I am developping a script that will allow me to find easily which columns are "only spaces", by creating a mask showing any column with anything other than a space.

To do this: I print each line "on top of each others" + modify spaces to "right-arrow". (an additional step would also be to ignore spaces seen after a "column farther than the beginning of the last title", but it is out of topic here)

I have trouble with the last step: how to get the resulting string (the output my terminal displays correctly), without every \r, \n, and <Esc>[C (= right arrow) that were used to create/interpret it.


$ PS2=""
$ Esc=$( printf '\033' )
$ Right=$( printf "${Esc}[C" )
$ ps | head -n 2
  PID TTY          TIME CMD
12415 pts/1160 00:00:00 bash
$ ps | head -n 2 | LC_ALL=C tr '!-~' '*'
  *** ***          **** ***
***** ******** ******** ****
$ ps | head -n 2 | LC_ALL=C tr '!-~' '*' | sed -e "s/ /${Right}/g" | while read line; do
    printf "%s\r" "$line"
  done ; printf '\n'
***** ******** ******** ****
  ## the line above is what I am looking for: 
  ## a "mask" of each column that at one point had a non-space character displayed in it
$ ( ps | head -n 2 | tr '!-~' '*' | sed -e "s/ /${Right}/g" | while read line; do
      printf "%s\r" "$line"
    done ; printf '\n'
  ) | cat -ve
  ## but of course, the terminal sees (and outputs) this instead, which contains all the "contructing characters"

How could I retrieve in a variable the string "as shown by my terminal", ie:

***** ******** ******** ****

(I tried to output that whole thing into a "while read -e finalstring; do print '[%s]' "$finalstring" ; done , but that finalstring still contains the whole "pre-interpretation" string, not the one "post-terminal interpretation")

A shorted way to ask the same question:

# if:
$ printf "aaaa\rbb\rc\n"
# displays:
$ var=$( printf "aaaa\rbb\rc\n" ) # $var will be: "aaaa\rbb\rc\n"
# how can I instead have the resulting "displayed string": "cbaa" in $var ?
  • Please be aware that the exemple shows as the final output the same thing as the last line of the input: this is just a coincidence... the final output is the accumulation of every non-space characters encountered on any line. The input I will use my script on will be several hundreds lines long, and I need the final output to show which columns only had spaces in them overall. Jul 29, 2022 at 13:15
  • You'd need a terminal emulator to interpret those sequences, several terminal emulators including screen, tmux or xterm can be told to dump the contents of their screen programmatically. Jul 29, 2022 at 15:58
  • Thank you @StéphaneChazelas for both comments, I will have a look at those. (I did a broad and long research before posting my question, and missed that one) Jul 30, 2022 at 9:08

2 Answers 2


It looks like you are just making an extra problem by adding terminal escapes to data that has none to begin with. If all you want is to visualize the blank columns from ps, you can use awk (which it seems you know) to simply accummulate an array indexed by column number, which is all blanks and gets a "*" where necessary. Eg:

awk '
{ len = length($0)
  if(len>max)max = len
  for(i=1;i<=len;i++) if(substr($0,i,1)!=" ") col[i] = "*"
END{ for(i=1;i<=max;i++) if(col[i])str = str "*"; else str = str " ";
     print str
  • You are right. I am trying to use something already there, and not go char by char on a potentially very large input. I am not sure which would be more efficient. I will probably go with your approach at first, but the question (how to get the displayed string by the terminal, instead of the construction of it) remains and is the one I am seeking an answer to. Jul 30, 2022 at 9:06

A simple ansi parser I used many years ago is the one provided with the Python module pexpect. (Sadly, this is now deprecated in favour of the pyte module, with which I am not familiar.)

This old module still seems to be provided as part of the package. You send it a stream of data and it writes the characters in the appropriate place in a 40x80 array, taking into account the ansi escape sequences.

The following example hooks into the scroll-up action and keeps a one-line array result of 80 characters with a "*" placed wherever a non-space character was scrolled up out of the screen. The str() call on the screen is changed to scroll-up the whole screen to flush it, and then just return this array. Just pipe your data into this program. This is just a start, as you might also want to supress updating the array when spaces overwrite non-space characters, and so on.

# https://unix.stackexchange.com/q/711733/119298
# extend ANSI by noting which cols get non-whitespace
import sys, re
from pexpect.ANSI import ANSI

class KeepScroll(ANSI):
    def __init__(self, r=24,c=80):
        ANSI.Log = None
        self.result = [" "] * self.cols

    def scroll_up(self):
        if self.scroll_row_start==1:
            for i, ch in enumerate(self.w[0]):
                if ch!=" ": self.result[i] = "*"

    # scroll up rest of screen
    def __str__(self):
        for i in range(self.rows): self.scroll_up()
        return "".join(self.result)

def parsevt100(data):
    scr = KeepScroll()
    return str(scr)

if __name__ == "__main__":

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .