/**
* Helper function of a helper function to convert charset for iptc values.
* @param string|array $data The IPTC string
* @param string $charset The charset
*
* @return string
*/
private static function convIPTCHelper($data, $charset)
{
if ($charset) {
MediaWiki\suppressWarnings();
$data = iconv($charset, "UTF-8//IGNORE", $data);
MediaWiki\restoreWarnings();
if ($data === false) {
$data = "";
wfDebugLog('iptc', __METHOD__ . " Error converting iptc data charset {$charset} to utf-8");
}
} else {
// treat as utf-8 if is valid utf-8. otherwise pretend its windows-1252
// most of the time if there is no 1:90 tag, it is either ascii, latin1, or utf-8
$oldData = $data;
UtfNormal\Validator::quickIsNFCVerify($data);
// make $data valid utf-8
if ($data === $oldData) {
return $data;
// if validation didn't change $data
} else {
return self::convIPTCHelper($oldData, 'Windows-1252');
}
}
return trim($data);
}
/**
* Do userComment tags and similar. See pg. 34 of exif standard.
* basically first 8 bytes is charset, rest is value.
* This has not been tested on any shift-JIS strings.
* @param string $prop Prop name
*/
private function charCodeString($prop)
{
if (isset($this->mFilteredExifData[$prop])) {
if (strlen($this->mFilteredExifData[$prop]) <= 8) {
// invalid. Must be at least 9 bytes long.
$this->debug($this->mFilteredExifData[$prop], __FUNCTION__, false);
unset($this->mFilteredExifData[$prop]);
return;
}
$charCode = substr($this->mFilteredExifData[$prop], 0, 8);
$val = substr($this->mFilteredExifData[$prop], 8);
switch ($charCode) {
case "JIS":
// JIS
$charset = "Shift-JIS";
break;
case "UNICODE":
$charset = "UTF-16" . $this->byteOrder;
break;
default:
// ascii or undefined.
$charset = "";
break;
}
if ($charset) {
MediaWiki\suppressWarnings();
$val = iconv($charset, 'UTF-8//IGNORE', $val);
MediaWiki\restoreWarnings();
} else {
// if valid utf-8, assume that, otherwise assume windows-1252
$valCopy = $val;
UtfNormal\Validator::quickIsNFCVerify($valCopy);
// validates $valCopy.
if ($valCopy !== $val) {
MediaWiki\suppressWarnings();
$val = iconv('Windows-1252', 'UTF-8//IGNORE', $val);
MediaWiki\restoreWarnings();
}
}
// trim and check to make sure not only whitespace.
$val = trim($val);
if (strlen($val) === 0) {
// only whitespace.
$this->debug($this->mFilteredExifData[$prop], __FUNCTION__, "{$prop}: Is only whitespace");
unset($this->mFilteredExifData[$prop]);
return;
}
// all's good.
$this->mFilteredExifData[$prop] = $val;
}
}
/** Function to extract metadata segments of interest from jpeg files
* based on GIFMetadataExtractor.
*
* we can almost use getimagesize to do this
* but gis doesn't support having multiple app1 segments
* and those can't extract xmp on files containing both exif and xmp data
*
* @param string $filename Name of jpeg file
* @return array Array of interesting segments.
* @throws MWException If given invalid file.
*/
static function segmentSplitter($filename)
{
$showXMP = XMPReader::isSupported();
$segmentCount = 0;
$segments = ['XMP_ext' => [], 'COM' => [], 'PSIR' => []];
if (!$filename) {
throw new MWException("No filename specified for " . __METHOD__);
}
if (!file_exists($filename) || is_dir($filename)) {
throw new MWException("Invalid file {$filename} passed to " . __METHOD__);
}
$fh = fopen($filename, "rb");
if (!$fh) {
throw new MWException("Could not open file {$filename}");
}
$buffer = fread($fh, 2);
if ($buffer !== "ÿØ") {
throw new MWException("Not a jpeg, no SOI");
}
while (!feof($fh)) {
$buffer = fread($fh, 1);
$segmentCount++;
if ($segmentCount > self::MAX_JPEG_SEGMENTS) {
// this is just a sanity check
throw new MWException('Too many jpeg segments. Aborting');
}
while ($buffer !== "ÿ") {
// In theory JPEG files are not allowed to contain anything between the sections,
// but in practice they sometimes do. It's customary to ignore the garbage data.
$buffer = fread($fh, 1);
}
$buffer = fread($fh, 1);
while ($buffer === "ÿ" && !feof($fh)) {
// Skip through any 0xFF padding bytes.
$buffer = fread($fh, 1);
}
if ($buffer === "þ") {
// COM section -- file comment
// First see if valid utf-8,
// if not try to convert it to windows-1252.
$com = $oldCom = trim(self::jpegExtractMarker($fh));
UtfNormal\Validator::quickIsNFCVerify($com);
// turns $com to valid utf-8.
// thus if no change, its utf-8, otherwise its something else.
if ($com !== $oldCom) {
MediaWiki\suppressWarnings();
$com = $oldCom = iconv('windows-1252', 'UTF-8//IGNORE', $oldCom);
MediaWiki\restoreWarnings();
}
// Try it again, if its still not a valid string, then probably
// binary junk or some really weird encoding, so don't extract.
UtfNormal\Validator::quickIsNFCVerify($com);
if ($com === $oldCom) {
$segments["COM"][] = $oldCom;
} else {
wfDebug(__METHOD__ . " Ignoring JPEG comment as is garbage.\n");
}
} elseif ($buffer === "á") {
// APP1 section (Exif, XMP, and XMP extended)
// only extract if XMP is enabled.
$temp = self::jpegExtractMarker($fh);
// check what type of app segment this is.
if (substr($temp, 0, 29) === "http://ns.adobe.com/xap/1.0/" && $showXMP) {
$segments["XMP"] = substr($temp, 29);
} elseif (substr($temp, 0, 35) === "http://ns.adobe.com/xmp/extension/" && $showXMP) {
$segments["XMP_ext"][] = substr($temp, 35);
} elseif (substr($temp, 0, 29) === "XMP://ns.adobe.com/xap/1.0/" && $showXMP) {
// Some images (especially flickr images) seem to have this.
// I really have no idea what the deal is with them, but
// whatever...
$segments["XMP"] = substr($temp, 29);
wfDebug(__METHOD__ . ' Found XMP section with wrong app identifier ' . "Using anyways.\n");
} elseif (substr($temp, 0, 6) === "Exif") {
// Just need to find out what the byte order is.
// because php's exif plugin sucks...
// This is a II for little Endian, MM for big. Not a unicode BOM.
$byteOrderMarker = substr($temp, 6, 2);
if ($byteOrderMarker === 'MM') {
$segments['byteOrder'] = 'BE';
} elseif ($byteOrderMarker === 'II') {
$segments['byteOrder'] = 'LE';
} else {
wfDebug(__METHOD__ . " Invalid byte ordering?!\n");
}
}
} elseif ($buffer === "í") {
// APP13 - PSIR. IPTC and some photoshop stuff
$temp = self::jpegExtractMarker($fh);
if (substr($temp, 0, 14) === "Photoshop 3.0") {
$segments["PSIR"][] = $temp;
}
} elseif ($buffer === "Ù" || $buffer === "Ú") {
// EOI - end of image or SOS - start of scan. either way we're past any interesting segments
return $segments;
} else {
// segment we don't care about, so skip
$size = wfUnpack("nint", fread($fh, 2), 2);
if ($size['int'] < 2) {
throw new MWException("invalid marker size in jpeg");
}
fseek($fh, $size['int'] - 2, SEEK_CUR);
}
}
// shouldn't get here.
throw new MWException("Reached end of jpeg file unexpectedly");
}
