Problem with files >2 GB

[Ivan-Perez]

Stream::getSize function is not working when opening files over 2 GB in size. I get a negative number for files between 2 and 4 GB. I'm on a 32-bit machine, so PHP internally uses a 32 bit signed integer type. Is there any workaround to get the real size?

Thanks!

[g105b]

Please take a look into the output of phpinfo(), most importantly upload_max_filesize and post_max_size. It might be that your php.ini is set to only process a certain maximum filesize.

Good to rule this out before looking into a Guzzle bug, and if you are looking to file a bug we will need to know more information from you, including how you can replicate the issue.

[sagikazarmark]

@g105b thanks, but it has nothing to do with upload_max_filesize or post_max_size. In this case there is no upload, only download. The problem is (as outlined by the issue reporter) the 32 bit machine.

The biggest integer you can represent in 32 bits (in two's complements): 2147483647
2 GB in bytes: 2147483648

Sorry for not answering earlier, I don't have a clear answer at this time.

One rough idea: you could count the bytes read manually in a stream wrapper as you read from the stream. When you overflow the integer limit, PHP will automatically cast the value to float for you, so unless you cast it to integer manually, you should get the real value.

[l0gicgate]

The following method is for calculating file size bigger than 2gb on a 32-bit instance.

This could possibly be adapted to Stream::getSize().

<?php
/**
 * Filesize method that works with files >2GB on 32bit PHP
 * 
 * @link http://php.net/manual/en/function.filesize.php#113457
 * @param string $file Path to file
 * @return int|float Size of file
 */
function realFileSize($file)
{
    $fp = fopen($file, 'r');
    
    $pos = 0;
    $size = 1073741824;
    fseek($fp, 0, SEEK_SET);
    while ($size > 1)
    {
        fseek($fp, $size, SEEK_CUR);
        if (fgetc($fp) === false)
        {
            fseek($fp, -$size, SEEK_CUR);
            $size = (int)($size / 2);
        }
        else
        {
            fseek($fp, -1, SEEK_CUR);
            $pos += $size;
        }
    }
    
    while (fgetc($fp) !== false)  $pos++;
    
    fclose($fp);
    
    return $pos;
}

[sagikazarmark]

Sorry, but this solution is rather a workaround and I would rather not add it to the core library. But the problem is searchable, so you can find it on the issue tracker. It's also trivial to turn this into a Stream decorator for easier usage.

[Tobion]

Also according to PSR-7 https://www.php-fig.org/psr/psr-7/ getSize must return an int. So we cannot return a float on 32 bit machines as it would conflict with the contract.

[Ivan-Perez]

    /**
     * Get the size of the stream if known.
     *
     * @return int|null Returns the size in bytes if known, or null if unknown.
     */
    public function getSize();

Yep, it should return an int. What about returning null in this case, as the size is unknown (not computable)? At least you wouldn't get a wrong size.

[Tobion]

Yes makes sense, Would be happy to see a PR.