How to use column with tput colors?

Question

Using tput colors seems to disturb the output of column. Why does this happen and more importantly is there a way around it?

$ {
  echo "$(tput bold)foo$(tput sgr0)" "$(tput bold)bar$(tput sgr0)"
  echo bar foo
} | column -t
foo  bar
bar                     foo

I have reproduced this issue on both macos and ubuntu 20.04 although it seems they both use BSD column

Answer 1

You can pipe to cat -v to see all the characters of this file stream. column calculates the width using them:

^[[1mfoo^[(B^[[m ^[[1mbar^[(B^[[m
bar foo

---

The right way is to separate any content creation from any formatting (like the MVC model). Note that column is adding spaces to produce a tab-like output, so it modifies the content. While tput adds instructions to the terminal. By separating these tasks, we are able to save the output to a file, or pass it from our formatting function to send it to console, or format it for other destination.

---

But this is also a good exercise for awk, to implement column -t excluding some characters from the width calculations. The regular expression I use below catches a lot of tput stuff, but I am almost sure not all of them.

$ cat tst.awk
BEGIN {
    OFS = "  "
}

{
    nr = NR
    for (i=1;i<=NF;i++) {
        f[NR,i] = $i
        gsub(/\x1B[^[:alpha:]]+[[:alpha:]]/,"",$i)
        if (length($i) > col[i]) col[i] = length($i)
    }
}

END {
    for (i=1;i<=nr;i++) {
        for (j=1;j<=length(col);j++) {
            printf "%-"col[j]"s%s", f[i,j], (j==length(col)? ORS : OFS)
        }
    }
}

and you pipe the first output to awk -f tst.awk

Answer 2

$ {
    echo -e "$(tput bold)foo$(tput sgr0)" "$(tput bold)bar$(tput sgr0) baz";
    echo bar foo;
  } |cat -v
^[[1mfoo^[(B^[[m ^[[1mbar^[(B^[[m baz
bar foo

The problem is on tput sgr0 that as you know by default tputs uses escapes codes based on the Terminal type taken from the environment variable TERM.

if you do:

infocmp -1 |grep -E 'bold|sgr0'
    bold=\E[1m,
    sgr0=\E(B\E[m,

You will see what bold and sgr0 codes are (those are non-ANSI escapes values) which as you see in output after cat -v those converted to ^[[1m for tputs bold and ^[(B^[[m for sgr0 and column command do the same and resolves to ANSI escape type codes but fails when there is (B and left in ^[(B^[[m and then that doesn't recognized in ANSI escapes and caused final output not expected column-wise correctly.

if you do:

infocmp -1 -T ansi |grep -E 'bold|sgr0'
        bold=\E[1m,
        sgr0=\E[0;10m,

you will find the equivalent ANSI escape codes for each; so you have two options to resolve it:

1. Replace tputs with ANSI escape codes above

   { 
     echo -e "\033[1mfoo\033[0;10m" "\033[1mbar\033[0;10m baz";
     echo bar foo;
   } |column -t
   

2. Pass Terminal type to tputs:

   {
     echo -e "$(tput bold)foo$(tput -T ansi sgr0)" "$(tput bold)bar$(tput -T ansi sgr0) baz";
     echo bar foo;
   } |column -t
   

Answer 3

I was running Ubuntu 18.04 and was puzzled by behaviour of column not acting as expected. As others have suggested, the problem is that the repo contains an antique.

The only way I found to obtain a modern version was by compiling from the util-linux repo.

It was difficult to figure this out because -V to obtain the version had not yet been added! There might be a more elegant way but I inferred the approximate version by calling up man column which has a date at the bottom "July 29, 2004". Compare the 2004 manpage to the 2019 manpage. While the former describes only four options, the latter has almost two dozen!

I can also say that on MacOS Catalina I get exactly the same thing. I was unable to locate a brew package that contains a more recent version of column but it might be hidden somewhere.

If you are unable to obtain access to a modern version, you should make sure to check the man page corresponding to whatever is present otherwise things will be frustrating!