In PHP, how to decompress a file on the fly that was compressed twice?

Question

I have a bigfile.gz.gz file that is… big. I would like to uncompress it on the fly. Ideally, this is what I have in mind:

$in = fopen('compress.zlib://compress.zlib://bigfile.gz.gz', 'rb');
while (!feof($in))
    print fread($in, 4096);
fclose($in);

However, compress.zlib:// cannot be chained that way:

PHP Warning:  fopen(): cannot represent a stream of type ZLIB as a File Descriptor in gztest.php on line 1

So I thought I'd combine gzopen() and compress.zlib:// together:

$in = gzopen('compress.zlib://bigfile.gz.gz', 'rb');
while (!gzeof($in))
    print gzread($in, 4096);
gzclose($in);

However, this only decompresses one level of gzip.

I tried probably 10 other methods, unfortunately gzopen() does not work with php://memory if it's been written to using fwrite() . And stream_filter_append(… zlib.inflate …) cannot read gzipped files.

This is the best I could come up with, but it spawns two system processes, which has undesirable overhead:

$in = popen('zcat bigfile.gz.gz | gunzip', 'rb');
while (!feof($in))
    print fread($in, 4096);
fclose($in);

Can someone suggest something better maybe?

Answer 1

It's possible to uncompress .gz files using the zlib.inflate filter. You just need to strip out the gzip header first. To do that on the fly, you have to deploy a custom filter:

<?php

class gzip_header_filter extends php_user_filter {

    private $filtered = 0;

    public function filter($in, $out, &$consumed, $closing) {
        while ($bucket = stream_bucket_make_writeable($in)) {
            if($this->filtered == 0) {
                $header_len = 10;
                $header = substr($bucket->data, 0, 10);
                $flags = ord($header[3]);
                if($flags & 0x08) {
                    // a filename is present
                    $header_len = strpos($bucket->data, "\0", 10) + 1;
                } 
                $bucket->data = substr($bucket->data, $header_len);
                $this->filtered = $header_len;
            }
            $consumed += $bucket->datalen;
            stream_bucket_append($out, $bucket);
        }
        return PSFS_PASS_ON;
    }
}

stream_filter_register('gzip_header_filter', 'gzip_header_filter');

$in = fopen('bigfile.gz.gz', 'rb');
stream_filter_append($in, 'gzip_header_filter', STREAM_FILTER_READ);
stream_filter_append($in, 'zlib.inflate', STREAM_FILTER_READ);
stream_filter_append($in, 'gzip_header_filter', STREAM_FILTER_READ);
stream_filter_append($in, 'zlib.inflate', STREAM_FILTER_READ);

while (!feof($in))
    print fread($in, 4096);
fclose($in);

?>

Note that the code above doesn't handle comments and other extra data that could be stored in the gz file.