Problem using `sed` to colorize `ls` output (Tenth back-reference `\10` not working)

Question

I had asked this question a while back: Is it possible to colorize permissions-part of the output of ls?

While the accepted answer in that question is OK, I now want to do the same thing without using any "external" language. I came up with the following using a bash function and sed.

The question is can you help me fix one bug and also help me make this code "better" ?

• Make better: I am repeating \([r-]\)\([w-]\)\([x-]\)\ three times. How to make this more succint?

• Fix bug: I am unable to get a back-reference to the tenth match \10. Does sed only support up to 9 back-references?

  

The relevant portion of the code (broken into multiple lines for purposes of better readability when posting in this question):

ls -la | sed "s
/^
\([d-]\)
\([r-]\)
\([w-]\)
\([x-]\)
\([r-]\)
\([w-]\)
\([x-]\)
\([r-]\)
\([w-]\)
\([x-]\)
/
`echo -e \"
$EMG\1
$EMY\2\3\4
$EMR\5\6\7
$EMC\8\9\10
$NONE\"`/"

For reference, the complete function I am using is:

domagic() {
  local NONE="\033[0m"    # unsets color to term's fg color

  # regular colors
  local K="\033[0;30m"    # black
  local R="\033[0;31m"    # red
  local G="\033[0;32m"    # green
  local Y="\033[0;33m"    # yellow
  local B="\033[0;34m"    # blue
  local M="\033[0;35m"    # magenta
  local C="\033[0;36m"    # cyan
  local W="\033[0;37m"    # white

  # emphasized (bolded) colors
  local EMK="\033[1;30m"
  local EMR="\033[1;31m"
  local EMG="\033[1;32m"
  local EMY="\033[1;33m"
  local EMB="\033[1;34m"
  local EMM="\033[1;35m"
  local EMC="\033[1;36m"
  local EMW="\033[1;37m"

  # background colors
  local BGK="\033[40m"
  local BGR="\033[41m"
  local BGG="\033[42m"
  local BGY="\033[43m"
  local BGB="\033[44m"
  local BGM="\033[45m"
  local BGC="\033[46m"
  local BGW="\033[47m"

  #do magic here
  ls -la | sed "s/^\([d-]\)\([r-]\)\([w-]\)\([x-]\)\([r-]\)\([w-]\)\([x-]\)\([r-]\)\([w-]\)\([x-]\)/`echo -e \"$EMG\1$EMY\2\3\4$EMR\5\6\7$EMC\8\9\10$NONE\"`/"
}
domagic
#to run: save as domagic.sh and run with: bash domagic.sh

Answer 1

Sed supports up to 9 backreferences, but observing that \2\3\4 go always together, as do \5\6\7 and \8\9\10, you can use only four groups, like in the following:

"s/^\([d-]\)\([r-][w-][x-]\)\([r-][w-][x-]\)\([r-][w-][x-]\)/`echo -e \"$EMG\1$EMY\2$EMR\3$EMC\4$NONE\"`/""

Answer 2

You're calling sed from bash, so you can use bash to avoid repeating the character classes. Everything inside double quotes gets expanded:

rwx='\([r-]\)\([w-]\)\([x-]\)'
ls -la | sed "s/^\([d-]\)$rwx$rwx$rwx/`echo -e \"$EMG\1$EMY\2\3\4$EMR\5\6\7$EMC\8\9\10$NONE\"`/"

But I don't think looking for drwx is quite right. What if the file is a symlink? That will start with l. And if it's a "sticky" directory (e.g. /tmp), that will end with t instead of x, etc.

Why not just

ls -la | sed "s/^\(.\)\(...\)\(...\)\(...\)/`echo -e \"$EMG\1$EMY\2$EMR\3$EMC\4$NONE\"`/""

Finally, if you change how you declare your colors, you don't need the echo -e:

domagic() {
  local NONE=$'\033[0m'    # unsets color to term's fg color

  # regular colors
  local K=$'\033[0;30m'    # black
  local R=$'\033[0;31m'    # red
  local G=$'\033[0;32m'    # green
  local Y=$'\033[0;33m'    # yellow
  local B=$'\033[0;34m'    # blue
  local M=$'\033[0;35m'    # magenta
  local C=$'\033[0;36m'    # cyan
  local W=$'\033[0;37m'    # white

  # emphasized (bolded) colors
  local EMK=$'\033[1;30m'
  local EMR=$'\033[1;31m'
  local EMG=$'\033[1;32m'
  local EMY=$'\033[1;33m'
  local EMB=$'\033[1;34m'
  local EMM=$'\033[1;35m'
  local EMC=$'\033[1;36m'
  local EMW=$'\033[1;37m'

  # background colors
  local BGK=$'\033[40m'
  local BGR=$'\033[41m'
  local BGG=$'\033[42m'
  local BGY=$'\033[43m'
  local BGB=$'\033[44m'
  local BGM=$'\033[45m'
  local BGC=$'\033[46m'
  local BGW=$'\033[47m'

  #do magic here
  ls -la | sed "s/^\(.\)\(...\)\(...\)\(...\)/$EMG\1$EMY\2$EMR\3$EMC\4$NONE/"
}