col produces incorrect output

Question

I'm trying to post-process the output of script into a more readable form, similar to Removing control chars (including console codes / colours) from script output, but I've noticed that col doesn't always work.

For instance,

$ cat -v uncolored 
foo^H^H^Hbfoo^H^H^Hafoo^H^H^Hr^M
$ col -bp < uncolored
baroo

Why doesn't col -bp output just bar? Where are the extra two os coming from?

Is your terminal's backspace character ^M or ^?? – DopeGhoti Jun 11 '18 at 22:40 — DopeGhoti, Jun 11 '18 at 22:40
The backspace character is ^H. – Jeffrey Jun 11 '18 at 23:49 — Jeffrey, Jun 11 '18 at 23:49

score 3 · Accepted Answer · edited Jun 17 '18 at 12:14

3

^H in this case is backspace, AKA dec/hex 8 or oct 10 or \b. All it is doing is moving the cursor; take this example:

$ printf 'bravo\10\10X'
braXo

We have moved the cursor back 2, but we only wrote over one letter, the v. We didn't write over the o, so it remains. If you want to get rid of the rest of the letters, you have to overwrite them with something, usually a space character:

$ printf 'bravo\10\10X '
braX

http://wikipedia.org/wiki/Backspace#%5eH

edited Jun 17 '18 at 12:14

Jeff Schaller

67,283
35
116
255

answered Jun 11 '18 at 22:47

Zombo

1
5
44
63

Ah, the uncolor script from the other post removes all the ^[[K controls which would've cleared the rest of the line, except col doesn't understand those either. – Jeffrey Jun 11 '18 at 23:48

score 1 · Answer 2 · answered Jun 11 '18 at 23:53

1

Here's a hacky workaround:

sed -re ':b; s,[^\x08]\x08,,g; tb'

:b: Label b
s,[^\x08]\x08,,g: Pair a non-backspace character with a backspace character and remove both
tb: If the previous s directive did something, jump back to label b

answered Jun 11 '18 at 23:53

Jeffrey

133
4

A better workaround is https://github.com/RadixSeven/typescript2txt, which actually emulates a terminal and processes backspaces and ANSI escape sequences. – Jeffrey Jun 13 '18 at 03:11

col produces incorrect output

2 Answers2

Linked