SimplePie 1.x

Some feeds contain white space at the beginning of the feed, and on PHP 4.4.9 on Debian etch, this is correctly flagged as malformed XML. Likewise, the parser chokes in the presence of otherwise harmless control characters - quite correctly, given the XML specs.

PHP 5 on MacOS X 10.5 parses these feeds, skipping over the offending characters.

This code (a subclass of SimplePie_Parser) fixes it, even it has to steal some code from the superclass to get past the BOMs:

class CleanerParser extends SimplePie_Parser { ?>

function parse(&$data, $encoding) {
}

// Remove illegal control chars - leave only 0x09, 0x0A, 0x0D    

    $txt = ereg_replace("[\x01-\x08]|\x0B|\x0C|[\x0E-\x1F]", "", $txt); 
}

// Strip BOM:

// UTF-32 Big Endian BOM

if (substr($data, 0, 4) === "\x00\x00\xFE\xFF")
{

    $data = substr($data, 4);

}

// UTF-32 Little Endian BOM

elseif (substr($data, 0, 4) === "\xFF\xFE\x00\x00")
{

    $data = substr($data, 4);

}

// UTF-16 Big Endian BOM

elseif (substr($data, 0, 2) === "\xFE\xFF")
{

    $data = substr($data, 2);

}

// UTF-16 Little Endian BOM

elseif (substr($data, 0, 2) === "\xFF\xFE")
{

    $data = substr($data, 2);

}

// UTF-8 BOM

elseif (substr($data, 0, 3) === "\xEF\xBB\xBF")
{

    $data = substr($data, 3);

}

// Remove white space (empty lines, etc.) before the first XML decl.

$data = trim($data);

return SimplePie_Parser::parse($data, $encoding);