XMLReader::isValid

1. If you validate against relax-ng, no need to call $xml->setParserProperty(XMLReader::VALIDATE, true);2. Be aware that $xml->isValid() will return validity for currently active node (ie. node currently positioned using $xml->read()). It won't check validity of your entire tree at once, but rather on a step by step basis

Be aware that $xml->isValid() will return validity only for currently active node, so you have to loop trough the nodes.Here is a example how to validate a entire XML file against a XSD schema:<?php$xmlReader = new \XMLReader();$xmlReader->open('./example.xml');$xmlReader->setParserProperty(\XMLReader::VALIDATE, true);$xmlReader->setSchema('./schema.xsd');\libxml_use_internal_errors(true);$msgs = [];while ($xmlReader->read()) {    if (!$xmlReader->isValid()) {        $err = \libxml_get_last_error();        if ($err && $err instanceof \libXMLError) {            $msgs[] = \trim($err->message) . ' on line ' . $err->line;        }    }}if ($msgs) {    throw new \Exception("XML schema validation errors:\n - " . implode("\n - ", array_unique($msgs)));}?>

I encountered some problems to use isValid method with xml2assoc function.I use this way to validate the entire xml file and put it into an associative array.$xml = new XMLReader();if (!$xml->xml($xml_string, NULL, LIBXML_DTDVALID)) {  echo "XML not valid: load error";  exit();}libxml_use_internal_errors(TRUE);$xml_array = xml2assoc($xml);$arErrors = libxml_get_errors();$xml_errors = "";foreach ($arErrors AS $xmlError) $xml_errors .= $xmlError->message;if ($xml_errors != "") {  echo "XML not valid: ".$xml_errors;  exit();}//all ok

This comment is only partially correct:
"isValid() always returns false unless you enable checking for validity by $reader->setParserProperty(XMLReader::VALIDATE, true);"
This enables DTD checking, but you can also check by using RelaxNG (see setRelaxNGSchema() and setRelaxNGSchemaSource()).

And also, this is NOT correct:
"If you just need to check if XML file is well formed, successful loading into XMLReader object is usually enough."
It is not enough. Pull parsers operate on stream and if you have a large enough file they will not know it is well formed until it is read to the end. If you need to know if it is well formed or/and valid, read it till the end or validation error (you can use next() for fast reading if you don't care about contents).