Take the 2-minute tour ×
Unix & Linux Stack Exchange is a question and answer site for users of Linux, FreeBSD and other Un*x-like operating systems. It's 100% free, no registration required.

Is there a standard tool which converts an integer count of Bytes into a human-readable count of the largest possible unit-size, while keeping the numeric value between 1.00 and 1023.99 ?

I have my own bash/awk script, but I am looking for a standard tool, which is found on many/most distros... something more generally available, and ideally has simple command line args, and/or can accept piped input.

Here are some examples of the type of output I am looking for.

    1    Byt  
  173.00 KiB  
   46.57 MiB  
    1.84 GiB  
   29.23 GiB  
  265.72 GiB  
    1.63 TiB  

Here is the bytes-human script (used for the above output)

awk -v pfix="$1" -v sfix="$2" 'BEGIN { 
      split( "Byt KiB MiB GiB TiB PiB", unit )
      uix = uct = length( unit )
      for( i=1; i<=uct; i++ ) val[i] = (2**(10*(i-1)))-1
   }{ if( int($1) == 0 ) uix = 1; else while( $1 < val[uix]+1 ) uix--
      num = $1 / (val[uix]+1)
      if( uix==1 ) n = "%5d   "; else n = "%8.2f"
      printf( "%s"n" %s%s\n", pfix, num, unit[uix], sfix ) 

Update  Here is a modified version of Gilles' script, as described in a comment to his answer ..(modified to suit my preferred look).

awk 'function human(x) {
         s=" B   KiB MiB GiB TiB EiB PiB YiB ZiB"
         while (x>=1024 && length(s)>1) 
               {x/=1024; s=substr(s,5)}
         xf=(s==" B  ")?"%5d   ":"%8.2f"
         return sprintf( xf"%s\n", x, s)
      {gsub(/^[0-9]+/, human($1)); print}'
share|improve this question
It looks like here we have a new standard tool in the making :) –  Gowtham Jul 26 '12 at 18:57
@Gowtham - your wish may have come true! See my answer below or blog.frankleonhardt.com/2015/… –  FJL Mar 23 at 11:05

12 Answers 12

up vote 18 down vote accepted

No, there is no such standard tool.

Since GNU coreutils 8.21 (Feb 2013, so not yet present in all distributions), on non-embedded Linux and Cygwin, you can use numfmt. It doesn't produce exactly the same output format (as of coreutils 8.23, I don't think you can get 2 digits after the decimal points).

$ numfmt --to=iec-i --suffix=B --padding=7 1 177152 48832200 1975684956

Many older GNU tools can produce this format and GNU sort can sort numbers with units since coreutils 7.5 (Aug 2009, so present on modern non-embedded Linux distributions).

I find your code a bit convoluted. Here's a cleaner awk version (the output format isn't exactly identical):

awk '
    function human(x) {
        if (x<1000) {return x} else {x/=1024}
        while (x>=1000 && length(s)>1)
            {x/=1024; s=substr(s,2)}
        return int(x+0.5) substr(s,1,1)
    {sub(/^[0-9]+/, human($1)); print}'

(Reposted from a more specialized question)

share|improve this answer
Okay, thanks. About your script, I basically really like it. There are a few things which caught my attention: (1) var s should be have leading B. Also this string is easily changed to IEC Binary notation. (2) It skips the 1000-1023 range in favour of 1<next size> (easily changed) (3) It does not have decimal values (which I do want). Again this is easily changed. When displaying 2 decimal places, the %f format causes a round-up to the <next size> for values 1019-1023; but it's not worth a workaround ..I've posted a modified version in my answer, for general reference. –  Peter.O Jul 27 '12 at 8:17
I found your awk script is off by a letter, if you shift them over it works without further modification: s=" kMGTEPYZ" –  A.Danischewski Mar 22 at 12:11

As of v. 8.21, coreutils includes numfmt:

numfmt reads numbers in various representations and reformats them as requested.
The most common usage is converting numbers to / from human representation.


printf %s\\n 5607598768908 | numfmt --to=iec-i

Various other examples (including filtering, input/output processing etc) are presented HERE.

In addition, as of coreutils v. 8.23, numfmt can process multiple fields with field range specifications similar to cut, and supports setting the output precision with the --format option

numfmt --to=iec-i --field=2,4 --format='%.3f' <<<'tx: 180000 rx: 2000000'
tx: 175.782Ki rx: 1.908Mi
share|improve this answer
numfmt is a newly added tool to coreutils package from coreutils-8.21 onwards . –  Zama Ques Mar 24 '14 at 10:50

This is a complete rewrite inspired by Peter.O's modified version of Gilles' awk script.


  • Fixes Peter.O's bug where he looks for a string of >1 character where he should be looking for one >4 characters. Due to that bug, his code doesn't work for ZiB units.
  • Removes the very ugly hardcoding of a long string of space-separated unit sizes.
  • Adds command line switches to enable/disable padding.
  • Adds command line switches to go from base-1024 (KiB) to base-1000 (KB) notation.
  • Wraps it all in an easy to use function.
  • I place this in the public domain and welcome widespread use.


bytestohuman() {
    # converts a byte count to a human readable format in IEC binary notation (base-1024), rounded to two decimal places for anything larger than a byte. switchable to padded format and base-1000 if desired.
    local L_BYTES="${1:-0}"
    local L_PAD="${2:-no}"
    local L_BASE="${3:-1024}"
    BYTESTOHUMAN_RESULT=$(awk -v bytes="${L_BYTES}" -v pad="${L_PAD}" -v base="${L_BASE}" 'function human(x, pad, base) {

         while (x>=base && length(s)>1)
               {x/=base; s=substr(s,2)}

         xf=(pad=="yes") ? ((s=="B")?"%5d   ":"%8.2f") : ((s=="B")?"%d":"%.2f")
         s=(s!="B") ? (s basesuf) : ((pad=="no") ? s : ((basesuf=="iB")?(s "  "):(s " ")))

         return sprintf( (xf " %s\n"), x, s)
      BEGIN{print human(bytes, pad, base)}')
    return $?

Test Cases (if you want to look at the output):

bytestohuman 1; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 500; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1023; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1024; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1500; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000000000; echo "${BYTESTOHUMAN_RESULT}.";

bytestohuman 1 no 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 500 no 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1023 no 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1024 no 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1500 no 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000 no 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000 no 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000 no 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000 no 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000000 no 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000000000 no 1000; echo "${BYTESTOHUMAN_RESULT}.";

bytestohuman 1 yes; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 500 yes; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1023 yes; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1024 yes; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1500 yes; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000 yes; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000 yes; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000 yes; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000 yes; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000000 yes; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000000000 yes; echo "${BYTESTOHUMAN_RESULT}.";

bytestohuman 1 yes 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 500 yes 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1023 yes 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1024 yes 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 1500 yes 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000 yes 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000 yes 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000 yes 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000 yes 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000000 yes 1000; echo "${BYTESTOHUMAN_RESULT}.";
bytestohuman 150000000000000000000 yes 1000; echo "${BYTESTOHUMAN_RESULT}.";


share|improve this answer

There are a couple of perl modules on CPAN: Format::Human::Bytes and Number::Bytes::Human, the latter one being a bit more complete:

$ echo 100 1000 100000 100000000 |
  perl -M'Number::Bytes::Human format_bytes' -pe 's/\d{3,}/format_bytes($&)/ge'
100 1000 98K 96M

$ echo 100 1000 100000 100000000 |
  perl -M'Number::Bytes::Human format_bytes' -pe 's/\d{3,}/
   format_bytes($&,bs=>1000, round_style => 'round', precision => 2)/ge'
100 1.00k 100k 100M

And the reverse:

$ echo 100 1.00k 100K 100M 1Z |
  perl -M'Number::Bytes::Human parse_bytes' -pe '
100 1024 102400 104857600 1.18059162071741e+21

NOTE: the function parse_bytes() was added in version 0.09 (2013-03-01)

share|improve this answer

Via linux - Is there a command line calculator for byte calculations? - Stack Overflow, I found about GNU Units - though without examples on the SO page; and as I didn't see it listed here, here is a small note about it.

First, check if the units are present:

$ units --check-verbose |grep byte
doing 'byte'

$ units --check-verbose |grep mega
doing 'megalerg'
doing 'mega'

$ units --check-verbose |grep mebi
doing 'mebi'

Given that they are, do a conversion - printf format specifiers are accepted to format the numeric result:

$ units --one-line -o "%.15g" '20023450 bytes' 'megabytes'  # also --terse
    * 20.02345
$ units --one-line -o "%.15g" '20023450 bytes' 'mebibytes' 
    * 19.0958499908447
$ units --one-line -o "%.5g" '20023450 bytes' 'mebibytes' 
    * 19.096
share|improve this answer

Actually, there is a utility that does exactly this. I know cos it was me wot wrote it. It was written for *BSD but ought to compile on Linux if you have the BSD libraries (which I believe are common).

I've just released a new version, posted here:


It's called hr, and it will take stdin (or files) and convert numbers to human-readable format in a way that is (now) exactly the same as ls -h and so on, and it can select individual feeds in lines, scale pre-scaled units (e.g. if they're in 512-byte blocks convert them to Mb etc), adjust column padding, and so on.

I wrote it a few years ago because I thought trying to write a shell script, although intellectually interesting, was also utter madness.

Using hr, for example, you can easily get a sorted list of directory sizes (which come out in 1Kb units and need shifting before converting) with the following:

du -d1 | sort -n | hr -sK

While du will produce -h output, sort won't sort by it. The addition of -h to existing utilities is a classic case of not following the unix philosophy: have simple utilities doing defined jobs really well.

share|improve this answer

Recursified version of Gilles Awk solution (from above):

awk 'function human(x) {
      if (x[1]>=1000) { x[2]++; human(x); }
    {a[1]=$1; a[2]=0; human(a); print a[1],substr("kMGTEPYZ",a[2]+1,1)}'
share|improve this answer
user@host:/usr$ alias duh="du -s -B1 * | sort -g | numfmt --to=iec-i --format='%10f'"
user@host:/usr$ duh


 4.0Ki games
 3.9Mi local
  18Mi include
  20Mi sbin
 145Mi bin
 215Mi share
 325Mi src
 538Mi lib

Unfortunately I can't figure out how to get two decimals accuracy. Tested on Ubuntu 14.04.

share|improve this answer

Here is a way to do it almost purely in bash, just needs 'bc' for the floating point math.

function bytesToHR() {
        local SIZE=$1
        local UNITS="B KiB MiB GiB TiB PiB"
        for F in $UNITS; do
                local UNIT=$F
                test ${SIZE%.*} -lt 1024 && break;
                SIZE=$(echo "$SIZE / 1024" | bc -l)

    if [ "$UNIT" == "B" ]; then
        printf "%4.0f    %s\n" $SIZE $UNIT
        printf "%7.02f %s\n" $SIZE $UNIT


bytesToHR 1
bytesToHR 1023
bytesToHR 1024
bytesToHR 12345
bytesToHR 123456
bytesToHR 1234567
bytesToHR 12345678


   1    B
1023    B
   1.00 KiB
  12.06 KiB
 120.56 KiB
   1.18 MiB
  11.77 MiB
share|improve this answer

AFAIK there is no such standard tool to which you can pass text and it returns a human readable form. You may be able to find a package to accomplish the said task for your distro.

However, I do not understand why you may need such a tool. Most packages that give a related output, usually have a -h or equivalent switch for human-readable output.

share|improve this answer
For purposes of understanding: Human-readable means just that; readable by humans. The various different units of size shown by the tools you mention are not intended for programatic calculations, for which uniformity of units is essential. Working with bytes, which are always integers, is the only way bash can do any arithmetic with them. So... calculate in Bytes... report to in Human, eg. "You are about to permanently delete 3 files, totalling 2.44 GiB. Continue? –  Peter.O Jul 26 '12 at 16:13
I think this should be part of your question. Looks to me like you've got the problem solved. Good luck. –  shellter Jul 26 '12 at 22:20
A common application is to generate numbers of bytes for sorting, and convert to human-readable units after sorting. –  Gilles Jul 26 '12 at 23:44

Short and sweet, shell only solution:

convertB_human() {
for DESIG in Bytes KB MB GB TB PB
   [ $NUMBER -lt 1024 ] && break
   let NUMBER=$NUMBER/1024
printf "%d %s\n" $NUMBER $DESIG

It doesn't show the decimal potion.

The let VAR=expression is Korn-ish. Substitute with VAR=$(( expression )) for Born-again-ish.

share|improve this answer
This solution introduces a ton of error as the / 1024 always rounds, I am sure that you dont want to round up 1.5 TiB to 2 TiB. –  Geoffrey Aug 6 at 2:43

The answer to your question is yes.

While the output format isn't exactly to your specification, the conversion itself is easily done by a very standard tool (or two). The ones to which I refer are dc and bc. You can get a segmented report by altering their output radices. Like this:

{   echo 1024 o           #set dc's output radix
    echo 1023 pc          #echo a number then print + clear commands
    echo 1024 pc
    echo 1025 pc
    echo 8000000 pc
} | dc

...which prints...

 1023                    #1 field 1023 bytes
 0001 0000               #2 fields 1k 0b
 0001 0001               #2 fields 1k 1b
 0007 0644 0512          #3 fields 7m 644k 512b or 7.64m

I use dc above because it is a personal favorite, but bc can do the same with different syntax and adheres to the same format rules as specified by POSIX like:

  • bc obase

    • For bases greater than 16, each digit shall be written as a separate multi-digit decimal number. Each digit except the most significant fractional digit shall be preceded by a single space. For bases from 17 to 100, bc shall write two-digit decimal numbers; for bases from 101 to 1000, three-digit decimal strings, and so on. For example, the decimal number 1024 in base 25 would be written as:

    01 15 24

    and in base 125, as:

    008 024

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.