How to determine the Content-Length of a gzipped file?

Question

Right now I'm trying to serve CSS and JS files from a server that won't let me enable mod_gzip or mod_deflate. So I wrote a small PHP script to compress with GZIP and return to the user.

Example Code:

$filename = "style.css";

if (!file_exists($filename) || !($info = stat($filename))) {
  header("HTTP/1.1 404 Not Found");
  die();
}

header("Date: ".gmdate("D, j M Y H:i:s e", time()));
header("Cache-Control: max-age=2592000");
header("Last-Modified: ".gmdate("D, j M Y H:i:s e", $info['mtime']));
header("Etag: ".sprintf("\"%x-%x-%x\"", $info['ino'], $info['size'], $info['mtime']));
header("Accept-Ranges: bytes");
header("Cache-Control: Expires ".gmdate("D, j M Y H:i:s e", $info['mtime']+2592000));
header("Content-Type: text/html");

ob_start("ob_gzhandler");
echo file_get_contents($filename);
ob_end_flush();

I'm having two problems right now. The first is, I'm having trouble determining the resulting size of the compressed file to inform the browser of the content-length. Normally, I would include this line:

header("Content-Length: ".$info["size"]);

But, if I do, the browser hangs while trying to wait for more data. Is there a way to calculate the total size? Or should I ignore this header directive.

The other issue is, whenever I view this PHP file in Firefox, it tries to have me download the result. In Chrome, it just displays it like I would expect. Any suggestions?

Edit: Thanks to SoapBox, I replaced the end of the code with this:

header("Content-Encoding: gzip");
$compressed = gzencode(file_get_contents($filename), 5);
header("Content-Length: ".strlen($compressed));
die($compressed);

This works great for the content-length! But I'm still getting Firefox to download the file instead of display it. :(

Edit Again: Here is the modified end-of-code code, courtesy of Cletus.

// Start buffered output
ob_start();
// Check for gzip capability
if (stripos($_SERVER['HTTP_ACCEPT_ENCODING'], "gzip") !== false) {
  ob_start("ob_gzhandler");
  echo file_get_contents($filename);
  ob_end_flush();
} else
  echo file_get_contents($filename);

// Write the content length
header('Content-Length: '.ob_get_length());
ob_end_flush();

I'm going to start a new question to figure out why Firefox keeps trying to download the file.

Answer 1

You need to first do the entire gzip and measure the result (either holding the contents in memory, or writing them to disk as you compress and then stat'ing the gzipped file), then write the Content-Length header and then send the file contents.

Or use chunked transfer encoding.

Answer 2

To solve your firefox issue, I think you need to include header( "Content-Encoding: gzip" ); so that the browser knows to decompress the content.

As for the content length, you can try just leaving this value off, or try to figure out a way to use "Transfer-Encoding: chunked" (you can't jsut send this header, you need to format the data specially for it). It is possible that ob_end_flush automatically enables chunking.

I recommend you get wireshark and capture what your php script is sending and compare it to a properly behaving server to see what headers, etc are missing.

Answer 3

The problem here is that to know the content length you need to know whether or not the client supports gzip encoding and you've delegated that decision by using ob_gzhandler. From HTTP Headers:

ob_start();
ob_start('ob_gzhandler');

  ... output the page content...

ob_end_flush();  // The ob_gzhandler one

header('Content-Length: '.ob_get_length());

ob_end_flush();  // The main one

Complete version:

$filename = "style.css";

if (!file_exists($filename) || !($info = stat($filename))) {
  header("HTTP/1.1 404 Not Found");
  die();
}

header("Date: ".gmdate("D, j M Y H:i:s e", time()));
header("Cache-Control: max-age=2592000");
header("Last-Modified: ".gmdate("D, j M Y H:i:s e", $info['mtime']));
header("Etag: ".sprintf("\"%x-%x-%x\"", $info['ino'], $info['size'], $info['mtime']));
header("Accept-Ranges: bytes");
header("Cache-Control: Expires ".gmdate("D, j M Y H:i:s e", $info['mtime']+2592000));
header("Content-Type: text/css"); // note: this was text/html for some reason?

ob_start();
ob_start("ob_gzhandler");
echo file_get_contents($filename);
ob_end_flush();
header('Content-Length: '.ob_get_length());
ob_end_flush();

This is much better than taking on the gzip encoding issue yourself.