Convince grep to output all lines, not just those with matches

You can try:

perl -MTERM::ANSIColor -nle '/pattern/ ? print colored($_, 'color') : print' test

Not very portable however, and even if Perl is installed, you may need to download another module. In addition it will color the entire line, not just the search word.

ack --passthru --color string file

for Ubuntu and Debian, use ack-grep instead of ack

ack-grep --passthru --color string file

I have the following function that I use for such things:

highlight () {
    perl -pe "s/$1/e[1;31;43m$&e[0m/g"
}

Internally it looks kind of ugly, it’s nice and easy to use, like so:

cat some_file.txt | highlight some_word

or, for a slightly more real-world example:

tail -f console.log | highlight ERROR

You can change the colors to anything you like (which might be hard with grep–I’m not sure) by changing the 1 and 31 and 43 (after e[) to different values. The codes to use are all over the place, but here’s a quick intro: the 1 bolds the text, the 31 makes it red, and the 43 gives a yellow background. 32 or 33 would be different colors, and 44 or 45 would be different backgrounds: you get the idea. You can even make it blink (with a 5) if you’re so inclined.

This doesn’t use any special Perl modules, and Perl is nearly ubiquitous, so I would expect it to work just about anywhere. The grep solution is very clever, but the –color switch on grep is not available everywhere. For instance, I just tried this solution on a Solaris box running bash, and another running ksh, and my local Mac OS X machine running zsh. All worked just fine. Solaris choked on the grep --color solution, however.

Also, ack is awesome, and I recommend it to anyone who hasn’t yet discovered it, but I’ve had some issues installing it on a few of the many servers I work on. (I forget why: I think related to Perl modules it required.)

Since I don’t think I’ve ever worked on a Unix box that didn’t have Perl installed (with the exception of embedded-type systems, Linksys routers, and such) I’d say this is pretty much a universally useable solution.

There’s a much easier way to do this for GNU grep but I don’t think it’s portable (i.e., BSD grep):

In a pipe:

cat <file> | grep --color=always -z <query>

On a file:

grep --color=always -z <query> <file>

Credit goes to Cyrus’s answer here.

A sed version,
works on both bash and ash.

#highlight
hsed(){
    local pattern="$1"
    shift
    local r=`echo -e 'e'[31m`
    local c=`echo -e 'e'[0m`
    sed "s:${pattern//:/:}:$r$c:g" "$@"
}

Another way to do this properly and portably with grep (besides using two regexes with alternation as in the accepted answer) is via the null pattern (and respectively null string).
It should work equally well with both -E and -F switches since, per the standard:

-E
    Match using extended regular expressions. 
    [...] A null ERE shall match every line.

and

-F
    Match using fixed strings.
    [...] A null string shall match every line.

So it’s simply a matter of running

grep -E -e '' -e 'pattern' infile

and respectively

grep -F -e '' -e 'string' infile

OP asked for grep, and that is what I RECOMMEND; but after trying hard to solve a problem with sed, for the record, here is a simple solution with it:

sed $'s/main/E[31m&E[0m/g' testt.c

or

cat testt.c | sed $'s/main/E[31m&E[0m/g'

Will paint main in red.

• E[31m : red color start sequence
• E[0m : finished color mark
• & : the matched pattern
• /g : all words in a line, not just the first
• $'string' is bash strings with escaped characters interpreted

Regarding grep, it also works using ^ (begin of line) instead of $ (end of line). Example:

egrep "^|main" testt.c

And just to show this crazy alias that I DO NOT RECOMMEND, you can even let the open quotes:

alias h='egrep -e"^|'
h main" testt.c
cat testt.c | h main"

all work! 🙂 Don’t worry if you forget to close the quote, bash will remember you with a “continuing line character”.

None of the answers given so far do provide a portable solution.

Here is a portable¹ shell function I already posted in a closed as duplicate question that doesn’t require non standard tools or non standard extensions provided with perl, ack, ggrep, gsed, bash and the likes but only needs a POSIX shell and the POSIX mandatory utilities sed and printf:

grepc()
{
  pattern=$1
  shift
  esc=$(printf "33")
  sed 's"'"$pattern"'"'$esc'[32m&'$esc'[0m"g' "$@"
}

You can use it that way:

grepc string_to_search [file ...]

The highlight color can be adjusted by using one of these codes in the sed command argument (32m being green here):

30m black
31m red
33m yellow
34m blue
35m magenta
36m cyan
37m white
7m reverse video

¹ As long as your terminal supports ANSI colors escape sequences.

Instead of using Grep, you can use Less:

less file

Search like this: /pattern Enter

This will:

1. scroll to first matching line
2. output every line starting from there on
3. highlight all matches

To scroll to next matching line: n

To scroll to previous matching line: N

To toggle highlighting: Esc u

Also you can change the highlighting color if you like.

ripgrep

Use ripgrep with its --passthru parameter:

rg --passthru pattern file.txt

It’s one of the fastest grepping tools, since it’s built on top of Rust’s regex engine which uses finite automata, SIMD and aggressive literal optimizations to make searching very fast.

--passthru – Print both matching and non-matching lines.

Another way to achieve a similar effect is by modifying your pattern to match the empty string. For example, if you are searching using rg foo then using rg "^|foo" instead will emit every line in every file searched, but only occurrences of foo will be highlighted. This flag enables the same behavior without needing to modify the pattern.

If you’re interested in the return value of grep, then GNU sed is also the way:

sed '/pattern/h; ${p;x;/./Q0;Q1}'

You can add to the script modifications to the line, e.g.:

sed '/pattern/h; ${p;x;/./Q0;Q1}; s/pattern/<>/g'

This checks that pattern appears on the current line, and stores it in the hold buffer if it is. At the end of the file ($), the last line is printed (p), then the hold space and pattern space are swapped (x). Next, the pattern space is tested for emptiness (/./), and if it is nonempty, exit with code 0 (Q0) otherwise with code 1 (Q1).

An alternative is /pattern/{:a $q1; n; ba}, which loops after detecting pattern, but changing the current line can get tricky with this. This works by:

• /pattern/: the following clause will only be executed when the pattern is detected on a line.
• :a … ba: this is defining a label a and branching to it; as is, this is an infinite loop.
• n: prints the current line and loads the next one.
• $q1: when reaching the end of file ($), exit with code 1.

By default, sed will exit with code 0, so with this expression, if the pattern is detected, the rest of the file is dumped and an exit code of 1 is produced.

Similar to the perl answer, but I use it with multiple expressions separated by |:

tail -F somefile.log | perl -pe 's/bfirstwordb/e[1;31m$&e[0m/g|s/bsecondwordb/e[1;32m$&e[0m/g'

You and add more words by append multiple | (i.e., |s/bmorewordsb/e[1;32m$&e[0m/g) and on What color codes can I use in my bash PS1 prompt? you can see more bash the colors to use.