If I have really long output from a command (single line) but I know I only want the first [x] (let's say 8) characters of the output, what's the easiest way to get that? There aren't any delimiters.


7 Answers 7


One way is to use cut:

 command | cut -c1-8

This will give you the first 8 characters of each line of output. Since cut is part of POSIX, it is likely to be on most Unices.

  • 7
    Note that cut -c selects characters; cut -b or head -c selects bytes. This makes a difference in some locales (in practice, when using UTF-8). Oct 24, 2010 at 22:07
  • 1
    You also don't have to specify the start index in this case. Saying cut -c-8 will select from character 1 to 8.
    – Sparhawk
    May 9, 2014 at 5:08
  • @Steven, cut's equivalent on Windows is?
    – Pacerier
    Aug 25, 2015 at 13:06
  • Also command | dd bs=8 count=1 2>/dev/null. Not saying it's shorter or superior. Just another alternative.
    – dubiousjim
    Sep 24, 2015 at 3:50
  • 1
    @Gilles, but note that with current versions of GNU cut, cut -c works like cut -b (that is, it doesn't work properly for multi-byte characters). Aug 9, 2016 at 13:49

These are some other ways to get only first 8 characters.

command | head -c8

command | awk '{print substr($0,1,8);exit}' 

command | sed 's/^\(........\).*/\1/;q'

And if you have bash

echo ${var:0:8}
  • 3
    I think the following sed formulation is a bit easier to read: command | sed 's/\(.\{8\}\).*/\1/' or if your sed supports it: command | sed -r 's/(.{8}).*/\1/'; Otherwise, +1
    – Steven D
    Oct 24, 2010 at 4:48
  • 2
    Good stuff, but note that head -c counts bytes, not characters. Similarly, among the major Awk implementations, only GNU awk handles multi-byte characters correctly - FreeBSD Awk and Mawk do not.
    – mklement0
    Jul 5, 2015 at 17:30

Another one liner solution by using parameter expansion

echo ${word:0:x}

EG: word="Hello world"
echo ${word:0:3} or echo ${word::3} 
o/p: Hel

EG.2: word="Hello world"
echo ${word:1:3}
o/p: ell
  • 1
    You can also use a variable holding the length, e.g.: x=8; echo ${word:0:$x} instead of hard-coding the integer.
    – Cometsong
    Apr 25, 2019 at 14:58
  • worth noting this will not be possible in ksh88, only 93 Apr 29, 2020 at 1:45
  • 1
    @Cometsong Testing with the Bash shell that came with "Git for Windows", it looks like you don't need to prefix x with the $ sign in this case: x=8; echo ${word:0:x} will work the same.
    – AJM
    Mar 26, 2021 at 11:10
  • The source you're linking to does not describe the construct you used.
    – zrajm
    Mar 12 at 19:44

If you have a sufficiently advanced shell (for example, the following will work in Bash, not sure about dash), you can do:

read -n8 -d$'\0' -r <(command)

After executing read ... <(command), your characters will be in the shell variable REPLY. Type help read to learn about other options.

Explanation: the -n8 argument to read says that we want up to 8 characters. The -d$'\0' says read until a null, rather than to a newline. This way the read will continue for 8 characters even if one of the earlier characters is a newline (but not if its a null). An alternative to -n8 -d$'\0' is to use -N8, which reads for exactly 8 characters or until the stdin reaches EOF. No delimiter is honored. That probably fits your needs better, but I don't know offhand how many shells have a read that honors -N as opposed to honoring -n and -d. Continuing with the explanation: -r says ignore \-escapes, so that, for example, we treat \\ as two characters, rather than as a single \.

Finally, we do read ... <(command) rather than command | read ... because in the second form, the read is executed in a subshell which is then immediately exited, losing the information you just read.

Another option is to do all your processing inside the subshell. For example:

$ echo abcdefghijklm | { read -n8 -d$'\0' -r; printf "REPLY=<%s>\n" "$REPLY"; }
  • 1
    If you just want to output the 8 chars, and don't need to process them in the shell, then just use cut.
    – dubiousjim
    Sep 8, 2012 at 14:04
  • Good to know about read -n <num>; small caveat: Bash 3.x (still current on OS) mistakenly interprets <num> as a byte count and thus fails with multi-byte characters; this has been fixed in Bash 4.x.
    – mklement0
    Jul 6, 2015 at 1:41
  • This is a great and useful answer. Much more general than the others.
    – not2qubit
    Oct 25, 2019 at 10:08
  • On my git bash, I have the "-N" flag, which reads exactly N chars until EOF or timeout. Isn't that what you try to achieve your "d" flag ? May 3, 2022 at 8:16
  • @Itération122442 yes but as I wrote "I don't know offhand how many shells have a read that honors -N as opposed to honoring -n and -d."
    – dubiousjim
    May 4, 2022 at 9:43

This is portable:

a="$(command)"             # Get the output of the command.
b="????"                   # as many ? as characters are needed.
echo ${a%"${a#${b}}"}      # select that many chars from $a

To build a string of variable length of characters has its own question here.


I had this problem when manually generating checksum files in maven repository. Unfortunately cut -c always prints out a newline at the end of output. To suppress that I use xxd:

command | xxd -l$BYTES | xxd -r

It outputs exactly $BYTES bytes, unless the command's output is shorter, then exactly that output.

  • another method to take off cut's trailing newline is to pip it into: | tr -d '\n'
    – Cometsong
    Apr 25, 2019 at 15:00

How to consider Unicode + UTF-8

Let's do a quick test for those interested in Unicode characters rather than just bytes. Each character of áéíóú (acute accented vowels) is made up of two bytes in UTF-8. With:

printf 'áéíóú' | LC_CTYPE=en_US.UTF-8 awk '{print substr($0,1,3);exit}'
printf 'áéíóú' | LC_CTYPE=C awk '{print substr($0,1,3);exit}'
printf 'áéíóú' | LC_CTYPE=en_US.UTF-8 head -c3
printf 'áéíóú' | LC_CTYPE=C head -c3

we get:


so we see that only awk + LC_CTYPE=en_US.UTF-8 considered the UTF-8 characters. The other approaches took only three bytes. We can confirm that with:

printf 'áéíóú' | LC_CTYPE=C head -c3 | hd

which gives:

00000000  c3 a1 c3                                          |...|

and the c3 by itself is trash, and does not show up on the terminal, so we saw only á.

awk + LC_CTYPE=en_US.UTF-8 actually returns 6 bytes however.

We could also have equivalently tested with:

printf '\xc3\xa1\xc3\xa9\xc3\xad\xc3\xb3\xc3\xba' | LC_CTYPE=en_US.UTF-8 awk '{print substr($0,1,3);exit}'

and if you want a general parameter:

printf 'áéíóú' | LC_CTYPE=en_US.UTF-8 awk "{print substr(\$0,1,$n);exit}"

Question more specific about Unicode + UTF-8: https://superuser.com/questions/450303/unix-tool-to-output-first-n-characters-in-an-utf-8-encoded-file

Related: https://stackoverflow.com/questions/1405611/how-to-extract-the-first-two-characters-of-a-string-in-shell-scripting

Tested on Ubuntu 21.04.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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