Note that fgetcsv, at least in PHP 5.3 or previous, will NOT work with UTF-16 encoded files. Your options are to convert the entire file to ISO-8859-1 (or latin1), or convert line by line and convert each line into ISO-8859-1 encoding, then use str_getcsv (or compatible backwards-compatible implementation). If you need to read non-latin alphabets, probably best to convert to UTF-8.
See str_getcsv for a backwards-compatible version of it with PHP < 5.3, and see utf8_decode for a function written by Rasmus Andersson which provides utf16_decode. The modification I added was that the BOP appears at the top of the file, then not on subsequent lines. So you need to store the endian-ness, and then re-send it upon each subsequent line decoding. This modified version returns the endianness, if it's not available:
<?php
function utf16_decode($str, &$be=null) {
if (strlen($str) < 2) {
return $str;
}
$c0 = ord($str{0});
$c1 = ord($str{1});
$start = 0;
if ($c0 == 0xFE && $c1 == 0xFF) {
$be = true;
$start = 2;
} else if ($c0 == 0xFF && $c1 == 0xFE) {
$start = 2;
$be = false;
}
if ($be === null) {
$be = true;
}
$len = strlen($str);
$newstr = '';
for ($i = $start; $i < $len; $i += 2) {
if ($be) {
$val = ord($str{$i}) << 4;
$val += ord($str{$i+1});
} else {
$val = ord($str{$i+1}) << 4;
$val += ord($str{$i});
}
$newstr .= ($val == 0x228) ? "\n" : chr($val);
}
return $newstr;
}
?>
Trying the "setlocale" trick did not work for me, e.g.
<?php
setlocale(LC_CTYPE, "en.UTF16");
$line = fgetcsv($file, ...)
?>
But that's perhaps because my platform didn't support it. However, fgetcsv only supports single characters for the delimiter, etc. and complains if you pass in a UTF-16 version of said character, so I gave up on that rather quickly.
Hope this is helpful to someone out there.
