I've come upon this sed command and I cannot figure out what it is doing. I understand that it is changing a file in place using -i, that it is using a script -e and that the script is $a\, but what is this script doing?

sed -i -e '$a\' filename
  • Looks like it's doing the same thing as GNU awk '1' file - accept an input file with a terminating newline (valid POSIX text file) or without a terminating newline (invalid POSIX text file) and produce an output file that does have a terminating newline (i.e. is a POSIX text file) whether it was present in the input or not.
    – Ed Morton
    Dec 14, 2021 at 16:23

2 Answers 2


As others have said, e.g. in How to add a newline to the end of a file?, with GNU sed (and some other implementations), $a\ adds a newline to the end of a file if it doesn’t have one.

Why it does this isn’t so clear, and the documentation doesn’t explain it. However, examining the source code does...

Let’s start with the documentation. $a\ is a variant of a, exploiting a special case of a GNU extension:

As a GNU extension, the a command and text can be separated into two -e parameters, enabling easier scripting:

$ seq 3 | sed -e '2a\' -e hello

$ sed -e '2a\' -e "$VAR"

The way a is implemented in sed is with a queue of text to append, tracked in an append_queue. When the time comes to process this queue, in a function called dump_append_queue, the first step is

output_missing_newline (&output_file);

which adds a missing newline if necessary — to ensure that the appended text will be added to separate lines, not to the end of the current line. Then the contents of the queue are appended.

With sed -i '$a\', the missing newline is added if necessary, and then the queue is processed — but the queue is empty, so no additional change is made.


That adds a newline to the last line, if it is absent:

Create a file with no trailing newline

$ printf 'line1\nline2' > file
$ od -c file
0000000   l   i   n   e   1  \n   l   i   n   e   2

Let's see what that sed does:

$ sed -i '$a\' file
$ od -c file
0000000   l   i   n   e   1  \n   l   i   n   e   2  \n

What happens if we run it again? Does it add another newline?

$ sed -i '$a\' file
$ od -c file
0000000   l   i   n   e   1  \n   l   i   n   e   2  \n


  • Could you also explain why it works and especially why it doesn't add a second newline if one is present?
    – terdon
    Dec 14, 2021 at 16:07
  • Not really. The a command is documented at gnu.org/software/sed/manual/html_node/Other-Commands.html, but it does not particularly illuminate this edge case. Dec 14, 2021 at 16:11
  • 1
    Strictly speaking, that's probably undefined in the same way as other similar ones are: the POSIX text says for sed that "The input files shall be text files.", and text files are supposed to have the final newline. (also "...many utilities only produce predictable or meaningful output when operating on text files.", which gives a general cop-out.). But it does seem to work with GNU, Busybox and the *BSD sed on my Mac (though that last one adds the newline with about any program.)
    – ilkkachu
    Dec 14, 2021 at 16:26
  • Some implementations add a newline to the last line, if it is absent (I need to stress this out since not all systems are GNU or Macs). For example, busybox sed '$a\' file always adds a newline to the file, at least on version 1.34.1 from 2021-11-19.
    – Quasímodo
    Dec 14, 2021 at 17:21
  • @Quasímodo, of course not all systems are GNU or Macs, those are just the ones I had available for testing. If anyone wants to test with something else, the results could be useful. Anyway, like I said, it appears undefined, as anything with sed when given a non-text file as input, so yes, it would have to depend on the implementation.
    – ilkkachu
    Dec 14, 2021 at 19:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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