How do I tell DOMDocument->load() what encoding I want it to use?

Question

I search for and process XML files from elsewhere, and need to transform them with some XSLTs. No problem. Using PHP5 and the DOM library, everything's a snap. Worked fine, up till now. Today, funky characters were in the XML file -- "smart" quotes from Word, it looks like. Anyways, DOMDocument->load complained about them, saying that they weren't UTF-8, and to specify the encoding.

Lo and behold, the encoding is not specified in these XML files. If I add in 'encoding="iso-8859-1"' to the header, it works fine. The rub is I have no control over these XML files.

Reading the file into a string, modifying its header and writing it back out to another location seems to be my only option, but I'd prefer to do it without having to use temporary copies of the XML files at all. Is there any way to simply tell the parser to parse them as if they were iso-8859-1?

Answer 1

Does this work for you?

$doc = new DOMDocument('1.0', 'iso-8859-1');
$doc->load($xmlPath);

Edit: Since it appears that this doesn't work, what you could do instead is similar to your existing method but without the temp file. Read the XML file from your source just using standard IO operations (file_get_contents() or something), then perform whatever changes to the encoding you need (iconv() or utf8_decode()) and then use loadXML()

$myXMLString = file_get_contents($xmlPath);
$myXMLString = utf8_decode($myXMLString);
$doc = new DOMDocument('1.0', 'iso-8859-1');
$doc->loadXML($myXMLString);

Answer 2

I haven't found a way to set the default encoding (yet) but maybe the recover mode is feasible in this case.
When libxml encounters an encoding error and no encoding has been explicitly set it switches from unicode/utf8 to latin1 and continues parsing the document. But in the parser context the property wellFormed is set to 0/false. PHP's DOM extension considers the document valid if wellFormed is true or the DOMDocument object's attribute recover is true.

<?php
// german Umlaut ä in latin1 = 0xE4
$xml = '<foo>'.chr(0xE4).'</foo>';

$doc = new DOMDocument;
$b = $doc->loadxml($xml);
echo 'with doc->recover=false(default) : ', ($b) ? 'success':'failed', "\n";

$doc = new DOMDocument;
$doc->recover = true;
$b = $doc->loadxml($xml);
echo 'with doc->recover=true : ', ($b) ? 'success':'failed', "\n";

prints

Warning: DOMDocument::loadXML(): Input is not proper UTF-8, indicate encoding !
Bytes: 0xE4 0x3C 0x2F 0x66 in Entity, line: 1 in test.php on line 6
with doc->recover=false(default) : failed

Warning: DOMDocument::loadXML(): Input is not proper UTF-8, indicate encoding !
Bytes: 0xE4 0x3C 0x2F 0x66 in Entity, line: 1 in  test.php on line 11
with doc->recover=true : success

You still get the warning message (which can be suppressed with @$doc->load()) and it will also show up in the internal libxml errors (only once when the parser switches from utf8 to latin1). The error code for this particular error will be 9 (XML_ERR_INVALID_CHAR).

<?php
$xml = sprintf('<foo>
    <ae>%s</ae>
    <oe>%s</oe>
    &
</foo>', chr(0xE4),chr(0xF6));

libxml_use_internal_errors(true);
$doc = new DOMDocument;
$doc->recover = true;
libxml_clear_errors();
$b = $doc->loadxml($xml);
$invalidCharFound = false;
foreach(libxml_get_errors() as $error) {
    if ( 9==$error->code && !$invalidCharFound ) {
     $invalidCharFound = true;
     echo "found invalid char, possibly harmless\n";
    }
    else {
     echo "hm, that's probably more severe: ", $error->message, "\n";
    }
}