灵活的解决方案:

function size($size, array $options=null) {

    $o = [
        'binary' => false,
        'decimalPlaces' => 2,
        'decimalSeparator' => '.',
        'thausandsSeparator' => '',
        'maxThreshold' => false, // or thresholds key
        'suffix' => [
            'thresholds' => ['', 'K', 'M', 'G', 'T', 'P', 'E', 'Z', 'Y'],
            'decimal' => ' {threshold}B',
            'binary' => ' {threshold}iB',
            'bytes' => ' B'
        ]
    ];

    if ($options !== null)
        $o = array_replace_recursive($o, $options);

    $base = $o['binary'] ? 1024 : 1000;
    $exp = $size ? floor(log($size) / log($base)) : 0;

    if (($o['maxThreshold'] !== false) &&
        ($o['maxThreshold'] < $exp)
    )
        $exp = $o['maxThreshold'];

    return !$exp
        ? (round($size) . $o['suffix']['bytes'])
        : (
            number_format(
                $size / pow($base, $exp),
                $o['decimalPlaces'],
                $o['decimalSeparator'],
                $o['thausandsSeparator']
            ) .
            str_replace(
                '{threshold}',
                $o['suffix']['thresholds'][$exp],
                $o['suffix'][$o['binary'] ? 'binary' : 'decimal']
            )
        );
}

var_dump(size(disk_free_space('/')));
// string(8) "14.63 GB"
var_dump(size(disk_free_space('/'), ['binary' => true]));
// string(9) "13.63 GiB"
var_dump(size(disk_free_space('/'), ['maxThreshold' => 2]));
// string(11) "14631.90 MB"
var_dump(size(disk_free_space('/'), ['binary' => true, 'maxThreshold' => 2]));
// string(12) "13954.07 MiB"

伪代码:

$base = log($size) / log(1024);
$suffix = array("", "k", "M", "G", "T")[floor($base)];
return pow(1024, $base - floor($base)) . $suffix;

这是Kohana的实现，你可以使用它:

public static function bytes($bytes, $force_unit = NULL, $format = NULL, $si = TRUE)
{
    // Format string
    $format = ($format === NULL) ? '%01.2f %s' : (string) $format;

    // IEC prefixes (binary)
    if ($si == FALSE OR strpos($force_unit, 'i') !== FALSE)
    {
        $units = array('B', 'KiB', 'MiB', 'GiB', 'TiB', 'PiB');
        $mod   = 1024;
    }
    // SI prefixes (decimal)
    else
    {
        $units = array('B', 'kB', 'MB', 'GB', 'TB', 'PB');
        $mod   = 1000;
    }

    // Determine unit to use
    if (($power = array_search((string) $force_unit, $units)) === FALSE)
    {
        $power = ($bytes > 0) ? floor(log($bytes, $mod)) : 0;
    }

    return sprintf($format, $bytes / pow($mod, $power), $units[$power]);
}

下面是一个使用log10的选项:

<?php

function format_number(float $d): string {
   $e = (int)(log10($d) / 3);
   return sprintf('%.3f', $d / 1e3 ** $e) . ['', ' k', ' M', ' G'][$e];
}

$s = format_number(9012345678);
var_dump($s == '9.012 G');

https://php.net/function.log10

这是Chris Jester-Young的实现，我见过的最干净的，结合了php。net的和一个精度参数。

function formatBytes($size, $precision = 2)
{
    $base = log($size, 1024);
    $suffixes = array('', 'K', 'M', 'G', 'T');   

    return round(pow(1024, $base - floor($base)), $precision) .' '. $suffixes[floor($base)];
}

echo formatBytes(24962496);
// 23.81M

echo formatBytes(24962496, 0);
// 24M

echo formatBytes(24962496, 4);
// 23.8061M

下面是Drupal format_size函数的简化实现:

/**
 * Generates a string representation for the given byte count.
 *
 * @param $size
 *   A size in bytes.
 *
 * @return
 *   A string representation of the size.
 */
function format_size($size) {
  if ($size < 1024) {
    return $size . ' B';
  }
  else {
    $size = $size / 1024;
    $units = ['KB', 'MB', 'GB', 'TB'];
    foreach ($units as $unit) {
      if (round($size, 2) >= 1024) {
        $size = $size / 1024;
      }
      else {
        break;
      }
    }
    return round($size, 2) . ' ' . $unit;
  }
}