nodes to
// This will actually insert the unicode non-breaking space, so we'll have
// to massage our output at the HTML byte-string level later.
if ($nbsp) {
domSpacesToNBSP($newNode->firstChild->firstChild);
}
// This is some terrible shotgun hacking; for some reason, the above code
// will sometimes put our desired nodes two levels deep, but in other
// cases, it'll descend one less level. The proper solution would be
// to sort out why this is, but for now, just detecting which of the
// two happened seems to work well enough.
if (isset($newNode->firstChild->firstChild->childNodes)) {
$nodes = $newNode->firstChild->firstChild->childNodes;
} elseif (isset($newNode->firstChild->childNodes)) {
$nodes = $newNode->firstChild->childNodes;
} else {
$nodes = array();
}
// And now we convert our target nodes to HTML.
// Because we don't want any of the wrapper nodes to appear in the
// output, we'll have to convert them one by one and concatenate the
// HTML.
$result = '';
foreach ($nodes as $node) {
$result .= Maid::renderFragment($node);
}
if ($nbsp) {
$result = str_replace(html_entity_decode(' '), ' ', $result);
}
// Restore previous libxml settings
libxml_disable_entity_loader($prevEntityLoaderDisabled);
libxml_use_internal_errors($prevInternalErrors);
return $result;
}
