Colored grep output with highlighted pattern

Question

I do have the following find-file script:

find . -type f -iname "*$1*" -printf '%P\n' \
| GREP_COLORS="sl=0;33;49:ms=1;34;49" grep --color=always '^\|[^/]*$' \
| grep -i --color=auto "$1"

which does:

• colorize the path part
• colorize the pattern in the base name part

Problem: after the pattern, the colour is reset

For the base name part, it could be entirely white, that'd solve the problem (though I did not find the way to change this).

For the path part, it'd stay a problem, as you can see in the above screenshot: the colour is reset, hence we don't see anymore where the file path really ends!

Is there a solution to this?

EDIT -- There were 3 solutions to this!! All below proposals do work, they have light differences, but they do answer the question. I choose the one which highlights all occurrences of the pattern, and is almost a one-line, but honestly choosing was tough, as they're quite equivalent…

EDIT -- One wished improvement to this: that the find results don't get blocked until analysis, that is that the output gets flushed and processed line per line. Is this possible?

Answer 1

maybe something like this?

command:

pattern='oRg'; find . -type f -iname "*$pattern*" -printf '%P\n' \
| GREP_COLORS="sl=0;33:mt=1;34" grep --color=always '[^/]*$' \
| GREP_COLORS="sl=1;34" grep --color=always -iP "$pattern(?=[^/]*$)" \
| GREP_COLORS="sl=0;33" grep -i "$pattern" --color

remove the last line if you don't want highlight the pattern in dirname part.

see grep's Environment Variable GREP_COLORS section for details.

Answer 2

You could do it with zsh's builtin glob operators. That would have several benefits:

• fix this issue easily
• work with pathnames that contain newline characters
• make it easy to highlight the pattern in the basename only
• work with wildcards (in your approach, find and grep interpret the pattern differently)
• get you a sorted list
• work even on non-GNU systems (-printf, -iname, --color are all non-standard extensions).

Maybe something like:

#! /bin/zsh -
pattern="(#i)${1?Please specify a pattern}"

set -o extendedglob

typeset -A find_file_color
find_file_color=(
  dirname  $'\e[0;33;49m'
  basename $'\e[1;34;49m'
  match    $'\e[1;33;44m'
  reset    $'\e[m'
)

colorize_file() {
  local file=${1-$REPLY}
  case $file in
    (*/*)
      REPLY=$find_file_color[dirname]$file:h$find_file_color[reset]/;;
    (*)
      REPLY=
  esac
  REPLY+=$find_file_color[basename]${${file:t}//(#m)$~pattern/$find_file_color[match]$MATCH$find_file_color[basename]}$find_file_color[reset]
}

print -rC1 -- **/*$~pattern*(ND.+colorize_file)

Note that it builds and sorts the whole list before passing to print for printing. So you'll only start getting some output once all the files have been found. To print them as they are found (but then we'd need to give up on sorting), you could have the glob qualifier function print the colorized file instead:

#! /bin/zsh -
pattern="(#i)${1?Please specify a pattern}"

set -o extendedglob

typeset -A find_file_color
find_file_color=(
  dirname  $'\e[0;33;49m'
  basename $'\e[1;34;49m'
  match    $'\e[1;33;44m'
  reset    $'\e[m'
)

colorize_file() {
  local file=${1-$REPLY}
  case $file in
    (*/*)
      REPLY=$find_file_color[dirname]$file:h$find_file_color[reset]/;;
    (*)
      REPLY=
  esac
  REPLY+=$find_file_color[basename]${${file:t}//(#m)$~pattern/$find_file_color[match]$MATCH$find_file_color[basename]}$find_file_color[reset]
  print -r -- $REPLY
  false # don't bother adding the file to the glob expansion
}

: **/*$~pattern*(ND.+colorize_file)

Answer 3

1) a example how to colorize text

# creation of 3 escape sequences
C_1=$(echo -ne "\033[30;47m" ) ;
C_2=$(echo -ne "\033[32;41m" ) ;
C_3=$(echo -ne "\033[37;44m" ) ;
C_N=$(echo -ne "\033[0m"  )  ;

#simple example for using it
echo $C_1 Hello in C_1 $C_N normal ;
echo $C_2 Hello in C_2 $C_N normal ;
echo $C_3 Hello in C_3 $C_N normal ;

Explanation of values are here : https://en.wikipedia.org/wiki/ANSI_escape_code#3/4_bit

In your case because you are looking for simple pattern i can use only one sed command .

PAT="tot" ; 
find ./ -iname  \*$PAT\*  -printf '%P\n' | \
    sed "s/^\(.*\/\)\{0,1\}\([^\/]*\)\($PAT\)\([^\/]*\)\$/${C_1}\1${C_2}\2${C_3}\3${C_N}\4/i"

\1 is the full path ending with the last /
\2 is the pattern of the filename before the pattern
\3 is the pattern
\4 is after the pattern