I typically use column to convert input into a table, eg:

$ echo 'a\tb\tc\nd\te\tf' | column -t -s $'\t'
a  b  c
d  e  f

However it collapses empty columns eg:

$ echo 'a\tb\tc\nd\t\tf' | column -t -s $'\t'
a  b  c
d  f

Rather than printing an empty column when there are consecutive delimiters. This is what I would like, using column or otherwise:

a  b  c
d     f
  • 1
    Instead of echo try printf – Romeo Ninov Jan 23 '19 at 11:53
  • 3
    The questioner is clearly using a shell where -e is implied. We have a whole family of duplicate Q&As about that, starting from unix.stackexchange.com/questions/65803 . But that is not the focus of this question, which is about the column command. Imagine that the input of the column command is whatever is necessary to yield TAB-delimited fields within LF-delimited records, some of which are empty. – JdeBP Jan 23 '19 at 12:14

If you use GNU column:

By default, the column command will merge multiple adjacent delimiters into a single delimiter when using the -t option; this option disables that behavior. This option is a Debian GNU/Linux extension.

printf 'a\tb\tc\nd\t\tf\n'  | column -t -n -s $'\t'


a  b  c
d     f

If GNU column is not available, you can use sed to add a space (or something else, e.g. a -) between the tabs:

printf 'a\tb\tc\nd\t\tf\n'  | sed -e ':loop; s/\t\t/\t-\t/; t loop' | column -t -s $'\t'
| improve this answer | |
  • Thanks for the sed suggestion! I'm using Mac OS X so don't have GNU column or GNU sed, so this is what ended up working for me: printf 'a\tb\tc\nd\t\tf\n' | sed -$'s/\t\t/\t-\t/g' | column -t -s $'\t' – tekumara Jan 24 '19 at 3:21

For those without GNU column (e.g. MacOS): be careful about first and last column. If the first column is not the longest, there would be a shift. The loop in the script didn't seem to be needed. I suggest the following pipe: | sed $'s/^\t/-\t/;s/\t\t/\t-\t/g;s/\t$/\t-/' | column -t

paste \
<(printf "a\nb\nc") \
<(printf "1\n2\n3\n4444\n5\n6") \
<(printf "d\ne\nf\ng\nh\ni\nj\nijkl\nk\nl") \
<(printf "X") \
| sed $'s/^\t/-\t/;s/\t\t/\t-\t/g;s/\t$/\t-/' | column -t
a  1     d     X
b  2     e     -
c  3     f     -
-  4444  g     -
-  5     h     -
-  6     i     -
-  -     j     -
-  -     ijkl  -
-  -     k     -
-  -     l     -
| improve this answer | |
  • (1) We prefer each answer to be self-contained. Your post does provide enough new information that I support keeping it as a separate answer. However, it does talk about the other answer; you should explicitly say, from the start, that you are critiquing another answer. (2) I don’t understand what you’re saying about “If the first column is not the longest”; can you explain that more clearly? (3) Unless I’m overlooking something, your answer fails when there are three tabs in a row. That’s why the other answer used a loop. – Scott Jun 28 '19 at 15:33

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.